Professional Documents
Culture Documents
Combinatorial Chemistry
Drugs, Catalysts, Materials
Edited by K. C. Nicolaou,
R. Hanko, and W. Hartwig
Combinatorial Chemistry
2000, ISBN 3-527-30186-0
G. Jung (Ed.)
Combinatorial Chemistry
Synthesis, Analysis, Screening
F. Zaragoza D
orwald
A. Beck-Sickinger, P. Weber
Contents
Volume 1
Volume 2
Index 1099
xxvii
List of Authors
Part I
General Aspects
1
Combinatorial Chemistry in Perspective
K. C. Nicolaou, R. Hanko, and W. Hartwig
1.1
Introduction
distinct advantages over the latter. First, in solid-phase synthesis, large excesses of
reagents can be used to accelerate reactions and to drive them to completion; these
reagents can then conveniently be removed at the end by ltration and washing.
Furthermore, and because of easy separation, solid-phase chemistry can be auto-
mated through robotics to a higher degree than solution chemistry can. Most im-
portantly, solid-phase chemistry can be applied to the elegant and powerful split-
and-pool synthesis strategy for combinatorial chemistry. Despite its dramatic
contribution to increasing eciency over traditional methods, high-throughput
parallel synthesis remains a laborious task. Thus, combinatorial chemists quickly
recognized the benets of automation as a crucial component of combinatorial
chemistry. Besides the synthetic sequence, purication and characterization of
compounds are also important aspects of combinatorial chemistry. The develop-
ment of high-throughput chromatography methods to support automated parallel
synthesis and of high-performance liquid chromatographymass spectrometry
(HPLC/MS) systems to direct the collection of the desired products are only the
beginning of such high-throughput methods. Progress in this area is moving
quickly and will be indispensable to any serious combinatorial chemistry eort.
1.2
Brief History of Combinatorial Chemistry
1.3
Applications of Combinatorial Chemistry
With the advent of combinatorial chemistry, the traditional one molecule at a time
approach to drug discovery was severely shaken. The initial euphoria of the early
1990s, however, was based to a considerable extent, on faulty grounds. The idea of
synthesizing a myriad of compounds randomly, often as mixtures, sounded like a
dream to many chemists of biotechnology and pharmaceutical companies. Soon
thereafter, however, the principles of combinatorial chemistry for small organic
molecules crystallized on a more pragmatic platform. The prevailing approach
today is that based on both solution-phase and solid-phase chemistry applied in
parallel or split-and-pool formats and directed at discrete and high-purity com-
pounds [25, 26]. Initial-phase combinatorial chemistry is applied to discover lead
compounds rapidly which are then subjected to lead optimization to produce drug
candidates. The last part of the process is the domain of the medicinal chemists,
who may also practice combinatorial strategies to achieve their goals. Thus,
smaller focused libraries are carefully designed and synthesized, either in parallel
or by the split-and-pool strategy using solution- or solid-phase chemistry. Combi-
natorial chemistry has, therefore, penetrated the laboratories of medicinal chem-
ists who recognized its power in delivering the targeted compounds in a much
faster way, and in acceptable quantities and purities. In similar ways, academic
laboratories have adopted and rened combinatorial techniques in their quest for
libraries of compounds needed for chemical biology studies [27].
The capability of combinatorial chemistry to produce large numbers of com-
pounds rapidly is a powerful tool not only for chemical biology and drug discovery
but also for a host of other research endeavors. Indeed, this philosophy and these
combinatorial processes have been successfully applied to reaction optimization,
the discovery of new materials [28] and the development of new catalysts [29].
In pioneering work, Fuchs and coworkers [30] reported in 1984 the use of auto-
mation to optimize reaction conditions with multiple variables. Reddington and
Sapienza [31] reported in 1998 results from a highly parallel, optical screening
method to discover novel electrocatalysts. Such practices are currently gaining
wide popularity in industry for the optimization of process chemistry. The rst
1.4 Outline of the Book 7
1.4
Outline of the Book
The following chapters in this series will address the various aspects of combina-
torial chemistry in order to facilitate further advances in the eld as well as to aid
the reader in his or her practice of combinatorial chemistry.
The book is divided into three parts. The rst chapters (15) serve as the in-
troductory part and build the foundation on which the next chapters are based.
The following part (Chapters 520) deals with basic reaction mechanisms. The
aim is to describe thoroughly the repertoire of combinatorial chemistry in solution
and on solid support and to provide the reader with a critical overview of the best
methods and conditions found so far. As an introduction, each chapter briey dis-
cusses the applicability of the reactions dealt with for library synthesis in general
and for solution-phase or solid-phase chemistry in particular. This entry is fol-
lowed by the detailed reection of each reaction type including the most recent
developments/achievements. Based on mechanistic considerations, emphasis is
put on the suitability of a given reaction for library synthesis using either solid-
phase or solution-phase chemistry.
Generally, synthetic examples from all techniques are presented (solution-phase
parallel synthesis, solid-phase chemistry, solid-supported reagents). In most cases
examples originate from syntheses of large libraries.
The last part of the book (Chapters 2135) describes special topics such as ap-
plications, design and instrumentation as well as aspects beyond pure synthetic
8 1 Combinatorial Chemistry in Perspective
Acknowledgments
This book would not have been possible without the dedicated work of the editing
team, especially Berthold Hinzen and Tobias Wunberg. We would also like to
thank P. Golitz and his team for their conceptual contributions and their support
to ensure the ambitious scheduling of this book.
References
2
Introduction to Combinatorial Chemistry
David L. Coen and Joachim E. A. Luithle
2.1
Combinatorial Chemistry in Drug Discovery a Perspective
The categorical imperatives of modern drug discovery are to produce better clinical
candidates that are less prone to failure at a late-stage, and to do this more rapidly
than the industry performance standards of the past two decades would predict
and at a cost that is responsive to social and political pressures on drug prices.
Combinatorial chemistry is nothing less than a cornerstone technology in the re-
alization of these imperatives. Its function is basically that of a high-output engine,
providing very large numbers of well-designed, well-made, high-quality com-
pounds for high-throughput evaluation as potential drug candidates. In fact, this
ability to leverage productivity has prompted researchers in other elds, such as
catalysis and material science, to adopt and adapt this engine to their needs as
well.
The purpose of this chapter is to provide an overview of the basic principles of
combinatorial chemistry and to outline the operating principles associated with its
most widely practiced forms. Subsequent chapters provide more detailed accounts
of how the various types of synthetic chemistry are utilized, of specic combinato-
rial chemistry technologies, and of specic applications.
By way of background one can compare how a typical medicinal chemist of 1975
would set about making a compound for testing with the way that the same
chemist would approach the task today. First of all, each decision about what com-
pound to make was typically embedded in the program category in which the
chemist was a participant. The program description frequently included a struc-
tural component, such as muscarinic cholinergics, cephalosporin antibiotics,
benzodiazepine anxiolytics, or tricyclics, and the chemist would have had con-
siderable expertise in the synthetic aspects, structureactivity relationships, the lit-
erature and patents, and, to some extent at least, the pharmacology of the particu-
lar compound class. Each new hypothesis was based on a blend of past experience,
recent papers and/or patents in the eld, and on the latest biological testing re-
sults, and was then refracted through knowledge of how to make the compounds
of interest and what was available to make them from. If the synthesis worked, a
15-g sample (huge by todays standards) would be made and rigorously puried
by column chromatography or recrystallization. Meticulous analyses using, as a
minimum, thin-layer chromatography (TLC), mass spectrometry (MS), infrared
(IR), nuclear magnetic resonance (NMR), ultraviolet (UV), and combustion analy-
sis determined if the sample was suitable for registration and biological testing.
A compound data sheet that included analytical results, melting point, solubility
data, and other information would typically be prepared, reviewed, approved, and
led before a sample could be sent for testing. Using this process, a chemist with
one or two assistants was deemed productive if 50200 new compounds a year
were forthcoming, depending on the complexity of the structural class.
In contrast, the drug designer of year 2000 is most likely a hybrid medicinal,
combinatorial, computational, and analytical chemist with a working knowledge
of molecular and cellular biology. He or she spends little time in the library, lots of
time at a computer, and much of the remaining time meeting with the project
team to review results and plan activities. Also, in comparison with the handful
of laboratory appliances considered essential 25 years ago, todays chemistry labo-
ratory is replete with complex (and expensive) productivity-enhancing equipment,
making it possible for the 200 compounds of yesteryear to be made in an auto-
mated synthesizer during the hour spent in a meeting.
The decision as to which compound to make next has been replaced by a design
process regarding which set (or library) of compounds to make next. Routine syn-
thesis of single compounds now occurs mainly in the later stages of lead opti-
mization. The design process ows not from a compound class associated with
a particular drug discovery program, but from the biomolecular targets associated
with the program. This fact underlies the critical relationship between modern
medicinal chemistry and structural biology/computational chemistry. As an illus-
tration, a 1970s program on inuenza drugs might be centered on aminoadaman-
tane derivatives (antiviral compound class), whereas a modern program might be
focused on inhibitors of neuraminidase (viral protein target). The discovery phase
of the program is driven by the target and by knowledge of its structure and natu-
ral ligand in the case of a receptor or by its mechanism of catalysis and natural
substrate in the case of an enzyme.
Overall, it can be stated that the integration of biomolecular target-driven com-
putational design methods, combinatorial techniques for compound production,
and high-throughput biological screening technologies has, in principle, resulted
in a huge increase in productivity in medicinal chemistry. The role of combinato-
rial chemistry in modern pharmaceutical research and development (R&D) is per-
haps best shown in Figs. 2.1 and 2.2, which depict as triads the major components
of both the overall R&D process and its discovery phase.
2.2
Key Issues
The basic principles of combinatorial chemistry are often tied to, and sometimes
confused with, the origins of combinatorial chemistry. The founding (but awed)
12 2 Introduction to Combinatorial Chemistry
Fig. 2.1. The essential elements of modern pharmaceutical research and development.
principle was that, given a suciently large and diverse set of compounds to test,
the discovery of an ideal drug for any given disease state would be statistically un-
avoidable. This principle was frequently popularized in metaphorical terms, such
as the likelihood that any lock could be opened if a suciently large number of
keys are tried, or that the pharmacological richness of natural products could be
easily accessed and expanded through the laboratory equivalent of a rain forest.
The anticipated surge in overall pharmaceutical R&D productivity has not ma-
terialized and this optimistic view of combinatorial chemistry has largely been
abandoned. The contemporary view is more pragmatic and generally conforms to
the principles outlined below.
1 The synthesis of many compounds simultaneously is more ecient than the synthesis
of a single compound. In terms of simple time utilization, this principle was
4 While synthetic organic chemistry evolved primarily in terms of serial processing, its
productivity can be greatly enhanced by introducing parallel processing. Strong em-
phasis needs to be put on the importance of parallel processing as an explicit
component of experimental design. The synthesis plan must comprehend the
fact that hundreds or thousands of compounds are being synthesized simulta-
neously using hundreds or thousands of reagent combinations in each syn-
thetic step. Each unit operation and each operating parameter such as solvent
used, temperature, and level of agitation should encompass the conditions
needed to bring each reagent combination to complete reaction while mini-
mizing side-reactions and product decomposition. Sample tracking and in-
process control data must also be managed at very high volume.
5 Laboratory automation, robotics, and mechanical devices enabling the simultaneous
performance of multiple tasks are essential to combinatorial chemistry. The range of
enabling tools goes from simple devices such as multichannel hand pipettes to
high-performance, fully integrated systems such as the Irori NanoKan system.
This issue will be revisited, but the basic principle is that the successful prac-
tice of combinatorial chemistry requires some level of commitment to and in-
vestment in laboratory automation.
6 Robust process chemistry is required to assure the desired outcome. In the pharma-
ceutical world, the importance of process chemistry lies at the two extremes.
For single compound synthesis in a medicinal chemistry laboratory, a good
process may be on the nice to have list, but as long as the synthesis scheme
aords some of the desired compounds, intermediate and endstage puri-
cations will overcome deciencies in the process. For bulk pharmaceutical
production, the process is everything: it must be extremely ecient, fully vali-
dated, and conform to regulated manufacturing practices. The importance of
process chemistry in combinatorial synthesis is closer to the bulk production
context.
For large compound libraries, the opportunity for product purication is
very limited (although technologies for doing this are evolving nicely). Library
members with poor-quality or outright failures are highly undesirable as they
waste time and resources in biological testing laboratories. The best resolution
of this dilemma is to develop the synthesis scheme to be used for production of
a library into a robust, well-dened process. This requires careful optimization
of reaction conditions, validating the individual members of the reagent sets,
developing reliable analytical methods, and dening the process in something
equivalent to a standard operating procedure.
7 Electronic tracking and control systems are critical components of combinatorial
chemistry. Every chemist has been exposed to the tedium of labeling samples by
hand, lling out analytical request forms, and completing compound data
sheets. Preserving the identity of samples and information about them by
linkage to laboratory notebook pages is a cumbersome system that works when
a few hundred samples per year are involved but it cannot function when indi-
vidual experiments are producing thousands of individual compounds. The
data management aspect of combinatorial chemistry requires access to elec-
tronic tracking and control systems for sorting and labeling samples (e.g. with
2.3 Combinatorial Synthesis 15
machine-readable barcodes), for generating structure lists (SD les) that link
sample codes to compound structures, and for collecting and processing ana-
lytical data and linking these data to sample codes and structure les. In addi-
tion to these sample and data management applications, electronic control sys-
tems embedded in the operating software of automated synthesis systems are
necessary to ensure that each in-process material is in the right place at the right
time for each step in order to ensure that the library plan is faithfully executed.
8 Analysis and quality control procedures are just as important in combinatorial
chemistry as in other forms of synthetic chemistry. The shortcomings of early ex-
pressions of the combinatorial concept included the misperception that quality
was not important. In fact many libraries were deliberately prepared as com-
pound mixtures. It was taken for granted that biologically active samples could
be separated, deconvoluted, or in some way dealt with after hits were detected.
Two things resulted. The fact that synthesis without analysis is a prescription
for poor science was rearmed and condence in combinatorial chemistry as
a productive discovery tool developed rather slowly. At its current state of
development and acceptance, combinatorial chemistry is expected to provide
samples of individual compounds with a purity level [by high-performance
liquid chromatography (HPLC)] of at least 80%. The identity of compounds is
routinely veried by mass spectrometry, using electronic comparison of proba-
ble molecular ion peaks with calculated molecular weights. Combined liquid
chromatography/mass spectrometry (LC/MS) is preferable as it will conrm or
deny that the major peak in an HPLC trace corresponds to the design intent for
that particular library member. An aspect of quality control which needs fur-
ther development is the application of in-process control procedures (IPCs)
which are basic and routine in multistep synthesis of single compounds.
9 Since the purpose of combinatorial chemistry is to facilitate the discovery of useful
compounds, combinatorial syntheses must be reproducible and scalable. This is not
to say that the reaction sequence and mode of synthesis (solid or solution
phase) must eventually serve to provide multikilogram supplies, but the com-
pounds in a library have little value if interesting compounds cannot be con-
veniently resynthesized in small amounts for closer examination.
10 As a nal basic principle it should be understood that combinatorial chemistry is a
productivity-enhancing tool for chemists engaged in pharmaceutical research,
agrochemicals, catalysis, and materials science any eld where the prepara-
tion and testing of new compositions of matter are the essential elements of
discovery. However it does not displace or supplant established elds of syn-
thetic chemistry such as medicinal chemistry its best use is to leverage the
productivity of existing elds.
2.3
Combinatorial Synthesis
2.3.1
Solid-phase Combinatorial Synthesis
Solid-phase combinatorial chemistry relies on the fact that the molecule under
construction is attached via a linker to a polymeric carrier (bead) (Fig. 2.4). This
immobilization allows a simple separation of intermediates and nally of the
product from reagents and soluble byproducts. To guarantee the ecacy of this
principle, the bead and linker have to be stable under the reaction conditions. Fur-
thermore, the bond between the target product and the linker has to be cleaved
selectively under mild conditions without destroying the product. Thus, after
simple ltration, nothing but the liberated pure product AB is obtained.
Fig. 2.4. Solid-phase chemistry. After reaction the product is liberated and ltered o.
but introduce several more restrictions. Also the temperature that can be applied is
restricted: 100 C is generally at or beyond the upper limit for most solid-phase
systems room temperature to 80 C being the comfort zone. It should be noted,
however, that very low temperature operations involving, for example, reactive
enolates generated at 78 C are more easily handled in a directed split-and-pool
solid-phase mode (fewer batches). Chemistry involving heterogeneous catalysts
such as palladium on charcoal or solid reagents such as manganese dioxide are
intrinsically incompatible with solid-phase synthesis methods. Furthermore, the
solid-phase chemistry literature is still limited, and only a relatively small propor-
tion of the synthetic solution-phase repertoire has been adapted to solid phase.
Therefore the development of a suitable methodology for the synthesis of large
libraries can be time-consuming (several months). Another drawback is the di-
culty to analyze the outcome of a given reaction. On-bead analysis is still not sat-
isfying, although NMR and IR as well as MS techniques have been improved. In
most cases, reaction products have to be cleaved from the support and are ana-
lyzed by normal methods (HPLC, NMR).
2.3.1.2 Automation
As pointed out before, the ltration of the reaction mixture, e.g. the separation of
solid-supported products from starting materials and reagents, is the key process
in solid-phase chemistry. Because of its simplicity, this process can easily be auto-
mated and run in parallel. Thus, the rst robotic synthesizers were those for solid-
phase chemistry: many reactions were set up in parallel by automated dispensing
stations and the normally very tedious work-up procedure was reduced to simple
ltrations. This worked ne as long as the library size was not too large (<1000
compounds). Larger libraries are dicult to prepare by this process. A much more
ecient way to prepare large libraries is the so-called split-and-combine protocol.
reaction the product is selectively cleaved from the bead. The identity of the prod-
uct can subsequently be determined by conventional analytical methods analyzing
the tagged bead. W. Clark Still invented in 1993 the rst binary encoding using
haloaromatic reagents for tagging (see Chapter 5). Several more possibilities, such
as tagging with oligonucleotides or peptides, can be used for the chemical encod-
ing of single beads.
When encapsulated in porous containers, aliquots of beads traverse the split-
and-pool protocol with the same outcome as a single bead all the beads in each
container carry the same compound at every stage. Today radiofrequency tagging
is the most prominent encoding technology. This method is based on a small radio
transponder, which is attached to the synthetic platform (e.g. a polypropylene re-
actor). Upon activation this device emits a given radiofrequency, thus making the
identication possible. Development of chemically passive tagging systems for
solid-phase synthesis vastly enhanced the utility of the split-and-pool protocol. The
enormous gain-in-function oered by split and pool can now be used to produce
huge libraries of pure, single compounds in multimilligram quantities with min-
imal synthetic eort. For example, the production of a 160,000-compound library
can be achieved in just 80 discrete synthetic operations if the library design em-
braces four sets of building blocks with 20 members in each set. Automated sort-
2.3 Combinatorial Synthesis 19
ing and washing stations have been built, but handling became so easy that small
libraries can be prepared with minimal automation.
2.3.1.4 Cost
Modest capital investments are required for tag-reading/sorting devices, apparatus
for cleavage and delivery of compounds to plates, and centrifugal evaporation. The
instrumentation requirements for product quality control (QC) are, of course, the
same as in solution-phase chemistry. Solid-phase methods foster proigate use of
reagents to drive reactions to completion, as mentioned before. Added to these
considerations is the fact that speciality resins, hi-tech linkers, and single-use
resin containers can add substantial material costs to library production.
2.3.2
Solution-phase Combinatorial Synthesis
chemistry (Fig. 2.8). The versatility of this approach has been proved by Ley
et al. in the total synthesis of two natural products, (G)-oxomaritidine and (G)-
epimaritidine.
Another possibility to deal with the problem of excess reagents in solution-phase
combinatorial chemistry is the use of uorous tagging and extraction with uori-
nated solvents (Fig. 2.9). Substrate A is attached to a moiety which is polyuo-
rinated. Because of the uorous tag the product of the reaction is exclusively so-
luble in the uorous solvents. After reaction with excess quantities of reagent B
an extraction with a polyuorinated solvent is performed. Reagent B is more solu-
ble in the organic phase and can be separated from the product. Finally the uo-
rous tag is cleaved o, removed by another extraction and the pure product can be
isolated.
2.4
Conclusion
Solid-phase chemistry has made tremendous progress during recent years, and
today highly complex natural products can be prepared on solid supports (see
Chapter 3). Nevertheless, method development is slow and often fails to produce
the compounds needed, e.g. in medicinal chemistry optimization programs. Thus,
careful validation of the possible outcome of solid-phase chemistry has to be given
the highest priority before the development of methodology is started. However,
for the preparation of large libraries with only minor time constraints, solid-phase
chemistry remains the method of choice. On the other hand, solution-phase
chemistry is much more exible and quicker. However, the production of large
libraries is very tedious owing to the dicult purication of the products.
Polymer-supported reagents and scavengers represent a versatile addition to
solid-phase organic synthesis and parallel solution-phase chemistry. The combina-
tion of these reagents oers exciting possibilities. The methods described in this
chapter are interesting and fascinating but, so far, they do not enable the chemist
to exploit all the synthetic routes that he or she might think of. Thus, there is a
need for more solid-supported reagents, linkers, scavengers, and, especially, solid-
phase methodology.
Bibliography
Diversity in Drug Discovery, John Wiley Gordon, J. Med. Chem. 1994, 37,
& Sons, New York, 1998. 12331251.
8 K. Burgess, Solid-Phase Organic 16 E. M. Gordon, R. W. Barrett, W. J.
Synthesis, John Wiley & Sons, New Dower, S. P. A. Fodor, M. A. Gallop,
York, 2000. J. Med. Chem. 1994, 37, 13851401.
9 D. Obrecht, Solid-Supported 17 G. Lowe, Chem. Soc. Rev. 1995, 24,
Combinatorial and Parallel Synthesis 309382.
of Small-Molecular-Weight Com- 18 Dedicated issue, Account. Chem. Res.
pound, Pergamon Press, Oxford, 1996, 29, 111170.
1998. 19 F. Balkenhohl, C. von dem
10 W. Bannwarth, E. Felder, Combi- Bussche-Hunnefeld, A. Lansky, C.
natorial Chemistry. A Practical Zechel, Angew. Chem. Int. Ed. Engl.
Approach, Wiley-VCH, Weinheim, 1996, 35, 22882337.
New York, 2000. 20 P. H. H. Hermkens, H. C. J.
11 M. R. Pavia, T. K. Sawyer, W. H. Ottenheijm, D. Rees, Tetrahedron
Moos, Bioorg. Med. Chem. Lett. 1993, 1996, 52, 45274554.
3, 387396. 21 R. E. Dolle, Mol. Diversity 1997, 2,
12 W. H. Moos, G. D. Green, M. R. 223236.
Pavia, Annu. Rep. Med. Chem. 1993, 22 P. H. H. Hermkens, H. C. J.
28, 315324. Ottenheijm, D. Rees, Tetrahedron
13 K. S. Lam, M. Lebl, Methods: A 1997, 53, 56435678.
Companion to Methods in Enzymology 23 R. E. Dolle, Mol. Diversity 1998, 3,
1994, 6, 372380. 199233.
14 K. D. Janda, Proc. Natl. Acad. Sci. 24 D. P. Curran, Angew. Chem. Int. Ed.
USA 1994, 91, 1077910785. 1998, 37, 11751196.
15 M. A. Gallop, R. W. Barrett, W. J. 25 R. E. Dolle, K. H. Nelson, Jr,
Dower, S. P. A. Fodor, E. M. J. Comb. Chem. 1999, 1, 235282.
http://www.5z.com/divinfo/
http://www.combinatorial.com/
24
3
Solid Phase and Soluble Polymers for
Combinatorial Synthesis
Rainer Haag, Andre Hebel, and Jean-Francois Stumbe
3.1
Introduction
Polymeric supports have revolutionized organic synthesis in the past decade and
have become a major driving force for laboratory automation and combinatorial
chemistry in general. While polystyrene (PS)-based ion exchange resins have been
known since the 1950s, the rst solid-phase peptide synthesis on modied PS
microbeads was reported by Merrield in 1963 [1]. Until the late 1980s, when the
rst nonpeptidic molecule libraries were reported [2], polymeric supports were
mainly used for special applications, such as peptide and oligonucleotide synthe-
sis. In these cases the polymeric support had to be stable only toward two repeti-
tive reaction conditions (coupling and deprotection). Therefore, some of these
original supports are of limited use for parallel multistep synthesis of small or-
ganic molecules, generally known as combinatorial chemistry. In the last decade a
rapidly increasing number of new polymeric supports, crosslinked (insoluble) [3
5] and noncrosslinked (soluble) polymers [610], have been described and used for
combinatorial synthesis. Although solid-phase synthesis on PS-based resins ex-
hibits a number of problems due to the heterogeneous nature of reactions and the
low concentration of functional groups (typically a1.5 mmol substrate/g polymer)
[11], no other polymeric support has yet found the same broad application for
combinatorial synthesis [12]. However, the reader should be aware of the fact that
there is no polymeric support for general application in organic synthesis. Every
polymer has its drawbacks (e.g. chemical stability, polarity) and hence is stable
only within a certain range of reaction conditions.
In this chapter, we will describe the structure and the properties of polymeric
supports as well as the eects of dierent spacers (Fig. 3.1). Spacer molecules,
compared with linker molecules, are used to provide more accessible linker func-
tionalities and to modify the properties of the polymer backbone (e.g. polarity,
swelling characteristics). Linker functionalities, which also inuence the materials
properties, will be the subject of the next chapter. In the following sections the two
major classes of polymeric supports solid and soluble polymers will be further
classied by their chemical structures and the polymer topologies. A special focus
will be put on the polymer properties, such as stability, swelling property, reactive
site distribution, loading capacity and the range of applications in organic synthe-
sis.
3.2
Solid-phase Supports
3.2.1
Polystyrene-based Resins
Fig. 3.2. Relation between mesh size and bead diameter in micrometers.
pected results described in this work were recently revised by further studies using
confocal raman spectroscopy [23] and confocal uorescence spectroscopy of thin
slices in order to avoid uorescence quenching [25]. It is now obvious that the
distribution of active sites is essentially uniform throughout polystyrene resins, as
shown already by Merrield [21]. However, for large beads, diusion processes can
certainly play an important role and lead to an unequal reactivity of functional
groups located closer to the core and the shell of the individual bead. Based on geo-
metrical considerations, one should also be aware of the fact that, in microbeads
with a diameter of 100 mm, 50% of the active sites are within the rst 10 mm of the
outer shell [23].
Scheme 3.1. Synthesis and structure of interconnected pores. (1) Porogen and network
macroporous resins: (a) polymer network start to phase separate; (2) porogen phase
forming; (b) porogen phase acts as pore removed to yield pores (hatched area,
template; (c) dry macroporous resin with large crosslinked polymer; dots, porogen phase).
As mentioned above the generation of pores can take place in two ways: non-
crosslinked polymers as well as organic solvents [28, 31]. For instance, if linear
3.2 Solid-phase Supports 29
Fig. 3.3. Dierent chemical species present in the technical grade DVB mixture.
weight alcohols. The pore size and the presence of channels interconnecting these
cavities allow solvents to diuse quickly in and out of these pores.
One drawback of these heterogeneous structures is the very low accessibility of
the solvents and reagents to the highly crosslinked areas, leading to limited load-
ing capacities with a typical range of 0.81.0 mmol g1 [34, 35]. However, some
commercial ion exchange resins, also used for organic reagents, have loading ca-
pacities up to 4.5 mmol g1 [26]. Generally, macroporous resins display lower
reactivities than microporous swollen beads. In contrast to microporous resins,
they show high resistance toward osmotic shock (see Fig. 3.5b) [20], but they can
be brittle when not manipulated carefully [13].
Fig. 3.5. ESEM pictures demonstrating the of a swollen resin (24 h, 300 rpm, 20 C,
mechanical stability of 2% crosslinked DMF); (d) after mechanical shaking of a
Merrield resins under dierent conditions: (a) swollen bead (7 days, 60 rpm, 20 C, DMF).
after swelling in THF and drying; (b) after The ESEM pictures were taken by Dr. Rudiger
osmotic shock (swelling in THF, washing with Landers, Materials Research Center, University
MeOH and drying); (c) after magnetic stirring of Freiburg, Germany.
lysts [28] and as adsorbents [37]. Other interesting applications involve their use as
chromatographic materials and for continuous ow synthesis in columns [28, 38,
39].
The loading capacity is controlled by the yield of the electrophilic aromatic sub-
stitution in method 1 or by the amount of 4-(chloromethyl)-styrene incorporated
in the starting organic phase for copolymerization in method 2. Typical loading
values are between 0.2 and 4.0 mmol g1 , and a loading capacity of 1.5 mmol g1
(for the most commonly used Merrield resins) corresponds approximately to 20%
substituted aromatic groups. Higher loading Merrield resins (4 mmol g1 ) have
also been used in organic synthesis [47]; however, the highest possible loading
6.55 mmol g1 , which corresponds to 100% chloromethylstyrene, would not be
useful in practice [12].
For microporous resins the exact degree of crosslinking and the nature of the
crosslinker are even more important than for macroporous resins, owing to the
severe eect on the swelling properties. The most common microporous resins are
12% crosslinked, but resins with less crosslinkage have also been studied [48].
They are mechanically weak and consequently easily subject to damage. However,
increased reaction rates have been observed for these more exible polymer net-
works. It is important to keep in mind that the DVB used for crosslinking is usu-
ally a technical grade product with a composition that can vary from one batch
to another (see Fig. 3.3), which inuences the properties of the beads. The con-
sequences are that varying amounts of crosslinking agents are incorporated in
resins depending on the dierent polymerization batches, which generally leads to
relatively high error values (G0.5%) in the degree of crosslinking indicated in the
catalogues of many suppliers.
Other crosslinkers are ethyleneglycol dimethacrylate (EGDMA) (4), N,N-
methylene bisacrylamide (6) (MBA), trimethylolpropane trimethacrylate (8)
(TRIM), and, more recently, novel crosslinkers have been introduced such as 1,4-
bis(vinylphenoxy)-butane (5) [49], and bis(vinylphenoxy)-polyethylene glycol (PEG)
(7) [50, 51] which present the advantage of having a strong inuence on the swell-
ing properties due to the increased exibility between the two crosslinking units
(Fig. 3.4).
In addition, styryl-terminated dendrimers have been introduced as novel poly-
mer crosslinkers (9). They consist of 816 peripheral styryl units attached to aryl end
branches of dendritic TADDOL or BINAP ligands and these were copolymerized
3.2 Solid-phase Supports 33
Mechanical and chemical stability of microporous resins Care has to be taken when
working with microporous resins in dierent solvents. Osmotic shock can occur
when a preswollen bead (in a good solvent) is introduced into a bad solvent.
The beads start to shrink rapidly under expulsion of the good solvent and are
subjected to stress. This leads to mechanical damage or at least to non-negligible
modications of the structure, as shown in Fig. 3.5b (compare with the surface of
the original Merrield resin in Fig. 3.5a).
Another problem that has to be encountered when conducting a reaction on
swollen microporous resins is that they can break if agitation conditions are too
vigorous. Figure 3.5c shows an example of beads after a solid-phase reaction con-
ducted with a magnetic stirrer. The scanning electron microscope (SEM) picture
reveals that most of the beads are broken compared with the beads collected after
a reaction conducted on a mechanical shaker (Fig. 3.5d). The consequence for
broken resin beads is the clogging of the lters used during their purication.
34
Solvent Merrield Tentagel S ROMP PS denrimers PAP CLEAR POEPOP SPOCC POEPS-3
resins (1% spheres
crosslinked) Polyamide Arylether Mn (PEG)F1500 g mol C1
dendrimer dendrimer
3.2.2
Polystyrene Hybrid Supports
The main drawback of the rst technique is the possibility of additional cross-
linking due to the fact that PEG macromolecules are bifunctional and conse-
quently that the two end chains can both react with the resin [56]. This, however,
can be overcome if monoprotected PEGs are used.
PEGylated supports are far more polar than Merrield resins and hence they
swell in a broad range of solvents from apolar aprotic to polar protic solvents
(Table 3.2) [53]. However, the loading of TentaGel9 resins is relatively low (loading
values range from 0.15 to 0.40 mmol g1 for PEG35 chains) compared with the
higher loading capacities that can be reached for Merrield resins. A similar type
of PEGylated resin known as ArgoGel9 has slightly higher loading capacities (up
to 0.5 mmol g1 for PEG35 chains) because it has two PEG chains that are grafted
onto every active site [61, 62].
PSPEG hybrid resins opened many new possibilities in combinatorial syn-
thesis by allowing the use of protic solvents. Even on bead, screening is possible
with these resins. However, the user should be aware of some drawbacks, such as
bleeding, which is dened as the loss and the release of PEG grafts when treated,
for example, with strong acids [63, 64]. It has been reported for instance that tri-
uoroacetic acid (TFA), which is commonly used for cleavage of protecting groups,
can also cleave the benzylicCH2 OPEG bond [65]. This problem can be overcome
by using a TentaGel9 resin with the PEG spacer attached to the polystyrene back-
bone via an alkyl linkage. This hybrid resin is not sensitive to acids or bases [34,
57, 62, 64]. Other drawbacks of such PSPEG hybrid resins originate from the
nature of the PEG polymer, which is hygroscopic. Hence, the presence of large
amounts of PEG (up to 70%) render the beads more sticky and more dicult to
manipulate in some cases than regular Merrield resins [20].
In order to overcome the decrease of loading capacity due to bleeding of the
PEG spacers, a similar type of resin has been designed (e.g. NovaGel9) with active
sites located on the PS backbone [64] rather than at the end of the PEG-grafted
chains as in TentaGel9 and ArgoGel9 resins. This results not only in a good
swelling behavior in both apolar and polar solvents due to the PEG chains (PEG
content @ 50%) but also in a slightly higher loading capacity (0.7 mmol g1 ). In
addition, eventual bleeding of the PEG grafts does not decrease the loading capac-
ity and only a slight modication of the swelling properties will occur. However,
the advantage of improved kinetics and better on-bead analytics (e.g. nuclear mag-
netic resonance) due to the highly exible polyether spacers are lost in this ap-
proach. It should be pointed out again that all PEGylated resins are sensitive to-
ward oxidation and hence bleeding will occur if not properly handled.
styrene beads and norbornene derivatives [66]. In a rst step the immobilization of
the ruthenium catalyst (Grubbs catalyst) on vinyl polystyrene (11) can be achieved
via insertion of ruthenium into the styryl double bond (Scheme 3.2). The PS-
supported ruthenium alkylidene (12) shows good stability under normal atmo-
spheric conditions when dried. In the presence of solvents (e.g. dichloromethane),
however, the catalyst becomes inactive within a period of 5 h. Treatment of the PS-
supported catalyst (12) with an excess of a norbornene derivative (13) yields a
ROMP-based polymer (14), so-called ROMP spheres (Scheme 3.2).
This resin shows high loading capacities (up to 3.0 mmol g1 ), but the swelling
properties are lower than for conventional resins, good solvents include THF
and DCM (see Table 3.2). The utility of these high-loading supports was exempli-
ed by a palladium-catalyzed coupling reaction between a resin-bound bromo-
benzoate and aryl zincates in a THF/MeOH (4:1) mixture. Also, crosslinked
ROMP-based polymers have been used as supports for catalysts [67] and high-
loading amine scavengers [68].
Rasta silanes Silyl linkers (see Chapter 4.3.5) oer a broad range of mild
and chemospecic cleavage conditions and are frequently used in SPOS. A high-
loading silyl-functionalized PS hybrid resin, so-called Rasta silane, has recently
been introduced for solid-phase chemistry [69]. The required silyl substituted
styrene monomer (16) can be prepared by lithiumhalogen exchange of p-
bromostyrene (15) and subsequent quenching with the appropriate dialkylchlo-
rosilane. A living free radical polymerization initiated by heating TEMPO-methyl
resin (17) with the silyl styrene (16) under solvent-free conditions at 130 C leads
to the formation of Rasta silanes (18) (Scheme 3.3).
Rasta resins are characterized by high loading levels between 1.6 and 3.8
mmol g1 with regard to the silicon linker. These values are signicantly higher
than for conventional silicon-linker resins, which show loading capacities between
0.5 and 1.6 mmol g1 . Easy conversion of the silyl units to reactive silyl chloride or
triate by standard synthetic methods allow the immobilization of alcohols and
phenols for solid-phase synthesis. It has been shown that these Rasta silanes allow
fast diusion of reagents throughout the resin beads [70].
After coupling of a Rink amide linker [73] onto resin (21), the derivatized PS
dendrimer resin (22) with a loading capacity of 0.9 mmol g1 , with regard to the
Fmoc (9-uorenylmethoxycarbonyl) amino groups, is obtained. Repeating this
procedure yields the second generation [G2] (23). This exhibits a slightly higher
loading capacity of 1.17 mmol g1 . It is important to mention that the loading per
gram of resin does not increase substantially because of the mass of the attached
linker. However, the loading per bead increases geometrically from one generation
3.2 Solid-phase Supports 39
to the next. These PS dendrimer resins have successfully been tested in the syn-
thesis of small peptides [72].
The synthesis of symmetrical tri-branching dendrimers on aminomethyl poly-
styrene macrobeads has also been described [74]. These [G2] dendritic resins (Fig.
3.7) oer loadings of 30.6 nmol per macrobead (0.8 mmol g1 ). A split-and-mix
synthesis (e.g. SPPS, Suzuki-coupling) on a gram of these resin beads (@ 27,000
beads g1 ) provides the number of compounds and the amounts necessary for a
single bead screening approach (including 1 H-NMR!). The obtained beads, espe-
cially the PS[G2] dendrons (24), show higher swelling values in polar solvents
such as methanol (MeOH) and dimethyl sulfoxide (DMSO) than the originally
used aminomethyl resin (see Table 3.2).
Some aryl ether dendrimers can be synthesized directly on hydroxymethyl poly-
styrene (25). Mitsunobu reaction with 3,5-bis(acetoxymethyl)-phenol (26) gives the
branched precursor (27) (Scheme 3.5) [75]. After deprotection and repetition of the
sequence a [G3] dendrimer resin was obtained with a loading capacity of 2.85
mmol g1 (3 nmol per bead). In unpolar solvents such as DCM and THF the den-
drimer resin (28) swells better than in MeOH or water.
The development of high-loading immobilized catalysts is another interesting
application for PS dendrimer resins. An example is given by the rhodium complex
with dendritic phosphine ligands that are anchored onto a PS resin (Fig. 3.8) [76].
Dierent dendrimer generations have been prepared and tested as catalysts for the
hydroformylation reaction of several olens. The PS[G2] dendrimer (29) shows
higher reactivities than the rst generation and does not decrease for up to ve
cycles. The origin of this increased reactivity might be due to the higher density of
ligands on the outer core and cooperative eects.
High internal phase emulsion polymers: PolyHIPE PolyHIPE structures are formed
by polymerization of a high internal phase emulsion with styrene and divinylben-
zene [77]. The resulting material consists of an extremely porous and rigid matrix,
which provides a scaold onto which functional polymers (e.g. polyacrylamide
gels) can be chemically bound (Scheme 3.6). The polyacrylamide gel is prepared
3.2.3
Other Crosslinked Polymeric Supports
The obtained PAP resin (39) has a loading capacity of 0.7 mmol g1 (amino
groups), which can be used for SPPS. In either its protonated or acylated form,
this resin exhibits high swelling values in water, methanol, CH2 Cl2 and CHCl3 .
Several applications of PEGA resins have been reported for SPPS [87].
but they should also be suitable for limited application in SPOS, especially when
protic solvents are required.
time the loading capacity drops. Typical loading capacities for POEPOP resins
are between 0.4 and 0.7 mmol g1 . In comparison with typical polystyrene-based
TentaGel9 resins (see Table 3.1), this is still a relatively high value [90]. POEPOP
supports have been successfully employed for a solid-phase synthesis of peptide
isoesters by nucleophilic reactions [91].
3.2.4
Inorganic Supports
In the following section, we give a short overview of some inorganic supports that
are suitable for SPOS. A good summary was published recently [4, 26]. Dierent
forms of silica gel are very often used for SPOS. They oer a number of advan-
tages and are now commercially available in various derivatized forms. These sup-
ports are rigid and do not swell or contract in the solvents employed in synthetic
cycles. Reagents and products can easily diuse into and out of the pores. Typical
pore sizes are between 2 and 10 nm, and the surface area is between 200 and 800
m 2 g1 . Although higher coupling rates in peptide synthesis were obtained on
silica than on polystyrene [93], silica gel did not become an established support
for solid-phase organic synthesis. In the meantime, silica gel has been partially
replaced by controlled pore glass (CPG) as an inorganic support [38, 9498]. The
main advantages of CPG, compared with silica gel, are its more regular particle
size and shape, and greater mechanical stability. Functionalization with long-chain
alkylamine groups leads to exceptional properties: rigid, nonswelling, mechanical
stability within a series of polar solvents, and chemical stability over the whole pH
range. Furthermore, CPG is unreactive to a broad variety of nucleophiles and elec-
trophilic reagents. Its stability during heating and its excellent solvation and ow
properties make it the support of choice for automated solid-phase synthesis. Typ-
ical loading capacities, however, are rather low (0.1 mmol g1 ).
3.3
Soluble Polymeric Supports
3.3.1
Separation Techniques for Soluble Polymeric Supports
Limitations Unsuitable for Suitable only for Unsuitable for Dierent solubilities
nal cleavage nal cleavage multistep required
step step or removal syntheses
of polymeric
reagents
3.3 Soluble Polymeric Supports 49
3.3.2
Terminal Functionalized Linear Polymeric Supports
Linear polymers that carry functionalities only on their chain ends (e.g. PEGs),
have been frequently used as soluble polymeric supports and are listed in Table 3.4
(Entries 1 and 2) [610]. Because of the limited number of functional groups (one
or two per chain) these materials have rather poor loading capacities (typically
<0.5 mmol g1 ) and products coupled to the polymeric support are more dicult
to analyze (e.g. by NMR) than higher loading supports. On the other hand, these
soluble polymers do not dramatically change their physical properties upon func-
tionalization. This is extremely important if precipitation is used for separation.
The most widely used soluble polymeric support in organic synthesis is mono-
methylated polyethylene glycol (typically MPEG 5000), which is soluble in many
organic solvents and easily precipitated in nonpolar solvents (e.g. diethyl ether) [7,
105]. It was introduced by Bayer and Mutter in the 1970s for peptide synthesis and
was rediscovered by Janda and others as a soluble support for organic synthesis in
the 1990s [6, 8, 9]. Owing to the linear topology of this polyether, it contains only
one reactive functionality and hence exhibits a rather poor loading capacity (0.2
mmol g1 ). In order to increase the loading capacity, bifunctional PEGs endcapped
with benzylether dendrons have been prepared and used for the synthesis of
b-lactams [106, 107]. Another approach is the coupling of PEG arms onto a multi-
functional core to give a PEG-star polymer. Such multiarm PEG-star polymers
have been recently introduced as supports for organic synthesis based on pentae-
rythritol [108], hyperbranched polyglycerol (see Section 3.3.4) [109] and cyclo-
triphosphazene [110]. The advantages of these polyethers are their high chemical
stability and the good reactivity of the functional groups in homogeneous phase.
The achieved loading capacities of these multifunctional PEG polymers (a 1.0
mmol g1 ), however, are only marginal better than the commercially available
mono- and bifunctional PEG derivatives (0.2 and 0.4 mmol g1 respectively).
3.3.3
Polyfunctional Linear Polymeric Supports
Tab. 3.4. Selected linear soluble polymeric supports for organic synthesis.
5 Polyacrylamide (14.1) C
6 Polyvinylalcohol 10.0 A, B, C, D
(22.7)
8 Cellulose (18.5) B
10 Poly(styrene-b-vinyl-alcohol) (6.8) D
polymers have been used for various organic synthesis applications as well as for
polymer-supported reagents. However, the physical properties (i.e. solubility) of
these materials change dramatically upon functionalization and can be problem-
atic in some cases for broad application in combinatorial synthesis.
3.3 Soluble Polymeric Supports 51
3.3.4
Dendritic Polymeric Supports
exibility of the branches, the terminal 1,2-diol groups show excellent accessibility
and can be used directly as linker groups for various applications in organic syn-
thesis [104]. In addition the characterization of these dendritic polyethers by stan-
dard analytical techniques (e.g. NMR) is much more rapid than the monofunc-
tionalized PEG derivatives owing to the high loading capacity of functional groups
[104].
3.3.5
Microgels
3.4
Conclusions
After more than 10 years of intensive research on the preparation and the applica-
tion of new polymeric supports in combinatorial chemistry many new polymeric
materials have been developed and evaluated. However, there is still no polymeric
support for general application in combinatorial chemistry. Every polymer has its
drawbacks (e.g. chemical stability, polarity, loading capacity). Therefore, a poly-
meric support has to be carefully selected for the synthetic problem that needs to
be solved. In solid-phase organic synthesis several new supports with better swell-
ing properties in a wider range of organic solvents and higher loading capacities
have been introduced. Among these are several new PS hybrid resins as well as a
new family of crosslinked PEGs. In terms of material stability, inorganic supports,
i.e. controlled pore glass, cannot be beaten. However, the width of synthetic appli-
cations is very limited so far and will depend on the chemical stability of the
spacer and linker molecules used. In addition, these supports suer a heterogenic
nature and rather low loading capacities.
Soluble polymeric supports, like solid supports, have had a similar revival over
the past decade. In terms of stability, aliphatic polyethers and noncrosslinked poly-
styrene are among the most promising candidates. Dendritic and linear polyfunc-
tional soluble polymers have by far the highest loading capacities and show great
potential as supports for reagents and catalysts in combinatorial synthesis because
of their homogeneous reaction conditions. The use of these soluble polymers for
References 55
References
36 A. Guyot, Pure Appl. Chem. 1988, 60, Makromol. Chem. 1983, 184, 2603
365376. 2617.
37 J. M. J. Frechet, Makromol. Chem., 59 S. A. Kates, B. F. McGuiness, C.
Macromol. Symp. 1993, 289, 7071. Blackburn, G. W. Griffin, N. A.
38 H. Koster, F. Cramer, Liebigs Ann. Sole, G. Barany, F. Albericio,
Chem. 1974, 946958. Biopolymers 1998, 47, 365380.
39 C. McCollum, A. Andrus, Tetra- 60 P. Wright, D. Lloyd, W. Rapp, A.
hedron Lett. 1991, 32, 40694072. Andrus, Tetrahedron Lett. 1993, 34,
40 T. Balakrishnan, W. T. Ford, J. Appl. 33733376.
Polym. Sci. 1982, 27, 133138. 61 O. W. Gooding, S. Baudart, T. L.
41 K. W. Pepper, H. M. Paisley, M. A. Deegan, K. Heisler, J. W. Labadie,
Young, J. Chem. Soc. 1953, 40974105. W. S. Newcomb, J. A. Porco Jr, P. V.
42 R. P. Pinnel, G. D. Khune, N. A. Eikeren, J. Comb. Chem. 1999, 1,
Khatri, S. L. Manatt, Tetrahedron 113122.
Lett. 1984, 25, 35113514. 62 J. A. Porco, T. Deegan, W.
43 R. Arshasy, G. W. Kenner, A. Devonport, O. W. Gooding, K.
Ledwith, Makromol. Chem. 1976, 177, Heisler, J. W. Labadie, B. Newcomb,
29112918. C. Nguyen, P. V. Eikeren, J. Wong,
44 W. T. Ford, S. A. Yacoub, J. Org. P. Wright, Mol. Diversity 1997, 2,
Chem. 1981, 46, 819821. 197206.
45 S. Mohanraj, W. T. Ford, Macro- 63 V. Smali, N. J. Wells, G. J. Langley,
molecules 1986, 19, 24702472. M. Bradley, J. Org. Chem. 1997, 62,
46 Q. Sheng, H. D. H. Stover, 49024903.
Macromolecules 1997, 30, 67126714. 64 J. H. Adams, R. H. Cook, D.
47 S. P. Raillard, G. Ji, A. D. Mann, Hudson, V. Jammalamadaka, M. H.
T. A. Baer, Org. Process Res. Dev. 1999, Lyttle, M. F. Songster, J. Org. Chem.
3, 177183. 1998, 63, 37063716.
48 S. Rana, P. White, M. Bradley, J. 65 E. Bayer, W. Rapp in: Poly(ethylene
Comb. Chem. 2001, 3, 915. glycol) chemistry: Biotechnical and
49 A. R. Vairo, K. D. Janda, J. Comb. Biomedical Applications, Harris, M.
Chem. 2000, 2, 579596. (ed.), Plenum Press, New York, 1992,
50 S. Itsuno, I. Moue, K. Ito, Polym. p. 325.
Bull. 1989, 21, 365370. 66 A. G. M. Barrett, S. M. Cramp, R. S.
51 M. Renil, M. Meldal, Tetrahedron Roberts, Org. Lett. 1999, 1, 1083
Lett. 1996, 37, 61856188. 1086.
52 P. B. Rheiner, H. Sellner, D. 67 M. R. Buchmeiser, K. Wurst, J. Am.
Seebach, Chem. Eur. J. 2000, 6, 3692 Chem. Soc. 1999, 121, 1110111107.
3705. 68 T. Arnauld, A. G. M. Barrett, S. M.
53 R. Santini, M. C. Griffith, M. Qi, Cramp, R. S. Roberts, F. J. Zecri,
Tetrahedron Lett. 1998, 39, 8951 Org. Lett. 2000, 2, 26632666.
8954. 69 C. W. Lindsley, J. C. Hodges, G. F.
54 S. Pickup, F. D. Blum, W. T. Ford, Filzen, B. M. Watson, A. G. Geyer,
M. Periyasami, J. Am. Chem. Soc. J. Comb. Chem. 2000, 2, 550559.
1986, 108, 39873990. 70 S. R. McAlpine, C. W. Lindsley, J. C.
55 D. W. V. Krevelen, Properties of Hodges, D. M. Leonard, G. F.
Polymers, 3rd edn, Elsevier, New York, Filzen, J. Comb. Chem. 2001, 3, 15.
1990. 71 V. Swali, N. J. Wells, G. J. Langley,
56 E. Bayer, Angew. Chem., Int. Ed. Engl. M. Bradley, J. Org. Chem. 1997, 62,
1991, 30, 113128. 49024903.
57 B. D. Park, H. I. Lee, S. J. Ryoo, Y. S. 72 A. Mahajan, S. R. Chabra, W. C.
Lee, Tetrahedron Lett. 1997, 38, 591 Chan, Tetrahedron Lett. 1999, 40,
594. 49094912.
58 H. Hellermann, H. W. Lucas, J. 73 H. Rink, Tetrahedron Lett. 1987, 28,
Maul, V. N. R. Pillai, M. Mutter, 37873790.
References 57
4
Linkers for Solid-phase Synthesis
Stefan Brase and Stefan Dahmen
4.1
Introduction
ylic acids and amines, which are essential for peptide synthesis, has been studied
extensively. However, the synthesis of nonoligomeric compound libraries requires
more versatile linkers (see Sect. 4.5) [13].
A particular linker should resist the entire synthesis sequence without bias to
the diversity or structure of the target compound library and without limiting the
chemical methodology. None of the reaction conditions used for the construction
of the building blocks should lead to a premature cleavage (orthogonality principle;
Sect. 4.3.8). Beyond that, the linker must be cleavable o the resin as mildly as
possible to insure that the product will not be aected.
In this chapter, linker types classied according to their structures, and cleavage
strategies are presented. In general, linkers for all kinds of building blocks (pep-
tides, oligosaccharides, nucleic acids and small molecules) are covered.
4.2
General Linker Structures
4.2.1
Immobilization of Molecules
some cases a simple and rapid mixing of reagents is sucient to drive the attach-
ment to completion, in other cases tedious monitoring is necessary. Additionally,
in some cases the attachment proceeds under similar conditions as the detach-
ment, for example the formation of ketals, where excess reagents are required to
drive the reaction to completion.
In a second general method for attachment, the building block can be coupled
to give the preformed linker in solution. The fragment thus formed, which is
called the handle (see for example Fig. 4.2), can then be activated for attachment
onto the resin using the coupling strategies outlined above. This method requires
an additional step in solution, which can be advantageously used for an increase of
purity. Especially in cases where the activated linker tends to decompose (e.g. silyl
chlorides) or can be formed only in moderate yields and purities, this handle ap-
proach can be favorable. However, in automated synthesis, the required solution-
phase step is clearly a drawback.
4.2.2
Spacers
The linker either can be attached directly to the resin or can be located further
from the polymeric backbone using spacers. These bifunctional constructs either
can be built sequentially or can be attached via a handle approach. The spacer acts
as a connection to give the building block more mobility and hence the kinetics for
a given reaction are superior to the corresponding resin without spacers [14]. Fur-
thermore, larger spacers considerably change the physical properties, such as
swelling behaviors. This has been demonstrated with Tentagel resin [15]. However,
one must take into account both that the use of spacers requires an additional
synthetic step usually connected with a decrease of loading capacity and that the
spacer has to be as robust as the linker toward the reaction conditions performed
on the bead; Tentagel resin, in particular, leads to bleeding.
For the characterization of compounds attached to a polymeric support by nu-
clear magnetic resonance (NMR) spectroscopy, however, long spacers are an ad-
vantage, because they increase the mobility of the substrate and reduce the line
broadening usually observed in NMR spectra of polymers.
In most cases, a clear distinction between linker and spacer is not easy. The
linker is the minimal part of the resin required for the functional cleavage (for the
62 4 Linkers for Solid-phase Synthesis
silyl linkers this is the silyl group, for the trityl linkers this is the triphenylmethyl
moiety, for the triazene linker this is the 1-aryltriazenyl group, and so on).
The spacer therefore is the part between the linker and the resin as depicted in
Fig. 4.1 and described in Chapter 3 of this book.
4.2.3
Functionalized Linkers as Analytical Constructs
Functionalized spacers or linkers can play an important role for the determination
of loading and reaction conversion. The NMR-active uorinated linkers are espe-
cially suitable in this regard [16].
4.3
Linker Families
Usually, linkers are categorized according to the kind of functional group or sub-
strate class they are able to selectively immobilize (linkers for carboxylic acids,
alcohols, amines, and so on). As there are various types of linkers available for
solid-phase synthesis, many of them belong to certain well-established classes of
protecting groups (Table 4.1) and therefore can be grouped into families. The
members of each family have certain reactivity patterns in common.
While a linker presents the chemical structure essential for the loading and
cleavage of a particular functional group, a linker system provides the whole pro-
tocol for the attachment to and cleavage from the resin.
However, because of an increased demand for the exible anchoring of mole-
cules, other families of linkers such as sulfur linkers, triazenes linkers, among
others, have emerged. Nevertheless, the largest number of linkers developed so far
are based on benzylic-type groups.
Tab. 4.1. Common protecting groups and the analogous linker families.
4.3.1
Benzyl-type Linkers Including Trityl and Benzhydryl Linkers
Benzyl-type linkers are the most common anchoring groups for various kinds of
functionalities. In particular, esters, amides, amines, alcohols, and thiols can be
immobilized by this linker family.
As the pioneering works of Merrield [22] and Wang [23] are based on this
linker type, they represent the starting point of modern linker development. Ben-
zylic linkers are typically cleaved by strong acids (e.g. triuoroacetic acid, TFA),
which causes a protonation and subsequent elimination. A nucleophilic scavenger
usually quenches the resonance-stabilized cation thus formed. An increase of acid
lability can thus be achieved by stabilization of this intermediate by, for example, o-
and p-substitution of methoxy groups onto the ring [24]. This has been demon-
strated in the development of the SASRIN resin (11; super acid-sensitive resin: 1%
TFA cleavable) [25, 26] with one additional alkoxy group related to the Wang resin
and the HAL linker (13; hypersensitive acid labile: 0.1% TFA cleavable) [27] having
two additional alkoxy groups. In addition, benzyl-type linkers might be cleaved by
ammonolysis [28], light (Sect. 4.4.3; see Table 4.15), metal salts [29] (see Scheme
4.8), and oxidation reagents (e.g. Wang resin with H2 O2 [30] or dichlorodicyano-
benzoquinone (DDQ) [31]). The introduction of nitro groups onto benzyl-type
linkers leads to photolabile systems (see Sect. 4.4.3).
The prototype of a functional group with an appropriate breakable bond is the
Wang resin (6), which contains a 4-hydroxybenzyl alcohol linker moiety. The ben-
zyl alcohol hydroxyl group can be functionalized using either electrophilic or nu-
cleophilic substrates (Scheme 4.1) to give a benzylic linkage. It is very stable in a
whole set of reactions, but can be cleaved by acids such as triuoroacetic acid or
HF. Acids, alcohols, esters, and amides can be obtained as products after cleavage.
The Rink resin is particularly useful for the attachment of various functionalities
such as primary amines (see for example [82]; Scheme 4.2) and is commercially
available from various sources. The loading can be achieved via the Rink chloride
(29) or triate (30) (Table 4.2).
A linker that is particularly suitable for peptide amides and cyclopeptides is the
PAL linker and the backbone amide linker (BAL) concept [83] (Scheme 4.3).
64 4 Linkers for Solid-phase Synthesis
Scheme 4.2. The Rink resin as a linker for primary amines according to Garigipati [82].
Trityl resins (3843) (Table 4.2) are especially suitable for the immobilization of
nucleophilic substrates such as acids, alcohols, thiols, and amines. They are quite
acid sensitive and can be cleaved with acetic acid, for example; this feature is use-
ful in cases where acid-labile protecting groups are used. The stability of trityl
resin can be tailored using substituted arene rings, as demonstrated with the
chlorotrityl resin, which furnishes a more stable linker than the trityl resin itself.
Furthermore, the steric hindrance prohibits the formation of diketopiperazines
during the synthesis of peptides. The orthogonality towards allyl-based protective
groups was demonstrated in the reverse solid-phase peptide synthesis of oligopep-
tides [84] (Scheme 4.4).
Tab. 4.2. Structures and properties of benzyl-type linkers including trityl and benzhydryl linkers (excluding
photolabile linkers: see Sect. 4.4.3, Table 4.15).
Rink acid (X OH: 27); Rink Cleavage for esters: Widely used
amide (X NH2 : 28) [66, 67]; 10% AcOH, CH2 Cl2 , anchor for
Rink chloride (X Cl: 29); 20 C, 1.5 h; 1% TFA amines
Rink triate (X OTf: 30) [66]; cleavage for
amides: TFA [66];
20% TFA, CH2 Cl2
[68]
Rink amide AM (31) (RAM) Similar to the
(rink amide aminomethyl) Rink resin above
Besides benzyl-type linkers, other arylmethyl moieties can serve as linkers. Re-
cently, a new backbone amide linker has also been devised using indole chemistry
[85] (Scheme 4.5).
4.3.2
Allyl-based Linkers
A second, to the benzyl linkers related class is the family of the allyl-based linkers
(Table 4.3). They have been used for the attachment of carboxylic acids, which can
4.3 Linker Families 69
Structure Reference
[86]
[86]
[93]
[94]
be detached using metal catalysis. The advantages of linker cleavage under palla-
dium catalysis are the mild reaction conditions and the orthogonality (Sect. 4.3.8)
to various protecting groups. Kunz et al. [86, 87] developed the rst and simplest
linker to use the p-allyl detachment strategy. Starting from 2-bromocrotonic acid,
attachment to an amino group on a resin and further reaction with the cesium salt
of an appropriate protected amino acid or peptidic structure yields the HYCRAM
(hydroxycrotonylamide) resin [88]. The allylic cleavage proceeds with Pd(PPh3 )4
and morpholine or hydroxybenzotriazole [89]. The readily available (hydroxycrotyl-
oligoethyleneglycol-n-alkanoyl) HYCRON linker [9092] is based on a similar con-
cept; however, in this case, a handle comprising an amino acid and a preformed
linker has been used to minimize the risk of racemization upon cleavage. A higher
stability towards unwanted nucleophilic cleavage was achieved than that observed
with the HYCRAM linker. The incorporation of b-alanine facilitates easier moni-
toring of the reaction. Several other similar constructs have been used for compa-
rable purposes [9396]. Recently, the semisynthesis of vancomycin on solid sup-
port was accomplished using an allylic linker [97].
An example for the use of the allyl linkers has been provided in the synthesis of
peptide nucleic acids (PNAs) [91] (Scheme 4.6).
70 4 Linkers for Solid-phase Synthesis
Scheme 4.6. Synthesis of PNAs using the HYCRON linker according to Seitz [91].
4.3.3
Ketal/acetal-based Linkers
Another class of linkers incorporating a ketal moiety are the Lezno diol-linked
aldehydes [101, 102] and ketones [103, 104]. A similar linker using dithianes is
also suitable for ketones [105] (Sect. 4.3.7.1).
4.3.4
Ester-, Amide-, and Carbamate-based Linkers
Ester and amide moieties are, apart from being used in the benzyl and allyl linker
types, also suitable for attachment. Various examples have demonstrated the use-
ful applications of these functional groups (Table 4.5). Basically, two dierent
strategies can be used (Fig. 4.3). For example, May et al. [115] recently introduced
thioesters as linkers for alcohols, ketones, and lactones.
Tab. 4.5. Overview of some ester- and amide-based linkers (see also Table 4.16).
Benzyl carbamates are also useful linkers for the synthesis of amines since they
are readily cleavable by palladium salts [17] (Scheme 4.8).
Carboxylic esters have been released by the action of alkoxides on ester resins. In
most cases, the cleavage has been performed using methoxide in methanol (e.g.
[116]; Scheme 4.9). Drawbacks to this method are the need to remove excess metal
salts and/or the aqueous work-up. Alternatively, after cleavage from solid support,
postsynthesis of methyl esters with diazomethane is also possible [121] (Scheme
4.10).
Scheme 4.8. Detachment of peptides from polymeric benzyl-type protecting groups [17].
4.3 Linker Families 73
Scheme 4.9. The use of esters as linkers for benzoic acid derivative by Kondo et al. [116].
Scheme 4.11. Detachment of peptides from hydrazide resins by Millington et al. [119].
4.3.5
Silyl Linkers
The electronic and steric properties of silicon compounds have been used in many
applications for the design and use of new linker types (Table 4.6).
The dierent applications can be divided as follows:
1 Direct attachment of building blocks on silylated resins: linker for alcohols [19,
20], traceless linkers for arenes (Sect. 4.5.5).
2 Use of the b-silicon eect for elimination reactions: e.g. the SEM (2-trimethylsi-
lylethoxymethyl) linker [21].
3 Silylated benzhydryl linkers: [69, 122].
74 4 Linkers for Solid-phase Synthesis
The silicon-based linkers, which are commercially available [132], were devel-
oped further by several groups [133] and have been used in the synthesis of vari-
ous systems for the traceless detachment of various arenes and heteroarenes [126
128, 133143]. They are also suitable as traceless linkers for allyl silanes to give
alkenes [144] (Sect. 4.5.5).
A recent traceless application was demonstrated in the synthesis of chrome-
nones (104) [145]. In this case, a mild cyclization method was used to circumvent
a premature cleavage for the support (Scheme 4.13). Similarly, heteroarenes are
also accessible [128] (Scheme 4.14).
Other linkers having silyl fragments are the silyl acid linker (SAC linker 21)
[56], the silyl amide linker (SAL linker 20) [55], the Pbs linker (silico Wang
linker) (98) [122], the Ramage silyl linker (15) [52], the SEM linker (99) [21], and
silylated benzhydryl linkers (32) [69].
76 4 Linkers for Solid-phase Synthesis
4.3.6
Boronate Linkers
Boronates have been used in various linker types either as linkers for diols [146] or
as precursors for metal-mediated cleavage (Table 4.7).
A boronic acid ester, which contains an aryl iodide moiety attached by an ap-
propriate tether, can act as an intramolecular arylation agent. Thus, Li and Bur-
gess [148] developed a polymer-bound precursor, the ensuing cleavage of which
furnished a macrocyclic constrained b-turn peptide mimic via biaryl coupling
(Scheme 4.15).
[146] Diols
4.3 Linker Families 77
Scheme 4.16. Traceless cleavage of boronic acid derivatives according to Pourbaix et al. [147].
4.3.7
Sulfur, Stannane- and Selenium-based Linkers
A set of modern linkers (Table 4.8) based on sulfur, stannane, and selenium chem-
istry can be found in the literature. Their popularity obviously stems from the fact
that these elements can favorably be tailored for use as fragile points of attach-
ment.
Aryl sulfonates can be used as linkers for arenes, as shown by Jin et al. [152].
These can be cleaved under reducing conditions to give the corresponding hydro-
carbons (130). In addition, this linker might be suitable as a multifunctional an-
chor, as proposed in a patent [166] (Scheme 4.17).
Scheme 4.18. Cleavage Suzuki coupling approach using sulfonium salts by Vanier et al. [154].
Photolabile sulfur linkers are based either on the relatively weak carbonsulfur
bond in thioethers [161] or on the photolytic decarboxylation of thiohydroxamic
acids to give methylindoles [120]. Also, a safety-catch linker (Sect. 4.5.1) for amines
is based on 2-(thiobenzyl)ethylcarbamates [151, 168]. The linkage is performed
with preformed handles containing ethenyloxycarbonyl-protected amines (138).
Attachment to thiomethylated polystyrene (139) was performed under conditions
involving radicals. The cleavage was carried out with an oxidizing agent, which
forms the retro-Michael substrate [151] (Scheme 4.20).
The utility of a thioacetal-based anchor (68) as a chemically robust linker for
the immobilization of ketones employed the commercially available (G)-alpha-
lipoic acid. The products were easily cleaved from solid support by treatment with
[bis(triuoroacetoxy)iodo]-benzene [105] (Table 4.4).
Recently, Nicolaou et al. [153] have shown that the reaction of alkenes in the
presence of dimethyldioxirane could be used for loading onto polystyrene sulfonic
acid resin. Subsequent cleavage with nucleophiles proceeded smoothly to give a
vast array of heterocycles.
4.3 Linker Families 81
Scheme 4.21. The stannane linker for Stille reactions according to Nicolaou et al. [163].
Based on a similar concept, a selenium linker has been described that is loaded
via a preformed handle [171]. Oxidation and thermal elimination give rise to the
formation of alkenes (153) (Scheme 4.23). Interestingly, the selenoxides (152) de-
compose at room temperature, whereby the corresponding sulfoxides fragment at
100 C.
The selenium bromide linker (154) (Scheme 4.24) is also the starting point for a
library of natural products and analogs of the benzopyran type through reaction of
prenylphenols [172174] (Scheme 4.24). Furthermore, this linker enables the syn-
thesis of medicinally interesting molecules [175].
Besides phenols, allyl anilines can be used in this sequence to produce indo-
lines. Depending on the substitution pattern, either traceless (! 162) (Sect. 4.5.5)
4.3 Linker Families 83
Scheme 4.24. The selenium linker in chromene synthesis according to Nicolaou et al. [172].
or cyclative (! 160) cleavage can be realized [176] (Scheme 4.25). The seleno
linker has also been used in the synthesis of 2-deoxy glycosides, orthoesters, and
allyl orthoesters [177]. Finally, the SEM linker for the attachment and detachment
of alcohols takes advantage of the substitution of phenylselenyl groups at seleno
acetal resins [21].
Scheme 4.25. The selenium linker in the indoline synthesis by Nicolaou et al. [176].
4.3.8
Triazene-based Linkers
Inspired by the use of triazenes in the total synthesis of vancomycin [178] and the
pioneering work of Moore and coworkers [179, 180] and Tour and coworkers [181]
in the synthesis of triazenes on a solid support and the nal detachment to give
iodoarenes (180-I), a whole set of triazene-based linkers (Table 4.9) has been de-
veloped [182].
The chemistry of diazonium salts provides tremendous opportunities for the
construction of a wide range of aromatic compounds. Triazenes, which have been
used as traceless linkers for arenes as shown in Sect. 4.5.5, provide both interest-
ing new possibilities for activation of the ortho-position of the arenes and are ideal
synthons for diazonium salts. Triazenes are stable toward light, moisture, and
bases; however, they are cleaved by Brnsted acids and certain Lewis acids to give
diazonium salts and amines.
84 4 Linkers for Solid-phase Synthesis
Iodoarenes [179]
Two linkers based on triazene chemistry have been developed. While the T1
linker system consists of 3,3-dialkyl-1-aryl triazene bound to a support via the alkyl
chain (Scheme 4.26), the T2 linker family is based on immobilized aryl diazonium
salts.
The triazene T1 linker has been successfully used as a linker for arenes (see
also Sect. 4.5.5). Until now, approximately 100 dierent anilines (168) have been
immobilized. Functionalization on the bead has been demonstrated extensively.
These immobilized diazonium derivatives are stable towards various reaction con-
ditions, such as alkyl lithium reagents, reducing agents, and oxidizing reagents.
However, acids cleave the triazenes to give the amine resin (170) and the modied
aryl diazonium salts (173).
The latter can be transformed into various dierent products giving modied
arenes in high yields and purities [purities > 9095% according to gas chroma-
tography (GC), NMR, and high-performance liquid chromatography (HPLC)] di-
rectly at the cleavage step [183] (Scheme 4.27).
The diazonium salts, for example, can be reduced to the hydrocarbon (180-H) in
THF with the aid of ultrasound [182]. The latter facilitates this reduction due to a
radical pathway. A new reagent for this reduction was found to be trichlorosilane
[184]. This is not only a source of traces of hydrochloric acid, which cleaves the
86 4 Linkers for Solid-phase Synthesis
triazene moiety, but, as a hydride donor, it is also able to reduce diazonium ions
cleanly (Scheme 4.27 and see Scheme 4.64).
As already shown by Moore and coworkers [180] and Tour and coworkers [181],
addition of methyl iodide to a triazene resin at elevated temperature (110 C) gives
rise to aryliodides (180-I) (Nu I) in excellent yields. Furthermore, aryl halides
(180-X) (X Cl, Br, I) are readily available by the action of lithium halides in the
presence of an acidic ion exchange resin or with the corresponding trimethylsilyl
halide at room temperature [190]. A mixture of acetic anhydride and acetic acid
produces phenol acetates (180-OAc) [186]. Although quite exible in the range of
possible electrophiles that may be employed, the most striking feature was the de-
velopment of a cleavage cross-coupling strategy [187]. Starting from modied tri-
azene resins, a one-pot cleavage cross-coupling reaction was conducted with two
equivalents of triuoroacetic acid in MeOH at 0 C to give a diazonium ion. In situ
coupling with various alkenes (182) in the presence of catalytic amounts (5 mol%)
palladium(II) acetate furnished the corresponding products (183) in excellent yield
and purities. Using palladium on charcoal as the catalyst has the advantage of de-
creasing palladium contamination, as well as providing a subsequent hydrogena-
tion option [187]. Multicomponent Heck reactions (domino Heck DielsAlder re-
action) are possible in this context and lead to further diversication [187]. In the
examples above, the diazonium group, upon cleavage from the resin, is lost as
dinitrogen. However, a suitable nucleophilic ortho-substituent favors cyclization
to give heterocyclic structures. Benzotriazoles, for example, are accessible from
o-aminoaryl-substituted triazenes [192]. Other heterocyclic systems such as cin-
nolines (178) are available by a cleavage Richter reaction strategy, which starts from
o-alkynylaryl triazenes. Cleavage was conducted with aqueous hydrogen chloride
or hydrogen bromide in acetone or dioxane at room temperature to produce vari-
ous cinnolines (178) in a library format with up to 95% yield and with a range of
purities between 60% and 95% [191]. In addition to these methods, reduction
of diazonium salts gives rise to the formation of hydrazines (177) [199], which are
important building blocks for the synthesis of other heterocyclic compounds.
Whereas the T1 linker involves the immobilization of a diazonium salt on an
amine resin, the T2 linker represents the reverse of this concept. Thus, an immo-
bilized diazonium salt (190) was prepared from Merrield resin (3) in two steps,
and subsequent additions of primary and secondary amines generated triazenes
(191). In addition, attachment of hydroxylamines, hydrazines, sulfoximines, and
phenols (to give azo coupling products) proceeds equally well (Scheme 4.28).
Secondary amines can be cleaved directly [193] or after modication from the
resin. Primary amines can be derivatized on the free NaH functionality and can
therefore be modied to an array of products [194, 195, 198]. Thus, ureas (198)
[194], thioureas (197) [195], guanidines (196) [195], and carboxamides (199) [194]
were prepared in excellent yields (Scheme 4.29).
While the cleavage of trisubstituted triazenes gives rise to the formation of sec-
ondary amines in excellent yields [193], the cleavage of disubstituted triazene (194)
gives rise to aliphatic diazonium salts [183, 196, 197]. The diazonium ion thus
formed undergoes substitution with the nucleophile present in the reaction mix-
4.3 Linker Families 87
ture. Therefore, alkyl halides (202), alcohols (201-OH), and alkyl esters (201-
OCOR) can be formed by cleavage with trimethylsilyl halides (X I, Br, Cl) or
carboxylic acids (X for example OAc, OTfa) [200] [183, 196, 197]. The regio-
selectivity of the cleavage can be explained by the presence of one tautomer of the
triazene in which the hydrogen atom is next to the arene ring. Overall, this reac-
tion sequence provides a substitution of an amino group with oxygen or a halogen
(Cl, Br, I).
Various immobilized diazonium salts have been prepared [201], thermo-analyti-
cally characterized [202], and used for the linkage of nucleophiles. The structure of
the salts clearly inuences the stability of the diazonium moiety. The thermally
88 4 Linkers for Solid-phase Synthesis
Electrophiles
Nucleophiles 0 0 0 0 0
Fluoride 0 0 0 0 0
hn () 0 0 0 0
Oxidative conditions 0 0 0 0 0/
Reductive conditions 0 0 0 0/
a (): specially designed linker; : cleavage; : partial cleavage;
0: no cleavage.
stable diazonium ion (194a) (Z Cl, Y CH2 O) [t1=2 (25 C) > 100 days] is also
capable of scavenging various nucleophiles (amines, phenols, and anilines) [198].
4.3.9
Orthogonality Between Linkers
The orthogonality of linkers is important for the design and execution of both
simple and complex reaction sequences performed on a solid support. Recently, an
entire set of innovative linkers and cleavage strategies has been disclosed, which
enables the full set of orthogonality to be produced.
As discussed above, each linker family is sensitive toward a certain spectrum of
cleavage conditions and is therefore stable to dissimilar conditions. Since most of
the linkers are based on well-established protecting groups, Table 4.10 can be used
for the determination of orthogonality. For example, benzyl-type linkers, which are
mostly cleaved by electrophiles and are stable toward nucleophiles, can be com-
bined with ester-based protective groups.
4.4
Cleavage
In this section, various methods and reagents for the cleavage of linkers are pre-
sented. In most cases, cleavage of linkers is conducted with protons. However,
other electrophiles, photones, oxidizing and reducing reagents, and nucleophiles
can be used in many cases.
4.4.1
Electrophilic and Nucleophilic Cleavage
Cleavage of linkers can be conducted with various kinds of electrophiles and nu-
cleophiles (Tables 4.11 and 4.12). The most popular cleaving reagent is triuoro-
4.4 Cleavage 89
Tab. 4.11. Typical electrophiles and Lewis acids used for detachment.
HF [203205]
HF Wang linker [33] Thiols
HF Anisole PAM resin [35] Carboxylic acids
HF Cresol p-Acyloxy BHA Amides
resin [62]
CF3 SO3 H [206]
TFA (0.1% to neat) CH2 Cl2 Various linkers
TFA p-Cresol [205, 207, 208]
TFA p-Cresol, Me2 S [205]
TFA Anisole (PhOMe)
TFA (25%) Et3 SiH [209]
TFA Ethanedithiol [210]
(EDT)
TFA iPr3 SiH [211]
TFA Et3 SiH [212, 213]
TFA PhSMe [214]
HCl Dioxane Ketals [103] Ketones
CF3 SO2 OSiMe3 [215]
HBr/Ac2 O [216, 217]
AcOH CH2 Cl2 [66]
AlCl3 CH2 Cl2 /MeNO2 [218]
Et2 AlCl [219]
Me3 SiCl [194]
acetic acid in various solvents and concentrations. Because of its low boiling point,
removal is readily achieved. Besides this reagent, various other acids have been
used. Anhydrous HF, quite a toxic reagent, or triic acid are required for more
stable linkers. A mild reagent is trimethylsilyl chloride, which solvolyzes slowly to
HCl and hexamethyldisiloxane.
Typical nucleophilically cleavable linkers are the REM (regenerated Michael ac-
ceptor) linker (203) [220225] and Dde (Dimethyldioxocyclohexylidene)ethyl group
(204) (ADCC anchor) [226228], which are linkers for tertiary and primary amines,
respectively (Fig. 4.4).
4.4.2
Oxidative/Reductive Methods
orthogonality with other cleaving methods, a drawback for the use of oxidative or
reductive reagents is the necessity to remove excess reagents or byproducts.
4.4.3
Photocleavable Linkers
Light-induced cleavage oers new possibilities for orthogonal use of linkers (Sect.
4.3.8) and acid- or base-labile protecting groups (Table 4.15).
The rst photolabile linker (215), which was based on the o-nitrobenzyl protect-
ing group, was developed by Rich and Gurwara [238, 247] for the synthesis of pro-
tected peptides (Scheme 4.30). This linker was developed further because in the
synthesis of the original linker, a nitration of Merrield resin was involved, thus
leading to nitration of excess phenyl rings. Therefore, mostly preformed handles
have been used. In all cases, upon ultraviolet (UV) photolysis the photo byproduct,
a nitrosobenzaldehyde, is also photoactive and causes a reduction in cleavage yield
from the support. To circumvent this problem, an additional methyl group was
introduced to give linkers, which lead to the photoreactive nitrosoactophenone
system [248]. Moreover, introduction of methoxy groups para to the nitro groups
(vanilline-type linker) [249] improves cleavage properties and the compounds are
typically released within 3 h in >90% yield and >95% purity under neutral con-
ditions [241, 250, 251]. Therefore, these linkers are biocompatible and are suitable
for Fmoc solid-phase peptide synthesis (Fmoc-SPPS) (however, see [252]). Various
modications and improvements [239, 241, 248, 253258] and applications [259]
of this linker type have been reported [260263].
Scheme 4.30. The prototype of a photolabile linker according to Rich and Gurwara [238].
Benzoins [264] and related b-keto systems (phenacyl esters) [245, 246, 265]
have been used as linkers for over 25 years now. A novel safety-catch linker (217)
(Sect. 4.5.1) based on the benzoin-protecting group has been utilized to anchor
carboxylic acids. The use of a dithiane-protected 3-alkoxy-benzoin allows for elabo-
ration of molecules, linked as esters to the secondary hydroxy function of the ben-
zoin, prior to deprotection of the dithiane and photolytic cleavage [244, 266]
(Scheme 4.31).
A new photolabile linker (126) based on a thiohydroxamic acid has been shown
to be an ecient traceless linker yielding an aliphatic CH bond upon photolysis
92 4 Linkers for Solid-phase Synthesis
[240] Amines
[241] Amines
[243] Alcohols
[161] Methylarenes
4.4 Cleavage 93
[162] Methylarenes
[120] Methylarenes
4.4.4
Metal-assisted Cleavage
promoted or -catalyzed reactions has some advantages over other cleavage meth-
ods. Since most protecting groups and functionalities are resistant towards palla-
dium complexes, a selective surgical removal is frequently possible. In addition,
intermediate p-allyl and s-arylpalladium complexes can in principle be used for
further derivatization with the use of appropriate linker types.
Boronates can be reduced to the corresponding hydrocarbons using silver salts,
making this linker as a traceless linker (Sect. 4.5.5) for arenes [147]. Similarly, aryl
sulfonates are reduced by the action of palladium or a reducing medium [166].
Allylic substitution reactions The cleavage of polymer-bound allyl esters with pal-
ladium catalysts provides a general access to p-allyl complexes, which can react
with various nucleophiles. This approach has been used in the development of p-
allyl-based linkers (Sect. 4.3.2).
Schurer and Blechert [273] reported on an eneyne cross metathesis (see also
Sect. 4.4.4.2) and a subsequent cleavage in the presence of various nucleophiles to
yield the corresponding functionalized dienes (224) (Scheme 4.32).
Heck reactions Cleavage by an ensuing Heck reaction was developed utilizing the
T1 triazene linker [187]. Upon cleavage with triuoroacetic acid, a diazonium ion
is formed which can couple to an alkene under palladium catalysis (Scheme 4.34).
The coupling proceeds well with simple terminal alkenes, styrenes, as well as di-
and even trisubstituted alkenes. The coupling with 1,3-cyclohexadiene eventually
yields a biaryl, apparently by a facile dehydrogenation of the primary coupling
product. The advantage of this process is clearly the possibility of using volatile al-
kenes (and alkynes) without contamination of any salt or other less volatile by-
products, particularly with the use of palladium on charcoal as the catalyst. In this
case, a subsequent hydrogenation is also possible [187].
Stille couplings The intermolecular Stille reaction of aryl halides with immobi-
lized stannanes (Scheme 4.21) provide coupling products in good yields, as dem-
onstrated by Kuhn and Neumann [169]. In addition, the stannylated resin pro-
duced in the cleavage coupling can be recycled. Although the products obtained
were not contaminated by any stannane, they were separated from an excess of the
reactive electrophiles that had to be applied in the cleavage-coupling step. The in-
tramolecular variant, which was used by Nicolaou et al. [163] to produce macro-
cyclic ring systems such as the natural product (S)-zearalenone (231), does not
have this drawback (Schemes 4.35 and 4.36).
of the boronic acid derivative to give homocoupling products, the need to apply
additional ligands, and the low volatility of the boronic acid derivative, a more or
less tedious work-up is required after these types of transformations. A few studies
have proven that certain functionalities, when generated during cleavage, may act
as leaving groups for a subsequent Suzuki reaction. One of these is the diazonium
group, which can be generated by cleavage of the triazene T1 linker. While the
Heck-type coupling with alkenes gives good yields of the desired products [187]
(Scheme 4.34), the analogous reaction with phenylboronic acid appears to be di-
cult because of work-up problems [187] (Scheme 4.37).
Arylmethyl(homobenzyl)ethylsulfonium salts [154] and aryl boronates [148]
have been used as precursors for a cleavage/Suzuki approach (Schemes 4.15 and
4.20).
4.4.5
Unusual Cleavage Methods
4.5
Linker and Cleavage Strategies
catch linkers (Sect. 4.5.1), cyclative cleavage strategies, and fragmentation reactions
have been presented.
Cleavage of linkers might be monofunctional or with functionalization of the
linking site, whatever is required (Fig. 4.5). In the latter case, which is also known
as the multifunctional cleavage strategy (see Sect. 4.5.6), the membership of library
compounds is multiplied by the number of building blocks or functional groups
that can be incorporated into the cleavage step. Hence, an anchoring group capa-
ble of functionalization and traceless linking is a versatile tool for enhancing di-
versity in a given system.
4.5.1
Safety-catch Linkers
The safety-catch consists of a linker that, during synthesis, is inert towards the
cleavage conditions and has to be activated (Table 4.16). Ellman and coworkers
[286288] have used this strategy in various applications, such as in the sulfona-
mide linker proposed by Kenner et al. [289]. Since safety-catch means the activa-
100 4 Linkers for Solid-phase Synthesis
tion of the linker before cleavage, such a system can be applied to monodirectional
linkers, such as traceless linkers (Sect. 4.5.5) or to multifunctional linkers (Sect.
4.5.6), as well as to cleavage-cyclization strategies.
Recent applications of the Kenner linker have been shown in the synthesis of
vinylsulfones [149] (Scheme 4.42) or in the synthesis of amides [294, 295].
The oxidation of suldes to sulfones [150] or the reduction in the opposite di-
rection [64, 65] are popular methods for the design of a safety-catch protocol.
A resin-bound thioether can be activated by oxidation to insure the nucleophilic
cleavage of a phenoxide moiety [150] (Scheme 4.43).
A safety-catch p-allyl-cleavable linker was developed for the synthesis of DNA on
solid support. Starting from a resin carrying an Alloc-protected amino group frag-
ment, conventional phosphoramidite chemistry was carried out to build up the
desired nucleotide (267). Removal of the Allyloxycarbonyl (Alloc) group under palla-
4.5 Linker and Cleavage Strategies 101
[290] Amides
[65] Esters
[293] Amides
[236] Diketopiperazines
Scheme 4.42. The Kenner safety-catch linker according to Overkleeft et al. [149].
Scheme 4.44. A safety-catch palladium activated linker by Lyttle et al. [296]. Nuc nucleotide.
4.5.2
Cyclative Cleavage (Cyclorelease Strategy)
In general, the starting material for cyclative cleavage is anchored to the resin
via a leaving group (Scheme 4.45). An internal nucleophile provides the ring clo-
sure by displacement of this leaving group either directly or after activation. Apart
from nucleophilic attack, cyclative cleavage can be achieved, for example, by Stille
(Scheme 4.35) or WittigHorner reactions [304, 305].
Since the intramolecular reaction is by far faster than any intermolecular step,
this strategy provides an additional purication step since only the cyclized struc-
tures are detached from the bead. Incomplete building blocks will remain on the
solid support.
Cyclative cleavage has to be distinguished from cases in which the cyclization
occurs in solution after cleavage because unsuccessful cyclization precursors re-
main in the liquid phase (Sect. 4.5.3).
In most cases, the precursor for the cyclization cleavage is linked via an ester
bond to solid support whereby the nucleophile is based on an amine functionality.
The product thus formed is therefore a cyclic amide or analog. Indeed, one of
the rst examples of this type was the pioneering benzodiazepine synthesis by
Camps and Castells in 1974 [306]. In this case, the benzodiazepine ring formation
proceeded by simultaneous cleavage from the bead. A vast list of examples has ap-
peared since that suggests various kinds of heterocyclic systems. Examples are lac-
tams, hydantoins [307319], thiohydantoins [315], oxazolidinones [320, 321]
(Scheme 4.46), diketopiperazines [322331], benzodiazepines and benzodiazepi-
nones [306, 318, 332334], pyrazolones [335, 336], diketomorpholines [323], tetra-
mic acids [337340], quinazolinediones [341], dihydropyrimidine-2,4-diones [342],
quinolinones [343], tetrahydrocarbolines [326], thiazoles [317], perhydrodiazepin-
ones [327], sulfahydantoins [344], and benzimidazoles [345].
Scheme 4.47. The phosphonium linker for the synthesis of indoles by Hughes [363].
A new safety-catch linker (Sect. 4.5.1), which is based on the lability of the dia-
zenyl carbonyl derivative, was developed for the synthesis of monoketopiperazines
(285) (piperazinones). After elaboration of the anchoring system, the dimethylben-
zyl group was removed by TFA and the hydrazide thus formed was oxidized by
copper(II) acetate furnishing a diazenyl group. This motif was attacked intra-
molecularly by the primary amine functionality, which resulted in cleavage of the
product from the resin [236] (Scheme 4.48). The potential of this linker class for
the synthesis of heterocyclic amidic structures is discussed in Chapter 22.
A cyclative approach for the synthesis of thiazoles (288) begins with the acylated
Rink amide resin (286) which has been transformed into the thioamide using Law-
esson reagent. The S-alkylation with a-bromo ketones (287) proceeds with con-
comitant cleavage from the resin [347] (Scheme 4.49).
The intramolecular Stille reaction proceeds similarly under cyclorelease con-
ditions (see above; see also Schemes 4.35 and 4.36).
4.5.3
Cleavage-cyclization Cases
4.5.4
Fragmentation Strategies
Scheme 4.53. The Reissert complex strategy used by Kurth and coworkers [351].
Recently, Gibson et al. demonstrated that arenes can be attached to a solid sup-
port using chromium arene complexes (301) [353; cf. 354] (Scheme 4.54). This
method even allows, at least from a theoretical point of view, the complete varia-
tion of the arene backbone; however, modication of the arene ring system might
be limited to a certain extent. It has already been shown that various other metals
or unsaturated molecules (e.g. alkynes with dicobaltoctacarbonyl fragments) can
lead to interesting examples for the design of new linkers [355]. Linkers based on
arene complexes have been patented [356, 357].
Scheme 4.54. The p-complexation for the linkage of arenes by Gibson et al. [353].
The fragmentation-cleavage strategy has been used for the synthesis of pyrazoles
(308) [358]. An a-silylated N-nitrosamide (304) has been rearranged via a thermal
silyl shift to an azomethine ylide (306), which then reacted with the dipolarophile
(305); subsequent fragmentation yielded the target pyrazoles (308) (Scheme 4.55).
4.5 Linker and Cleavage Strategies 109
Scheme 4.55. The synthesis of pyrazole (308) according to Komatsu and coworkers [358].
Further examples of the fragmentation strategy have also been published [359
362]. The semisynthesis of vancomycin clearly demonstrates the advantage of this
kind of linker strategy [97].
4.5.5
Traceless Linkers
The term traceless linker has led to ambiguous interpretations in the past. Many
authors have claimed their linkers to be traceless because the term has, in the
past, been quite fashionable even when the reported linker was used to immobi-
lize and release amines (which upon cleavage carry a hydrogen).
Traceless linking is nowadays considered to be leaving no functionality,
meaning for arenes and alkanes that only a CaH bond remains at the original po-
sition of attachment (Scheme 4.57). A broadening of this denition to OH or NH
groups is not useful, because otherwise every linker derived from polymeric pro-
tecting groups would have to be regarded a traceless linker.
(Scheme 4.59). So far, this linker has not been fully explored and is limited in its
range of functionalized arenes, since a phenyl substitution instead of biphenyl re-
sults in the formation of disuldes.
However, the most prominent anchors for traceless linkage for arenes (Table
4.17) are based on silyl linkers [124, 125, 369, 370] (Scheme 4.60) (Sect. 4.3.5). The
generation of a diverse benzodiazepine library by Plunkett and Ellman [131] has
shown clearly the advantages of this type of detachment since no additional func-
tionalities were retained in the nal molecules, which might bias the library.
Starting from an immobilized stannane, palladium-catalyzed coupling with acid
chlorides, deprotection of the aniline protecting group, acylation of the aniline
with a series of Fmoc amino acids, Fmoc deprotection, and cyclization aorded
resin-bound benzodiazepines, which were cleaved from the support using triuo-
roacetic acid. Improvements in the chemoselectivity of the cleavage step the silyl
linker produces a substantial amount of the silyl arene upon cleavage were ac-
complished using a germanium linker, which is more labile towards acids [131,
Scheme 4.60. The traceless silicon linker according to Veber and coworkers [124].
112 4 Linkers for Solid-phase Synthesis
[126]
363] (Scheme 4.61). For example, the synthesis of the silyl linker has been opti-
mized using preloaded handles to assist the coupling of the product to the resin.
Other arylsilyl linkers have also been used (Table 4.6) to facilitate loading, synthe-
sis, and/or detachment from the support [126128, 130, 134136].
Furthermore, silicon linkers can be used for the attachment for allylsilanes,
which can be cleaved to alkenes in a traceless fashion [144].
The use of sulfones as suitable anchoring groups for alkanes in soluble polymer
chemistry has been reported previously [155, 156] (Table 4.18). After oxidation of a
sulde to a sulfone (338), treatment of the latter with sodium/mercury gives rise to
the formation of the parent hydrocarbon (339) in high yields. However, aqueous
work-up is necessary to provide the pure product (339) [155] (Scheme 4.62).
Another traceless linker type was independently developed by the groups of
Nicolaou et al. [164] and Ruhland et al. [165]. Starting from lithiated selenium
polystyrene (146), readily available from metallated polystyrene and selenium re-
agents, reaction with iodoalkanes led to the smooth formation of alkylated com-
pounds. The cleavage can be conducted in such a way as to give alkenes (150)
114 4 Linkers for Solid-phase Synthesis
[163] Alkenes
[368] Ketones
Scheme 4.62. The sulfone traceless linker for alkanes according to Janda and coworkers [156].
4.5 Linker and Cleavage Strategies 115
upon treatment with hydrogen peroxide. Alternatively, tin hydride reduction leads
to the generation of alkanes (149) (Scheme 4.22).
Electron-poor aryl sulfonates are suitable candidates for oxidative palladium
insertion. Hence, immobilized phenol sulfonates (129) have been employed in a
palladium-catalyzed reductive cleavage using formic acid to yield arenes (130) with
overall traceless cleavage [152] (Scheme 4.17). It might be anticipated that this type
of linker is also suitable for functionalization (see below).
The phosphoruscarbon bond in phosphonium salts is readily cleavable by the
aid of a base in the absence of an aldehyde. Hence, the polymer-bound phospho-
nium salt (340) gives direct access to methylarenes (341). An interesting feature of
this linker is the fact that carbonyl compounds can be olenated, which leads to a
cleavage-olenation linker system [364, 365] (Scheme 4.63).
Scheme 4.63. The phosphonium linker for methylarenes according to Hughes [364].
Tin hydride reagents are versatile tools for the functionalization of alkenes
and alkynes. Based on this concept, Nicolaou and coworkers [163] developed a
polymer-bound tin hydride (142), which reacts in a hydrostannylation reaction
with alkynes to give alkenylstannanes (143). After further transformation, the lat-
ter undergo proteolytic traceless cleavage to yield unsubstituted alkenes (145) [163]
(Scheme 4.21).
The decarboxylation of appropriately substituted arenes [371] and alkanes [343,
368, 372375] has been used to generate the parent hydrocarbons. Since the neigh-
boring group eect is essential, limitation to special substrates is required.
One possible method for converting functionalized arenes into the correspond-
ing hydrocarbons is the reduction of diazonium compounds [182]. Hence, the syn-
thetic utility of the triazene linker as a traceless anchor for arenes has been dem-
onstrated by Brase et al. [182] using short reaction sequences. Thus, cinnamic
esters were synthesized in a sequence starting from the iodoarene resin (343).
Heck coupling with acrylates using palladium catalysis aords an immobilized
cinnamate. This can be detached either directly or by a sequence of transforma-
tions yielding to an allyl amine in a traceless fashion either using trichlorosilane
[184] or a HCl/THF mixture [182], to give the products 342 and 344, respectively,
in high yields and without further purication or aqueous work-up (Scheme 4.64).
Acyl aryl diazenes are known to fragment upon treatment with nucleophiles, a
strategy which was used for a linker for carboxylic acid derivatives [119] (Scheme
4.11). Waldmann and coworkers recently developed a traceless linker for arenes
based on this methodology [237]. Starting from a hydrazide resin (345), which
116 4 Linkers for Solid-phase Synthesis
Scheme 4.65. The hydrazide linker according to Waldmann and coworkers [237].
A new traceless photolabile linker has been published using a hydroxamic acid
derivative (83) [120].
The cleavage of specially designed polymeric benzyl-type protecting groups has
been achieved using heterogeneous palladium black. In these cases, the catalytic
hydrogenation furnishes methyl-substituted arenes as side products or targets. An
early example takes advantage of the properties of the MeO-PEG-type support for
the synthesis of di- and higher oligosaccharides (Scheme 4.66). It is interesting to
note that the DOX linker enables the cleavage of the PEG structure, leaving the p-
methylbenzyl group attached under certain conditions [376] (but see [29]) (Scheme
4.66).
Similarly, cleavage from polystyrene resins was achieved using homogeneous
palladium catalysis (palladium acetate) either with formate reduction [366]
(Scheme 4.67) or under an atmosphere of hydrogen [377] to yield methyl arenes.
The chemistry of traceless linkers is a fast-emerging eld in the intensively in-
vestigated area of solid-phase organic synthesis [ for reviews, see 183, 197, 378].
Although some confusion about the denition or classication has been related to
4.5 Linker and Cleavage Strategies 117
this linker type, and therefore a careful designation has to be made, it is now clear
that this anchoring mode will play an important role in the design and syntheses
of drug-like molecules.
4.5.6
Multifunctional Cleavage
eliminate or to be removed by for example scavenger resins, etc.) and should not
interfere with the functionalities of the library compounds.
The addition of nucleophiles such as amines, alkoxides, thiolates, and carbon
nucleophiles to carbonyl groups leading to modied carbonyl moieties has been
used widely for multifunctional cleavage. Hence, anchoring a Weinreb amide to
the resin and cleaving with Grignard reagents leads to the formation of ketones
[383]. Alternatively, thioesters provide access to amides or ketones, whereas sulfo-
nate esters such as 359 provide access to amines and thiols (360) [381] (Scheme
4.69), halides [382], azides [382], or acetates [382].
4.5.7
Linkers for Asymmetric Synthesis
The use of enantiomerically pure drugs has increased within the pharmaceutical
industry over recent years. It is assumed that the market was well over US$120
billion in the year 2000 [386, 387], and approximately 60% of all marketed drugs
are enantiomerically pure. Chiral building blocks have been used extensively in
solid-phase synthesis (Fig. 4.6) and have been incorporated into modern small-
molecule compound libraries.
However, asymmetric synthesis based on linking strategies, with either diaster-
eoselective or enantioselective methods, has been more or less neglected. The ob-
vious advantage of a linker, which induces stereoselectivity, is the ease with which
the auxiliary can be removed and recycled [388].
Pioneering work on asymmetric synthesis on an insoluble support was de-
scribed in 1972 when Kawana and Emoto [389, 390] reported the synthesis of an
atrolactic acid on a polymer containing the sugar 1,2-o-cyclohexylidene-a-d-xylofur-
anose (365). Shortly after this, Lezno and coworkers [391, 392] demonstrated the
synthesis of a-chiral cyclohexanone derivatives using polymer-bound imines (366).
4.6 Linkers for Functional Groups 121
These early examples have demonstrated that polymer-bound chiral auxiliaries are
suitable both for anchoring of organic molecules and for the induction of asym-
metry. Furthermore, the recycling of chiral auxiliaries can be conrmed.
In recent years, a series of auxiliaries has been immobilized on solid support,
including oxazolidinones (372) according to Evans [231, 393398], oxazoles (367)
[399], mono- (368) [400] and bisalkoxymethylpyrrolidines (369) [401], as well as
SMP/SAMP auxiliaries (370, 371) by Enders and coworkers [402, 403]. Alter-
natively, polymer-bound chiral sulnamides [404], sulfoximines by Hachtel and
Gais [405], imines [406], and amines (as galactosylamine) (Scheme 4.71) have
been used.
4.6
Linkers for Functional Groups
4.6.1
Linkers for Nitrogen Functionalities
Scheme 4.72. The Rink resin as a linker for primary amines [82].
Alternatively, benzyl carbamates can be attached to the solid support, and hy-
drogenolytic cleavage can be used to detach the molecules, which are then usually
left with a nitrogen functionality (cleavage of CaO bond) [29, 447]. The polymers
in these cases are formally immobilized Benzyloxycarbonyl Cbz (Z) groups. Inter-
estingly, TentaGel and polystyrene give the products in similar yields under iden-
tical conditions [17] (Scheme 4.73). Benzylic linkers can also be used advant-
ageously in the presence of other benzylic protecting groups, since they can be
removed in the same step [29].
Scheme 4.73. Detachment of peptides from polymeric benzyl-type protecting groups [17].
124 4 Linkers for Solid-phase Synthesis
An anchor for primary amines, which is cleavable under basic conditions (Table
4.12), is the dimedone-based Dde group (204) [226, 228, 425] (Fig. 4.8).
Anilines are released from the PAL linker, the Rink linker [82], the PhFl linkers
[437, 448], the p-benzyloxybenzylamine (BOBA) resin [32], and carbamate linkers.
The latter type has also been used as a safety-catch version [168] (Sect. 4.5.1).
A novel dialkylhydrazine linker (381), which is stable toward organometallic re-
agents, has recently been reported. A series of a-branched primary amines was
synthesized by attachment of various aldehydes, the resulting hydrazones (384)
were modied via 1,2-addition of organolithium reagents furnishing trisubstituted
hydrazines (385) which in turn were cleaved from the solid support to yield acy-
lated a-branched primary amines [449] (Scheme 4.74).
Carbamates are suitable linkers for secondary amines, which can be released by
diluted TFA [17, 345, 451, 452], as demonstrated in a Hantzsch dihydropyridine
synthesis by Breitenbucher and Figliozzi [409] (Scheme 4.75).
Scheme 4.76. The T2 linker for the synthesis of secondary amines on solid support [198].
Scheme 4.77. The REM linker for the synthesis of tertiary amines [223].
used [384]. Furthermore, various other linkers are suitable for tertiary amines
[224, 225, 465].
in high yields. However, owing to their instability, the diazonium ion is mostly
transformed in situ to a new functional group [467] (Scheme 4.82) (see also
Scheme 4.27).
Scheme 4.82. In situ cleavage of triazenes for the synthesis of diazonium salts [467].
Scheme 4.83. Azo coupling through immobilized diazonium salts according to Das et al. [469].
The THP (Table 4.4) [99] and other acetal linkers [100] are also suitable for the
attachment of indoles. Alternatively, heterocyclic amines with a free NaH func-
tionality can be linked via an acetal, which can be cleaved rst under acidic and
then under basic conditions [456] (Scheme 4.86).
4.6 Linkers for Functional Groups 131
Scheme 4.86. Synthesis of indoles (434) using a diol linker by Gmeiner and coworkers [456].
Indole synthesis can be realized on solid support via a sulfone linker. Advanta-
geously, the heterocyclic core was installed in this case via a palladium-catalyzed
process [302] (Scheme 4.87). Similarly, benzimidazoles can be released via a retro-
Michael addition from sulfonate resins [471] (Scheme 4.88).
Other examples for the synthesis of aromatic nitrogen heterocycles include the
synthesis of pyridines and tetrahydropyridines [452], benzimidazoles [471], isox-
azolines [472], isoquinolines via Reissert complexes (Scheme 4.53) [351, 352],
quinoxalinones [49], benzisoxazoles [473], and imidazoles on trityl linkers [78].
The second strategy is dedicated to cleavage conditions, which result in the in-
stallation of the heterocyclic core; this strategy obviously belongs to the cleavage-
132 4 Linkers for Solid-phase Synthesis
4.6.2
Linkers for Carbonyl Functionalities
Scheme 4.95. The solid-phase BMPSE group by Kurth and coworkers [483].
Other reports describe the use of arylhydrazides as linkers [119, 484] and also
other linkers [372, 373, 485489].
Scheme 4.96. The use of esters as linkers for benzoic acid derivative by Kondo et al. [116].
Scheme 4.98. Cleavage and carbonylation of the triazene T1 resin (465) [187].
Scheme 4.99. Synthesis of carboxylic esters via triazene T2 resins [183, 196, 197].
Lactones are accessible via thioesters [115], esters [346, 400], or amides [490]
with a suitably positioned internal nucleophile or electrophile. In addition, alkyl
acetates are also accessible from sulfonates upon cleavage with sodium acetate
[382].
Linkers for primary amides Primary amides can be synthesized on solid support
using, for example, the Rink resin [68, 494]. This support is suitable for various
reaction conditions, as outlined in Schemes 4.100 and 4.101.
Scheme 4.100. Detachment of primary amides from a Rink resin by Wang et al. [494].
4.6 Linkers for Functional Groups 137
Scheme 4.101. Synthesis of Nenitzescu products according to Wilson and coworkers [495].
Linkers for secondary amides Linking by the NaH of a secondary amide bond has
been developed, hence leading to so-called backbone amide linkers (BAL). Origi-
nally designed for the NaH protection of amide bonds to circumvent b-turns and
other problems during peptide synthesis, the Hydroxymethylbenzoic acid Hmb
group [496, 497] or the Silylmethylmethoxyhydroxybenzyl SiMB group [498] can
also serve as linkers for SPPS.
Although important in the synthesis of cyclopeptides, they also play an impor-
tant role in the synthesis of small molecule libraries containing amide moieties.
Barany and coworkers have described an application of a backbone amide linker
for the synthesis of oligopeptides [40, 83, 424, 499, 500] (Scheme 4.102), glycopep-
Scheme 4.102. Backbone amide linkage using the PAL linker by Barany and coworkers [83].
138 4 Linkers for Solid-phase Synthesis
tides, and peptide aldehydes [501] based on the peptide amide linker (PAL) con-
cept. Attachment of amines furnishes alkylated products (482), which in turn can
be acylated under peptide-coupling reaction conditions to give amidic structures
(484). The latter can be cleaved from the resin using acidic media to give amides
(485). This concept was used for the synthesis of benzodiazepine libraries by Ell-
man and coworkers [47] (Scheme 4.103), hapalosin mimetics [502], and further
rened to yield an alternative, TFA-stable benzylamine linker [503]. Benzylamine-
based linkers have encountered some problems during cleavage which also caused
the fragmentation of the whole linker. Furthermore, the reactivity is decreased be-
cause of steric hindrance [14, 241].
Recently, it has been shown that the BAL linker is also suitable as a safety-catch
linker (Sect. 4.5.1) for the synthesis oligosaccharides [454] (Scheme 4.104). Start-
ing from an aldehyde, reductive amination of a protected glucosamine and subse-
The T2 linker has recently been shown to be a versatile backbone amide anchor
[194]. Immobilized disubstituted triazenes were acylated with carboxylic acid an-
hydrides or chlorides to give amidic structures (482) [194] (Scheme 4.106).
A new backbone linker for b-lactams and secondary amides was presented using
a benzyloxyaniline linker, which is cleavable by ceric ammonium nitrate [508].
Secondary amides and ureas can be synthesized using supported hydroxyl-
amine, which can be cleaved by samarium(II) iodide [509].
Linkers for tertiary amides Linkers for tertiary amides are the ester anchors cleav-
able with secondary amines. In addition, the triazene T2 linker is also a precursor
for amides when cleaved by acid chlorides [193, 510].
140 4 Linkers for Solid-phase Synthesis
Linkers for carbamates Both the SASRIN and the indole linker are suitable for the
synthesis of carbamates, however the SASRIN linker can be cleaved with 5% TFA
in dichloromethane [41], whereas 50% TFA in dichloromethane is required for the
indole linker [85].
Linkers for ureas Urea derivatives, which are important biologically active com-
pounds and building blocks for organic syntheses, have been synthesized on solid
support using various strategies.
A useful linker for ureas is the phoxime resin (498) (phosgenated oxime resin)
[511, 512], which is cleavable at elevated temperature with amines to yield ureas
(Scheme 4.107). Other linkers for ureas are the SASRIN linker [41, 44] and other
systems [422, 513].
Scheme 4.107. The phoxime resin for the synthesis of urea derivatives [511].
The T2 linker has also being used for the attachment and modication of pri-
mary amines, yielding urea derivatives after mild cleavage (Scheme 4.108).
Aminosulfonylureas have been released from carbamate linkers using primary
and secondary amines (multifunctional: Sect. 4.5.6) [420]. The samarium(II)
iodide-promoted cleavage of acylated polymer-bound hydroxylamines gives general
access to urea derivatives [509].
Linkers for thioureas Thioureas have been synthesized using the T2 linker [194]
(Scheme 4.109).
Linkers for isothioureas Isothioureas have been detached from solid support
using the T2 linker under mild conditions [514] (Scheme 4.110).
Linkers for guanidines Guanidines are basic molecules with the capacity of form-
ing H-bonding interactions. They are therefore a promising class of potentially use-
4.6 Linkers for Functional Groups 141
Linkers for amidines Several techniques enable the synthesis of amidines on solid
support. In most cases, benzyl-type linkers such as the indole linker [85], the
Wang linker [525], or a carbamate system on the Wang linker [525] have been
used.
Linkers for hydroxamic acids Hydroxamic acids are important building blocks in
metalloproteinase inhibitors. Therefore, various linkers have been developed to sat-
isfy these requirements.
The Wang linker is, for example, suitable for the detachment of hydroxamic
acids as demonstrated in a cascade carbopalladation reaction [526] (Scheme
4.112).
Alternatively, the THP, the Wang resin [527, 528], the Rink linker [529], trityl
resin [530], PAL resin [528, 531], oxime resins [532], and others [285, 437, 533
537] are suitable linkers for hydroxamic acids.
T2 linker (Sect. 4.14.3.8) [466] (Scheme 4.113). Hydrazides can be cleaved from
solid support using the trityl linkers, for example.
Linkers for lactams The synthesis of b-lactams has been achieved using the hy-
droxyaniline linker [508] (Scheme 4.114).
4.6.3
Linkers for Ketones and Aldehydes
4.6.4
Linkers for Alcohols, Phenols, Ethers, and Ketals
Scheme 4.117. Synthesis of indolactam analogs according to Waldmann and coworkers [555].
ally left with an oxygen or nitrogen functionality (cleavage of CaO Bn and CaN
bond respectively) [29, 447]. The polymers in these cases are formally immobilized
Z or Cbz groups. Interestingly, TentaGel and polystyrene give similar yields under
identical conditions. Benzylic linkers can also be used advantageously in the pres-
ence of other benzylic protecting groups, since they can be removed in the same
step [29] (Scheme 4.119).
Scheme 4.119. Detachment of saccharides from polymeric benzyl-type protecting groups [29].
Finally, alcohols can be released from ester [117, 556], thioester [115], and amide
resins using reductive methods (Sect. 4.4.2, Table 4.13). Furthermore, the reaction
of Grignard reagents with thioesters has been reported to give tertiary alcohols
[115]. Other anchoring groups are enzyme-labile linkers [285] and uoride-
sensitive linkers [557]. Cyclic ethers (tetrahydrofurans) are accessible via an iodo-
lactonization approach [558, 559].
4.6.5
Linkers for Sulfur Compounds
Scheme 4.120. Synthesis of quinolones from avilylium salts according to Sato et al. [564].
with thiols [566]. However, other suitable linkers are the Wang linker, cleavable by
HF [33], Rink-type linkers [82, 567], thiocarbamates, cleavable by bases [568], dini-
troaryl linker, cleavable by nucleophiles [229]; and others [569, 570]. Because thiols
are prone to oxidation to give symmetrical disuldes, the latter were frequently
found during cleavage if oxygen had not been strictly excluded.
tached to a solid support using the corresponding benzyl alcohol resins and sul-
fonyl chlorides [575].
Scheme 4.123. A linker for sulfones according to Gais and Hachtel [405].
4.6.6
Linkers for Hydrocarbons
Linkers for hydrocarbons are important tools in combinatorial chemistry for the
synthesis of the lipophilic compounds required for modern drug research. As de-
scribed above, access to hydrocarbons can proceed via the formation of either a
CaC bond or a CaH bond. The latter strategy has been discussed in detail in
Sect. 4.5.5 (traceless linkers), since the introduction of a hydrogen atom clearly is
the prototype for this kind of linker.
4.6 Linkers for Functional Groups 149
4.6.7
Linkers for Aryl and Alkyl Halides
Aryl iodides have been synthesized by Moore et al. [179], starting from triazene
resin by the action of methyl iodide (Sect. 4.14.3.8) (Scheme 4.125). Aryl iodides,
bromides, and chlorides are also accessible from the triazene T1 linker using the
corresponding trimethylsilyl halide (Scheme 4.126) [190].
Scheme 4.126. Synthesis of aryl halides via the T1 triazene resin [190].
Scheme 4.127. Synthesis of alkyl halides via the triazene T2 linker [183, 196, 197].
4.6.8
Linkers for Heterocycles
Various methods are applicable in the synthesis of heterocycles [361, 362, 384,
584]. The cyclofragmentation of a certain class of sulfones leads to 3-arylbenzo-
furans [160].
4.6.9
Linkers for Reactive Intermediates
Reactive intermediates cleaved from a solid support can be used for a subsequent
functionalization. Thus, radicals, carbanions, and carbocations might then react
with additional building blocks. This multifunctional cleavage mode (Sect. 4.14.5.6)
has, for example, been used with the stannane, selenium (Sect. 4.14.3.7), and tri-
azene linkers (Sect. 4.14.3.8).
152 4 Linkers for Solid-phase Synthesis
4.6.10
Linkers for Other Functional Groups
4.7
Overview for Linkers for Functional Groups
Table 4.20 gives a short overview of the dierent linker families, as described in
Chapter X.
4.8
Conclusion, Summary and Outlook
In recent years, various new types of linkers have emerged. The design of a new
anchoring group can be essential for the success of a synthesis, especially for small
molecules on a solid support. Linker, cleavage conditions, and functional groups
are associated with each other. Therefore, the decision to use one specic linker
type has to be balanced with the requirements of the library to be synthesized.
Although the perfect or universal linker has not yet been developed, and
will prove unattainable, interesting new developments increase the exibility of
solid-phase synthesis by traceless (Sect. 4.5.5) and multifunctional cleavage (Sect.
References 153
References
G. Barany, J. Org. Chem. 1999, 64, Leznoff, Can. J. Chem. 1983, 61,
87618769. 14051409.
84 N. Thieriet, F. Guibe, F. Albericio, 104 R. Maltais, M. Berube, O. Marion,
Org. Lett. 2000, 2, 18151817. R. Labrecque, D. Poirier, Tetra-
85 K. G. Estep, C. E. Neipp, L. M. S. hedron Lett. 2000, 41, 16911694.
Stramiello, M. D. Adam, M. P. 105 C. M. Huwe, H. Kunzer, Tetrahedron
Allen, S. Robinson, E. J. Roskamp, Lett. 1999, 40, 683686.
J. Org. Chem. 1998, 63, 53005301. 106 G. T. Wang, S. Li, N. Wideburg, G.
86 H. Kunz, B. Dombo, Angew. Chem. A. Krafft, D. J. Kempf, J. Med. Chem.
Int. Ed. Engl. 1988, 12, 711712; 1995, 38, 29953002.
Angew. Chem. 1988, 100, 732734. 107 S. Q. Chen, K. D. Janda, Tetrahedron
87 H. Kunz, W. Kosch, J. Marz Lett. 1998, 39, 39433946.
(Orpegen GmbH), Patent No. 108 J. S. Koh, J. A. Ellman, J. Org. Chem.
US5214195, 1990. 1996, 61, 44944495.
88 C. Schumann, L. Seyfarth, G. 109 E. K. Kick, J. A. Ellman, J. Med.
Greiner, S. Reissmann, J. Pept. Res. Chem. 1995, 38, 14271430.
2000, 55, 428435. 110 G. Wess, K. Bock, H. Kleine, M.
89 T. Johnson, R. C. Sheppard, J. Kurz, W. Guba, H. Hemmerle, E.
Chem. Soc., Chem. Commun. 1991, Lopez-Calle, K. H. Baringhaus, H.
16531655. Glombik, A. Enhsen, W. Kramer,
90 O. Seitz, H. Kunz, J. Org. Chem. Angew. Chem. Int. Ed. 1996, 35, 2222
1997, 62, 813826. 2224; Angew. Chem. 1996, 108, 2363
91 O. Seitz, Tetrahedron Lett. 1999, 40, 2366.
41614164. 111 W. H. Pearson, R. B. Clark, Tetra-
92 O. Seitz, C. H. Wong, J. Am. Chem. hedron Lett. 1997, 38, 76697672.
Soc. 1997, 119, 87668776. 112 S. E. Yoo, J. S. Seo, K. Y. Yi, Y. D.
93 B. Blankemeyer-Menge, R. Frank, Gong, Tetrahedron Lett. 1997, 38,
Tetrahedron Lett. 1988, 29, 5871 12031206.
5874. 113 J. H. Ryu, J. H. Jeong, Arch. Pharm.
94 K. Kaljuste, A. Unden, Tetrahedron Res. 1999, 22, 585591.
Lett. 1996, 37, 30313034. 114 M. Ramaseshan, J. W. Ellingboe,
95 F. Guibe, O. Dangles, G. Balavoine, Y. L. Dory, P. Deslongchamps,
A. Loffet, Tetrahedron Lett. 1989, 30, Tetrahedron Lett. 2000, 41, 47434749.
26412644. 115 P. J. May, M. Bradley, D. C.
96 X. H. Zhang, R. A. Jones, Tetra- Harrowven, D. Pallin, Tetrahedron
hedron Lett. 1996, 37, 37893790. Lett. 2000, 41, 16271630.
97 K. C. Nicolaou, N. Winssinger, R. 116 Y. Kondo, T. Komine, M. Fujinami,
Hughes, C. Smethurst, S. Y. Cho, M. Uchiyama, T. Sakamoto, J. Comb.
Angew. Chem. Int. Ed. 2000, 39, 1084 Chem. 1999, 1, 123126.
1089; Angew. Chem. 2000, 112, 1126 117 L. F. Tietze, A. Steinmetz, Angew.
1130. Chem. Int. Ed. 1996, 35, 651652;
98 O. Seitz, H. Kunz, Angew. Chem. Int. Angew. Chem. 1996, 108, 682683.
Ed. 1995, 34, 803805; Angew. Chem. 118 U. Grether, H. Waldmann, Angew.
1995, 107, 901904. Chem. Int. Ed. 2000, 39, 16291632;
99 L. A. Thompson, J. A. Ellman, Tetra- Angew. Chem. 2000, 112, 1688
hedron Lett. 1994, 35, 93339336. 1691.
100 S.-e. Yoo, Y.-D. Gong, M.-Y. Choi, 119 C. R. Millington, R. Quarrel, G.
J.-s. Seo, K. Y. Yi, Tetrahedron Lett. Lowe, Tetrahedron Lett. 1998, 39,
2000, 41, 64156418. 72017204.
101 C. C. Leznoff, J. Y. Wong, Can. J. 120 J. R. Horton, L. M. Stamp, A.
Chem. 1973, 51, 37563764. Routledge, Tetrahedron Lett. 2000, 41,
102 C. C. Leznoff, S. Greenberg, Can. J. 91819184.
Chem. 1976, 54, 38243829. 121 P. Soucy, Y. L. Dory, P. Deslong-
103 Z. H. Xu, C. R. McArthur, C. C. champs, Synlett 2000, 11231126.
References 157
Int. Ed. 1998, 37, 25342537; Angew. Moore, J. Org. Chem. 1996, 61, 8160
Chem. 1998, 110, 26772680. 8168.
164 K. C. Nicolaou, J. Pastor, S. 180 J. K. Young, J. C. Nelson, J. S.
Barluenga, N. Winssinger, Chem. Moore, J. Am. Chem. Soc. 1994, 116,
Commun. 1998, 19471948. 1084110842.
165 T. Ruhland, K. Andersen, H. 181 L. Jones, J. S. Schumm, J. M. Tour,
Pedersen, J. Org. Chem. 1998, 63, J. Org. Chem. 1997, 62, 13881410.
92049211. 182 S. Brase, D. Enders, J. Kobberling,
166 B. Chenera (Smithkline Beecham F. Avemaria, Angew. Chem. Int. Ed.
Corporation), Patent No. WO PCT. 1998, 37, 34133415; Angew. Chem.
98/17695, 1995. 1998, 110, 36143616.
167 G. Marchand, J. F. Pilard, J. 183 S. Brase, S. Dahmen, Chem. Eur. J.
Simonet, Tetrahedron Lett. 2000, 41, 2000, 6, 18991905.
883885. 184 M. Lormann, S. Dahmen, S. Brase,
168 C. Garcia-Echeverria, Tetrahedron Tetrahedron Lett. 2000, 41, 3813
Lett. 1997, 38, 89338934. 3816.
169 H. Kuhn, W. P. Neumann, Synlett 185 S. Schunk, D. Enders, Org. Lett.
1994, 123124. 2000, 2, 907910.
170 R. Michels, M. Kato, W. Heitz, 186 M. Lormann, S. Brase, unpublished.
Makromol. Chem. 1976, 177, 2311 187 S. Brase, M. Schroen, Angew. Chem.
2320. Int. Ed. 1999, 38, 10711073; Angew.
171 H. E. Russell, R. W. A. Luke, M. Chem. 1999, 111, 11391142.
Bradley, Tetrahedron Lett. 2000, 41, 188 A. de Meijere, H. Nuske, M. Es-
52875290. Sayed, T. Labahn, M. Schroen, S.
172 K. C. Nicolaou, J. A. Pfefferkorn, Brase, Angew. Chem. Int. Ed. 1999, 38,
A. J. Roecker, G. Q. Cao, S. 36693672; Angew. Chem. 1999, 111,
Barluenga, H. J. Mitchell, J. Am. 38813884.
Chem. Soc. 2000, 122, 99399953. 189 F. Avemaria, V. Zimmermann, S.
173 K. C. Nicolaou, J. A. Pfefferkorn, Brase, Org. Lett. 2001, submitted.
H. J. Mitchell, A. J. Roecker, S. 190 S. Brase, M. Lormann, J. Heuts,
Barluenga, G. Q. Cao, R. L. Chem. Eur. J. 2001, in preparation.
Affleck, J. E. Lillig, J. Am. Chem. 191 S. Brase, S. Dahmen, J. Heuts,
Soc. 2000, 122, 99549967. Tetrahedron Lett. 1999, 40, 6201
174 K. C. Nicolaou, J. A. Pfefferkorn, 6203.
G. Q. Cao, Angew. Chem. Int. Ed. 192 M. E. P. Lormann, C. H. Walker, S.
2000, 39, 734739; Angew. Chem. Brase, Chem. Commun. 2001, sub-
2000, 112, 750755. mitted.
175 K. C. Nicolaou, G. Q. Cao, J. A. 193 S. Brase, J. Kobberling, D. Enders,
Pfefferkorn, Angew. Chem. Int. Ed. M. Wang, R. Lazny, S. Brandtner,
2000, 39, 739743; Angew. Chem. Tetrahedron Lett. 1999, 40, 2105
2000, 112, 755759. 2108.
176 K. C. Nicolaou, A. J. Roecker, J. A. 194 S. Brase, S. Dahmen, M. Pfeffer-
Pfefferkorn, G. Q. Cao, J. Am. korn, J. Comb. Chem. 2000, 2, 710
Chem. Soc. 2000, 122, 29662967. 717.
177 K. C. Nicolaou, H. J. Mitchell, K. 195 S. Dahmen, S. Brase, Org. Lett. 2000,
C. Fylaktakidou, H. Suzuki, R. M. 2, 35633565.
Rodrguez, Angew. Chem. Int. Ed. 196 a) C. Pilot, Matrise de Chimie
2000, 39, 10891093; Angew. Chem. 1999, RWTH Aachen/Universite
2000, 112, 11311135. Strasbourg. b) C. Pilot, S. Dahmen,
178 K. C. Nicolaou, C. N. C. Boddy, S. F. Lauterwasser, S. Brase, Tetra-
Brase, N. Winssinger, Angew. Chem. hedron Lett. 2001, 42, 91799181.
Int. Ed. 1999, 38, 20962152; Angew. 197 S. Brase, Chimica Oggi 2000, 18(9),
Chem. 1999, 111, 22302287. 1418.
179 J. C. Nelson, J. K. Young, J. S. 198 S. Dahmen, S. Brase, Angew. Chem.
References 159
511 M. A. Scialdone, Tetrahedron Lett. 531 K. Ngu, D. Patel, J. Org. Chem. 1997,
1996, 37, 81418144. 62, 70887089.
512 M. A. Scialdone, S. W. Shuey, P. 532 A. Golebiowski, S. Klopfenstein,
Soper, Y. Hamuro, D. M. Burns, J. Tetrahedron Lett. 1998, 39, 3397
Org. Chem. 1998, 63, 48024807. 3400.
513 E. E. Swayze, Tetrahedron Lett. 1997, 533 S. Hanessian, F. Xie, Tetrahedron
38, 86438646. Lett. 1998, 39, 737740.
514 S. Dahmen, S. Brase, unpublished. 534 U. Bauer, W. B. Ho, A. M. P.
515 K. Burgess, J. Chen in: Solid-Phase Koskinen, Tetrahedron Lett. 1997, 38,
Organic Synthesis. Burgess, K. (ed.), 72337236.
John Wiley & Sons, New York 2000, 535 G. E. Atkinson, P. M. Fischer, W. C.
S. 123. Chan, J. Org. Chem. 2000, 65, 5048
516 J. Chen, M. Pattarawarapan, A. J. 5056.
Zhang, K. Burgess, J. Comb. Chem. 536 M. F. Gordeev, H. C. Hui, E. M.
2000, 2, 276281. Gordon, D. V. Patel, Tetrahedron
517 D. S. Dodd, O. B. Wallace, Tetra- Lett. 1997, 38, 17291732.
hedron Lett. 1998, 39, 57015704. 537 S. M. Dankwardt, Synlett 1998, 761.
518 L. J. Wilson, S. R. Klopfenstein, M. 538 J. Nielsen, P. H. Rasmussen,
Li, Tetrahedron Lett. 1999, 40, 3999 Tetrahedron Lett. 1996, 37, 33513354.
4002. 539 D. W. Emerson, R. R. Emerson, S. C.
519 M. Patek, M. Smrcina, E. Nakani- Joshi, E. M. Sorensen, J. E. Turek,
shi, H. Izawa, J. Comb. Chem. 2000, J. Org. Chem. 1979, 44, 46344640.
2, 370377. 540 A. M. Murphy, R. Dagnino, Jr, P. L.
520 J. A. Josey, C. A. Tarlton, C. E. Vallar, A. J. Trippe, S. L. Sherman,
Payne, Tetrahedron Lett. 1998, 39, R. H. Lumpkin, S. Y. Tamura, T. R.
58995902. Webb, J. Am. Chem. Soc. 1992, 114,
521 M. Bonnat, M. Bradley, J. D. 31563157.
Kilburn, Tetrahedron Lett. 1996, 37, 541 T. Q. Dinh, R. W. Armstrong,
54095412. Tetrahedron Lett. 1996, 37, 11611164.
522 H. M. Zhong, M. N. Greco, B. E. 542 J. A. Fehrentz, M. Paris, A. Heitz,
Maryanoff, J. Org. Chem. 1997, 62, J. Velek, C. F. Liu, F. Winternitz, J.
93269330. Martinez, Tetrahedron Lett. 1995, 36,
523 L. Gomez, F. Gellibert, A. Wagner, 78717874.
C. Mioskowski, Chem. Eur. J. 2000, 6, 543 S. Kobayashi, I. Hachiya, M.
40164020. Yasuda, Tetrahedron Lett. 1996, 37,
524 D. P. Arya, T. C. Bruice, J. Am. 55695572.
Chem. Soc. 1998, 120, 66196620. 544 I. Vlattas, J. Dellurecio, R.
525 P. Roussel, M. Bradley, I. Dunn, I. I. Sytwu, J. Stanton,
Matthews, P. Kane, Tetrahedron Lett. Tetrahedron Lett. 1997, 38, 73217324.
1997, 38, 48614864. 545 B. J. Hall, J. D. Sutherland, Tetra-
526 R. Grigg, J. P. Major, F. M. Martin, hedron Lett. 1998, 39, 65936596.
M. Whittaker, Tetrahedron Lett. 1999, 546 C. Pothion, M. Paris, A. Heitz, L.
40, 77097711. Rocheblave, F. Rouch, J. A.
527 L. S. Richter, M. C. Desai, Tetra- Fehrentz, J. Martinez, Tetrahedron
hedron Lett. 1997, 38, 321322. Lett. 1997, 38, 77497752.
528 C. D. Floyd, C. N. Lewis, S. R. Patel, 547 D. R. Cody, S. H. DeWitt, J. C.
M. Whittaker, Tetrahedron Lett. 1996, Hodges, B. D. Roth, M. C.
37, 80458048. Schroeder, C. J. Stankovic, W. H.
529 S. L. Mellor, W. C. Chan, Chem. Moos, M. R. Pavia, J. S. Kiely
Commun. 1998, 20052006. (Warner-Lambert Co.), Patent Nos
530 S. L. M. C. Mellor, W. C. Chan, WO PCT 9408711; US 958,383; CAN
Tetrahedron Lett. 1997, 38, 3311 122:106536, 1994.
3314. 548 S. Chamoin, S. Houldsworth, C. G.
168 4 Linkers for Solid-phase Synthesis
5
Encoding Technologies
Thomas Kramer, Valery V. Antonenko, Reza Mortezaei,
Nicolay V. Kulikov
5.1
Introduction
ever, for combinatorial library syntheses before or during synthesis encoding (i.e.
encoding compound libraries at the time of their synthesis, rather than afterwards)
is far more ecient. Again, it is the synthetic history of a compound that is en-
coded rather than its chemical identity.
Furthermore, a major requirement of encoding during synthesis is that the tag
must invariably be connected to the compound itself (like a protecting group) or to
the polymer on which the compound is being synthesized. Consequently, the tags
need to be chemically inert under the reaction conditions used to synthesize the
library compounds, i.e. tag chemistry and library chemistry have to be orthogonal.
Since some compounds are screened for biological activity in an on-bead screening
assay (with the tag still attached), the tags should also not interfere with the bio-
logical properties of the library compound. Therefore, it is essential that tags have
to be chemically and biologically inert.
In the following sections a review of the existing methods for encoding combi-
natorial libraries will be given that divide encoding strategies into two categories:
chemical encoding and non-chemical (physical) encoding.
5.2
Chemical Encoding Methods
5.2.1
Oligonucleotide Tags
In 1992, Brenner and Lerner [3] were the rst to suggest in the literature the con-
cept of chemical encoding when they proposed a method for producing an oligo-
nucleotide-encoded peptide library. Addition of each amino acid to the polymeric
bead is followed by the attachment of two preselected oligonucleotides to a dier-
ent tag site on the same bead. Each base pair encodes one, and only one, amino
acid. In the split-and-pool strategy, the oligonucleotide chain grows in parallel
with the peptide chain; thus, each unique peptide sequence is encoded by a
unique oligonucleotide. The decoding process starts by amplifying the encoding
oligonucleotide using the polymerase chain reaction (PCR). To reveal the primary
172 5 Encoding Technologies
Gallop and coworkers from the Aymax Research Institute used the oligo-
nucleotide tagging strategy for the encoding of a library containing 823,543 hepta-
peptides [5]. The synthesis was carried out on 10-mm-diameter monodisperse
beads. The beads were used in sucient number to synthesize 200 copies of each
sequence in the library. The beads were made of polystyrene crosslinked with di-
vinylbenzene and derivated with a 1,12-diaminododecane linker. The capacity of
the resin was 10 mmol g1 , corresponding to 10 fmol per bead. The resin was fur-
ther functionalized with two chemical linkers: a mixture of 9-uorenylmethoxy-
carbonyl (Fmoc)Thr(t-Bu)OBt and succinimidyl activated ester of 4-O-(dimeth-
oxytrityl) oxobutyrate. The former is introduced to support the growing peptide
chain; the latter to support the synthesis of the encoding oligonucleotide. The re-
sulting construct contained a 20:1 ratio of peptides to oligonucleotides on the bead.
The structure of the construct is shown in Scheme 5.2.
The peptides were synthesized using Fmoc-protecting groups for a-amino func-
tionalities and t-butyl-type protection for the side-chains of amino acids. Oligo-
nucleotides were assembled using dimethoxytrityl-protected 3 0 -O-methyl-N,N-
diisopropyl phosphorimidates. To protect the encoded oligonucleotides from the
depurination side reaction upon triuoracetic acid (TFA)-mediated side-chain de-
protection of t-butyl-type protecting groups, the susceptible 2 0 -deoxyguanosine
(dG) unit was not used. 2 0 -Deoxyadenosine (dA) was substituted with 7-deaza-2
deoxyadenosine (c7dA), which is stable to TFA treatment. To make the encoding
5.2 Chemical-encoding Methods 173
5.2.2
Peptide Tags
5.2.3
Haloaromatic Binary Coding
tion as part of a construct with a photolabile linker (Scheme 5.3) at the expense of
the ligand synthesis sites. Each compound in the library is encoded by a limited
set of tags. The presence, as well as the absence, of each member of the set carries
information about the specic encoded structure. In a binary code, a set of n tags
can encode 2 n 1 dierent structures. For example, 20 tags can encode 20 20 1
1,048,575 dierent library members. After being released from the beads by pho-
tolysis, the tags are detected by capillary gas chromatography using electron cap-
ture detection (ECGC), a detection method that is particularly sensitive to heavily
chlorinated aromatics. The haloaromatic compounds were selected to ensure reli-
able and reproducible separation by ECGC, which is capable of detecting sub-
picomolar amounts of the tags. Consequently, the beads can be tagged at only 0.5
1% of the resin loading (0.51 pmol per bead) without detectable interference with
the library synthesis.
Later, Still and coworkers modied the original strategy by developing a new
type of tagging reagent TnC [9], presented in Scheme 5.4. A derivative of vanillic
acid (3-methoxy-4-hydroxybenzoic acid) was chosen as a linker. Synthesis of the
tagging reagent TnC begins with a Mitsunobu reaction of a tag alcohol Tn with
methyl vanillate, followed by LiOH hydrolysis of the methyl ester. This produces
free acid TnA, which is converted to acid chloride TnB. Excess of diazomethane
converts acid chloride TnB into the tagging diazoketone TnC. In the presence of
rhodium reagents Rh2 (OAc)4 or Rh2 (O2 CCF3 )4 TnC forms an acylcarbene, which
rapidly and cleanly reacts with benzene, forming a derivative of cycloheptatriene
Tn1. Benzene was used as a soluble analog of polystyrene resin. Derivatives of TnC
containing dierent numbers of methylene groups n have been prepared. The re-
searchers found that diazoketones TnC are stable solids and can be stored at room
temperature for months. They react easily with polystyrene resin, providing a
means for binary encoding. At the decoding stage, the tags are oxidatively cleaved
by ceric ammonium nitrate and analyzed by ECGC.
The acylcarbene tagging strategy does not require any specic functional group
for tag attachment, and the tags and linkers are generally compatible with a wide
5.2 Chemical-encoding Methods 175
5.2.4
Secondary Amine Binary Coding
In 1996, Gallop and coworkers developed another robust encoding strategy [11,
12], based on secondary amine tags, which are incorporated into a polyamide back-
bone. The secondary amine binary coding scheme utilizes an amine-based poly-
meric resin that is dierentially functionalized with sites for both ligand synthesis
and tag addition (Scheme 5.5). The ligand synthesis site is derivatized with a N-
Fmoc-protected photocleavable linker group [13]. This linker allows for the release
of the ligand from the resin by exposure to ultraviolet (UV) light. The amino group
of the tag site is protected with orthogonal to the Fmoc group functionalities (e.g.
Boc or Alloc). The tag site occupies only 10% of the total number of amino groups
on the resin. Each 130-mm-diameter bead of the TentaGel S resin, which is recom-
mended by the authors of this methodology, contains about 300 pmol of amino
groups. Therefore, the theoretical yield of the ligand is 270 pmol per bead. Such
176 5 Encoding Technologies
Scheme 5.5. Resin construct for secondary amine binary coding [11, 12].
Upon completion of the library synthesis, each bead is distributed into a sepa-
rate well of a microtiter plate. The ligands are released from the beads into the
assay medium by exposure to UV light at 365 nm. The tag residues remain coval-
ently attached to the beads. After screening, beads from the wells containing active
compounds are collected for decoding. The decoding process is shown schemati-
cally in Scheme 5.8, and begins with the acid hydrolysis of the beads in 6 N hy-
drochloric acid. Under these conditions all amide bonds are hydrolyzed, releasing
free secondary amine tags into the solution. After evaporation of the HCl, the
amines are converted into the corresponding dansyl derivatives by treatment with
dansyl chloride. Analysis of the resulting mixture of dansylated tags is carried out
Scheme 5.8. The decoding process for secondary amine binary coding.
178 5 Encoding Technologies
5.2.5
Mass Encoding
The code block can also consist of equimolar mixtures built from a set of isotopi-
cally labeled dipeptides: GlyGly, Gly[ 13 C]Gly, Gly[ 13 C]2 Gly, [ 13 C]Gly[ 13 C]Gly,
[ 13 C]Gly[ 13 C]2 Gly, and [ 13 C]Gly[ 13 C]2 Gly. These dipeptides incorporate equal to
their content in the mixture, because isotopic labeling does not aect chemical re-
activity. The resulting mass spectra of these mixture codes serve as bar codes.
In the ratio-encoding strategy, the encoded information is derived from just
two peaks in the mass spectrum. A mixture of a reagent, common to all members
of the library, is prepared from dierent ratios of the isotopic isomers of the
reagent. The resulting mass spectrum will reveal two distinct peaks corresponding
180 5 Encoding Technologies
to each of the isomers. The relative heights of the peaks provides the encoded in-
formation.
The mass encoding strategy was used by Wagner and coworkers [18] for the
synthesis of a peptoid library. A set of 20 14 N: 15 N ratio-encoded imidazoles was
synthesized to investigate pharmacokinetic applications of isotopic labeling.
5.3
Non-Chemical Encoding Methods
5.3.1
Positional Encoding
participate in a coupling reaction; the rest of the synthesis area remains protected
and intact. Synthesis continues by illuminating another part of the surface
through photolithographic mask B, followed by the next chain elongation reaction.
By repeating the photodeprotection and coupling steps, highly dense arrays, each
consisting of thousands of peptides, can by synthesized. Importantly, the primary
structure of each peptide in the array is suciently dened by the sequence of
coupling and photolysis steps, and by photolithography mask patterns. Therefore,
the structure can be easily deduced from the (x, y) coordinates of the peptide on
the slide. This eliminates the need for encodingdecoding procedures required by
some other combinatorial technologies. After completion of the synthesis, the syn-
thesis area is exposed to reagents necessary for the elimination of side-chain pro-
tecting groups. To assess the binding properties of all synthesized peptides, the
entire array is incubated with a uorescently labeled target molecule and scanned
using a stage-scanning confocal uorescence microscope. Sites, containing pep-
tides that bind to the target, become uorescent. Anity data on all peptides in the
entire array are obtained in one step.
The consumption of chemical reagents required for the synthesis of thousands
of peptides composing the array, together with the biological reagents necessary
for bioassay, is very small, because the capacity of the at glass surface is only 5
20 pmol cm2 . Biological reagents used in this technology are recoverable and can
be reused. Moreover, after performing an assay with one target molecule, the
bound target can be easily dissociated from the array (e.g. by treating it with 6 M
guanidine hydrochloride), making the array available for subsequent screening
with other targets. These arrays are reusable for at least 6 months.
With special (orthogonal) masking strategies the number of synthetic regions on
the glass surface can be increased until the limit of photolithographic resolution is
achieved (1020 mm). With this resolution, 250,0001,000,000 compounds can be
synthesized in 1 cm 2 . Routinely, 50-mm resolution is practiced and allows for the
production of 40,000 compounds in the same area.
Light-directed, spatially addressable synthesis is a powerful technology for gen-
erating chemical diversity. Unfortunately, the technique is limited to peptides, oli-
gonucleotides and other linear oligomeric structures.
HiTOPS system Over time, original approaches have been developed by many
companies to perform the synthesis of organic compounds on the footprint of a
96-well plate. The HiTOPS (high-throughput organic parallel synthesis) system
[29] (Scheme 5.11) utilizes a variety of 96-deep-well ltration microtiter plates
available from Polyltronics/Whatman.
The plates are made of polypropylene and other polymers, and are available with
a selection of dierent lters. The volume of each well is 2 mL and allows the use
of up to 50 mg of resin. For larger scale syntheses, several wells or even the entire
row or the entire column can be used for the preparation of the same compound.
Reactants are retained in wells by the positive pressure of an inert gas.
Scheme 5.11. HiTOPS system. Synthesis device (left) and cleavage device (right).
5.3 Nonchemical Encoding Methods 183
5.3.2
Non-Positional Encoding
devices have been used for years to tag laboratory mice subcutaneously, and more
recently these have been used to provide security for automobile ignition keys, to
secure building entrance identication, and for a variety of other functions. The
application of RF tags to combinatorial syntheses was reported in 1995 by two
groups, one working at IRORI [38] and the other at Ontogen [39]. The RF ID tag
is about the size of a ea stirbar and is encased in a thick-walled glass shell. As
shown in Scheme 5.12, the essential components are an antenna (the largest
component) and a microelectronic chip. Each chip has a unique, nonvolatile 40-bit
ID code laser-etched into it. With 40 bits available, a total of 2 40 (over 1 trillion)
unique ID codes is possible. The virtually inexhaustible range allows one to guar-
antee the uniqueness of all present and future RF tag ID codes. Additional bits are
used to perform extensive digital error detection, which prevents incorrect report-
ing. (Bar codes are a more common graphical embodiment of encoding, but of
course one that is substantially less information-rich.)
A transceiver controlled by a computer is used to interrogate and receive the ID
code of each RF tag. The transceiver antenna transmits a specially modulated, 125-
kHz electromagnetic eld. This eld is of very low energy and is not harmful.
When an RF tag is held within about 1 cm from the transceivers antenna, energy
is picked up by the RF tags antenna. A rectier in the chip converts this energy to
microwatt levels of direct current (DC) power, which is enough to power-up the
logic circuitry on the chip. In a very real sense, the RF tag is similar to a crystal
radio (which does not require an external power source), except that the device
serves as both receiver and transmitter. It is self-contained, in that the chip used
no internal batteries and has no external metallic connections.
A synchronization signal modulated onto the transceivers signal allows the chip
to respond with its ID code (a serial sequence of ID bits) and error-checking bits.
The time elapsed between placing the chip on the transceiver and seeing the ID
code on the interfaced computer screen is about 0.5 s.
If one could physically associate this chip with a compound undergoing syn-
5.3 Nonchemical Encoding Methods 185
thetic transformations, then the ID code on the chip would permit one to pick up
a sample at any point of the combinatorial synthesis and know which reaction(s)
it had already been through. Of course, just knowing the reaction history of a
sample is not equivalent to knowing the chemical concept of a sample, but it is
much better than nothing. If each reactor (compound-carrying unit) must be
present in a series of reaction asks containing lots of other reactors, then having
them tagged makes it possible to put them into the right ask for each step of the
synthesis. Moreover, if that tag can be read easily, then that movement of reactors
can be automated. Electronic tagging permits all of these benets.
How this works in practice is detailed as follows. After a compound has been
identied for which several hundred to several thousand derivatives would be of
value, a synthetic route is chosen that: (i) permits linkage to a solid-phase support;
(ii) utilizes reaction steps that appear possible to optimize to b90% yield; and (iii)
aords reagents in each step for which desirable variants can be purchased (or,
less optimally, can be made trivially). In the synthesis itself, one of the signicant
advantages of the microreactor approach becomes evident: one can use standard
laboratory glassware and equipment to accomplish the library synthesis. There is
no need for the automation of liquid-handling steps, and indeed there is no need
for automation at all until rather large libraries are desired.
IRORI oers several miniature reactors MicroKans9 and MicroTubes9
(Scheme 5.13). MicroKans9 is a small cylindrical container with mesh walls; the
internal volume of the container is 330 ml. In addition to the RF tag, the container
hold up to 30 mg of any commercially available resin.
5.4
Conclusion
References
Peptides, Polypeptides and Oligo- Angew. Chem. Int. Ed. 1995, 3, 2289
nucleotides, Macro-Organic Reagents 2291.
and Catalysts-1990. Epton, R. (ed.), 39 E. J. Moran, S. Sarshar, J. F.
SPCC, Birmingham 1990, pp. 453 Cargill, M. M. Shabaz, A. Lio, A. M.
459. M. Mjalli, R. W. Armstrong, J. Am.
37 N. K. Terrett, M. Gardner, Chem. Soc. 1995, 117, 1078710788.
D. W. Gordon, R. J. Kobylecki, 40 T. Czarnik, M. Nova, Chem. Britain
J. Steele, Chem.-Eur. J. 1997, 3, 1997, 33, 3941.
19171920. 41 X. Y. Xiao, C. Zhao, H. Potash, M.
38 K. C. Nicolaou, X. Y. Xiao, Z. P. Nova, Angew. Chem. Int. Ed. Engl.
Parandoosh, A. Senyei, M. Nova, 1997, 36, 780782.
190
6
Instrumentation for Combinatorial Chemistry
Marcus Bauser and Hubertus Stakemeier
6.1
Automation in Combinatorial Synthesis
6.1.1
General Remarks
Economic pressure to speed up the drug discovery process has had a huge impact
on all elds of medicinal chemistry [1], therefore automation has increasingly be-
come one of the main strategies to fulll this demand [2]. While automation was
successful in high-throughput screening (HTS) [3] and peptide synthesis, it played
a minor role in mainstream organic synthesis. Automated systems are recom-
mended for procedures that are highly predictable and repetitive [4]. However, or-
ganic chemistry is seldom dened in this way. The successful synthesis of organic
molecules depends strongly on the chemical properties of the reagents and react-
ants. Within the library production process solubility and reactivity of synthons
can be highly dierent, therefore it is very dicult to nd one protocol that works
for every building block. Since multiple parallel synthesis began, there has been a
wide range of dierent approaches and concepts for the design of automated sys-
tems to overcome these problems [57]. The rst attempts at automation were the
simple parallelization of commercially available reaction vessels. Secondly, reaction
blocks were designed and used in combination with existing liquid-handling sys-
tems. Another approach uses stand alone systems that mimic the typical action of
a chemist. Today, modern automated systems are modular workstation approaches.
An overview is given in Table 6.1.
6.1.2
Fully Automated Systems for Solid- and Solution-phase Synthesis
Automation
Type Unit Unit Unit Unit
Arm Cylindric arm Two multiprobe XYZ arms Two multiprobe XYZ arms Two multiprobe XYZ arms
Chemistry Solid phase Solid phase Solid phase Solid phase
Solution phase Solution phase Solution phase Solution phase
Reaction block (RB) Monomer rack Ares reactor
Array 20 (4 5) 8-, 16-, 40-, 96-well reactor 40, 96, 384 40, 96, 384
Number of possible RBs Up to 5 RBs Up to 4 RBs (Omega 384) 1 Multiple
Type Fixed during synthesis Fixed during synthesis Fixed during synthesis Fixed during synthesis
Reaction vessels
Number/volume 20/5 ml Solid phase: 96/3.5 ml; 40/9 384/0.5 ml; 96/2 ml; 40/6 384/0.5 ml; 96/2 ml; 40/6
ml; 16/15 ml; 8/30 ml ml ml
Solution phase: 96/6 ml;
40/14.5 ml
Material Polypropylene or glass Multiple-well teon reactor Teon block with fritted Teon block with fritted
syringes with frit or block with frits for solid wells wells
customized blocks phase
Reaction procedures
Reux (condenser module) Completely sealed reaction Completely sealed reaction
block block
Filtration Bottom Bottom Bottom
Evaporation in rv
Agitation Magnetic levitation stirring Orbital shaking Orbital shaking
Temperature range 80 to 160 C 70 to 150 C 70 to 150 C
Pressure Up to 150 psi Up to 150 psi
6.1 Automation in Combinatorial Synthesis
Work-up
Liquidliquid extraction
Solid-phase extraction
Analysis HPLC upgradable
Special features . Fully enclosed robot-arm- . Automatic on-board . Heating above boiling . Conductometric detection
based system (SCARA) cleavage of all products point is possible of phase boundry
with balance, vortexer, . Multiple segregation of . Multichannel solvent/ . Up to 10,000 parallel
vacuum centrifuge reactor waste for hazard reagent delivery reactions
. Flexible control software classication or reagent . Fully enclosed . Fully enclosed
6 Instrumentation for Combinatorial Chemistry
recovery
. Fully enclosed
Automation
Type Modular Unit Unit Modular
Arm XYZ arm Two XYZ arms XYZ arm (Gilson 222)
Chemistry Solid phase Solid phase Solid phase Solid phase
Solution phase Solution phase Solution phase
Reaction block (RB) Trident Reaction Cassette TM Calypso Reaction Block
Array (wells/vials) 48 (6 8) Microplate footprint e.g. 16 (8 2)
96 (8 12)
Number of possible RBs Up to 4 RBs Up to 4 RBs 7
Type Flexible during synthesis Fixed during synthesis Fixed during synthesis
Reaction vessels KANTM
Number/volume 48/5 m, 24/14 ml 6/50 ml; 12/25 ml; 24/10 112/13 ml; 64/27 ml; 32/75 Kans for 30 mg, 60 mg and
ml; 48/5 ml; 96/2 ml ml; 16/100 ml 300 mg resin capacity
are available
Material Glass vessels Multiple-well teon reactor Individual glass vials Polypropylene
block with frits
Reaction procedures
Reux (Sealed vessels) (Sealed vessels) Reux condensers for each (standard glassware)
vial
Filtration Top Bottom Parallel vessel to vessel
Evaporation in rv (online)
Agitation Orbital shaking Orbital shaking Orbital shaking
Temperature range 40 to 150 C 70 to 150 C 60 to 120 C
Pressure (upgradable up to 10 bar)
Inert atmosphere In rv In rv In rv In rv
Work-up
Liquidliquid extraction (Trident Processing (by volume)
Station)
Solid-phase extraction (Trident Processing
Station)
Analysis HPLC upgradable Online TLC; online HPLC
possible valve preinstalled
Special features . Fully enclosed . Modular bench layout . Reagent additions while . Modular, expandable
. Temperature and and format shaking or stirring and system
agitation speed exible heating or cooling . AutoSortTM10K
for each reaction cassette . Closure of the reactors by automatically sorts
. Rvs sealed with rotatable a ceramic valve allowing microreactors between
Teon valve for ecient evaporation reaction steps
(DMSO, DMF) . Sort 1000 microreactors
. Fully enclosed per hour
6.1 Automation in Combinatorial Synthesis
ISRA Mettler Toledo Myriad Core MultiSyn Tech Syro II Perkin-Elmer Solaris 530
system
Automation
Type Unit Modular Unit Unit
Arm Articulated arm Transfer of RBs by a Two XYZ arms XYZ arm (TECAN)
conveyor belt
Chemistry Solution phase Solid phase Solid phase Solid phase
Solution phase Solution phase
Reaction block (RB) Minitray
Array (wells/vials) 12 (2 6) 96; 60; 40 48 (6 8)
Number of possible RBs Multiple 1 Up to 4 RBs
6 Instrumentation for Combinatorial Chemistry
Type Flexible during synthesis Fixed during synthesis Fixed during synthesis
Reaction vessels
Number/volume 50/35 ml 12/10 ml 96/2 ml; 60/5 ml, 40/10 ml 48/10 ml
Material Glass vessels Glass vessels Glass or polypropylene Single glass reactors
syringes
Reaction procedures
Reux Closed reactors Closed rvs with (reux channel)
overpressure value
Filtration Bottom Bottom Top
Evaporation in rv (online)
Agitation Magnetic stirring Magnetic stirring, gas Magnetic levitation stirring Orbital shaking
bubbling
Temperature range 40 to 150 C 60 to 150 C 60 to 150 C 30 to 150 C
Pressure
Inert atmosphere In rv In rv In rv In rv
Work-up
Liquidliquid extraction (Allex TM )
Solid-phase extraction
Analysis Online HPLC
Special features . Detection of phase . Modular, expandable . Fully enclosed . Oine incubator
boundary via a camera system . Reagent and solvent available
system . Automated scheduling addition under agitation . Seven sensors to monitor
. Online error detection and simultaneous module possible critical instrument
operation functions
. Compatibility with the . Fully enclosed
Myriad Discoverer series
. Septumless rvs with twist
cap
Automation
Type Unit Unit Unit
Arm XYZ arm Cylindrical arm XYZ arm
Chemistry Solid phase Solid phase Solid phase
(Solution phase) Solution phase
Reaction block (RB) Stem Block
Array (wells/vials) 98 Customized
Number of possible RBs 1 Customized
Type Fixed during synthesis Fixed during synthesis Fixed during synthesis
Reaction vessels
Number/volume 96/10 ml Customized 96/2 ml; 60/5 ml, 40/10 ml
Material Glass vessels Glass vessels Teon, glass vessels
6.1 Automation in Combinatorial Synthesis
195
196
Reaction procedures
Reux Closed reactors Closed reactors Optional
Filtration Top Top
Evaporation in rv (online) Optional
Agitation Orbital shaking Orbital shaking Orbital shaking
Temperature range Ambient to 150 C Ambient to 150 C 80 to 150 C
Pressure
Inert atmosphere In rv In rv
Work-up
Liquidliquid extraction
6 Instrumentation for Combinatorial Chemistry
Solid-phase extraction
Analysis Optional
Special features . Oine incubation . Modular, expandable . Oine incubation possible
possible system . Reaction vessels are moved
. 96-channel ltration head . Fully enclosed between functional unit on
workbench
6.1 Automation in Combinatorial Synthesis 197
the typical workow of an organic synthesis every step can be fully automated via
specialized devices that work totally independently but are at the same time con-
nected and controlled through an intelligent and easy-to-use software program.
Therefore, robotic systems are one of the most exible solutions to laboratory au-
tomation, and established systems can be further developed or redesigned if nec-
essary. Once a robotic system is established it is subject to constant optimization,
which is in contrast to workstation approaches where xed hardware and software
subroutines are usually used. In customized robotic systems, the control software
is used with a wide range of devices from dierent suppliers, and subroutines for
the specialized stations can be easily redened. With exible control software, au-
tomated error identication mechanisms and correction tasks are possible that
allow an unattended round-the-clock operation. But maximum versatility comes
at a price, which is one of the main disadvantages of customized robotic systems.
The size of the instrument requires custom-built ventilation cabinets and safety
installations such as a re extinguisher system, therefore the timeline for the suc-
cessful implementation of an automated robotlab is often a critical point for
pharmaceutical companies. The reliability of the equipment is dicult to predict
and can only be tested during installation, which can be a time-intensive process.
The key components of robot-arm systems are industrial robots that are used
extensively in many areas of industry, such as the car industry, for a wide range of
dierent tasks. Some of the main suppliers of robots are CRS [9], Beckman
Coulter [10], Mitsubishi [11], and Zymark [12] (robot-arm features are shown in
Table 6.2).
Physical characteristics
Travel length Up to 5 m
Height to enclosure 138.7 mm
Height to saddle mounting 201 mm
surface
Weight approximately 50 kg/m
Performance specications
Speeds 0.010.9 m s1
Acceleration 3 m s2
Repeatability 0.08 mm
198 6 Instrumentation for Combinatorial Chemistry
well as documentation of the chemistry undertaken are also very important fea-
tures of the control software. Therefore, a software concept was developed that of-
fers complete and legally admissible documentation of all substances which are to
be prepared by the robotic system. The data generated can easily be transferred
into widely used electronic laboratory journals. The whole system has a footprint
of 5 m long by 2 m wide and is divided into three functional parts for the synthe-
sis, work-up, and analysis; the layout is outlined in Fig. 6.1.
The individual components for the synthesis are a needle XYZ-pipetting robot,
heating and cooling reaction blocks with magnetic stirrers, a rack for reagents, and
a rack for starting materials. These devices have been successfully used in many
dierent syntheses under various conditions. The largest part of the machine is
used for storage and work-up procedures. The main devices for work-up are a lter
station, a drying station, a balance, a centrifuge, a solvent evaporator, and a phase
separation and a phase boundary recognition station. The analytic part of the robot
is represented by a Shimadzu high-performance liquid chromatography/ultraviolet
(HPLC-UV) system. Via an interface it can be used for online analytical character-
ization of the synthesized products. The main features of the system are:
Zymark One of the most widely used robot arms in fully automated systems for
process development and organic synthesis is the Zymate XP arm [14], which has
been commercially available since the early 1980s. The example shown uses a
cylindrical Zymark robot with interchangeable eectors for performing dierent
manipulations (Fig. 6.3).
This system was developed as a result of collaboration between Zeneca and
Zymark (Runcorn, UK) [15]. Depending on the individual components, the main
features of the system are:
using one master computer. The multifunctional software can also be used for
planning, scheduling, and administration of data entries and results.
Their main objective was to develop an unattended automated system that allowed
parallel preparation of compounds in glassware reactors with online purication
and analysis in a compact workstation approach (Fig. 6.6).
Chemspeed delivers the system on a trolley (footprint 1.4 m 0.8 m) covered
with a hood. Because of this compact design, the very exible system can be used
in nearly every laboratory. Within one run, 80 parallel reactions (with an option of
112) can be performed that have the following features:
MultiSyn Tech Syro II MultiSyn Tech [22] oers manual, semiautomated, and fully
automated systems for combinatorial chemistry. The fully automated Syro system
204 6 Instrumentation for Combinatorial Chemistry
(Fig. 6.8) consists of two independent XYZ robot arms and a specially designed
reaction block. The system can be equipped with dierent reaction vessels. For
solid-phase synthesis the removable reaction vessels (number/volume: 96/2 ml;
60/5 ml; 40/10 ml) in glass or polypropylene with glass or PTFE frits are rec-
ommended. Owing to the reaction block design all manipulations can be per-
formed under inert gas atmosphere using reaction temperatures between 60 and
150 C.
The system has a specially designed agitation mechanism. Each reaction tube is
circumvented by electric coils that are used to generate a magnetic eld. The coils
are placed in a movable levitation plate, the center of the magnetic eld is about
6 mm above the frit of the reaction tubes. This special set-up ensures that, during
agitation, the resin is not crushed between the stirring bar and the frit (Fig. 6.9).
Perkin-Elmer Solaris 530 The Solaris TM 530 [25] organic synthesizer for auto-
mated combinatorial chemistry was introduced in late 1998 by PE Biosystems, a
division of the PE Corporation (Fig. 6.10).
The system is able to synthesize 48 discrete molecules per run in parallel. The
portable synthesis module contains an array of 8 6 reaction vessels with a vol-
ume of 10 ml per vial (Fig. 6.11). A reux channel is built in the chemically inert
module to provide reux conditions. The dual septa secures an inert atmosphere.
206 6 Instrumentation for Combinatorial Chemistry
Zinsser Sophas In 1998 Zinsser [26] launched the Sophas system, which is spe-
cially designed for solid-phase synthesis. The synthesizer uses a robotic XYZ arm
with four independent probes that are manufactured by Rosys [27]. All liquid-
handling tasks are controlled via a personal computer. The easy-to-use software is
very exible, can import data from any database, and allows customized layouts to
6.1 Automation in Combinatorial Synthesis 207
be dened. The automated workbench oers the opportunity to choose from a set
of movable reactors that range from 96-well plates to 25-ml reaction vessels.
Argonaut Trident Argonaut Technologies [29] was founded in 1994, and provides
systems for parallel solution- and solid-phase chemistry. The Trident [30] family
consists of a reaction cassette, a library synthesizer, a workstation, and a process-
ing station. The core of the system is the Trident Reaction Cassette TM that contains
48 5-ml vessels or 24 14-ml vessels which are made from glass. The reaction cas-
sette ts into every member of the Trident family and can be easily removed dur-
ing synthesis for spot checks on the reactions.
With one run, the fully automated Trident library synthesizer can prepare up to
192 reactions in parallel. The whole system is controlled via the Trident software,
which controls, for example, the temperature, solvent deliveries, or product collec-
tion. The synthesizer can handle up to four dierent reaction cassettes, and each
cassette position can be set to a dierent temperature and agitation speed.
208 6 Instrumentation for Combinatorial Chemistry
. liquidliquid extraction;
. solid-phase extraction;
. clean up with scavenger resins;
. reverse ltration;
. dry solid loading to open vessels;
. reformatting (e.g. from tubes to microplates).
Mettler Toledo Myriad Core System The Myrid TM Core System (MCS) [31] was de-
veloped by a consortium of pharmaceutical companies (SmithKline Beecham,
Pzer, BASF, Novartis, Merck, Takeda, and Chiroscience) and The Technology
Partnership [32]. In December 1998 the MCS was sold to Mettler Toledo [33].
The MCS was created as a synthesis system based on a series of robotic pro-
cessing modules combined with a unique reaction vessel design, fullling the fol-
lowing objectives:
Irori In 1996 Irori [34] introduced the AccuTag TM -100 Combinatorial Chemistry
System. With this system, large numbers of discrete compounds can be produced
using the directed sorting split-and-pool technique [35]. The reactions are per-
formed in single microreactors which are identied using miniature electronic
tags. The Irori Kan TM reactor family is specially designed for solid-phase synthesis.
Three dierent Kans the MicroKan TM , the MiniKan TM , and the MacroKan TM
with a resin capacity of up to 30, 60, and 300 mg are available, therefore com-
pounds can be synthesized in approximately 10-, 20- and 100-mg batches (Fig.
6.1 Automation in Combinatorial Synthesis 209
6.12). Another big advantage of this approach is that standard glassware can be
used for performing the library synthesis.
To achieve high throughput, the Kans are used together with the AutoSort-10K.
This workstation has been designed to extend the directed sorting technique for
use with libraries in the range of 1000 to 10,000 compounds. The AutoSort TM -10K
Microreactor Sorting System (Fig. 6.13) automatically sorts microreactors between
reaction steps.
The key features of the system are:
6.2
Purication of Combinatorial Libraries
In the early days of combinatorial chemistry [36], mixtures derived from solid-
phase split-and-mix libraries [37] as well as nonpuried compounds from solution-
210 6 Instrumentation for Combinatorial Chemistry
phase synthesis played an important role as test compounds for biological screen-
ing, basically because it was a very easy way to produce the promised numbers
of compounds. It was soon recognized that those compounds often led to false-
positive test results, and that the deconvolution of mixtures and extraction of bio-
logically active molecule in an HTS mode are dicult tasks [38]. This problem re-
sulted in the synthesis of single compounds fullling properties such as diversity,
drug likeness, and a high degree of purity [39]. Classical purication procedures
such as liquidliquid extraction and chromatography were automated, solid-phase
extractions with ion exchangers were adopted, and scavenger reagents for trapping
excess starting material or reagents were developed. Automated preparative re-
versed phase HPLC systems were set up to address high-throughput purication
issues.
6.2.1
Automated LiquidLiquid Extraction
method is only appropriate for very simple separation problems and can be per-
formed in an automated fashion by using a liquid handler. The method totally fails
when polar compounds have to be puried because emulsions can appear or prod-
ucts remain in the aqueous phase. Coupled to a solid-phase extraction system, au-
tomated liquidliquid extraction can be performed together with a 96-needle pi-
petting system such as Quadra 96 [41]. Separation problems, for example removal
of amines from a combinatorial library, were solved by applying several extraction
cycles. Dichloromethane and dilute HCl were mixed and separated through ltra-
tion using solid-phase diatomaceous earth. The phase separation was achieved via
vacuum ltration through a 96-well lterplate carrying a hydrophobic membrane
which held the aqueous phase back. Liquidliquid extraction can be easily auto-
mated by using pipetting workstations. Reaction mixtures are mixed with buers
and the upper or lower phase, depending on the density, is removed. Ecient
mixing can be achieved through redispensing, which can be repeated several
times. The use of a capacity sensor represents a more sophisticated approach. This
sensor type is available for several liquid handlers (Zinsser, Tecan). A drawback
of this method is when phase separation is not complete or undetectable, erratic
results are obtained.
6.2.2
Solid-phase Extraction
6.2.3
Normal Phase Chromatography
Tab. 6.3. Common stationary phases for reversed phase preparative HPLC.
Material Application
6.2.4
Reversed Phase Chromatography
Dimension (mm)/material/particle size (mm) Loading (mmol) Flow rate (ml min1 )
8 75/C18/5 10 814
20 50/C18/5 50 2540
30 125/C18/5 75100 5070
214 6 Instrumentation for Combinatorial Chemistry
6.2.5
Preparative HPLC-MS
The search for more powerful techniques for the purication of combinatorial li-
braries using HPLC led to collaboration between researchers from the pharma-
ceutical and analytical industries, namely PE Sciex and Micromass, with the goal
of developing preparative HPLC-MS [57, 58]. The advantage of this technique is
that the molecular mass of the target molecule triggers the fraction collector. This
allows the collection of the desired compounds and the online identication of the
target molecules. Furthermore, this method oers logistical benets. If just the
target molecules are collected, fraction collector capacity is not critical and down-
stream processing can be performed very eciently. This enables the technique to
work very eciently in terms of throughput. For setting up such a system integra-
tion of the mass spectrometer and a reliable software platform which ensures
fraction collecting and sample tracking are necessary. This application is now of-
fered from more vendors than the pioneers Micromass and PE Sciex, e.g. Merck-
Hitachi, Gilson in collaboration with Thermoquest and Shimadzu.
throughput. One of the main disadvantages is the Apple-based software and the
unstable fraction collector script used for tracking fractions. Besides lacking soft-
ware stability, it is also not easy to integrate an Apple-based system into a usually
Microsoft-based laboratory environment.
6.2.5.3 Merck-Hitachi
As already mentioned above, Merck-Hitachi oers a system that is designed for
this application. By using a special valve as an interface, it is possible to switch
between two ows running into the MS inlet. Therefore two preparative separa-
tions at a time can be performed and analyzed, thus doubling the throughput and
reducing hardware costs. Merck-Hitachi devices together with Gilson fraction col-
lectors are compatible HPLC components.
6.2.5.4 Shimadzu
Shimadzu oers a preparative HPLC-MS/UV system with a Gilson 215 fraction
collector. Besides being good HPLC equipment, a reliable and exible software
package is available. In addition to the software package operating the instrument,
a Shimadzu LIMS (LIMS laboratory information management system) system
is available which can also be used for documentation of data other than those
from the HPLC-MS.
6.3
Analysis of Combinatorial Libraries
Combinatorial syntheses need fast and reliable analyses in order to determine the
identity of products and intermediates in each step of a parallel reaction sequence.
6.3 Analysis of Combinatorial Libraries 217
Method Application
Consequently, analyses were adopted that operate at the same speed as combina-
torial methodologies when these methods became popular for synthesis. The
analysis of the nal products from solid- and solution-phase synthesis usually
compounds for biological testing are the same because similar purity criteria are
applicable. Because of the nature of solid-phase synthesis, the analysis of polymer-
bound intermediates is rather dicult. Special nuclear magnetic resonance (NMR)
and FT-IR methods (Table 6.5) have been developed for the purpose of character-
izing polymer-bound intermediates. In the case of FT-IR techniques, reection
measurements are usually used, e.g. ATR and DRIFTS [61]. These methods can be
performed directly on bead and no KBr solids have to be made, as is the case in
transmission experiments. In the case of NMR, magic angle spinning can be used
to suppress signals from the polymer matrix [62]. All these special analytical tech-
niques are usually used for developing the method owing to the fact that they are
very time-consuming and therefore have a fairly low throughput.
The minimal requirements for combinatorial libraries are usually purity, iden-
tity, and quantity.
6.3.1
Purity of Combinatorial Libraries
identication of the target molecules is carried out in a parallel fashion. Mild ion-
ization techniques such as ESI and APCI are used. They almost always show the
molecular ion peak, which can be directly taken from the synthesis protocol. Cycle
times are highly optimized. It is possible to perform 400 HPLC-MS runs per day
[66]. The resulting data have to be stored in an appropriate LIMS system. Tools are
available from hardware suppliers for the semiautomatic processing of these data.
Using these tools, further steps such as adaptation to the laboratory workow are
additional issues, and eorts to optimize these interfaces are necessary [67]. The
widespread use of HPLC-MS for purity evaluation gives a limited insight into the
actual purity of a product because only UV-absorbing side products and those on
RP-absorbable compounds can be detected. Salts and polymer fragments are usu-
ally not detected. One should take this feature into account when assessing the
purity of combinatorial libraries if they are not puried chromatographically before
biological testing. Many suppliers sell fast HPLC-MS applications especially de-
signed for library purity checks, and all have semiautomated software packages
available for data processing.
6.3.2
Identity of Combinatorial Libraries
72]. Cycle times below 10 s can be achieved even with standard equipment. Prob-
lems can arise because of the vast amounts of data, but software packages are avail-
able from all suppliers of ESI mass spectrometers for automated processing. The
implementation of these data into the workow of combinatorial laboratories is of
course another issue and is only achievable with programming resources which
can be limited to automated reformatting of Excel tables. FIA-ESI-MS does not
give any results concerning purity, and if mixtures are obtained ionization yields
can vary dramatically and lead to false results if the product is suppressed, e.g. by
basic impurities. In addition to FIA-ESI-MS, MALDI-MS (MALDI matrix aided
laser desorption ionisation) in an oine FIA mode can also be performed.
Through an automated process MALDI targets have been coated and prepared for
measurement [73]. This application is rather complex because it is impossible to
work directly from solution. The advantage however is that it is the only method
which gives a molecular peak from larger molecules such as proteins.
6.3.3
Quantication of Combinatorial Libraries
6.4
Automated Sample Processing
6.4.1
Sample Logistics
Tab. 6.6. Liquid handlers suitable for sample processing in parallel synthesis.
Fig. 6.15. Flow chart of sample logistics if preparative HPLC is used for purication.
6.4.2
Evaporation
Centrifuge Advantages/disadvantages
Vacuum centrifuges have the advantage that almost every solvent can be evapo-
rated, even those with very high boiling points such as dimethylsulfoxide (DMSO).
Evaporation times are usually fast and the samples are concentrated and centri-
fuged at once, which results in all the material ending up at the bottom of the vial.
Disadvantages are the thermal stress, which can lead to degradation of the prod-
ucts, and the possibility of imbalance due to the dierent concentrations of solvent
mixtures.
Several suppliers sell vacuum centrifuges for evaporation of HPLC fractions or
cleaving solutions from solid-phase chemistry. Usually, carrier formats are xed or
limited. The microplate footprint is the most common, which makes it necessary
to optimize the whole workow to this footprint. Advantages and disadvantages of
several commercially available instruments are shown in Table 6.7.
References
Part II
Synthetic Chemistry
7
Radical Reactions in Combinatorial Chemistry
A. Ganesan and Mukund P. Sibi
7.1
Introduction
been reported on solid phase. While this does not change the actual chemistry
compared with traditional solution-phase radical reactions, there are contexts in
which the solid-phase environment may be advantageous. Even in the swollen gel
phase, reactions are kinetically slower than in homogeneous solution-phase con-
ditions, which may be helpful in determining the relative partitioning of a radical
intermediate into various pathways. A second unique feature is the loading of poly-
meric resins, typically a1 mmol g1 of beads, thus eectively placing an upper
limit on the maximum concentration attainable with solid-phase reactions. This
enforced dilution can be useful for radical reactions, although gel-phase polymer
chains do exhibit signicant freedom of motion and certainly do not approach
innite dilution. A further variable is the spacer length between the substrate
and the matrix. Many reactions, including Merrields original peptide synthesis,
were carried out with the substrate attached directly to the polystyrene matrix.
Today, it is more common to include a linker [4] between the polymer and the
substrate. The linker can profoundly inuence chain mobility as well as the poly-
mer microenvironment where the reaction is taking place. Linkers are often used
for solid-phase radical reactions, although the reasons for selecting a particular
linker are seldom described. Finally, the ability to lter o reagents and byproducts
can certainly be a bonus for solid-phase radical reactions, especially those mediated
by tin reagents, whose removal is not always trivial in solution phase (see below).
7.2
Intramolecular Radical Additions to sp 2 and sp Carbon
Scheme 7.2. Aryl iodide cyclizations by Du and Armstrong [7, 8]. THF, tetrahydrofuran.
Scheme 7.3. Aryl iodide cyclizations by the Novartis group [10, 11].
The Novartis group has also studied [12] the radical cyclization of cyclohexene-
diols, immobilized by a ketal linkage on polystyrene (Scheme 7.4). The reaction of
14 gave the desired dihydrobenzofuran (15) and the uncyclized product of direct
reduction (16). Jia et al. [13] have reported a related cyclization with allylamine (17)
immobilized on polystyreneWang resin. The reaction was monitored by acetyla-
tion and cleavage to yield 18, as a mixture of free and Boc-protected amines. This
solid-phase synthesis of seco-CBI (18, R H), related to the pharmacophore of the
CC-1065 and duocarmycin class of cyclopropylindole antitumor antibiotics, has
potential for the preparation of analog libraries, and an example of further conver-
sion of resin-bound 18 to a polyamide has been presented.
A series of bromoacetals (19) (Scheme 7.5) linked to polystyrene was cyclized to
the corresponding acetals by Watanabe et al. [14]. Oxidative cleavage of the resin
using Jones reagent gave easy access to g-butyrolactones (20).
7.3
Intermolecular Radical Additions
Scheme 7.4. Aryl and vinyl iodide cyclizations by Berteina et al. and Jia et al. [12, 13].
feasibility of solid-phase radical reactions. Sibi and Chandramouli [15] reported the
rst examples of intermolecular radical allylations. The polymer-bound electro-
philic C-radicals generated from a-bromo esters (21) gave g; d-unsaturated acids
(22) (Scheme 7.6). Large excesses of allylstannane and AIBN were required for
good yields. Radical initiation at room temperature using Et3 B/O2 was less e-
cient. The yields were similar to those for solution-phase reactions, while reduc-
tion of 21 with tributyltin deuteride gave @93% deuterium incorporation, implying
<7% hydrogen atom transfer from the polymer matrix. Electron-withdrawing
groups at the 2-position of the allylstannane were also found to improve the reac-
tion yield, as in the formation of 23.
Enholm et al. have reported the rst example of a diastereoselective radical re-
action on a polymer (Scheme 7.7) [16]. A sugar-based auxiliary attached to a non-
crosslinked soluble polystyrene polymer (24) as the support was used in the
experiments. Allylation under standard conditions gave 25 in high yield and dia-
stereoselectivity (97%). Use of Lewis acids for the allylation was detrimental in that
chiral auxiliary cleavage was observed.
Although organotin reagents are the most popular reagents for generating radi-
cals in solid-phase reactions, other methods are also being explored. Zhu and Ga-
nesan have investigated [17] the conjugate addition of radicals generated from
Barton esters. Acrylic acid was immobilized as the ester (26) (Scheme 7.8) using
the Wang linker on both polystyrene and TentaGel resins. Conjugate radical addi-
tion and cleavage gave acids (27) in high yields. Also noteworthy, as in the tin-
mediated reactions, is the ability to use large excesses of reagent without problems
in product isolation. The yields with polystyrene resin were similar to solution-
phase results with methyl acrylate and appear to be relatively insensitive to the
nature of the radical: primary, secondary, or tertiary. Benzyl radical gave 27 in only
32% yield, presumably due to its highly stabilized nature. The TentaGel resin gave
consistently lower yields than polystyrene, suggesting interference from the poly-
ether spacer. When acrylic acid was immobilized as the amide on polystyrene
Rink resin, yields were also lower, which is consistent with the lower solution-
phase yields with acrylamide than with methyl acrylate.
An interesting cascade reaction was observed with the Barton ester (28) of
cyclopent-2-enylacetic acid. Radical generation and addition to the acrylate resin
results in electrophilic acyl-substituted radical 29, which can undergo chain trans-
fer to produce 30 (after cleavage and esterication). The major pathway is intra-
molecular 5-exo cyclization to provide radical 29a. The intermediate nucleophilic
radical 29a gave two dierent products depending on the termination step. Chain
transfer gave 31 as a minor product. Alternatively, addition of a second acrylate
followed by thiyl transfer gave 32 as the major product. Formation of 32 requires
the crosslinking of two polymer chains and shows that site isolation is not signi-
cant.
7.3 Intermolecular Radical Additions 233
Barton ester chemistry also featured in the work by Attardi and Taddei [18].
Irradiation of a solid-phase-bound Barton ester derived from aspartic acid (Scheme
7.9) gave the corresponding alkyl radical, which could be trapped by CBrCl3 pro-
vided it was in very large excess, to give the alkyl bromide (34) after cleavage. Also
presented were examples in which the alkyl bromide was displaced by an amine,
providing a route to the solid-phase synthesis of unnatural amino acids. Trapping
the alkyl radical by intermolecular conjugate addition to methyl acrylate or a nitro-
olen was also attempted, although desired products such as 35 were contami-
nated by 1020% of 36.
A route to a-amino acids by conjugate radical addition to the dehydroalanine
derivative (37) on solid phase has been reported by Yim et al. [19]. Organo-
mercurials as radical precursors (Giese method) gave better yields of 38 in contrast
to the standard conditions of reaction between alkyl halides with tributyltinhydride
(Scheme 7.10). The intermolecular addition of tosyl radicals to an unactivated al-
kene and alkyne (39) has also been reported [20], and the reaction has been found
to be quite sensitive to solvent toluene giving optimum results. The related free
radical addition of thiols to unactivated alkenes has been described by Plourde
et al. [21]. An inositol derivative was directly immobilized on carboxypolystyrene
234 7 Radical Reactions in Combinatorial Chemistry
resin and reacted with aliphatic thiyl radicals generated by AIBN, yielding 42
(Scheme 7.11) after cleavage. The reaction was not observed with arylthiyl radicals,
possibly because of their increased stability. Similarly, functionalization of an ami-
nocyclitol linked to the Wang resin to give 44 was accomplished.
Amines can be readily prepared by intermolecular radical additions to oxime
ethers. In a nice extension of its work in solution-phase chemistry, Naitos group
has published several papers on amine synthesis on solid support. Immobilized
glyoxylic oxime ether (45) underwent radical addition to give amino acid deriva-
tives (46) in moderate to good yields (Scheme 7.12) [22]. Preparation of enan-
tioenriched amino acids by this strategy has also been reported. Chirality was in-
corporated in the form of Oppolzers camphorsultam and the solid support was
attached on the oxime hydroxyl (47). Ethyl radical addition to 47 using either tri-
ethylborane or diethylzinc as radical initiators gave product (48a) with >95% dia-
stereoselectivity [23]. This was better than the analogous solution-phase reaction,
suggesting that the immobilized oxime is less reactive. The addition of isopropyl
7.3 Intermolecular Radical Additions 235
Scheme 7.12. Intermolecular oxime ether additions by Miyabe et al. [22, 23].
and cyclohexyl radicals by atom transfer from the corresponding iodide, to furnish
48b,c, has also been demonstrated. In these cases, the product was contaminated
by the competing addition of initiating ethyl radical, and this was avoided by using
a large excess of the alkyl iodide.
Pyrrolidines have been synthesized on solid phase by a combination of an inter-
molecular radical addition followed by an intramolecular oxime ether cyclization,
as exemplied in the preparation of 50 [24] and 52 [25] (Scheme 7.13). The solid-
phase reactions were sluggish with triethylborane as an initiator at room tempera-
ture, while the analogous solution-phase process was kinetically much faster. Rad-
236 7 Radical Reactions in Combinatorial Chemistry
Scheme 7.13. Intermolecular oxime ether additions by Miyabe et al. [24, 25].
iodides, and the tandem cyclization sequence with iodide (57) to aord bicyclic 58
has been accomplished, albeit in modest yield.
7.4
Functional Group Removal
More recently, several groups have explored the homolytic cleavage of resin-
bound selenides as a route to traceless synthesis. Nicolaou et al. [30] started with
lithiated polystyrene, which was trapped by dimethyl diselenide to furnish the
alkylselenide resin. Reaction with bromine generated selenylbromide resin (63)
(Scheme 7.16), which could be reduced to the selenide anion (64). This resin can
238 7 Radical Reactions in Combinatorial Chemistry
Scheme 7.16. Traceless selenide cleavage by Nicolaou et al. and Ruhland et al. [30, 31].
then be reacted with alkyl halides, followed by traceless reductive cleavage with
tributyltinhydride. A similar dealkylation was simultaneously reported by Ruhland
et al. [31], in which a chloro- and a bromoalcohol were immobilized by the sele-
nide resin (70). The free alcohol was then functionalized by a Mitsunobu reaction
prior to traceless cleavage. Nicolaou also described the use of resin 63 in pro-
moting the glycosidations of sugar glycals, resulting in a polymer-bound selenide
whose radical cleavage yields 2-deoxyglycosides. The Nicolaou group has since re-
ported a number of further applications with these selenide resins.
In the previous sections, many solid-phase radical reactions were described that
had an extended linker between the substrate and the polymer matrix, unlike the
Nicolaou and Ruhland examples. Fujita et al. were the rst to report [32] a trace-
less selenide cleavage with a linker. However, the yield from 73 (Scheme 7.17)
was poor, and the authors suggest that the tinhydride reagent is reacting with the
linker itself. A more robust ether-based selenide linker (76) has recently been de-
scribed by Li et al. [33], and an example of homolytic cleavage proceeded with good
yields.
7.5
Polymer-supported Reagents for Radical Chemistry
As was illustrated in the above sections, radical reactions are feasible on solid sup-
port and are relatively new entrants to the eld. In contrast, the use of polymer-
7.5 Polymer-supported Reagents for Radical Chemistry 239
Scheme 7.17. Traceless selenide cleavage by Fujita et al. and Li et al. [32, 33].
supported organic reagents has a much longer history in radical chemistry. Devel-
opments in this area can be attributed to the diculty in removal of organotin by-
products, since tin-derived compounds are used extensively for radical chemistry.
Use of polymer-supported reagents alleviates most of the diculties associated
with product purication. The tin-containing polymer can be easily removed by
simple washing.
7.5.1
Polymer-supported Tinhydrides
(Scheme 7.21). The reductions use only catalytic amounts of the polymeric tinhy-
dride and sodium borohydride as the co-reductant. Several examples (9094) of
chemoselective reductions have been reported. The tin contents in the products
were also determined and shown to be in the ppm range.
Curran and coworkers [39] have developed novel alternatives for polymer-
supported reagents. They have successfully demonstrated the use of very versatile
uorous reagents (95) for free radical chemistry [39]. The uorous tinhydrides
(Scheme 7.22) are commercially available and have the advantages of ease of prod-
uct purication by simple separation and high reaction rates since the reactions
are carried out in the solution phase.
242 7 Radical Reactions in Combinatorial Chemistry
7.5.2
Polymer-supported Allyl Stannane
7.5.3
Polymer-supported Reagents for Atom-transfer Reactions
Atom- and group-transfer radical reactions have enjoyed a lot of popularity in re-
cent years. Clark and coworkers [42] have prepared solid-supported catalysts for
atom-transfer radical cyclizations (Scheme 7.24). Functionalized aminopropyl sili-
cagel was coupled to pyridine 2-carboxaldehye to provide an orange solid (102).
The catalyst (103) was prepared by stirring a mixture of 102 and copper halide
in acetonitrile. Atom transfer cyclization of several N-allyl amides using 30 mol%
of the catalyst gave the product in high yields. The catalyst could be recovered by
simple ltration. Recently, a soluble copper catalyst supported on poly(ethylene)-
block-poly(ethylene glycol)-polymer (108) has been eectively used for atom-
transfer radical polymerizations [43]. Marsh and coworkers [44] have reported
Cu(I)-mediated radical polymerization of nucleoside monomers on silica support.
7.5 Polymer-supported Reagents for Radical Chemistry 243
7.5.4
Photochemical Generation of Radicals
7.6
Summary
References
5 For a review, see: B. K. Banik, Curr. Dahmen, Chem. Eur. J. 2000, 6, 1899
Org. Chem. 1999, 3, 469496. 1905.
6 A. Routledge, C. Abell, S. 28 M. J. Plunkett, J. A. Ellman, J. Org.
Balasubramanian, Synlett 1997, 61 Chem. 1995, 60, 60066007.
62. 29 I. Sucholeiki, Tetrahedron Lett. 1994,
7 X. Du, R. W. Armstrong, J. Org. 35, 73077310.
Chem. 1997, 62, 56785679. 30 K. C. Nicolaou, J. Pastor, S.
8 X. Du, R. W. Armstrong, Tetrahedron Barluenga, N. Winssinger, Chem.
Lett. 1998, 39, 22812284. Commun. 1998, 19471948.
9 For a review, see: S. Wendeborn, A. 31 T. Ruhland, K. Anderson, H.
De Mesmaeker, W. K. D. Brill, S. Pedersen, J. Org. Chem. 1998, 63,
Berteina, Acc. Chem. Res. 2000, 33, 92049211.
215224. 32 K. Fujita, K. Watanabe, A. Oishi, Y.
10 S. Berteina, A. De Mesmaeker, Ikeda, Y. Taguchi, Y. Synlett 1999,
Tetrahedron Lett. 1998, 39, 57595762. 17601762.
11 S. Berteina, S. Wendeborn, A. De 33 Z. Li, B. A. Kulkarni, A. Ganesan,
Mesmaeker, Synlett 1998, 12311233. Biotechnol. Bioeng. 2001, 71, 104106.
12 S. Berteina, A. De Mesmaeker, S. 34 a) U. Gerigk, M. Gerlach, W. P.
Wendeborn, Synlett 1999, 1121 Neumann, R. Vieler, V. Weintritt,
1123. Synthesis 1990, 458452; b) M.
13 G. Jia, H. Iida, J. W. Lown, Synlett Gerlach, F. Jordens, H. Kuhn,
2000, 603606. W. P. Neumann, M. Peerseim, J.
14 Y. Watanabe, S. Ishikawa, G. Takao, Org. Chem. 1991, 56, 59715972;
T. Toru, Tetrahedron Lett. 1999, 40, c) C. Bokelmann, W. P. Neumann,
34113414. M. Peterseim, J. Chem. Soc., Perkin
15 M. P. Sibi, S. V. Chandramouli, Trans. 1 1992, 3165.
Tetrahedron Lett. 1997, 38, 89298932. 35 a) G. Dumartin, M. Pourcel, B.
16 E. J. Enholm, M. E. Gallagher, S. Delmond, O. Donard, M. Pereyre,
Jiang, W. A. Batson, Org. Lett. 2000, Tetrahedron Lett. 1998, 39, 46634666;
2, 33553357. b) G. Dumartin, G. Ruel, J.
17 X. Zhu, A. Ganesan, J. Comb. Chem. Kharboutli, B. Delmond, M.-F.
1999, 1, 157162. Connil, B. Jousseaume, M. Pereyre,
18 M. E. Attardi, M. Taddei, Tetra- Synlett 1994, 952954; c) G. Ruel, N.
hedron Lett. 2001, 42, 35193522. K. The, G. Dumartin, B. Delmond,
19 A.-M. Yim, Y. Vidal, P. Viallefont, J. M. Pereyre, J. Organomet. Chem.
Martinez, Tetrahedron Lett. 1999, 40, 1993, 444, C18C20.
45354538. 36 P. Boussaguet, B. Delmond, G.
20 S. Caddick, D. Hamza, S. N. Dumartin, M. Pereyre, Tetrahedron
Wadman, Tetrahedron Lett. 1999, 40, Lett. 2000, 41, 33773380.
72857288. 37 a) A. Chemin, H. Deleuze, B.
21 R. Plourde Jr, L. L. Johnson Jr, Maillard, Eur. Polym. J. 1998, 34,
R. K. Longo, Synlett 2001, 439441. 1395; b) A. Chemin, H. Deleuze, B.
22 H. Miyabe, Y. Fujishima, T. Naito, J. Mailard, J. Chem. Soc., Perkin Trans.
Org. Chem. 1999, 64, 21742175. 1, 1999, 137142.
23 H. Miyabe, C. Konishi, T. Naito, 38 E. J. Enholm, J. P. Schulte, II, Org.
Org. Lett. 2000, 2, 14431445. Lett. 1999, 1, 12751277.
24 H. Miyabe, H. Tanaka, T. Naito, 39 a) D. P. Curran, S. Hadida, S.-Y.
Tetrahedron Lett. 1999, 40, 83878390. Kim, Z. Luo, J. Am. Chem. Soc. 1999,
25 H. Miyabe, K. Fujii, H. Tanaka, T. 121, 66076615; b) D. P. Curran, S.
Naito, Chem. Commun. 2001, 831 Hadida, J. Am. Chem. Soc. 1996, 118,
832. 25312532; c) I. Ryu, T. Niguma, S.
26 G.-H. Jeon, J.-Y. Yoon, S. Kim, S. S. Minakata, M. Komatsu, S. Hadida,
Kim, Synlett 2000, 128130. D. P. Curran, Tetrahedron Lett. 1997,
27 For a review, see: S. Brase, S. 38, 78837886.
246 7 Radical Reactions in Combinatorial Chemistry
8
Nucleophilic Substitution in Combinatorial and
Solid-phase Synthesis
Jan-Gerd Hansel and Stephan Jordan
8.1
Introduction
8.2
Nucleophilic Substitution at Aliphatic Carbons
8.2.1
General Remarks
8.2.2
Halogen Nucleophiles
Benzylic alcohols are converted into the corresponding benzyl chlorides or bro-
mides by reaction with phosphorus trihalides or under milder conditions with
triphenylphosphine (TPP) and tetrahalogenomethanes. The reaction has found
widespread application in the activation of benzyl alcohol-type linkers for subse-
quent attachment of nucleophilic substrates to a solid support [6].
The reaction was also applied to one of the masterpieces of multistep solution-
phase synthesis involving polymer-supported reagents and scavengers. During
Leys pyrrole synthesis, all six reaction steps to the highly diverse pyrroles used
these reagents. In those cases where the benzylic halides are not commercially
available, the alkyl halides used for N-alkylation are prepared from the corre-
sponding alcohols. In the bromination step, a reagent combination of polymer-
supported TPP and tetrabromomethane is used (see Scheme 8.2) [7].
8.2.3
Oxygen Nucleophiles
Carboxylate alkylation by alkyl halides can be used both on solid phase (see above)
and in solution [8] to give alkyl carboxylic acid esters. For optimum results, cesium
salts are employed. Alternatively, carboxylate alkylation can be performed under
Mitsunobu conditions using alcohols as electrophiles. For solid-phase applications
the common reagents TPP and diethyl azodicarboxylate (DEAD) can be used. In
solution-phase parallel synthesis, it is advantageous to use reagents that do not re-
quire a chromatographic product isolation. A combination of polymer-bound TPP
[9] and di-tert-butyl azodicarboxylate has been found to be particularly useful (see
Scheme 8.3) [10]. While the former reagent is ltered o easily, the latter (and its
byproduct di-tert-butylhydrazodicarboxylate) is readily converted into volatile com-
pounds by treatment with triuoroacetic acid (TFA).
Cesium carbonate in the presence of an alcohol and carbon dioxide can be alky-
lated by Merrield resin to give solid-supported carbonates (see Scheme 8.4). Add-
ing an amine instead of an alcohol gives rise to carbamate formation [11].
Alcohols are transformed into ethers by reaction with an alkylating agent in the
presence of a very strong base such as sodium hydride. The reaction, known as the
Williamson ether synthesis, is frequently used for attaching alcoholic substrates to
Merrield-type resins (see Scheme 8.5) [12].
The solid-phase Mitsunobu reaction has been employed for linker attachment
[16] as well as diversity generation [17]. In solution-phase chemistry, specially
designed reagents are used to facilitate product separation (see above). Thus,
polymer-bound TPP and DEAD derivatives that are easily ltered o are commer-
cially available. Also, ionophoric side-chains have been attached to both TPP and
DEAD to make them (and the corresponding byproducts) easily separable using
ion-exchange resins [18]. This technique is an example of a general concept called
phase tagging or phase labeling [19].
The solution-phase reaction of phenols with epoxides is catalyzed by polymer-
supported Co(salen) complexes [20]. Using chiral catalysts the corresponding 1-
aryloxy-2-alcohols are obtained in high yields, purities, and enantiomeric excesses.
8.2 Nucleophilic Substitution at Aliphatic Carbons 251
8.2.4
Sulfur Nucleophiles
8.2.5
Nitrogen Nucleophiles
The direct N-alkylation of primary amines is in general not the method of choice
for the synthesis of secondary amines because of potential of overalkylation. There-
252 8 Nucleophilic Substitution in Combinatorial and Solid-phase Synthesis
Intramolecular amine alkylation does not usually bear the risk of overalkyla-
tion and is an excellent way to close heterocyclic rings. An Epibatidine synthesis
using polymer-supported reagents provides an example that involves a mesylate
substitution [35]. Even strained rings such as aziridines can be formed (Gabriel
Cromwell reaction) [36].
Epoxide opening with amine nucleophiles is frequently used in combinatorial
chemistry since it leads to the attractive aminoalcohol substructure. When cata-
lyzed with Lewis acids, for example lithium perchlorate, the reaction proceeds
smoothly with a range of alkyl amines, anilines, and heteroaromatic amines. The
reaction is useful in solution-phase synthesis (see Scheme 8.12) [37] as well as in
solid-phase synthesis [38].
Other nucleophiles that are readily alkylated on the solid phase by alkyl halides
or sulfonates include hydroxyl amine [40], O-benzyl hydroxylamine [41], hydrazine
derivatives [42], azide ion [43], and sulfonamides [44].
Nitrogen compounds bearing hydrogens of sucient acidity can also be alky-
lated under Mitsunobu conditions (see above). Thus, a polymer-bound imidodicar-
bonate can be alkylated with primary and secondary alcohols using the standard
reagents TPP and DEAD. After cleavage o the resin, primary amines are obtained
(see Scheme 8.14) [45].
8.2.6
Ring-closing Reactions
8.3
Nucleophilic Substitution at Aromatic Carbons
8.3.1
General Remarks
8.3.2
Nitrogen Nucleophiles
From a historical point of view, the rst template for SN Ar reactions was cyanuric
chloride (2,4,6-trichloro-1,3,5-triazine), which is commercially available and inex-
pensive [50]. Compounds containing this core element have shown biological
activity, e.g. as herbicides (Atrazin).
A research group at ArQule showed that the selective and sequential derivati-
zation of cyanuric chloride could be achieved by simply controlling the reaction
temperature. The generality of the research groups method has been proven by
the solution-phase synthesis of a large combinatorial library of over 40,000 indi-
vidual compounds (see Scheme 8.15) [51]. The rst chloride substitution proceeds
at 20 C using N,N-diisopropylethylamine (DIPEA) as the base and acetonitrile
as the solvent. Even anilines react with the very reactive cyanuric chloride in the
proposed way. Usually, the arylation of the primary amine reduces its nucleophi-
licity strongly. Therefore, no bis-arylation occurs at this position and the second
chlorine atom can be substituted at room temperature. Owing to the relatively low
reactivity of the third position, heating is required for the last exchange and only
strong nucleophiles such as secondary amines give pure products and high con-
versions. Besides amines and anilines, also carbohydrates, dipeptides, amidines,
and a-ketoamides can be incorporated, giving access to more complex structures.
Upon screening of these compounds, a series of hits in the cardiovascular area has
emerged.
Using the cyanuric chloride template for solid-phase synthesis has become a
very ecient method for the production of large combinatorial libraries. This was
also demonstrated by the synthesis of a 12,000-compound library by Stankova and
Lebl [52]. The rst chlorine atom was selectively substituted by coupling the tem-
256 8 Nucleophilic Substitution in Combinatorial and Solid-phase Synthesis
plate to a resin loaded with amino acids (see Scheme 8.16). Taking advantage of
the decreasing reactivity of tri-, di-, and monochlorotriazines, various nucleophiles
(amines, anilines, hydrazines) were introduced at dierent temperatures [53].
Likewise, an 8000-membered library on a cellulose-based polymeric membrane
has been synthesized [54].
The reaction sequence can be extended to related starting materials such as 2,6-
dichloropurines, although the reaction conditions need to be harsher [55]. In a
representative example, the dichloropurine was treated with a primary amine at
elevated temperature. For the second substitution, reux conditions and ve equiv-
alents of amine were necessary. Excess amine was removed by the use of formyl-
polystyrene beads. The compounds could be benzylated at the N-9 position by an
alkylation protocol or by using the Mitsunobu reaction (see Scheme 8.17) [56].
THF, the resulting compound was cyclized with DIC and HOBt and the 1,5-
benzodiazepin-2-one was obtained. Alternatively, selective alkylation at the N-5 po-
sition adds further diversity to the library [66].
Using a-amino acids in the place of the b-amino acids, the [6,6]-fused ring sys-
tem of quinoxalin-2-ones has been accessed [67]. The synthetic strategy is an
adaptation of TenBrinks and coworkers solution-phase synthesis on solid phase
[68]. Variably substituted tetrahydroquinoxalin-2-ones can also be prepared based
on 4-uoro-3-nitrobenzoic acid. After substitution of the uorine with primary ali-
phatic amines at room temperature and reduction of the nitro group, double acy-
lation with chloroacetic anhydride has been shown to be the key step in the syn-
thesis [69].
Further ring contraction to [6,5]-fused systems such as benzimidazoles has
been the aim of other synthetic eorts in solution and in solid-phase synthesis.
Again, 4-uoro-3-nitroarenes were linked to a solid support via an ether linkage
[70] or via a carboxylic acid [71]. Commonly, both strategies use a SN Ar displace-
ment reaction of the uorine atom by an amine with subsequent reduction of
the nitro group. Whereas Phillips and Wie [70] achieved immediate cyclization by
condensation with benzimidates, a research group at Aymax acylated the inter-
mediate with an activated bromoacetic acetic acid rst (see Scheme 8.25). After
displacement of the bromide groups by nucleophiles, cyclization occurred upon
cleavage with a concomitant dehydration [72].
DMF, THF) and precipitates in particular solvents (e.g. diethyl ether) (see Scheme
8.26). Again, 4-uoro-3-nitrobenzoic acid was loaded onto the support and was
then allowed to react with a variety of amines. After reduction of the nitro group,
cyclization was achieved with trichlorophosgene.
into the enamine and cyclized using tetramethylguanidine (TMG). Many amines
can be incorporated by displacement of the uorine atom at C-7 before the prod-
ucts are cleaved o the resin [79].
8.3.3
Oxygen Nucleophiles
8.3.4
Sulfur Nucleophiles
The uoro-nitroarene motif is also the most preferred template for SN Ar for
sulfur nucleophiles. In many applications, a suitably protected form of cysteine as a
b-mercapto acid is reacted with 4-uoro-3-nitrobenzoic acid to form 1,5-benzothia-
zepin-4-ones, an important class of drugs in the treatment of cardiovascular dis-
orders. For example, 4-uoro-3-nitrobenzoic acid was treated with 1.5 equivalents
of 9-uorenylmethoxycarbonyl (Fmoc)-l-cysteine in DMF to be converted to the 2-
nitro-thioether (see Scheme 8.34) [85]. After the reduction of the nitro group and a
subsequent reductive alkylation (dicult because of the poor nucleophilicity of
some anilines), the resulting secondary anilines were cyclized to form the seven-
member thiazepine ring.
8.3.5
Macrocyclization Reactions
In this section the reaction for closing large rings by nucleophilic substitution of
activated arenes will be discussed. Cyclorelease reactions such as those used in the
well-established syntheses of hydantoins [88] or diketopiperazines [89] will not be
covered.
At rst sight, the closure of medium-sized or large rings by nucleophilic dis-
placement does not meet the demands of combinatorial synthesis. No new sub-
stituents are incorporated and therefore no diversication is achieved. On closer
inspection, the great importance of this reaction becomes obvious in view of the
immense dierence between the three-dimensional structure of a linear oligopep-
tide and the corresponding cyclic analog. So SN Ar macrocyclizations have become
an important part of solid-phase organic synthesis, especially for the preparation of
libraries of b-turn mimics.
Based on the experiences presented earlier in this chapter, suitably substituted
uoro-nitrobenzoic acids are the substrates of choice for intramolecular SN Ar re-
actions [90].
Small libraries of 14-membered macrocyclic diaryl ethers and thioethers can be
produced using a very similar procedure. Precursors are synthesized by the acyla-
tion reaction of solid-supported peptides with 3-uoro-4-nitro benzoic acid. The
substrates undergo cyclization by displacement of the uorine with the phenolic
oxygen of a tyrosine derivative [91] or with the thiol group of cysteines [92] under
exceptionally mild conditions (see Scheme 8.36). Diversity could be increased via
postmodication reactions of the nitro group.
Recently, Burgess and coworkers reported in detail on their research on libraries
of peptide turn mimetics [93]. The eects on nucleophilicity, product ring size,
resins, and other reaction conditions were examined. Optimized procedures to
produce 13- to 16-membered ring systems are described [93].
Larger rings are synthesized by the replacement of the uorine atom in three
regioisomeric uoro-nitrobenzoic acids to prepare analogs of tocinoic acid (see
Scheme 8.37) [94]. In this synthesis the amino group of lysine serves as an inter-
nal nucleophile for the closure of the macrocycle.
266 8 Nucleophilic Substitution in Combinatorial and Solid-phase Synthesis
References
9
Electrophilic Substitution in Combinatorial and
Solid-phase Synthesis
Jan-Gerd Hansel and Stephan Jordan
9.1
Introduction
9.2
Electrophilic Substitution at Aliphatic Carbons
9.2.1
Halogen Electrophiles
9.2.2
Nitrogen Electrophiles
9.2.3
Carbon Electrophiles
9.3
Electrophilic Substitution at Aromatic Carbons
9.3.1
General Remarks
Almost all kinds of SE reactions involving arene substrates are used in solution-
phase chemistry as a powerful synthetic tool. In solid-phase chemistry the use of
this reaction is limited owing to its incompatibility with electron-rich linkers. With
the exception of a few examples mentioned below, aromatic SE is restricted to the
functionalization of polystyrene-based supports [10].
Polystyrene can be brominated, nitrated, and acylated or alkylated applying
FriedelCrafts conditions in solvents such as carbon tetrachloride or nitrobenzene.
For example, in 1988 Ajayaghosh and Pillai [11] demonstrated the preparation of a
photosensitive resin using SE reactions (see Scheme 9.4). Commercially available
9.3.2
Halogen Electrophiles
There are only a few examples of arene brominations on solid support in the
literature. Using N-bromosuccinimide in dimethyl formamide at room tempera-
ture, electron-rich arenes such as thiophenes can be brominated (see Scheme 9.6).
Combination with a Stille coupling and reiteration of the reaction sequence leads
to oligothiophenes new materials with interesting optical and electronic prop-
erties [14].
A rare example of direct halogenation on a solid support has been reported for
phenols. The phenolic moiety of tyrosine undergoes iodination when treated with
bis-(pyridine) iodonium(I) tetrauoroborate (Ipy2 BF4 ) for no more than 10 min
(see Scheme 9.7) [15].
The only example in this chapter in which the leaving group in an SE reaction is
not a proton involves germanium-based linkers. These linkers have been devel-
oped for solid-phase synthesis as a means of traceless linkage (see Chapter 4).
However, reaction of germanium-linked substrates with bromine rapidly releases
the corresponding aryl bromides by ipso substitution of the germanium by bro-
mine (see Scheme 9.8). Aryl iodides can be prepared by the same method using
iodomonochloride [16].
9.3.3
Nitrogen Electrophiles
9.3.4
Carbon Electrophiles
A PictetSpengler reaction has been used as the key step during the synthesis of
indolyl diketopiperazine-based libraries of Fumitremorgin C analogs (see Scheme
9.13). This natural product, isolated by fermentation of the fungi Aspergillus fumi-
gatus, appears to be of interest in the area of central nervous system (CNS) and
cancer research and has therefore resulted in the preparation of some solid-phase
combinatorial libraries [22].
The reaction has also been applied to solution-phase synthesis, but has found
little application in library production to date [25].
In a Reissert-type reaction involving pyrrole derivatives, CaC bond formation is
achieved by aromatic SE on solid support using N-oxides (see Scheme 9.16). The
reaction also works with indoles and with enamines [26].
References
10
Elimination Chemistry in the Solution- and
Solid-phase Synthesis of Combinatorial
Libraries
Demosthenes Fokas and Carmen Baldino
10.1
Introduction
One of the challenges for organic chemists involved in the burgeoning eld of
combinatorial chemistry is to rediscover new uses and applications of old reactions
and subsequently adapt them to either solid- or solution-phase chemistry. Elimi-
nation reactions, which have been studied thoroughly and used extensively by the
chemistry community in the synthesis of several complex molecules, fall into this
category [1]. Among them, b-eliminations prevail in organic synthesis with nu-
merous applications in the preparation of olens. Although olens are versatile
and useful intermediates for combinatorial chemistry, the synthesis of libraries of
olenic substrates by adaptation of the classical b-elimination reaction has not re-
ceived much attention so far, presumably because of the lack of convergence in the
synthesis and diversity of the nal products. Instead, b-eliminations along with
other elimination reactions are gaining favor in solid-phase synthesis as a release
strategy of the desired products from solid support.
In this chapter we will address elimination reactions from a mechanistic
rational including: (1) b-eliminations, (2) conjugate eliminations, and (3) addition
elimination reactions. We will also discuss the utility of these transformations
in combinatorial chemistry. Olenation reactions such as the Wittig, Horner
Emmons, and ring-closing olen metathesis, which could fall into the addition
elimination category, will not be discussed herein since they will be addressed in
dierent chapters.
10.2
b-Eliminations in Combinatorial Chemistry
b-Elimination has been used for peptide synthesis since 1967. However, broader
applications have been limited, presumably because the cleavage conditions as de-
scribed needed to be tightly controlled [2]. The expanded utility of b-elimination
reactions in combinatorial chemistry and solid-phase synthesis has been realized
very recently with the advent of new linkers [3]. The design of traceless linkers that
leave no residue on the cleaved product enabled chemists to envision b-elimination
reactions as an eective release strategy of the desired products from solid sup-
port. The term traceless usually denes linkers that leave no obvious residue on
the cleaved molecule. A traceless linker is dened as one where a new CaH or
CaC bond is formed at the linkage site of the cleaved molecule. However, this def-
inition has been expanded to linkers that include other cleavage reactions. Al-
though there are several transformations in combinatorial chemistry that involve
an intermediate b-elimination step, we will focus on release-based b-eliminations
and their application to combinatorial chemistry.
10.2.1
The Hofmann Elimination Solid-phase Synthesis of Tertiary Amines
to give ammonium salt (4) introduces another site of diversity and activates the
linker for cleavage by a facile Hofmann elimination reaction. Then, NiPr2 Et at
room temperature liberates the tertiary amine (5) into solution and regenerates the
acrylate resin (2). Similarly, tertiary amines with the general structure 5 and with
three sites of diversity can result from a resin-bound equivalent of ammonia (6),
which can be derived from the coupling of 9-uorenylmethoxycarbonyl (Fmoc)-b-
alanine to hydroxymethylpolystyrene resin followed by Fmoc deprotection. Since
the resin linker 2 is regenerated after elimination of the product and is function-
alized via a Michael reaction, this resin was referred to as a REM (regenerated Mi-
chael acceptor) resin.
The REM resin system can be used in the monoalkylation of diamines without
the use of protecting groups (Scheme 10.2). For example, piperazine can be added
to resin 2 to give the monoalkylated derivative 7, which can then be acylated or al-
kylated cleanly at the second nitrogen [7]. Treatment of resin-bound piperazine (7)
with isocyanates or alkyl halides can generate compounds of the general structure
8 or 9 respectively. Quaternization and elimination provides disubstituted piper-
azine adducts 10 and 11 in high purity. A library of 125 piperazines with the gen-
eral structure 11 was prepared and a few of the compounds were found to be active
against d-opioid receptors [8]. The high purity of substrate 11 demonstrates that
although quaternization at the undesired nitrogen in piperazine (9) is possible, it
does not reduce the purity of the nal product since only the desired product can
be cleaved from the resin.
The overall yield for the three-step sequence ranges from 40% to 80% and is
substrate dependent. The purity of the isolated products is good because only the
desired tertiary resin-bound amine will be susceptible to the elimination condi-
tions. The method works well when the quaternization step can be conducted at
ambient temperature using reactive alkyl halides (i.e. allyl, benzyl, and methyl).
The quaternization step only requires heat when less reactive alkylating agents are
282 10 Elimination Chemistry in the Solution- and Solid-phase Synthesis of Combinatorial Libraries
Scheme 10.3. Other REM resins used in the synthesis of tertiary amines.
mann elimination, upon treatment with Me 2 NH, to furnish vinyl sulfone resin 20
and release in solution the desired tertiary amine with general structure 19 in a
25% average yield for the six-step sequence. However, all of the nal compounds
had to be puried by reversed phase high-performance liquid chromatography
(HPLC) owing to a minor 3-chlorobenzoic acid impurity. This problem can be cir-
cumvented by extended washing of the activated resin with 2 N HCl in tetrahy-
drofuran (THF), presumably exchanging the anions of the ammonium salt for
chloride. The advantage of the sulde safety-catch resin is the decreased sensitivity
of the alkylated product to b-elimination, which could be utilized to increase the
yield or extend the scope of alkylating agents to less activated systems. However,
oxidation of the sulde resin prior to amine synthesis would be necessary for more
oxidation-sensitive systems.
tonitrile) or samarium iodide (in THF) also resulted in highly ecient cleavage
delivering tertiary methylamines of high purity in good overall yields.
The quaternization works only with methyl triate as the alkylating reagent,
thus limiting the diversity of the produced library. Although other triates pre-
pared in situ via AgOTf/alkyl halide exchange gave the corresponding quaternary
salts in solution, this approach is unsuitable for the solid-phase sequence because
of interference of reaction kinetics from the silver halide precipitate. Both routes
are amenable to solution-phase parallel synthesis by careful selection of reaction
conditions and work-up methods starting from O-benzylhydroxylamine. Although
this hydroxylamine resin (22) is more stable than the ester REM resin (2) toward a
variety of nucleophilic reagents, it limits a library design to tertiary methylamines.
Furthermore, this resin cannot be recycled once it is used an advantage of the
REM resins.
Scheme 10.6. Proposed mechanism for the two-resin system-promoted Hofmann elimination.
The basic resin then desalts the amine to catalyze the b-elimination. Alternatively,
it may also be due to trace amounts of base, present either from previous steps or
from trace dimethylamine in the DMF. It was found that yield was a function of
reaction time with an optimal length of 18 h for a 14-member library production.
Identical results generating highly pure products were obtained using a depro-
tected Rink amide resin, albeit at lower yield [13].
A novel Wang resin-bound piperazine base (26) that resembles Murphys two-
resin system was introduced by Yamamoto et al. (Scheme 10.7) [14]. It was used
successfully (2 equiv., loading 1.62 mmol g1 ), in the absence of any other external
base, to cleave N-aryl-N 0 -benzylpiperazines from the resin by treating the quater-
nary ammonium compounds with resin 26 in CH2 Cl2 for 16 h at room tempera-
ture. However, the caveat of this reagent and the ion-exchange resin is that they
complicate the reusability of the REM resin since at the end of the sequence both
resins are mixed together.
MacroKans TM [16]. The MacroKans TM were placed in a glass peptide vessel, which
was then sealed under a slight positive pressure of ammonia gas. Products were
isolated cleanly in good yields after evaporation, resin sorting, and washing with
CH3 CN or dimethylsulfoxide (DMSO). This is a particularly suitable method of
parallel processing for the synthesis of large libraries, thus minimizing or elimi-
nating the impurities due to the cleavage reagent.
10.2.2
b-Elimination on Selenyl Resins
Indeed, the newly formed double bond can serve as a starting point for the gen-
eration of secondary libraries or more focused libraries, thus introducing addi-
tional elements of diversity. For example, epoxidation of the released benzopyran
31 followed by ring opening of the intermediate epoxide 32 with a variety of nu-
cleophiles provides access to a new series of benzopyran derivatives with the gen-
eral structure 33. Additionally, benzopyrans with the general structure 34 can
result from further elaboration of the secondary hydroxyl group of 33 with a series
of electrophiles [20c]. Taking into consideration the current advances in asymmet-
ric epoxidation of olenic substrates, this sequence could provide entry to chiral
benzopyran libraries starting from chiral benzopyran epoxides [23].
The loading to the selenium solid support is compatible with a great variety of
prenylated phenols except for substrates with electron-withdrawing groups adja-
cent to the prenyl group or adjacent to the phenol hydroxyl group participating in
the cyclization step. Therefore, these scaolds have to be loaded with the electron-
withdrawing groups masked. The elimination of the selenoxide resin tether pro-
ceeds smoothly at room temperature and seems to be independent of the substi-
tution pattern of the benzopyran scaold.
A similar resin-bound selenoxide elimination on 2-seleno carbohydrates (35)
was utilized by Nicolaou et al. in the synthesis of a small library of carbohydrate
orthoesters representing novel regions of the potent antibiotic everninomicin
(Scheme 10.9) [24]. Oxidation of glycoside 35 with meta-chloroperbenzoic acid (m-
CPBA) gave the corresponding resin-bound selenoxide, which underwent a ther-
mal syn-elimination. This thermal selenoxide elimination to the intermediate ke-
tene acetal 36 introduces the desired functionality for the formation of orthoesters
37 and 38 and can release the desired products in a traceless manner from solid
support. Orthoester 37 is formed from glycoside 35a and 2,3-allyl orthoester 38 is
formed from deprotected glycoside 35b. However, unlike the solution-phase selen-
oxides, it was observed that the resin-bound selenoxide was more prone to elimi-
nate at room temperature and therefore necessitated the use of lower temperatures
in the oxidation step. Thus, treatment of selenide 35 with m-CPBA in CH2 Cl2 at
78 C, followed by rapid ltration and transfer to a sealed tube, was found to give
the best results. Although the chemistry has been developed in both solution- and
solid-phase chemistry, it is well suited to the solid-phase synthesis of novel semi-
synthetic everninomicins and other carbohydrate libraries.
10.2.3
b-Elimination on Sulfone Resins
Scheme 10.10. Release of C-terminal peptides via a b-elimination from a sulfone resin.
Apart from being used successfully in the area of peptide chemistry, sulfone
linkers have gained favor in the solid-phase synthesis of small molecules. For
example, 4-aminobenzenesulfonamides with the general structure 44 were pre-
pared from 2-mercaptoethanol resin 42 (Scheme 10.11) [27]. Resin 42 was con-
verted to sulfone 43, which underwent a facile b-elimination to release the desired
10.2.4
b-Elimination on Silyl Resins
b-Elimination was utilized in the cleavage of several silyl amide linkers (SAL) and
trimethylsilylethyl ester linkers for the facile release of peptide fragments from the
solid support (Scheme 10.14). Stabilization of a carbocation by a b-trialkylsilyl
group, as shown in intermediate 52, seems to facilitate the release of C-terminal
290 10 Elimination Chemistry in the Solution- and Solid-phase Synthesis of Combinatorial Libraries
Scheme 10.14. Release of peptides via a b-silyl elimination on silyl amide linkers.
amides (53) from silyl resin 51 [30]. b-Elimination of the trialkylsilyl group neu-
tralizes the transient carbocation to give a stable styrene derivative (54). These silyl
amide linkers gave improved yields of C-terminal tryptophan amides over conven-
tional linkers since an irreversible alkylation of the tryptophan indole nucleus
by such carbocations is suppressed. However, acid scavengers (1,2-ethanedithiol/
phenol/thioanisole, 5:3:2) were needed as the styrene moiety is sensitive to proto-
nation. Therefore, purication of the nal product is required in order to remove
the scavenger byproducts.
Similarly, linker 55 was designed to be cleaved by a b-elimination mechanism
based on the 2-(trimethylsilyl)ethylester protecting group (Scheme 10.15) [31].
Fluoridolysis or dilute acid cleavage enabled the preparation of protected peptide
fragments such as 56. C-Terminal tryptophans or prolines could be successfully
anchored with this linker and no undesired alkylation or diketopiperazine forma-
tion was observed upon cleavage.
Scheme 10.15. Release of peptides via b-silyl elimination on 2-(trimethylsilyl)ethyl ester linkers.
Scheme 10.16. Release of olens from an allyl silane resin via a b-silyl elimination.
10.2.5
b-Elimination on Fluorenyl Resins
Scheme 10.17. Release of C-terminal peptides via a b-elimination from a uorenyl resin.
10.2.6
b-Elimination on 2-(2-Nitrophenyl)ethyl Resins
10.2.7
Radical-based b-Eliminations
carboxylic acids and peptides with high yields (6593%) and purities in various
solvents such as THF, CH2 Cl2 , dioxane, and DMSO by irradiation with light above
320 nm. The photo byproducts are either volatile (CO and isobutene/isobutane) or
inert resin-bound acetone. The linker proved to be stable upon treatment with
acids and bases and to be compatible with many reagents and reaction conditions,
such as palladium-catalyzed cross-coupling and epoxidation, with broad applicabil-
ity in combinatorial chemistry.
10.3
Conjugate Eliminations
10.3.1
1,6-Conjugate Eliminations
10.3.2
1,4-Conjugate Eliminations
mediate carbinol. All cleavage products are remarkably clean since the mechanism
only allows for the release of products generated through intermediate 85. The
limitation of this novel fragmentation pathway lies in the necessity for epoxides
derived from 2-hydroxybenzophenones, since both aryl groups are required for the
regioselective epoxide opening.
10.4
AdditionElimination Reactions
10.4.1
AdditionElimination on Vinylogous Systems
(93) (loading 0.7 mmol g1 ) eliminates the need for any further purication [45].
Thus, the resin-bound enaminone 94b could be easily separated from the excess
reagents and byproducts of the reaction mixture by a simple ltration. On-resin
cyclization of the enaminone 94b released the benzopyrone 95 in good yields and
good purities and regenerated the piperazinyl resin. This resin capture would fa-
cilitate a one-pot conversion of silylated salicylic acids to benzopyrones without re-
quiring any intermediate purication steps an attractive feature for the synthesis
of large libraries. Furthermore, this method can be applied to the synthesis of
benzopyrones with no residual functionalities required for linkage to the solid
support. The solid-supported capture reagent can be regenerated and recycled for
additional rounds of resin capture.
the reaction conditions, including ZnCl2 , AlCl3 , Et2 AlCl, TiCl 4 , BF3 Et2 O, and
Yb(OTf )3 . However, the water-tolerant Yb(OTf )3 gave the highest yield. The nal
products were obtained after cleavage with triuoroacetic acid (TFA)/CH2 Cl2 (1:1)
in relatively high yields and purities.
10.4.2
Cycloreversions
Pyrroles 99a were isolated in 40% overall yield with high purities, according to
Mjalli and coworkers protocol, after release from solid support. Armstrong and
coworkers four-step protocol, although not optimized, gave pyrroles 99b with 4
17% yields. Although this chemistry is amenable to solution-phase synthesis, low
yields of the tetra-substituted pyrroles are usually observed [50]. This may be at-
tributed to the substitution pattern of the acetylenic dipolarophile and to the ten-
dency of munchnones to self-condense. Furthermore, it would require the use of
the corresponding N-acyl-N-alkyl-a-amino acids as building blocks in order to in-
crease the purity of the nal products.
300 10 Elimination Chemistry in the Solution- and Solid-phase Synthesis of Combinatorial Libraries
The cleavage rate from solid support, at similar temperatures, is highest in polar
protic solvents, which could allow for the cleavage to be carried out directly in
aqueous media. The chemistry could be amenable to solution-phase parallel syn-
thesis since furans were recovered in good yields but would require a special work-
up or purication to remove the rhodium catalyst. The released isocyanate could
be eectively scavenged with a resin-bound amine.
10.4.2.4 1,2-Diazines
Access to a small library of functionalized 1,2-diazines can be provided by the in-
verse electron demand DielsAlder reactions of 3,6-substituted-1,2,4,5-tetrazines
on solid phase (Scheme 10.30) [55a]. Treatment of an immobilized azadiene (111)
with a variety of electron-rich olens in dioxane at room or elevated temperatures
gave bicyclo intermediate 112, which underwent a cycloreversion with loss of N2
and concomitant loss of HX to give resin-bound diazine 113. Removal of the Boc
group rst, followed by cleavage from solid support under basic conditions, gave
the corresponding 3-amino-6-thiomethyl-1,2-pyridazines (114) in good to moderate
yields. The chemistry can be extended with the azadienes bearing a sulfone group
in the 6-position. The sulfone substrate is more reactive than the corresponding
thiomethyl substrate in a DielsAlder reaction as more ecient conversion is
generally achieved with the less reactive alkynes. A wide range of electron-rich di-
enophiles can be used which permits the introduction of two diversity elements on
10.5
Summary
References
11
Addition to CC Multiple Bonds
(Except for CC Bond Formation)
Adrian L. Smith
11.1
Introduction
The last decade has witnessed the birth and maturation of combinatorial chemistry
as a technique for synthesizing large numbers of compounds. Much of the early
emphasis was on synthesizing large (mixtures) libraries for screening and lead
generation, and this was almost exclusively based upon solid-phase chemistry.
However, the technologies developed during this work were also applicable to the
more traditional medicinal chemistry lead optimization process, and this has led to
a rethink in the ways in which the medicinal chemist can most eciently optimize
a lead series, both in vitro and in vivo. Whilst this initially resulted in many people
developing solid-phase synthetic routes for the parallel synthesis of arrays of single
compounds (whether it be to make 10 or 10,000 compounds), the co-development
of postsynthesis sample-handling techniques (particularly in the area of purica-
tion) has signicantly broadened the scope of chemistries which can be used. In
particular, the initial advantages of solid-phase chemistry (primarily related to pu-
rication) are now less compelling in many cases where solution-phase chemistry
will require less chemistry development time and where crude synthetic products
can be puried in an automated fashion.
Today, the term combinatorial chemistry is loosely used to describe a very
broad range of techniques including solid- and solution-phase chemistry and the
synthesis of discrete (single) compound arrays and mixtures. Whilst the dierent
techniques each have their pros and cons, their unifying factor when properly
applied is the ability to address a particular problem (e.g. a medicinal chemistry
problem) by making and testing a larger and more diverse set of compounds than
would have been the case by more traditional methods. Therefore, although a large
number of publications have appeared in the area of solid-phase chemistry, the
purpose of this chapter is not to act as a comprehensive review of solid-phase
chemistry per se, but rather to highlight key areas in which chemists may prac-
tically utilize these techniques in chemical transformations involving overall
addition to carboncarbon multiple bonds. It should be noted that certain trans-
formations such as carboncarbon bond formation, heterocycle formation, cyclo-
11.2
Addition to C=C Double Bonds
Additions to CbC double bonds fall into two broad main categories dependent
upon the electronic nature of the double bond. The main focus of this section will
deal with electrophilic addition to isolated double bonds, whilst nucleophilic 1,4-
addition to conjugated a; b-unsaturated systems will be briey touched upon at the
end.
The reactivity of the isolated CbC double bond arises from the nucleophilic na-
ture of the p-bond, with the majority of reactions involving some form of electro-
philic addition. This may result in formal oxidation (e.g. epoxidation) or reduction
(e.g. hydrogenation). It may be expected that more electron-rich double bonds will
generally exhibit greater reactivity than electron-decient double bonds in the ab-
sence of overriding steric factors. Whilst the CbC double bond is also frequently
encountered in radical reactions, this is usually implicated in CaC bond formation
and is outside the scope of this chapter. This section aims to summarize some of
the key transformations which can be eected by addition to the CbC double bond,
emphasizing applicability to combinatorial chemistry.
11.2.1
Epoxidation and Subsequent Epoxide Opening
conversion with a minimal excess of reagent. As such, it is suitable for use in both
solid- and solution-phase chemistry. It should be noted that there is often little
diastereoselectivity with this reagent. Scheme 11.2 shows an example of an epox-
idation carried out on solid phase [2]. Here, the author experienced instability of
the urethane linker to the generated 3-chlorobenzoic acid byproduct and buered
the reaction with NaHCO3 . Another side-reaction sometimes observed is open-
ing of the epoxide by the generated 3-chlorobenzoic acid, and in these cases more
success may be possible with reagents such as dimethyldioxirane [3]. It is also
possible to use a range of solid-supported reagents for epoxidation (see Table 11.1)
[4].
The two-step processes shown in Scheme 11.1 are very well suited to solution-
phase parallel synthesis. Generally, the intermediate (2) is prepared in bulk, puri-
ed, and then split into individual reactors for parallel ring opening by a range of
nucleophiles. For reactions with amines to give 3, we nd a small excess of the
amine (typically 1.5 equiv.) in isopropyl alcohol heated at 65 C for 16 h usually
gives clean conversion, even with relatively non-nucleophilic amines such as ani-
line. It is important that no epoxide remains in nal products, since this can give
misleading data in biological assays by acting as a suicide inhibitor. It is there-
fore generally preferred to use the epoxide as the limiting reagent. For reactions
with secondary amines to give tertiary amine products, it may be possible to scav-
enge excess amine with an appropriate resin [9]. Alternatively, we nd automated
preparative high-performance liquid chromatography (HPLC) to be a convenient
method for purication.
Examples of larger solution-phase combinatorial mixtures libraries are known
utilizing a similar approach. A library of over 6000 b-amino alcohols was prepared
using LiClO4 in acetonitrile and 1.2 equiv. of amine [10]. In such cases, care is
needed to validate the reactivity of reagents properly in order to avoid misleading
bioassay data since purication is more dicult. A solid-phase mixtures library of
308 11 Addition to CC Multiple Bonds (Except for CC Bond Formation)
Oxone8/NaHCO3/THF/H2 O
60 h 7090% 2 equiv. of dioxirane 6
resin used. Resin is
CHCl3 prepared by treatment
of polystyrene 2-
oxoalkyl resin with
Oxone8 and NaHCO3
in THF/H2 O and
isolated prior to use.
Shelf stability of
dioxirane resin not
reported
4h 80% AG 50W-X8 ion 7
exchange resin (Bio-
THF/D
Rad), converted to
persulfonic acid with
potassium persulfate
4h 5095% Resin prepared by 8
treatment of 1%
THF/40 C crosslinked
carboxypolystyrene
with 85% H2 O2 in
MeSO3 H for 16 h.
Peracid resin stable to
storage at 20 C.
THF is vastly superior
to CH2 Cl2 as solvent
Scheme 11.3. Epoxidation, epoxide opening, and lactonization during cleavage from resin.
11.2.2
Dihydroxylation
The dihydroxylation of an olen is most eciently carried out using osmium tetr-
oxide, giving rise to vicinal diols (Scheme 11.5) [14]. The catalytic variant of the
reaction is usually employed owing to the toxic and volatile nature of the reagent,
310 11 Addition to CC Multiple Bonds (Except for CC Bond Formation)
and the facile regeneration of the catalyst with oxidants such as N-methylmorpho-
line N-oxide (NMO), potassium ferricyanide, hydrogen peroxide, or tert-butyl hy-
droperoxide. Osmium(III) chloride is sometimes used as the osmium source in
the catalytic reaction, being less volatile than the tetroxide and therefore easier to
weigh out safely. It is converted in situ to the tetroxide by the oxidant. High levels
of enantioselectivity can be achieved in the reaction using catalytic amounts of os-
mium tetroxide in the presence of cinchona alkaloid derivatives [15].
Osmium tetroxide is a very mild reagent and compatible with many functional
groups, making it suitable for combinatorial synthesis, at least in principle. There
have, however, been few reported uses in library synthesis. For solution-phase li-
brary synthesis, perhaps the most practical methods will involve the use of solid-
supported osmium tetroxide [4] since this will minimize handling problems for
the toxic osmium reagent, allowing recovery by simple ltration. One such exam-
ple is osmium tetroxide microencapsulated in polystyrene [16]. This work high-
lighted one potential practical problem associated with such reagents. The dihy-
droxylation is usually carried out in acetonewater or tert-butanolwater mixed
solvent systems. Not only does this raise the possibility of solubility issues for re-
actants, but also solvent/reactant access within the solid-supported reagent may be
restricted since polymers such as polystyrene are poorly solvated and not swelled
well by these solvents. In this work, acetonitrile was added as a cosolvent in order
to achieve good conversion. Good results have been obtained in the area of sup-
ported asymmetric dihydroxylations. The OsO4 copolymer of 20 [an analog of
(DHQ)2 -PHAL] behaves well under standard conditions [t-BuOH/H2 O (1:1) at 10
C using K3 Fe(CN)6 -K2 CO3 (3 equiv.) as secondary oxidant], giving high yields and
enantioselectivities (Scheme 11.6) [17]. The copolymer is easily recovered by ltra-
tion.
11.2.3
Oxidative Cleavage
11.2.4
Electrophilic Addition of A--X
Perhaps most use has been made of organoselenium reagents in this area of re-
search, providing mild and extremely versatile reactivity [28, 29]. A resin-bound
314 11 Addition to CC Multiple Bonds (Except for CC Bond Formation)
version of phenylselenyl bromide (43) has been developed which provides a conve-
nient and odorless way of handling this toxic and smelly reagent (Scheme 11.15)
[30]. This reagent can also be converted to polymer-supported phenylselenyl phtha-
limide (44), which provides a convenient method for hydration of olens (Scheme
11.16) [30]. A polymer-supported version of phenylselenyl cyanide is readily pre-
pared from Merrield resin and potassium selenocyanate [31].
11.2.5
Hydrogenation
for reduction of alkenes (and alkynes) on solid phase, and excellent conversions
have been observed using benzene sulfonylhydrazide in DMF at 100 C as diimide
source [40].
11.2.6
Hydrometallation
mary alcohol 63. IBX oxidation and a subsequent BaylissHillman reaction using
PhSeLi as nucleophile provided the key intermediate 64 which, upon oxidation to
the corresponding ketone, underwent an intramolecular DielsAlder reaction.
Cleavage from the resin then gave the polycyclic compound 65 in 35% overall
yield.
11.2.7
1,4-Addition to a; b-Unsaturated Carbonyl Systems
The reactions described above all make use of the electron-rich nature of isolated
carboncarbon double bonds and their consequent reactivity toward electrophiles.
By contrast, when the carboncarbon double bond is conjugated to a carbonyl
or sulfonyl group, the electronic character of the double bond changes, making
it susceptible to 1,4-nucleophilic addition (also known as Michael addition in
the case of carbon nucleophiles). Pioneering work in this area of combinatorial
chemistry is shown in Scheme 11.23, in which secondary amines 68 add to resin-
bound acrylate (67 ! 69), are quaternized with alkyl halides, and then undergo
a Hofmann-type b-elimination in the presence of triethylamine to give tertiary
amines 72 [43]. When primary amines 68 are used in this sequence (R2 H), a
reductive amination step has been used prior to quaternization to prepare terti-
318 11 Addition to CC Multiple Bonds (Except for CC Bond Formation)
ary amine libraries (72) with three points of diversity [43, 44]. Subsequent work
has shown that replacement of triethylamine with Amberlite weakly basic ion-
exchange resin in DMF simplies work-up procedures since it overcomes the ne-
cessity to remove triethylamine salts from cleaved products, and this procedure
has been used to generate large libraries of trisubstituted amines (> 10,000 com-
pounds) [44]. A similar approach has been employed utilizing a vinylsulfone resin
in place of the acrylate resin 67 [45].
11.3
Addition to C C Triple Bonds
The acetylenic group displays a versatile range of reactivities for use in com-
binatorial chemistry. Its use in cycloaddition reactions, palladium-mediated reac-
tions, and radical reactions is well documented and covered here in the relevant
chapters. Prominent examples on solid phase include their use in the palladium-
mediated syntheses of heterocyclic templates such as indoles [4649] and benzo-
furans [50].
The discussion above (Sect. 11.2.5) on hydrogenation of olens applies equally
to the hydrogenation of acetylenes the diimide reduction of an acetylene on solid
phase is one of the few reported examples [40]. Hydrometallation is another area
References 319
References
12
Addition to CarbonHetero Multiple Bonds
Philipp Grosche, Jorg Rademann, and Gunther Jung
12.1
Introduction
12.2
Additions to CN Double Bonds in sp2 Systems
CN double bonds are the structural motif of various functional groups. This sec-
tion refers to CN double bonds in sp2 systems such as imines, iminium ions, acyl-
iminium ions, and N-oxides of nitrogen heterocycles (pyridine, quinoline, iso-
quinoline). The reacting CN double bond can be isolated; however, the reactions
are often performed as a three-component reaction consisting of an amine, a car-
bonyl compound and the attacking nucleophile, with in situ formation of the reac-
tive CN double bond.
12.2.1
Attack by Hydride (Reductive Alkylation)
Scheme 12.1
324 12 Addition to Carbon--Hetero Multiple Bonds
Scheme 12.2
12.2.2
Addition of Carbon Nucleophiles
component where the imine is formed in situ. In this case, a dehydration agent
such as trimethyl orthoformate is normally added to support imine formation and
to prevent degradation of the ketene silyl acetals by H2 O. This Mannich-type reac-
tion has attracted a lot of interest in solution-phase as well as in solid-phase
chemistry.
A typical procedure in solution applies equimolar (or nearly equimolar) amounts
of imine (or amine and aldehyde) and silyl component and 0.050.3 mol%
Ln(OTf )3 . The reaction is performed in dichloromethane (DCM), acetonitrile
(ACN), or mixtures of both solvents at rt. The suitability of this reaction for library
synthesis was exemplied by the generation of a small library of 40 2,3-dihydro-4-
pyridones using Danishefskys diene as the silyl component and Yb(OTf )3 as the
Lewis acid catalyst [18]. Byproducts and unreacted aldehyde were removed with
polyamine resin and the products were obtained in good to excellent purities and
yields. Although this reaction was often described as an aza-DielsAlder reaction
in the literature, other authors favored a tandem MannichMichael mechanism [19].
Another solution-phase approach utilizes a polymer-bound scandium catalyst
(polyallyl-scandium-triylamide ditriate) (Scheme 12.3). The reaction is conducted
as a three-component reaction, with silyl enolates, ketene silyl acetals, and cya-
notrimethyl silane as nucleophiles [20]. Treatment with cyanotrimethyl silane
as the nucleophile aords a-aminonitriles in a Lewis acid-catalyzed variation of
the classical Strecker synthesis. This reaction is performed in dichloromethane/
acetonitrile (2:1) at rt for 19 h. When ketene silyl acetal is used, MgSO4 is added
to prevent decomposition of the ketene silyl acetal by traces of water. A cation-
exchange resin has also been used for the promotion of the imino-aldol reaction
[21].
In the imino-aldol reaction at least one asymmetric C atom is generated. Thus, a
procedure for enantioselective addition is of special interest. Catalysis of the reac-
tion with polymer-supported palladium BINAP m-hydroxo complex (BINAP 2,2 0 -
bis(diphenylphorphino)-1,10 -binaphthyl) yields an ee of 81% [22].
Scheme 12.3
This Mannich-type reaction is very suitable for combinatorial chemistry and has
also been adapted to solid-phase chemistry, thus pursuing both possible strategies:
the immobilization of the imine and the immobilization of the silyl component.
Polymer-supported silyl thioketene acetals obtained by silylation of immobilized
thioesters were converted to b-aminoesters by treatment with imines in the pres-
ence of Sc(OTf )3 using dichloromethane as solvent [23, 24]. Alternatively, a three-
component protocol can be employed (Scheme 12.4) [25]. In both cases reductive
cleavage of b-amino acid thioesters with LiBH4 yields b-amino alcohols.
326 12 Addition to Carbon--Hetero Multiple Bonds
Scheme 12.4
In the second approach, a polymer-supported imine was treated with silyl enol
ethers or silyl ketene acetals in the presence of Yb(OTf )3 in dichloromethane to
yield b-amino ketones and esters respectively [26, 27]. Instead of imines, polymer-
supported acyl hydrazones can also be used as substrates [28].
2,3-Dihydro-4-pyridones were synthesized by tandem MannichMichael reaction
in good yields. Resin-supported imines were reacted with Danishefskys diene in
THF using Yb(OTf )3 (Scheme 12.5) [29] or ZnCl2 [30] as catalyst.
Scheme 12.5
Scheme 12.6
be linked to the solid support. Typically, the components in solution are employed
in large excess. Polymer-bound N-substituted amino acids have been treated with
glyoxylic acid and boronic acids in dichloromethane at rt for 18 h, and the reaction
was repeated for additional 60 h. The carboxylic acids obtained were further modi-
ed by coupling with amines. After cleavage from the resin, products were ob-
tained in good purity and yield [33, 34]. This reaction was also carried out at 50 C
for 12 days in DMF/1,2-dichloroethane (DCE), and these reaction conditions have
also been successfully applied to immobilized phenyl boronic acid, glyoxylic acid
(Scheme 12.7), and proline or piperazine [35].
Scheme 12.7
Scheme 12.8
12.2.2.4 Reaction with Grignard Reagents, Lithium Organyles, and Zinc Organyles
Imines and iminium species are converted to amines upon treatment with
Grignard reagents, lithium reagents, and zinc organyles. This reaction is often
used in solution-phase chemistry and has been adapted to solid-phase chemistry.
Resin-immobilized aldimines, derived from condensation of Rink amide poly-
styrene resin with aldehydes, reacted with Grignard reagents at 60 C or lithium
reagents at 78 C to 20 C to yield after cleavage primary amines in good to ex-
cellent purity (Scheme 12.9) [15]. Since the aldimines of ammonia are unstable,
similar approaches to primary amines in solution phase need protecting groups
at the N atom to improve the stability. In this example, the function is fullled by
the resin. If the imines are not immobilized via the N atom, secondary amines are
obtained after cleavage [39, 40]. Enantiopure immobilized aldimines have been
converted diastereoselectively to homoallylamines upon treatment with allyl zinc
reagents in THF. The latter were obtained from allylbromide and zinc using
CeCl3 7H2 O as an additive [41].
Scheme 12.9
Scheme 12.10
12.2 Additions to CN Double Bonds in sp2 Systems 329
Scheme 12.11
Scheme 12.12
Scheme 12.13
Scheme 12.14
Scheme 12.15
12.2.3
Addition of Nitrogen Nucleophiles
Scheme 12.16
12.2.4
Addition of Phosphorus Nucleophiles
Scheme 12.17
Scheme 12.18
12.2.5
Reactions with Oxygen Nucleophiles
There are only few addition reactions of O-nucleophiles to CN double bonds re-
ported in solid-phase chemistry. 1,2-Aminoalcohols were cyclized with aldehydes
to form oxazolidines via an imine intermediate. This reaction has been exploited
for the immobilization of aldehydes. Polymer-bound serine or threonine was
treated with an aldehyde in 1% N,N-diisopropylethylamine (DIPEA)/MeOH for 2 h
at 60 C. This oxazolidine linker is stable to the conditions of Fmoc peptide syn-
thesis and is cleaved by heating with 5% AcOH/water for 30 min at 60 C [69].
A similar approach was chosen for the synthesis of disubstituted 1,3-
oxazolidines [70]. 1,2-Aminoalcohols linked to the polymeric support were con-
densed with aldehydes in TMOF at rt to yield the corresponding imines that were
converted to N-acyl-1,3-oxazolidines by treatment with acyl chlorides, isocyanates,
or isothiocyanates (Scheme 12.19). Cleavage of the Wang-PS resin was performed
using DDQ and aorded 1,3-substituted 4-(40 -formylphenoxymethyl)oxazolidines
in good yields. In this example, the cleavage using 1% TFA led to the decomposi-
tion of the heterocycles. Thus, an oxidative method was required for smooth isola-
tion of the products.
Scheme 12.19
12.2.6
Addition of Sulfur Nucleophiles
Thiols can readily add to imines, which is exploited in combinatorial chemistry for
the formation of several heterocycles, namely thiazolidinones, metathiazanones,
334 12 Addition to Carbon--Hetero Multiple Bonds
thiazolines, and benzothiazoles. Imines react with mercapto acids under addition
of the thiol followed by acylation of the resulting N,S-acetal. a-Mercaptocarboxylic
acids are used for the synthesis of thiazolidinones; b-mercaptocarboxylic acids for
the synthesis of metathiazanones. The latter reaction is most conveniently per-
formed as a three-component condensation of a primary amine, an aldehyde, and
a mercapto carboxylic acid, but a stepwise approach is also possible.
Solid-phase procedures of this reaction [71, 72] start from immobilized amines
that are treated with excess aldehyde and mercapto carboxylic acids in THF at 70 C
for 2 h (Scheme 12.20). Molecular sieves or TMOF are used to remove H2 O that
formed during the condensation as well as during the cyclization step [72].
Whereas thiazolidinones are obtained in high purities, the formation of meta-
thiazanones is more problematic. Besides immobilized amines, immobilized
glyoxylic acid has also been used as a substrate [73].
Scheme 12.20
12.3
Additions to CN Double Bonds in sp-Systems
12.3.1
Additions to Carbodiimides
Scheme 12.21
12.3.2
Reaction of Isocyanates and Isothiocyanates
The chemistry of these two groups is very similar. There is a vast number of ar-
ticles describing the addition of nucleophiles to isocyanates and isothiocyanates.
336 12 Addition to Carbon--Hetero Multiple Bonds
Scheme 12.22
Scheme 12.23
12.3 Additions to CN Double Bonds in sp-Systems 337
Scheme 12.25
338 12 Addition to Carbon--Hetero Multiple Bonds
12.3.3
Addition to CS Double Bonds in sp2 Systems
Thioamides and thioureas are the most frequently used synthons bearing CS
double bonds. Owing to its high nucleophilicity, the sulfur is readily alkylated. The
alkylated species is often used as an intermediate to facilitate subsequent substitu-
tion with amines to aord amidines and guanidines. Furthermore, the alkylated
species can be exploited for the synthesis of heterocycles such as thiazoles from
thioamides and a-halo ketones. These reactions are well established in solution
and have also been employed successfully on solid phase.
The alkylation of benzimidazol-2-thiones with various alkyl halides was per-
formed using MeO-PEG-OH as soluble polymeric support. The reaction proceeds
in dichloromethane using Et3 N as a base and yields the 2-alkylthiobenzimidazoles
in good to high purity [91].
S-Alkylation has been conducted on solid phase using, for example, immobilized
benzimidazolthiones and quinazoline-2-thioxo-4-ones as substrates. The reaction is
typically performed in DMF or N-methyl-2-pyrrolidone using a tertiary base and
various alkyl or aryl halides as alkylating agents [92, 93].
As already mentioned, S-alkylations can be a key step in heterocycle syntheses.
Thiazoles are formed by the Hantzsch synthesis by reaction of resin-bound thio-
amides with a-halo ketones [94]. 2,4-Diaminothiophenes are obtained by alkylation
of immobilized a-cyano thioamides with a-halo ketones followed by cylization (see
Sect. 12.4.1).
Formation of guanidines from thioureas via S-alkylation and substitution with
amines has been performed using methyl iodide, Mukaiyamas reagent, or carbo-
diimides as alkylating agents. Merrield resin can also act as an alkylation reagent
and yields immobilized S-alkyl isothioureas. These are typically cleaved as guani-
dines when treated with amines [95, 96]. For carbodiimides, the alkylated species
is not isolated but is directly substituted with amines [97, 98], whereas the alky-
lation products of methyl iodide [99, 100] or Mukaiyamas reagent are isolated
(Scheme 12.26) [101].
Scheme 12.26
12.3.4
Reaction of CS Double Bonds in sp Systems
Scheme 12.27
12.4
Additions to CN Triple Bonds (Cyanides, not Isocyanides)
12.4.1
Addition of Carbon Nucleophiles
Nitriles can be attacked by CH acidic compounds. This feature was exploited in the
synthesis of aminothienopyridine libraries in a solution-phase approach [105, 106].
Treatment of 3-cyanopyridine-2-thiones with acceptor-substituted bromomethylene
compounds and aqueous KOH in DMF at rt resulted in S-alkylation. Cyclization
was carried out again with aqueous base to yield aminothienopyridines (Scheme
12.28) [106].
340 12 Addition to Carbon--Hetero Multiple Bonds
Scheme 12.28
Scheme 12.29
12.4.2
Addition of Nitrogen Nucleophiles
Scheme 12.30
12.4 Additions to CN Triple Bonds (Cyanides, not Isocyanides) 341
was attached to the support. a-Formyl nitriles are accessible by reaction of immo-
bilized benzylnitrile with Brederecks reagent [bis-(dimethylamino)-t-butyloxy-
methane]. Cyclization occurs upon treatment with hydrazines and heating at 70 C
in 10% AcOH/EtOH for 5 h. The 5-aminopyrazoles were obtained in high purity.
In a similar approach, a; b-unsaturated nitriles were cyclized with hydrazines
using NaOEt as a base to yield 3-amino-2-pyrazolines in high purity. Heating at 70
C in ethanol for 24 h was necessary for complete conversion [111].
Kaisers oxime resin was used for the introduction of a hydroxylamine group
to cyanouorobenzenes. Cleavage of the hydroxylamine and cyclization was per-
formed in a one-step procedure using TFA/5 N HCl to yield 3-aminobenzis-
oxazoles in high purity (Scheme 12.31) [112, 113]. An example with the in situ
generation of the hydroxylamine nucleophile is the reduction of a polymer-bound
2-nitro-benzylnitrile with tin(II) chloride dihydrate. Hydroxylamine, an intermedi-
ate of the reduction, is trapped by cyclization with the nitrile to aord 2-amino-1-
hydroxyindoles [114].
Scheme 12.31
Guanidines can also act as the attacking N-nucleophile. This feature has been
exploited for the solid-phase synthesis of 2,4-diaminoquinazolines. The guanidines
were synthesized using acylisothiocyanate resin and 2-aminobenzonitriles as
building blocks. Cleavage and cyclization occurs during heating with TFA/H2 O
(95:5) at 80 C for 16 h. The cleavage step is repeated once. Following this proce-
dure, diaminoquinazolines were obtained in high purity and good yield [115].
There are only a few examples of nucleophilic conversions of nitriles that occur
without cyclization. One example is the aminolysis of polymer-bound nitriles with
hydroxylamine hydrochloride and DIPEA in 2-methoxyethanol at 85 C for 16 h to
provide amide oximes with a quantitative yield (Scheme 12.32). The products ob-
tained were used in the synthesis of oxadiazoles [116]. Another example is the re-
action of a-(benzotriazol-1-yl)acetonitril with amines in 2-methoxyethanol at 7580
C. The formed amidines were not isolated but were directly treated with resin-
bound chalcones to yield aminopyridines that had high purity after cleavage.
Scheme 12.32
342 12 Addition to Carbon--Hetero Multiple Bonds
12.4.3
Addition of Sulfur Nucleophiles
Scheme 12.33
References
13
Chemistry of the Carbonyl Group
Tobias Wunberg
13.1
Introduction
The carbonyl group is one of the most important functional groups in organic
chemistry. It is found in various structural classes, each one having its own char-
acteristic and fascinating chemistry. The wealth of chemical transformations
involving carbonyl groups covers virtually all forms of organic reactions, thus
making the carbonyl group an unreplacable tool in synthetic organic chemistry.
Additionally, this structural unit is extremely important in medicinal chemistry.
Seventy per cent of known drugs (CMC database vers. 94.1) contain carbonyl
groups in various manifestations [1]. This chapter deals with the application of
carbonyl group chemistry to the generation of libraries. It will not include CaC
single bond-forming reactions, enolate chemistry, or reductions and oxidations in-
volving carbonyl groups. These reactions are dealt with in other chapters.
13.2
Chemistry of the Carbonyl Group and Combinatorial Chemistry
13.3
Chemistry of Carboxylic Acids
13.3.1
C(O)--X Bond-forming Reactions: General Remarks
Formation of amides
There is a mechanistic requirement for the transformation of carboxylic acids into
amides: the introduction of a suitable leaving group instead of the OH group prior
to the reaction with a nucleophile (amine). Two principal ways of activation are
feasible: replacement of the OH group with better leaving groups (e.g. acid hal-
ides) or transformation of the OH moiety into a suitable leaving group (e.g. active
esters, anhydrides). While planning the synthesis of a library one should keep in
mind that the choice between solution and solid phase may cause limitations
which have their origin in the combination of the specic demands of resin-based
chemistry, the ecacy of coupling reagents, and the formation of acid chlorides.
A plethora of methods and conditions has been described for the formation of
acid chlorides in solution phase and most of them are applicable to parallel syn-
thesis and library production. Consequently, the reaction of acid chlorides with
aliphatic amines is a standard reaction and has been used in numerous pre-
parations of libraries in solution. In our hands, CH2 Cl2 or tetrahydrofuran (THF)
as solvent and NEt3 or DIEA as base are the standard methods. Even aromatic
amines can be smoothly converted into amides under these conditions. Further-
more, this protocol is also suitable for the formation of amides on solid phase with
348 13 Chemistry of the Carbonyl Group
In addition to the use of acid chlorides, chemists have been successfully trans-
forming carboxylic acids into amides using coupling reagents. A large variety of
reagents has been developed which provide chemists with a tool-box for the syn-
thesis of large and complex structures [10]. Numerous examples demonstrate the
ecacy of these methods for library synthesis in solution phase as well as in solid
phase. In the latter case, both starting materials can be bound to the resin: activa-
tion of a polymer-bound acid followed by addition of the amine, or the addition of
an excess acid plus coupling reagent to a polymer-bound amine.
The reaction of active esters with aliphatic amines generally leads to the forma-
tion of the desired amide in high yield and purity. However, the formation of ani-
lides using active esters as acylating agents poses diculties and only the most re-
active coupling reagents may give satisfactory results (see Table 13.1). Therefore,
the use of acid chlorides is necessary when anilines are used as nucleophiles.
The best conditions for a particular acylation are highly dependent on the steric
and electronic nature of both the amine and carboxylic acid. There is no general
rule as to which reagent is the best for a particular reaction and each case has to be
optimized with respect to coupling reagent, base, solvent, etc. Nevertheless, there
are some general guidelines for the choice of an appropriate coupling reagent:
Acids Amines
In the early days of combinatorial chemistry, these methods were used for the
construction of peptide libraries on solid phase leading to potent, bioactive lead
structures or elucidation of binding motifs [22]. Today, the application of coupling
reagents for amide synthesis remains an essential part of combinatorial chemistry.
Amino acids have gained popularity as readily available, multifunctional templates
with a high degree of diversity and biological importance, and they are often used
as valuable building blocks for the library synthesis of small molecules [23].
For parallel amide synthesis in solution phase, polymer-bound variations of the
well-known coupling reagents have been shown to be powerful tools for avoiding
350 13 Chemistry of the Carbonyl Group
This methodology is routinely used for the synthesis of large libraries (10,000
members) for lead discovery or for targeted libraries, e.g. optimization of factor Xa
inhibitors [30].
Formation of ureas using isocyanates. The solvent of choice for the reaction of
isocyanates with amines is CH2 Cl2 . The reaction usually is completed within 12
h (Scheme 13.5). Prolonged reaction times have been reported to lead to lower
product yields [34].
Tab. 13.2. Comparative scavenging of nucleophiles in DCM (20 C) or DCE (60 C).a
Scheme 13.9. Activation of secondary amines with COCl2 for solid-phase synthesis of ureas.
General remarks
For both solution- and solid-phase chemistry, esters and urethanes commonly
serve as protecting groups for carboxylic acids and amines, respectively [41]. Addi-
tionally, they are one of the most common ways to attach a molecule to a poly-
meric support. In the context of library synthesis, esters and urethanes may of
course occur as part of the targeted molecule [42].
Formation of esters
A wealth of methods exist for the esterication of carboxylic acids. Almost all of
them are suitable for parallel synthesis. Among these methods, DIC/DMAP (Steg-
lich esterication) [43], 2,4,6-trichlorobenzoyl chloride (Yamaguchi esterication)
[44], or 2,6-dichlorobenzoyl chloride [45] and the reaction of alcohols with acid
chlorides or anhydrides are the most prominent ones in solution phase. In partic-
ular, the Steglich procedure seems to be suitable for automation and paralleliza-
tion since polymeric carbodiimides are readily available (Scheme 13.11) [46].
Conversion of polymer-bound alcohols using these methods is as ecient as
in solution phase. Owing to the lack of methods for acid chloride formation on
solid phase (see Section 13.3.1), esterication of immobilized carboxylic acids is
354 13 Chemistry of the Carbonyl Group
mostly realized using either the Steglich or the Yamaguchi procedure. Neverthe-
less, the condensation with primary or secondary alcohols using DIC/DMAP or
2,4,6-trichlorobenzoyl chloride has been well described. Additionally, the combina-
tion MSNT/NMI has been reported to be superior for the coupling of immobilized
aromatic carboxylic acids to primary and secondary alcohols compared with the
Steglich or Yamaguchi esterication (Scheme 13.12) [47].
Hydrolysis of esters
In solution, esters are normally hydrolyzed using hydroxides such as LiOH,
NaOH, or KOH in aqueous (50%) MeOH or THF [41]. These procedures are suit-
able for parallelization. Adaptation of these methods to solid phase sometimes re-
quires slight modication of experimental details. Because of the extremely poor
swelling properties of PS-based resins in water and alcohols, hydrolysis requires
addition of solvents with better swelling properties usually THF. NaOH or LiOH
in THF/H2 O (4:1 v/v) or THF/MeOH/H2 O (3:1:1 v/v) at room temperature (rt) or
at 50 C generally give good conversions (Scheme 13.13) [48].
Formation of carbamates
There are two standard ways for the formation of carbamates: rst, from amines
and chloroformates (Scheme 13.13); second, from alcohols and isocyanates or car-
bonyl insertion compounds. In contrast to the reaction of amines with isocyanates,
13.3 Chemistry of Carboxylic Acids 355
13.3.2
Transformation of Carboxylic Acids into Other Functional Groups
Despite its synthetic potential, the Tebbe olenation has only rarely been used
for parallel synthesis. The starting point for a library of thiazoles on solid phase
was the conversion of polymer-bound esters into the corresponding enol ethers
using Tebbe reagents in toluene/THF at room temperature. For the following on-
resin functionalization, the authors elegantly combined the principles of solid- and
solution-phase chemistry (Scheme 13.18): in a rst step, the enol ether is bromi-
nated on the resin, yielding an a-halo ketone equivalent. Formation of thiazoles via
Hantzsch synthesis then simultaneously led to cleavage of the heterocycles from
the resin. Removal of excess thiourea was accomplished by applying a polymeric a-
halo ketone as a scavenger resin.
13.4
Reactions of Aldehydes and Ketones
13.4.1
Reactions of Carbonyl Groups with CaH Acidic Compounds
General aspects
The Wittig and the HornerEmmons olenations are well-established synthetic
transformations for the generation of double bonds. Parallelization of these ex-
tensively used reactions should be straightforward. However, in the literature there
are hardly any descriptions of libraries synthesized in solution. Probably, reactions
358 13 Chemistry of the Carbonyl Group
have been carried out in a parallel fashion but have not specically been reported
as libraries. Some reports deal with ways to simplify the work-up, e.g. by using
polymeric triphenylphosphine [55]. The situation is dierent for solid-phase syn-
thesis. The reaction of aldehydes or ketones with phosphonium ylides yielding
CaC double bonds was among the rst reactions to be used for the synthesis of
nonpeptidic molecules on solid phase [56]. Numerous examples prove the value of
Wittig and the HornerEmmons olenations for solid-phase synthesis.
Double bonds are versatile starting points for diversication and are, thus, inter-
esting building blocks for combinatorial chemistry. The double bonds obtained by
the Wittig reaction have been used for various purposes, e.g. in olen metathesis
reactions [57] or cycloadditions [58] (Scheme 13.20).
For solid-phase synthesis, two strategies have been pursued: generation of the
ylide in solution and adding it to a polymer-bound aldehyde, as well as the oppo-
site direction generation of the ylide on solid phase and subsequent addition of
the aldehyde to the resin. It is important to note that the choice between these two
possibilities has consequences for the reaction conditions. If the ylide is generated
on solid phase, KOtBu/THF has been reported to be better than NaHMDS/THF
for phosphonium salt-derived ylides [58a]. Additionally, the reaction time for the
generation of immobilized ylides is shorter (560 min) than in solution phase (up
to 3 h). Hydrolysis of these sensitive intermediates can be a signicant problem
unless the resin is rigorously dried prior to use, e.g. azeotroping it with benzene.
Owing to the highly basic conditions of the reaction, the choice of the linker is
also crucial. Even though the Wang linker has been employed previously [see 65],
most examples use more base-stable linkers, e.g. amide linkage [59], trityl-ethers
[56], and carbamates [60]. To avoid ester hydrolysis of substrates bound to Wang
resin, the following procedure has been recommended (Scheme 13.21): generation
of the phosphonate anion, removal of excess base under inert conditions, and ad-
dition of the aldehyde dissolved in 60% cyclohexane in THF [61].
The reasoning behind the cyclohexane/THF mixture was that the less-polar
solvent mixture should suppress ester hydrolysis, while allowing the Horner
Emmons condensation to proceed. The authors successfully synthesized acrylic
acid derivatives from aldehydes containing aliphatic, aromatic, and basic function-
alities. The reaction generally was completed within 2 days.
The undesired cleavage from the resin during ylide generation can also be
avoided by attaching the ylide to the resin via the phosphonate moiety. In this case,
the HornerEmmons reaction has been used for cleaving the molecule from the
resin (Scheme 13.22) [62]. This approach was used for the library synthesis of cy-
clic ketones such as (dl)-muscone: a polymer-bound methylphosphonate was de-
protonated with n-BuLi and quenched with an alkenyl ester. The double bond was
used for subsequent introduction of an aldehyde moiety. The following intra-
molecular olenation (cyclorelease) led to the cyclic enones which were further
derivatized in solution phase.
Recent examples
The use of commercially available polymer-bound PPh3 for Wittig reactions com-
bines ideas from solid-phase synthesis and the application of polymer-supported
reagents (Scheme 13.23) [63]. On the one hand (as a polymer-supported reagent) it
oers the advantage that the byproduct, triphenylphosphine oxide, remains at-
tached to the resin. On the other hand (similar to the function of a linker in solid-
phase synthesis) it immobilizes benzyl halides.
Starting from 2-nitrobenzylbromide, after immobilization and chemical mod-
ications, the products may be cleaved o from the support by an intra- or inter-
molecular Wittig reaction. The intramolecular Wittig reaction yields indoles,
whereas the intermolecular pathway gives access to stilbenes [64].
Double bonds have also been used for the synthesis of peptidomimetics. In this
context, they serve as nonhydrolyzable rigid mimetics of the three-dimensional
structure of the amide bond. The HornerEmmons reaction has been employed as
part of a two-step procedure for solid-phase synthesis of b; g-unsaturated d-amino
acids that mimic dipeptides (Scheme 13.24) [65]. This concept combines both the
use of the double bond as an amide bond mimetic as well as the use of the double
bond for diversication.
13.4.2
Reductive Amination
Scheme 13.25. Imine formation on solid support using TMOF as dehydrating reagent.
Imines formed in the presence of TMOF may be isolated and hence can be used
for various purposes besides reduction to amines, e.g. for cycloadditions [71].
In cases where the solubility of the amine component in neat TMOF is low,
cosolvents such as THF or dimethylformamide (DMF) can be used. As a rule of
thumb, the reaction is slower with polymer-bound aldehydes and the amine in so-
lution than with the opposite situation.
A given combination of building blocks has consequences for the reaction con-
ditions depending on the reactivity of the amine and carbonyl components (Table
13.3). The reaction of aliphatic amines and aromatic aldehydes reliably yields the
desired imine within short reaction times at room temperature. Other scenarios
may result in side-reactions or incomplete conversions. Aromatic amines also
form imines, but, depending on their electronic nature, this reaction may require
elevated temperatures and elongated reaction times. Imines derived from aliphatic
aldehydes tend to form the tautomeric enamines. The reactivity of ketones is
strongly inuenced by their structure. Cyclic ketones react nicely whereas acyclic
ketones often show hardly any conversion [72].
Tab. 13.4. Optimized reaction conditions for reductive aminations using NaCNBH3 .
Reaction conditions have been investigated for all reagents in order to achieve
maximum conversion and to suppress overalkylation. The classical reagent for re-
duction, NaCNBH3 [76], has been optimized with respect to aldehydes and amines
for both solution- and solid-phase chemistry (Table 13.4) [77].
13.4.2.4 Applications
reaction conditions for reducing the imine was crucial. Standard methods failed
(e.g. 110 equiv. R-CHO/NaCNBH3/1% HOAc in DMF). However, adjusting the
solvent system from anhydrous DMF to aqueous THF gave clean monoalkylation
with aliphatic and aromatic aldehydes as well as with ketones.
The presence of Ti(OiPr)4 requires that the reaction is carefully run under inert
conditions in order to avoid precipitation of TiO2 . The feasibility of this protocol
was demonstrated by the synthesis of a 8448-member library of benzopyrans [81].
References
14
Oxidation Except CC Double Bonds
Henning Steinhagen
14.1
Introduction
14.2
Oxidation of Alcohols to Aldehydes and Ketones
Conversion of alcohols are among the most used and reliable oxidation reac-
tions on solid phase. In general, many protocols from the classic repertoire
Scheme 14.1. Oxidation step in the synthesis of peptide aldehydes of the (S)-MAPI type [4f ].
Scheme 14.2. Oxidation using the aldol strategy in the solid-phase synthesis of epothilone A [5].
bound oxidants available that are suitable for oxidation of alcohols in solution
to the corresponding aldehydes and ketones. Poly(vinyl pyridinium dichromate)
(PVPDC) is an inexpensive, easy-to-use, recyclable reagent for the oxidation of dif-
ferent alcohols, including primary, secondary and allylic alcohols [13]. The latter
can be also oxidized using polymer-supported ammonium perchromate, although
the reagent fails with saturated alcohols [14]. Silica-based chromium reagents have
also been reported as easy to prepare and use for the oxidation of several alcohols
[15]. Many other convenient-to-use reagents such as KMnO4 on kieselguhr [16] and
ammonium chlorochromate on silica [17] have been reported to oxidize a variety
of alcohols. Ammonium chlorochromate proved especially useful for the oxida-
tion of benzoins to benzils. Nevertheless, the main problems of chromium-based
reagents remain selectivity and toxicity issues and the mostly stoichiometric
use. This draw back could be overcome by a very useful reagent for the oxidation
of several alcohols, including primary, benzylic and secondary alcohols: polymer-
supported perruthenate (PSP) [18]. The reagent can be used stoichiometrically or
in catalytic amounts with molecular oxygen or an N-oxide as cooxidant. Using
molecular oxygen as a cooxidant also avoids the need for conventional work-up
(Scheme 14.5) [18b]. Another recently introduced useful oxidant is a TEMPO-
derived oxammonium resin [19]. This powerful oxidant cleanly converts primary
alcohols to aldehydes and also secondary alcohols to ketones (Scheme 14.6).
14.2.1
Examples of the Oxidation of Polymer-bound Primary Alcohols to Aldehydes
Bradley and coworkers [4f ] have described an interesting example using SO3/
pyridine for the oxidation of polymer-bound primary alcohols in the synthesis of
peptide aldehydes based on the human immunodeciency virus (HIV) protease
inhibitor (S)-MAPI. Peptide aldehydes are an important class of transition-state
analogs of dierent proteases and have been of considerable interest since their
initial discovery in natural products [20]. The eciency and reliability of the oxi-
dation step was dramatically improved by the incorporation of a small polyethylene
glycol (PEG) spacer between the linker and the solid support (polystyrene resin).
372 14 Oxidation Except CC Double Bonds
In the course of the synthesis, the polymer-bound primary alcohol 1 was oxidized
by SO3/pyridine to the peptide aldehyde 2, which could be readily cleaved using
triuoroacetic acid (TFA).
An interesting example of the Swern oxidation of a primary alcohol in a multi-
step sequence on solid support (Merrield resin) was demonstrated by Nicolaou et
al. [5] in the synthesis of the anticancer agents epothilones A and B. In the course
of the synthesis, the TBS-protected alcohol 3 was deprotected with HF/pyridine
and subsequently oxidized to the aldehyde 4 (> 95% yield). The aldehyde was
essential for further conversion to 5 in a stereoselective aldol reaction. After three
additional steps [1, esterication; 2, metathesis upon cleavage; 3, epoxidation with
methyl-(triuoro-methyl)dioxirane], epothilone A was obtained.
14.2.2
Examples of the Oxidation of Polymer-bound Secondary Alcohols to Ketones
Miller at al. presented a simple and ecient way for the preparation of carboxy-
pyrrolinones on solid phase [9a]. The pyrrole system is an interesting potentially
bioactive motif for both agrochemicals and pharmaceutical drugs. In this ap-
proach, Wang resin-bound malonic acid [22] 8 was coupled with an amino alcohol
in the presence of 1-hydroxybenzotriazole hydrate (HOBt)/DIC, providing the hy-
droxyamide 9 in high yield. For the oxidation step to 10, several oxidizing reagents
were tested. Under Swern conditions the conversion was not complete because the
Swern reagent is unstable at temperatures above 10 C. SO3/pyridine is stable at
room temperature, but the reagent was not eective for all of the examples shown,
especially N-aryl-substituted compounds. The most broadly applicable reagent was
CrO2 (Ot-Bu)2 generated in situ according to Lezno et al. [9b]. The oxidation of
9 to 10 proceeded successfully for most substrates, and after ring closure with
lithium diisopropylamide (LDA)/ZnCl2 and subsequent cleavage with TFA, the
carboxypyrrolinones 11 could be obtained.
14.2.3
Examples of the Oxidation of Alcohols by Polymer-bound Reagents
Scheme 14.7. Proposed mechanism for the oxidation of alcohols using oxammonium resin [19].
14.3
Oxidation of Polymer-bound Aldehydes to Carboxylic Acids
Compared with the oxidation of alcohols to aldehydes and ketones, there are rela-
tively few reports of the direct oxidation of aldehydes to carboxylic acids under
solid-phase conditions. This is because the corresponding alcohols serve as more
stable and easily protectable functional groups in both solid and liquid phase.
Nevertheless, there have been some reports of the oxidation of polymer-bound al-
dehydes to carboxylic acids including the use of NaClO2 [25] and MCPBA [26] as
oxidants. An impressive example was demonstrated by Nicolaou et al. in the solid-
phase synthesis of a combinatorial library of sarcodictyins (Scheme 14.9) [25a]. Sar-
codictyins were discovered in 1987 in a Mediterranean coral [27a] and their potent
antitumor activity (Taxol-like mode of action) was recognized by Ciomei et al. in
1997 [27b]. In the example given, the tricyclic core system 25 was constructed
via multistep synthesis on Merrield resin. After deprotection of 25 using TBAF,
the primary alcohol was oxidized stepwise rst to the unsaturated aldehyde 26
with DMP. After subsequent oxidation to the corresponding carboxylic acid with
NaClO2 , esterication using MeOH/DCC led to the methyl ester 27. The stepwise
procedure gave optimal chemical yields (@95%).
376 14 Oxidation Except CC Double Bonds
14.4
Oxidation of Sulfur-containing Compounds
14.4.1
Examples of the Oxidation of Polymer-bound Suldes to Sulfoxides and Sulfones
There are in general two reasons for the oxidation of a sulfur-containing com-
pound on solid phase. Either the oxidized form displays some interesting biologi-
cal or chemical activity, or the oxidation procedure is part of a cleavage strategy in
which, for example, the sulfoxide function is cleaved by elimination or the sulfone
is directly displaced with a nucleophile. A typical example of a cleavage procedure
using a sulfone was presented by Gayo and Suto in the synthesis of substituted
aminopyrimidines (Scheme 14.10) [29b]. The solid-supported (TG Thiol resin)
14.4 Oxidation of Sulfur-containing Compounds 377
14.5
Oxidation of Selenium- and Phosphorus-containing Compounds
14.5.1
Examples of the Oxidation (Cleavage) of Selenides to Selenoxides on Solid Support
loaded using a threefold excess of 39. The further strategy in the course of con-
structing a combinatorial library contained various chemical manipulations (in
general transformation of R2 to R5). These transformations included annulations,
glycosidations, aldol condensations, and various coupling reactions, yielding the
products represented by formula 41. The selenium oxidation step readily cleaved
the products, yielding 42 in high purity and eciency with the introduction of a
double bond.
14.6
Oxidative Formation of Heterocycles on Solid Support
14.7
Oxidative Coupling and Cleavage Reactions on Solid Support
There are relatively few reports of oxidative coupling reactions on solid support.
Nevertheless, some quite complex reaction sequences have been successfully per-
formed. Phenols have been coupled to cyclic biarylethers using phenolic oxidation
employing thallium(III) salts [57]. Biaryls have been formed by oxidative coupling
using 1,3-dinitrobenzene (see Scheme 14.15) [58] and vanadium-based oxidants
[59]. Oxidative homocouplings of ortho-hydroxystyrenes yielding carpanones have
been performed using PhI(OAc)2 , as described in Scheme 14.14 [60a].
Scheme 14.14. Oxidative coupling strategy for the synthesis of carpanones [60].
In contrast to oxidative couplings, there have been many reports of cleavage re-
actions under oxidative conditions (compare Section 14.7.2). Most of these reports
directly refer to linker cleavage strategies (see Chapter 4.5). An interesting applica-
tion is the preparation of peptide aldehydes using the cleavage of amino alcohols
[61] and diols [62] with NaIO4 (see Scheme 14.16).
Other applications of oxidative cleavage procedures include arylethers which have
been cleaved using cerium ammonium nitrate (CAN) as an oxidant [63]. Benzyl-
ethers have been cleaved from solid support (Wang resin) using DDQ oxidation
[64] and benzylamines which have been cleaved from Merrield p-benzyloxy-
benzylamine (BOBA) resin by DDQ [65]. Aryl hydrazides have been used as linker
systems and can be oxidatively cleaved using dierent conditions including
Cu(OAc)2 and NBS [66].
14.7 Oxidative Coupling and Cleavage Reactions on Solid Support 381
14.7.1
Examples of Oxidative Coupling Reactions on Solid Support
In the rst step, the attached biarylic compound 47 was converted to the inter-
mediary cyclic cuprate 48 using lithiation and treatment with copper cyanide. Sub-
sequently, the biaryl ten-membered ring 49 was formed upon exposure to the oxi-
dant 1,3-dinitrobenzene (1,3-DNB). After cleavage from solid support using HF/
pyridine the product 50 could be isolated in a diastereomeric ratio of 6:1 (P/M). A
disadvantage of the solid-phase approach was the decreased diastereomeric ratio,
which was better (20:1) in the analog reaction in solution phase.
382 14 Oxidation Except CC Double Bonds
14.7.2
Examples of Oxidative Cleavage Reactions on Solid Support
Scheme 14.16. Synthesis of a-oxo aldehydes using formyl transfer from the linker system [62].
References
15
Reductions in Combinatorial Synthesis
Christopher P. Corrette and Conrad W. Hummel
15.1
Introduction
Reductions are one of the most fundamental and widely used transformations in
organic chemistry. The reduction of a wide variety of functional groups has been
exploited in both solid- and solution-phase combinatorial library synthesis. There
are a number of reviews that cover combinatorial chemistry, including reductions
[1]. This chapter is divided into two main sections. The rst section covers reduc-
tions where the substrate is support bound. The second section covers reductions
where the substrate remains in solution (resin-bound reagents and catalysts, catch-
and-release purication, and uorous chemistry). In general, the combinatorial
chemistry of peptides has been omitted in our coverage.
15.2
Solid-phase Reductions
15.2.1
Aldehyde Reductions
the Merrield resin [2a], the Wang resin [2b], and an arylsulfonate ester resin [6].
They have also been used as a key step in the preparation of a variety of other
resins. The preparation of the acid-labile resins 2 [7] and 4 [8] each required
NaBH4 -mediated aldehyde reductions (Scheme 15.1). The improved synthesis of
SASRIN resin, described by Katritzky et al. [9], and the synthesis of a photolabile
6-nitrovanillin-based resin, described by Zehavi and Patchornik [4], also required
aldehyde reductions. The use of an acid wash to remove excess reagent is inappro-
priate in the case of the acid-labile resin preparations. In some of these cases, an
alcoholic solvent wash suces.
15.2.2
Ketone Reductions
As with aldehyde reductions, the reagent of choice for simple ketone reductions on
a solid support is NaBH4 [10]. However, various other reducing agents have been
described in the literature to carry out this transformation. These include the
borohydride reducing agents diisoamylborane [11] and LiBH4 [12]. The latter was
used in the preparation of an acid-labile xanthone-based resin described by Sieber.
NaBH4 in combination with diethyl methoxyborane has also been used to reduce
the diketone 5 [13] to diol 6 (Scheme 15.2). Following boron oxidation and re-
moval, this diol was cyclized with dilute acid to d-lactone 7.
Asymmetric reductions of ketones bound to resin are also possible. The silane re-
agents diphenylsilane and a-naphthylphenylsilane in the presence of the catalyst
prepared from m-dichlorotetraethylenedirhodium(I) and () or ()-DIOP asymmet-
rically reduced a polymer-bound ketone [14]. The resulting secondary alcohol, of
unknown optical purity, was part of a resin-bound asymmetric rhodium-containing
hydrogenation catalyst that was used to reduce the double bond of a-N-acylami-
noacrylic acids. However, the optical purities (@ 70% ee) of the resulting amino
acids were comparable to those obtained from the analogous homogeneous catalyst.
Diastereoselective ketone reductions can be achieved by a variety of methods. An
example of a diastereoselective ketone reduction is found in Patersons library syn-
thesis of polyketides [15]. To increase the degree of diversity in this library, the re-
duction of b-hydroxy ketone intermediates employed two dierent sets of condi-
tions, allowing access to both syn- and anti-1,3-diols. The former was obtained with
95% diastereoselectivity (ds). Treatment of the b-hydroxy ketone with (c-Hex)2 BCl
and triethylamine delivered a boron aldolate, which could be reduced with LiBH4
aording the syn derivative with 95% ds. The 1,3-anti-diol was generated with 97%
ds via the anti-selective EvansTishchenko reduction protocol, using stoichiometric
quantities of SmI2 and propanal, followed by a LiBH4 -mediated ester cleavage. Pat-
erson also points out that attempts to eect the anti reduction with Me4 NBH(OAc)3
resulted in lower diastereoselectivity (75%).
Other examples of both chelation- and nonchelation-controlled diastereoselec-
tive ketone reductions were described by Ellman in his synthesis of an aspartyl
protease inhibitor library [16]. To achieve chelation-controlled reduction, the a,b-
dialkoxy ketone 8 was treated with Zn(BH4 )2 in diethyl ether and THF at 20 C
(Scheme 15.3). Following deprotection and resin cleavage, the triol product was
peracetylated. Analysis by gas chromatography (GC) showed that the syn product
had been delivered in 6080% ds. Following separation of this mixture by prepa-
rative high-performance liquid chromatography (HPLC), the major diastereomer
was found to have a 90% ee, which demonstrated that only a small degree of race-
mization had occurred during an eight-step sequence that included reduction. The
anti compound was obtained with 74% ds, utilizing the nonchelation-controlled
conditions of l-selectride in THF at 75 C. In another example of substrate-
controlled reduction, Schlessinger was able to reduce ketone 10 stereoselectively
with NaBH4 (Scheme 15.4) [17].
15.2.3
Ester Reductions
NaBH4 -mediated ester reduction in ethanol and water has also been demon-
strated; however, this protocol may not be general as esters are typically inert to
this reagent.
Esters are common points of attachment for substrates to solid supports. Cleav-
age can be eected by saponication or by reduction, aording acids or alcohols,
respectively, as the nal products. For reductions, both DIBAL [21] and LAH [22]
have been described in the literature. Tietze and coworkers have demonstrated the
use of DIBAL in the two examples shown in Scheme 15.8, where concomitant re-
duction of a diester [21c] and a b-ketoester [21d] have led to diols 21 and 23, re-
spectively. Kurth and coworkers have also used this reagent to reduce the resin-
bound b-hydroxy ester 24, thus providing diol 25 [21a], which is a representative of
a 27-compound library (Scheme 15.9). Kuster and Scheeren have reductively cleaved
the a-amino ester 26 from Wang resin to obtain the primary amino alcohol 27
using LAH (Scheme 15.10) [22].
Resin-bound acid chlorides have also been reduced to alcohols. Goldwasser and
Lezno have shown that diacid chlorides can be loaded onto 2% crosslinked poly-
styrene resin to give ester-linked acid chlorides. The resins can then be treated
with amines or hydride reducing agents to give amides or alcohols, respectively
[23]. In the latter case, the most successful reducing reagents were NaBH4 and
NaCNBH3 (4 equiv., in THF at ambient temperature), which delivered the desired
products in 67% and 65% yields, respectively. However, following basic cleavage
and esterication, it was found that these products were contaminated with the
dimethylester byproduct arising from incomplete reduction.
15.2.4
Mixed Anhydride Reductions
A mild method employed to obtain alcohols from carboxylic acids consists of using
a mixed anhydride intermediate. In their search for an optimized phosphomannose
isomerase inhibitor, researchers at Aymax coupled ten dierent symmetrical an-
hydrides and diacids to an immobilized 2-aminoindane-2-carboxamide to deliver
acids of the general structure 29 (Scheme 15.11). This set of acids was treated with
isobutylchloroformate and triethylamine in THF, followed by NaBH4 in water, to
deliver the primary alcohols 30 [24]. Further diversication led to a library of 600
analogs (60 pools of ten compounds). Chemists at Signal have also utilized this
reduction protocol to deliver pyrimidine 33 following Mitsunobu alkylation and sub-
sequent cleavage (Scheme 15.12) [25].
15.2 Solid-phase Reductions 393
15.2.5
Thioester Reductions
In their Evans aldol approach to polyketide libraries, Reggelin et al. have also
utilized thioesters and their reduction as part of the synthetic strategy [28]. They
rst converted resin-bound oxazolidinones 42 to thioesters 43 via a three-step
protocol involving hydrolysis, thioester formation, and alcohol protection. They
next reduced the thioester with LiBH4 and reoxidized the resulting alcohol to
aldehyde 45 with DessMartin reagent (Scheme 15.15). This aldehyde could then
be used in a subsequent Evans aldol reaction. Unlike the examples given above,
this route was designed to leave the substrate on the resin following thioester
reduction.
15.2.6
Weinreb Amide Reductions
of excess reagents. Reggelin et al. solved this problem in the solid phase by mak-
ing use of Weinreb amides, which deliver aldehydes upon reduction in standard
solution-phase chemistry. In their approach to polyketide libraries, they demon-
strated that the imides 46 can be converted to Weinreb amides under standard
conditions (Scheme 15.16). Following alcohol protection, these amides can be se-
lectively reduced with DIBAL to deliver the support-bound aldehydes 48 [29]. In
1998, Reggelin et al. replaced this sequence with another (see Section 15.2.5) that
goes through a thioester instead of the Weinreb amide. However, this change was
due to diculties arising from steps other than reduction [28].
Two other groups have developed Weinreb amide-based linkers that give alde-
hydes in solution upon reductive cleavage [30]. Salvino et al. modied a resin that
had been used for the synthesis of hydroxamic acids to give an O-linked Weinreb
amide linker [30d], whereas Martinez and coworkers synthesized an N-linked ver-
sion [30a]. The synthesis of C-terminal amino acid aldehydes has been successfully
demonstrated by both groups, providing clean products in low to moderate yields
using LAH as the reductant. Schemes 15.17 and 15.18 show the synthesis of Boc-
Phe-H using both linkers. In the case of the Martinez linker, this support has also
been applied to the synthesis of C-terminal peptide aldehydes [30b] and side-chain
aldehydes [30c]. The peptide analogs were produced without noticeable epimeri-
zation and the side-chains were reduced in high yield with LiAl(OtBu)3 H. The
O-linked resin was also employed to deliver ethyl ketones by treating inter-
mediate Weinreb amides with ethyl Grignard [30d]. Dinh and Armstrong also intro-
duced an N-linked Weinreb amide resin, which delivered ketones in low to
moderate yields upon treatment with Grignard reagents [31]. However, attempts to
reductively cleave these amides with LAH produced aldehydes in very low yields
(< 20%).
15.2.7
Sulfur Reductions
application, Lam and coworkers used NaBH4 to reduce a disulde bond in their
solid-phase approach to 1,4-benzothiazin-3(4H)-one derivatives [35].
In an example of linker preparation, Sucholeiki and coworkers have utilized
a b-mercaptoethanol-mediated reduction to synthesize the photolabile resin 62
(Scheme 15.21) [36]. Zhao and Janda also utilized a disulde reduction in the syn-
thesis of the thiol linker shown in Scheme 15.22 [37]. Treatment of a mixture of
disuldes generated in a previous alkylation step with DTT in water led to the de-
sired resin 64 in excellent yield. Following construction of the targeted substrate
on solid phase, oxidation of the thiol with KHSO5 in water gave a sulfone that was
cleaved with 5% Na/Hg in methanol and DMF to yield 66 [38]. The solvent mix-
ture of methanol and DMF (1:8) was crucial for this successful reductive cleavage
[39]. The use of additional methanol or THF as a replacement for DMF gave lower
yields and required the use of larger quantities of Na/Hg. Zhao and Janda have
also pointed out that, for reactions using soluble polyethylene glycol (PEG)-based
polymers, isopropanol is an excellent choice for polymer precipitation, and leading
to products of higher purity [38].
Scheme 15.22. Preparation and use of Jandas soluble polymeric thiol linker.
15.2.8
Selenium Reductions
The ease with which selenium can be reduced and oxidized is well understood,
however the toxicity and odor of selenium compounds has limited their use. Nic-
olaou et al. recognized that by attaching selenium to a solid support these draw-
398 15 Reductions in Combinatorial Synthesis
backs would be mitigated, and the positive attributes of selenium reagents and
substrates could then be exploited [40]. They have shown that supported selenium
reagents make excellent traceless linkers because loading of substrates and sub-
sequent cleavage to simple alkyl groups is carried out with equal ease. As shown in
Scheme 15.23, reduction of the selenium bromide 67 with LiBH4 delivers lithium
selenide 68, which then reacts with substrates containing an alkyl halide. Follow-
ing substrate modication (not shown), the carbonselenium bond can be reduced
with Bu3 SnH to give an alkyl group with no vestige of the linker remaining.
Alternatively, Nicolaou et al. have shown that these selenides can be treated with
oxidizing reagents to give an alkene.
15.2.9
Quinone Reductions
The redox chemistry of quinones has been used to prepare solid-phase linkers for
peptide synthesis [43]. The quinone can be easily reduced with NaBH4 or Na2 S2 O3
to provide the corresponding dihydroquinone. Treatment of the reduced product
with TBAF results in phenoxide displacement at the C-terminus of the attached
peptide, thus releasing the molecule from the resin. In initial eorts, the phenolic
oxygen forms a lactone and releases the peptide functionalized at the C-terminus
(Scheme 15.25A) [43a]. An improved approach is shown in Scheme 15.25B, in
which the peptide is provided without functionalization at the C-terminus via SN 2
displacement by the phenol [43b].
15.2.10
Amide Reductions
The reduction of amides to amines is a useful technique for the generation of di-
versity in combinatorial synthesis. Either separately, or in combination with the
complementary reductive amination techniques of aldehydes described in Section
15.2.12.4, an enormous number of amines and polyamines can be accessed from
the diverse set of amides and polyamides that can be generated via a combinatorial
approach. Further utility arises from the ability to increase diversity by modica-
tion of the product amines through any number of common approaches.
The reducing agents most routinely employed are BH3 THF [44], BH3 pyr [45],
or BH3 SMe 2 [46], although Red-Al [47] and LiBH4 [48] have also been used. Bor-
ane reductions are easily applied to resin-bound amides since any excess reducing
agent and resulting inorganic byproducts are easily washed away from the polymer
support. An important factor for such reductions is the cleavage of the borane
400 15 Reductions in Combinatorial Synthesis
Scheme 15.26. Iodine work-up for borane reduction of amides on solid phase.
Several articles have been published on the use of amide reductions for the gen-
eration of combinatorial libraries [44, 45a,c]. Dipeptides are typically constructed
on resin, acylated at the N-terminus, and exhaustively reduced with BH3 THF or
BH3 pyr to give polyamines. This methodology has been used to generate hydan-
toins and cyclic ureas, thus generating libraries from libraries [44a]. This se-
quence has also been used to construct a 100,000-compound library of polyamines
and bicyclic guanidines (Scheme 15.27) [44b].
15.2.11
Carbamate Reductions
There are several examples of carbamate reductions wherein the carbonyl is ex-
haustively reduced to an N-methylamine. All of these examples use aluminum hy-
15.2 Solid-phase Reductions 401
dride reagents. For example, Liu and Ellman used Red-Al-mediated carbamate re-
duction to prepare chiral N-methyl-2-pyrrolidinemethanol ligands [47]. Carbamate
linkers have been used as latent methylamines in solid-phase synthesis. Formation
of the carbamate on resin followed by LAH reduction generated the desired pri-
mary and secondary amines [50]. A carbamate linker has also been used to gen-
erate substituted N-methylpyrrolidines via reductive cleavage with LAH (Scheme
15.28) [51]. Xiao et al. used Na2 S2 O4 to reduce an O-piperidinyl carbamate linker,
thus revealing a secondary amine that underwent a cyclizative cleavage with a
pendant ester to release a desired alcohol [52].
15.2.12
Reductive Amination
Generally carbonyl reactivity follows the order aldehyde > cyclic ketone g acyclic
ketone. Steric considerations are important in reductive amination reactions, thus
hindered inputs may require forcing conditions for adequate results. For example,
hindered amino acid esters such as valine [53a] have been reported to be poor
402 15 Reductions in Combinatorial Synthesis
inputs for reductive amination reactions as have methyl amines [55]. Secondary
amines can require repetitive cycles [56] of reductive amination, or very long reac-
tion times [56, 57] to obtain good conversion.
Electronic factors can also be important in reductive amination reactions, as evi-
denced by the higher reactivity of alkyl amines versus anilines. The more reactive
nature of alkyl amines and aldehydes can render them prone to bis-alkylation [44b,
58]. Reductive aminations can also be sensitive to solvent choice. Brown and co-
workers found that the use of aqueous THF was crucial for the successful reduc-
tive amination of a Rink amide resin [59].
Racemization of amino acid esters as amine inputs during reductive amina-
tion is another concern for library synthesis [60]. Racemization is expected to oc-
cur by equilibration during imine formation [61]. In order to suppress racemiza-
tion, imines can be formed in the presence of a reducing agent, thus avoiding
equilibration [58a, 61].
There are several methods for monitoring the progress of solid-phase reductive
amination. Qualitative nitrogen analyses such as a (Kaiser) ninhydrin test can
quickly and easily assess reaction progress [62]. Infrared spectroscopy can also be
used to detect consumption of the carbonyl and/or imine components in a reduc-
tive amination process by their characteristic CbO (17401720 cm1 ) and CbN
(@1650 cm1 ) stretches [63].
Polymer-supported amines have also been used for reductive aminations with
aldehydes. There are several examples where Rink amide resin has been reduc-
tively alkylated [59b, 62b, 75a]. Katritzky et al. formed a variety of imines (both
electron-rich and -poor aromatic aldehydes) with TMOF and reduced them with
LiBH4 [75a]. Rivero and coworkers also used a two-step reductive amination pro-
cedure to prepare a library of 500 macrocycles (Scheme 15.34) [62b]. Benzyloxy-
Tab. 15.1. Recent libraries using reductive amination (since January 2000).
sive study of conditions, thus emphasizing the importance of varying reaction con-
ditions for successful reductive aminations. Barany and coworkers prepared a set
of 3,4-dihydro-1,4-benzothiazines using a one-pot alkylationintramolecular reduc-
tive amination to prepare the thiazine core [84]. A group at Abbott has prepared an
antibiotic screening library of 70,000 macrolides [58b]. The synthesis involved
three reductive amination steps (aldehydes with a primary amine and two second-
ary amines) to introduce diversity and began with a preconstructed macrolide core.
15.2.13
Azide Reductions
DMF [97]. A modied Staudinger reaction has been applied to the preparation of
amides using a solid-supported glycoazide in a one-pot procedure (Scheme 15.36)
[98].
problem, aqueous PPh3 provided the amine in good yield without any azide dis-
placement and allowed the synthesis of an array of hydroxybisamides. Nicolaou
et al. also used an aqueous PPh3 -mediated azide reduction methodology to provide
an amine for diversication in the preparation of a library of 50 sarcodictyins
[104].
Tab. 15.2. Recent libraries utilizing azide reduction (since January 1999).
15.2.14
Nitro Group Reductions
cally accompany nitro reduction with SnCl2 , which presents a potential problem
with acid-sensitive polymer supports. Addition of a small amount of buer such
as sodium acetate often remedies this situation [35, 115]. DMF is the solvent of
choice for tin reductions, but N-formylation of o-diaminobenzenes generated from
2-amino-substituted nitrobenzenes and subsequent cyclization to benzimidazoles
has been observed as a side-reaction [116]. This result can be avoided by employ-
ing other solvents such as NMP or N-methylmorpholine (NMM). Tin reductions
often require heating and can benet from exclusion of oxygen [117].
Tin reductions have been used in a number of library syntheses. In an early ex-
ample, the tin-mediated reduction of solid-supported substituted nitrobenzenes led
to anilines that were derivatized to provide a small library of phenols following
cleavage (Scheme 15.42) [118]. The synthesis of a library of 3,4,5-substituted 1,5-
benzodiazepin-2-ones began with a tin-mediated nitro reduction on polymer
support (Scheme 15.43) [119]. Sequential hydrolysis and intramolecular amide
coupling provided the benzodiazepine core.
Scheme 15.42. Meyers approach to a phenolic library via a tin-mediated nitro reduction.
15.2.15
Imine Reductions (not Reductive Amination)
Tab. 15.3. Recent libraries utilizing nitro reduction (since January 1999).
tion and subsequent reduction, see Section 15.3.1.2. Imine reductions have been
used in the synthesis of compounds on solid support as well as in linker activation
prior to cleavage (see below).
BischlerNapieralski cyclization products have been prepared on solid phase,
and the resulting cyclic imines were reduced with NaBH3 CN to provide tetrahy-
droisoquinolines (Scheme 15.47) [133]. When NaBH4 was used in this application,
the dihydroisoquinoline was cleaved at the esterresin linkage. An indolenine in-
termediate, generated via a Fischer indole reaction, was reduced with NaBH4 in the
synthesis of a small library of spiroindolines [134]. Resin-bound imines of amino
acids, prepared from transimination with N-H ketimines, have been reduced with
NaBH3 CN in an approach to a library of hydantoins [135]. In a new linker appli-
cation, a phenanthridine was reduced with NaBH4/BH3 THF [136]. The desired
acid was subsequently released via oxidative cleavage.
15.2.16
Nitrile Reduction
Conti and coworkers have reported the reduction of a nitrile on solid support
[137]. An aromatic nitrile was reduced with BH3 SMe 2 in diglyme at 80 C to pro-
vide a benzylamine. The resulting molecule was then released from the resin
using an a-chloroethyl chloroformate methanol activation-cleavage strategy.
15.2.17
NN and NO Bond Reductions
Samarium diiodide has been used to cleave NaO bonds in a hydroxylamine trace-
less linker application [138]. Recently, a report was published that described both
nitrosamine and hydrazine reductions on solid phase for the preparation of an ar-
ray of a-substituted primary amines (Scheme 15.48) [139]. DIBAL reduction of the
nitrosamine to the corresponding hydrazine followed by addition of an aldehyde
gave the resin-bound hydrazone. Nucleophilic addition and subsequent borane re-
duction of the resulting derivatized hydrazine provided the target amines in mod-
414 15 Reductions in Combinatorial Synthesis
Scheme 15.48. Preparation of a-substituted primary amines via hydrazine reductive cleavage.
est yields. This approach has also been used to prepare chiral hydrazones and the
corresponding chiral amine products with modest enantioselectivity (5086% ee)
[140].
15.2.18
Miscellaneous Reductions
15.3
Solution-phase Reductions
15.3.1
Supported Reagents
tional groups. Itsuno and coworkers attached optically active prolinol to polysty-
rene to give 152 and treated this product with BH3 THF to derive an enantio-
selective reducing agent (Figure 15.1) [151]. This reagent reduced prochiral ketones
to secondary alcohols in good optical purity. The highest optical yield (80%) was
obtained with a 1% crosslinked reagent (with 50% functionalization), which was
20% higher than that obtained by the solution-phase control. Following hydrolysis
of the reaction mixture with 2 M HCl, the polymer was collected via ltration.
Borane regeneration allowed this reagent to be used two more times. Itsuno and
coworkers also attached amino alcohols to a polymer through a pendant aromatic
group (Figure 15.1) [152]. An acetophenone oxime was reduced with the reagent
derived from this polymer-bound amino alcohol and NaBH4/ZrCl 4 or BH3 THF
[153]. The optical purity of the product was only 35% ee; however, the reagent
could be recycled.
Adjidjonou and Caze have also synthesized polymer-bound amino alcohols that
were combined with NaBH4 to reduce acetophenone [154]. These reagents deliv-
ered the product with modest enantioselectivity (up to 75% ee), which was much
more enantioselective than the product obtained from a solution-phase control ex-
periment (12% ee). Frechet et al. derived ligands from ephedrine and polystyrene
resin and utilized them in the LAH-mediated reduction of acetophenone [155].
The enantiomeric excess of the product was 79% when a lightly loaded insoluble
polymer species was utilized in the presence of an achiral phenol. The minimally
substituted resin allows the chiral amino alcohols to act independently from one
another and allows the hydride to access all of these units fully, thus providing
higher enantioselectivity.
Borane-based reagents
In 1977, Gibson and Bailey introduced the rst solid-supported borohydride
exchange resin (BER) [156]. It should can be noted that, following use, this reagent
can be collected by ltration and regenerated. Early studies with this reagent
focused on carbonyl reductions and related chemoselectivities, which were found
to be better than those produced by NaBH4 in solution [157]. It was understood
that this dierence in selectivity was due in part to the slower reaction kinetics of
the support-bound reagent.
416 15 Reductions in Combinatorial Synthesis
Aldehyde and ketone reductions. In 1983, Yoon et al. studied the chemoselectivity
of carbonyl reductions in a series of competitive reduction experiments with
BER (no additives) [157a]. Their results showed that aldehydes were reduced in
preference to ketones. More interesting were their observations that there was se-
lectivity between aldehydes and between ketones. Aromatic aldehydes were prefer-
entially reduced in the presence of aliphatic aldehydes. Benzaldehydes with para-
substituted electron-withdrawing groups were reduced preferentially to those with
para electron-donating groups. It was also shown that unhindered ketones were
reduced in preference to hindered ketones. In a separate study by Yoon et al., it was
also shown that the addition of CuSO4 to BER increased the diastereoselectivity of
the reduction of norcamphor to norborneol (endo/exo 94:6 vs. 82:18) [158]. The
reduction of ketones and aldehydes has also been carried out using zinc [161]
and zirconium [162] borohydride reagents immobilized on polyvinylpyridine. The
zinc-based reagent is completely inert toward ketones; however, addition of FeCl3
gives low to moderate yields of ketone reduction products. The solid-supported
zirconium borohydride reduces both aldehydes and ketones in the absence of an
additive. Further, it has been shown that the BER-Ni(OAc)2 system fully reduces
aromatic aldehydes to toluene derivatives in high yield regardless of aromatic sub-
stitution [160b]. A hindered equivalent of BER, which diastereoselectively reduces
ketones to secondary alcohols, has recently been introduced by Smith et al. [169].
Studies on the reduction of a,b-unsaturated aldehydes and ketones have also
been carried out using these reducing agents. BER selectively adds hydride in a
1,2-fashion to these substrates, delivering allylic alcohols in high yield [157b]. The
same properties are exhibited by the zirconium reagent [162]; however, the zinc
reagent [161] shows chemoselectivity in that it reduces aldehydes without aecting
ketones. Sim and Yoon showed that addition of 0.1 equiv. of CuSO4 to BER under
standard conditions (5 equiv. BER, methanol, room temperature) fully reduced a,b-
unsaturated systems to saturated alcohols [158]. However, if the amount of BER
was reduced to 2 equiv., the saturated ketone was isolated [158]. Despite these re-
sults, Ley et al. recently published a report describing the isolation of the allylic
alcohol from a BER-CuSO4 -mediated a,b-unsaturated ketone reduction [167a]. In
their synthesis of (G)-epimaritidine, Ley et al. successfully utilized BER-CuSO4
and BER-NiCl2 to carry out the 1,2-reduction of an a,b-unsaturated ketone [167a].
It should be pointed out that the structural complexity of the substrate in the Ley
synthesis is much greater than that of Sim and Yoon. In the report of the synthesis
418 15 Reductions in Combinatorial Synthesis
of (G)-epibatidine, Ley and coworkers also used the parent BER to carry out ketone
reductions in high yield [170].
Ester and acid chloride reductions. The reduction of fully oxidized carbons has
also been studied, but to a much lesser extent. Esters, for example, seem to be
inert to these exchange resins even when transition metal salts are employed. Acid
chlorides, on the other hand, have been reduced to both aldehydes and alcohols
depending on the resin used. Simple long-chain acid chlorides have been selec-
tively reduced to aldehydes in high yield by passage through a column of BER
[171]. Depending on the reaction conditions, Tamami and Goudarzian have shown
that polymeric Zn(BH4 )2 can deliver either the alcohol or the aldehyde, however
the products are not obtained cleanly [161b]. For example, if phenylacetyl chloride
is treated with Zn(BH4 )2 in hot THF, a 70:20 mixture of the alcohol and aldehyde
is recovered. If the reaction is run at room temperature in CH2 Cl2 , a 25:65 mixture
is obtained. Tamani and Lakouraj have also demonstrated that high yields of
clean alcohol can be obtained by using another polymeric zinc borohydride, poly-h-
(pyrazine)zinc borohydride, in THF at ambient temperature [172]. In Ley and
coworkers synthesis of (G)-epibatidine, the rst step involved an aromatic acid
chloride reduction to an alcohol mediated by BER [170].
Epoxide reductions. The reduction of epoxides has also been studied. BER with
CuSO4 does not react with aliphatic epoxides, yet cleanly reduces styrene oxide to
ethylbenzene [158]. Despite requiring additional quantities of reagents (10 equiv.
BER and 0.5 equiv. CuSO4 ), a-methylstyrene oxide and b-methylstyrene oxide
also gave the fully saturated alkylphenyl derivatives upon reduction. Supported
Zn(BH4 )2 was capable of reducing both aliphatic and styrenyl derivatives, how-
ever this reagent did not give fully reduced products. Instead, a mixture of the
more and less substituted alcohols was obtained, with the former predominating
[161b]. The poly-pyrazine zinc reagent was inert toward both types of epoxides
[172].
Halide reductions. Sim and Yoon looked at the reduction of alkyl and aryl halides
in detail. BER-CuSO4 was found to be inert toward simple alkyl and aryl chlorides,
while readily reducing primary and secondary alkyl bromides as well as aryl bro-
mides and iodides [158]. It should be noted that activated chlorides (benzylic or a
to an ester) can be reduced by this system. These chemoselectivities were demon-
strated by performing competition experiments. For instance, 1-bromo-4-chloro-
butane was readily reduced to 1-chlorobutane (95%) and p-bromochlorobenzene
was cleanly reduced to chlorobenzene (99%). Since aryl bromides required heat to
be eectively reduced, while aryl iodides did not, it was possible to selectively re-
duce p-bromoiodobenzene to bromobenzene at ambient temperature with a 97%
yield. Yoon et al. have also shown that BER-Ni(OAc)2 has nearly the same selectiv-
ity prole as BER-CuSO4 , and that this nickel-based system can be used to reduce
1-octyl tosylate to octane in 95% yield provided that NaI is present [160a].
15.3 Solution-phase Reductions 419
Azide reductions. BERs and combinations with nickel or copper salts are eec-
tive at reducing alkyl and aryl azides [158, 174]. In an early application of BERs,
both aryl and arylsulfonyl azides were reduced in methanol to amines and sulfo-
namides [175]. BER-Ni(OAc)2 has been used to reduce a variety of azides [174].
Tamami and Lakourajs piperazine-based zinc reagent can reduce both aryl and
alkyl azides to amines [172]. Tamami and Goudarzians pyridine-based version re-
duces aryl and arylsulfonyl azides but does not react with alkyl azides [161b].
Tin-based reagents
Polymeric tin hydrides are capable of reducing a number of functional groups, in-
cluding carbonyls, alkyl halides, and alcohols [179]. The last are reduced through
the intermediacy of a phenylthionocarbonate, according to the methodology set
forth by Barton [179b,c]. The main advantage of these reagents over tributyltin-
hydride (TBTH) is in the work-up. Separations to remove toxic tin byproducts are
avoided as the tin species can be easily removed by ltration.
In 1975, Crosby and coworkers introduced the rst of the supported tin re-
agents, a polystyrene-based n-butyldihydridotin species [179a]. This reagent
directly links a tin atom to the phenyl ring of the polystyrene backbone. In 1993,
Neumann and Petersheim published an optimized preparation for a polystyrene-
based monohydridotin reagent that utilizes a two-carbon linker between the tin
and aromatic backbone of the polymer [180]. Since aromatic tin bonds can be
420 15 Reductions in Combinatorial Synthesis
labile, this aliphatic carbon-linked tin reagent was believed to be more stable
than Crosbys reagent. Dumartin et al. introduced tin reagents with 3- and 4-carbon
linkers that more closely resemble the structure and reactivity of TBTH [179d].
These tin reagents have been used to carry out carbonyl reductions in high yield,
including the reduction of both aliphatic and aromatic aldehydes and ketones. It
has also been shown that chemoselectivity can be achieved with these reagents, as
alkyl halides have been reduced in the presence of ketones [179a]. Neumann and
coworkers demonstrated the feasibility of alcohol deoxygenation by utilizing the
Barton protocol. This methodology required the conversion of an alcohol to a phe-
nylthionocarbonate, which was then reduced with a solid-supported tin reagent to
give the saturated alkyl compound [179b,c]. Neumann and coworkers have also
applied this reagent to the reductive cyclization of o-alkenyl halides [181].
Scheme 15.55. Synthesis of amines via aza-Wittig reaction and imine reduction.
15.3.2
Supported Catalysts
Hydrogenations
Of the homogeneous asymmetric catalysts designed to carry out the reduction of
prochiral ketones with molecular hydrogen, perhaps none has garnered more
attention than the BINAP-Ru catalyst designed by Noyori [191b]. It should not
be surprising therefore that this catalyst system has been chosen for exploitation
by a number of research groups interested in heterogeneous catalysis. At least two
15.3 Solution-phase Reductions 423
dierent approaches have been used to incorporate the BINAP structure into a
polymer.
An approach chosen by a group from Oxford Asymmetry involved attaching
this C-2-symmetric ligand to an existing polymer with the attachment point distal
from the active site phosphine atoms [193]. This goal was accomplished by mono-
functionalizing the ligand at the 6-position with an alkyl carboxylic acid and then
coupling this group to aminomethyl polystyrene resin. The resulting non-C-2-
symmetric resin-bound ligand 179 (Figure 15.3) was then treated with a ruth-
enium(II) complex and methanolic HBr in acetone to give the active hydrogena-
tion catalyst. The catalyst (1.7 mol%) was added to a methanol/THF solution of the
substrate, which was then treated with 10 atm of hydrogen and heated to 70 C.
Reduction of the b-ketoester, methyl propionylacetate, was complete in 18 h with
an enantioselectivity of 97%. This heterogeneous catalyst was similar in activity
and selectivity to the parent homogeneous BINAP-Ru catalyst. Further, these data
show that the loss of C-2 symmetry is not detrimental to the parent catalysts se-
lectivity. Perhaps more important is that this catalyst was easily recovered and
reused two more times with only minimal losses in yield and enantioselectivity.
Another approach that has been used to incorporate BINAP into a polymer was
carried out by the Lemaire group [194]. This approach involved copolymerization
of a 6,6 0 -dimethylamine BINAP ligand with 2,6-tolylene diisocyanate to give a C-2-
symmetric BINAP polymer 180 (Figure 15.3). This noncrosslinked polymer was
soluble in DMF and DMSO, yet insoluble in the typical hydrogenation solvent
methanol. Utilizing conditions similar to those described above, Lemaire and co-
workers were able to completely reduce methyl propionylacetate in 14 h (0.1 mol%
catalyst, 40 atm., 50 C) to the desired b-hydroxyester in 98% ee. This catalyst was
also recovered and reused up to four times without any loss in activity or selectiv-
ity. Lemaire and coworkers utilized the same polymer in the presence of chiral
diamines [191c] to reduce simple ketones (lacking proximal heteroatoms),
such as substituted acetophenones to alcohols [195]. However, the enantiomeric
excesses of the products varied between 58% and 96%. It was also shown that
the absolute conguration of the added diamine is crucial to retain good enantio-
selectivity.
424 15 Reductions in Combinatorial Synthesis
Transfer hydrogenations
The replacement of molecular hydrogen by hydrogen donors is an issue of prac-
tical importance in the eld of catalytic asymmetric reduction (because of safety
concerns). As was the case for standard homogeneous hydrogenations, the Noyori
laboratory has made some of the most signicant contributions in this area.
Noyori and coworkers introduced the (1S,2S)- and (1R,2R)-N-( p-toluenesulfonyl)-
1,2-diphenylethylenediamine (TsDPEN) ligands, which carry out hydride transfer-
mediated ketone reductions in high yields and enantioselectivities when com-
plexed with ruthenium [191d].
Both the Oxford Asymmetry [197] and Lemaire [198] groups have incorporated
this ligand into polymers, using handles on the aromatic sulfonyl portion of the
ligand as the linkage point to the resin (Figure 15.4). Each group adopted a strat-
egy similar to the one they took in forming the BINAP polymers, described above.
The Oxford group attached the ligand via an amide bond to preformed polymers
(PS and PEG-PS) whereas the Lemaire group took a copolymerization approach.
The Lemaire group did not concern itself with producing a linear C-2-symmetric
polymer as they had previously, because the parent TsDPEN ligand is not C-2
symmetric. They copolymerized styrene and a TsDPEN ligand, equipped with a
vinyl group, in both the presence and absence of divinylbenzene, thus producing
both a crosslinked and a linear polymer.
Both groups studied the reduction of acetophenone; however, each group took
their own approach to optimize the reaction conditions. The Oxford group focused
on the variation of the hydride source, polymer, and solvent, while keeping the
transition metal constant [197]. The Lemaire group varied the polymer and transi-
tion metal, while keeping the hydride source and solvent constant [198]. Regard-
less of which polymeric ligand (PS or PEG-PS) was used in the catalyst prep-
aration with [RuCl2 ( p-cymene)]2 , the Oxford group encountered diculties with
isopropanol as the hydrogen donor. In the case of ligand 181 (Figure 15.4), the
activity of the catalyst and the optical purity of the products were acceptable; how-
ever, catalyst recycling failed. In the case of ligand 182, both the conversion and the
enantioselectivity observed were low with the initial use of the catalyst. To circum-
vent these problems a switch was made from isopropanol to a mixture of formic
acid and triethylamine (5:2). This combination led to successful reductions using
either ligand. The catalyst formed from ligand 182, in neat HCO2 H:Et3 N, gave
product in 97% ee with 95% conversion in 28 h and could be reused once without
any loss in ee. The catalyst formed from ligand 181 required a cosolvent to deliver
favorable results. Addition of either DMF or CH2 Cl2 resulted in enantiomer ex-
cesses of 94% or better with a reasonable degree of conversion (> 60% at 18 h).
This catalyst was also successfully subjected to recycling.
Although the Lemaire group varied both the transition metal and the polymer
in their eorts to nd a heterogeneous transfer hydrogenation catalyst, there was
little dierence in activity and selectivity between their crosslinked and non-
crosslinked polymers. From these results, they chose to focus on the signicance of
the transition metal [198]. Both Ir(I) and Ru(II) complexes were used in the prep-
aration of the catalysts. The iridium catalyst was prepared by combination of the
polymeric TsDPEN ligand 183 and [Ir(I)(COD)Cl]2 in an isopropanolic solu-
tion of KOH, whereas the preparation of two ruthenium catalysts (from either
[Ru(benzene)Cl2 ]2 or [Ru( p-cymene)Cl2 ]2 ) required heat (70 C) and the replace-
ment of KOH with triethylamine. Of these, the best results were found utilizing
the iridium-based catalyst, which gave 96% conversion to the S-alcohol with 94%
ee after 72 h. Unfortunately, the reuse of this catalyst led to poor results in terms
of activity and selectivity. The ruthenium-based catalysts, on the other hand, were
much less selective (3164% ee), but were able to be reused up to four times.
For comparative purposes it is interesting to note that when both groups em-
ployed their ligand with [RuCl2 ( p-cymene)]2 , a crosslinked polymer, and isoprop-
anol (as solvent and hydride source), the optical purity of the alcohol produced was
similar (84% ee for Lemaire and 90.5% ee for Oxford); however, the activities were
quite dierent. After 48 h, the former group saw just 23% conversion while the
latter group saw 88% conversion after 18 h. It must be noted that the catalyst load
(2.5% vs. 1% respectively) and usage of KOH (presence vs. absence respectively)
were dierent.
In another eort to identify a heterogeneous catalyst system capable of carrying
out asymmetric reductions, the Lemaire group has copolymerized dialdimine
ligands 184, 185 (Figure 15.5) with varying amounts of polystyrene and/or DVB
[199]. The iridium-based catalysts formed from the resulting ligands were used in
the isopropanol-mediated transfer hydrogenation of acetophenone. Although the
level of activity for these catalysts was high, the enantiomeric excess of the prod-
ucts obtained were never greater than 70%. Catalyst recycling suered losses in
426 15 Reductions in Combinatorial Synthesis
Fig. 15.5. Dialdimines and diamine used in the preparation of transfer hydrogenation catalysts.
both activity and selectivity. It is interesting to note that ruthenium and cobalt
failed to catalyze the reduction, and that a 71% crosslinked polymer gave higher
optical purity than both 15% and 100% crosslinked polymers. In another example
from the Lemaire group, diamine 186 (Figure 15.5) was copolymerized with both a
diacid chloride and a diisocyanate to give a poly(amide) and a poly(urea), respec-
tively [200]. Utilizing the rhodium-based catalysts prepared from these ligands, the
reduction results were less than optimal. The poly(amide) gave product in only
28% ee and the poly(urea) resulted in a product of only 60% ee. The latter catalyst
could be recycled at least twice.
Also worthy of mention are the Lemaire groups eorts directed toward catalyst
formation using molecular imprinting [201]. In an application of this method-
ology, this group copolymerized a preformed diaminerhodium complex with dii-
socyanate in the presence of the compound to be imprinted (the alcohol product)
optically pure 1-(S)-phenylethanol. Once the polymer was formed and the alcohol
was washed away, an imprint of the product was left in the catalytic site, which
allowed for binding of acetophenone (or a similar substrate) and biased reduc-
tion to the desired products. In practice, the imprinting eect was found to be
real, yet small. The enantiomeric excess of the product from acetophenone reduc-
tion increased modestly, from 33% to 43%, from the polymer catalyst formed in
the absence of the template to the one formed in the presence of the template.
Both of these optical purities were lower than those obtained using the diamine in
a homogeneous control reaction (55%). A drawback to this method is that it does
not allow for the reduction of a diverse set of ketones as the substrates must have a
similar structure to the imprinted molecule.
Borane-based reductions
A third major area of research directed at the heterogeneous asymmetric catalysis
of prochiral ketone reductions is focused on borane-based catalysts. Successful
solution-phase asymmetric reductions using chiral oxazaborolidines, described by
Itsuno et al. [151153] and Corey et al. [192], have prompted much of this re-
search.
Some of the early work carried out by Itsuno et al. involved covalent attachment
15.3 Solution-phase Reductions 427
alcohol, not the boron atom, acts as the point of attachment to the polymer.
The catalyst 189 is formed by combination of the polymer amino alcohol with
BH3 SMe 2 in THF (Figure 15.7). Aryl ketone reduction is then carried out by
treatment with the catalyst and stoichiometric quantities of BH3 SMe 2 in THF.
The resulting secondary alcohols are obtained in enantiomeric excesses ranging
from 89% to 98%, which compares favorably with nonpolymeric results. Un-
fortunately, the products still have to be puried by distillation or chromatography.
In early 2001, Zhao and coworkers reported the preparation of catalyst 190
(Figure 15.8) [206]. Unlike the amino alcohol described above, the nitrogen of this
ligand is attached to the resin via a sulfonamide bond. Product ee values were
good to excellent for the reduction of aromatic ketones and moderate for alkyl ke-
tones when this catalyst was employed. The combination of NaBH4 and Me3 SiCl
(or BF3 OEt2 ) was used as the stoichiometric reductant. Although the catalyst
could be reused up to three times, a regeneration step was required.
The asymmetric reduction of ketones has also been carried out using zinc com-
plexes of polynaphthyl ligands. These catalysts have been shown by the Pu group
to mediate the catecholborane reduction of prochiral ketones [207]. Although the
reduction of arylmethyl ketones gave products in good yield with ee values as high
as 80%, the reductions of alkyl and branched methyl ketones were much less suc-
cessful. After quenching the reaction, the homogeneous polymer was precipitated
by addition of methanol. Reuse of this catalyst also required a regeneration step.
been introduced which uses these tin reagents in catalytic quantities in the pres-
ence of stoichiometric amounts of sodium borohydride.
Enholm and Schulte developed a noncrosslinked polymer (191) that is soluble
in a number of organic solvents (Figure 15.9) [208]. This reagent can easily be
removed from a reaction mixture by precipitation with methanol. Alkyl and aryl
halide reductions have been carried out in N,N-dimethylacetamide (DMA) with 1.5
equiv. of NaBH4 , 0.1 equiv. of 191, and AIBN, as initiator. Because the reactions
are homogeneous, the reaction rates are faster than those found with insoluble
polymeric catalysts, with completion typically observed in just a few hours at
80 C.
Dumartin and coworkers demonstrated the utility of a polymeric tin iodide (192)
and compared it with the reducing capabilities of Neumanns tin chloride reagent
(193) (Figure 15.10, see Section 15.3.1.2) [181, 209]. In the comparative analysis of
1-bromoadamantane reduction, 0.050.9 equiv. of polymer 192 or 193 was combined
with NaBH4 (2.5 equiv.), AIBN (0.1 equiv.), and substrate in ethanol and heated to
65 C for 12 h. When 0.2 equiv. of the catalyst 192 was used, adamantane was ob-
tained in 93% GC yield, while 0.5 equiv. of Neumanns reagent gave only 40% of
the same product. Dumartins group also showed that catalyst 192 produces very
low levels of tin pollution and can be reused.
Bergbreiter and Walker introduced a catalytic tin halide polymer that reduced
alkyl and aryl bromides and iodides when combined with NaBH4 and catalytic
quantities of a crown ether [210]. Blanton and Salley extended this methodology by
attachment of both the crown ether and tin chloride to the same lightly crosslinked
polymer [211]. Although this polymeric co-catalyst showed lower activity than the
soluble catalyst controls, it showed a marked increase in activity (48%) over con-
trols with one catalyst supported and the other in solution. It appears that this
lightly crosslinked polymer was suciently mobile to allow the two catalysts to in-
teract. More recently, Deleuze and coworkers introduced an insoluble maleimide-
based polymer for catalytic tin reductions [212]. The reduction of 1-bromoadaman-
tane was successfully demonstrated, but the high reaction temperature required
(95 C) caused signicant leaching of tin.
15.3.3
Unsupported Reagents Using Catch-and-release Purication
15.3.4
Fluorous Chemistry
All of the reductions described above required that either the substrate or the re-
agent be attached to a polymer support; however, a new method is emerging that
allows both reactants to remain in solution which takes advantage of the fact that
highly uorinated reagents are immiscible in both standard organic and aqueous
phases at ambient temperature, yet are miscible in organic solvents at elevated
temperatures. This solubility prole simplies product isolation and purication
by making it possible to separate products from byproducts by straightforward ex-
tractive work-ups. Curran and coworkers has shown that uorous chemistry is
ideally suited to carry out tin-based reductions [215]. They demonstrated that per-
uorinated tin reagents can reduce a number of functional groups including sele-
nides, alkyl halides, nitro groups, xanthates, and aldehydes. Alkyl halides have
been reduced with both stoichiometric and catalytic quantities of the tin hydride
reagent using NaCNBH3 as the stoichiometric reductant in the latter case. Curran
References 431
and coworkers also described reductive additions and cyclizations of alkyl and aryl
halides to alkenes.
15.4
Conclusions
Reductions have been of enormous synthetic utility in both supported and unsup-
ported combinatorial applications. While solid-phase organic synthesis has pro-
vided many examples of reductions over the last few decades, the area of solution-
phase combinatorial synthesis has emerged and has grown rapidly more recently.
As the introduction of new solid-supported reagents and catalysts continues, the
ability of those involved in the drug discovery process to both generate and opti-
mize lead compounds should increase.
References
16
Cycloadditions in Combinatorial and
Solid-phase Synthesis
Markus Albers and Thorsten Meyer
16.1
Introduction
Cycloadditions are one of the most ecient reactions for the synthesis of isocyclic
and heterocyclic compounds in organic chemistry. The two most widespread are
the DielsAlder reaction ([4 2]) and the 1,3-dipolar cycloaddition ([3 2]). The
DielsAlder reaction is not only suitable for synthesis of carbocyclic but also for
N- or O-heterocyclic six-member rings (hetero-DielsAlder). The 1,3-dipolar cyclo-
additions are often used to synthesize ve-member aza- or azoxaheterocycles. In
addition to these, there is a whole string of other, less common cycloadditions
such as the [2 2] cycloaddition for the formation of four-member rings or the
[6 3] cycloaddition. The reactions usually proceed smoothly and only in some
cases is a moderate application of heat required. Owing to the pericyclic mecha-
nism two bonds are formed simultaneously in a usually stereo- and regiospecic
way, which can be promoted by Lewis acid catalysis [1]. Therefore, cycloadditions
are extremely valuable for the generation of stereogenic centers, especially during
natural product synthesis.
In this chapter the application of cycloadditions to combinatorial chemistry is
discussed. Almost all concepts and strategies are based on solid phase, with only a
few based on solution-phase chemistry. One reason behind this is the requirement
for full conversions. Transformations on solid support have signicant advantages
over those in solution. Most important is the simple removal of nonresin-bound
byproducts and excess of reagents, which is necessary to allow completion of the
cycloaddition within a reasonable time, by simply washing the resin with an appro-
priate solvent. In this way the products have high purity without a time-consuming
and expensive chromatographic purication. Solution-phase combinatorial chem-
istry is only applicable if the excess component is volatile, which allows its removal
in vacuo, or it carries an additional functional group, such as an amine, alcohol,
ketone, or aldehyde, so that scavengers can be used for its capture.
Many types of cycloadditions have been applied to solid phase, including the use
of a plethora of dierent dienes, dienophiles, dipoles, dipolarophiles, or olens.
These are either resin-bound or used as reagents, and are inter- or intramolecular
16.2
[4B2] Cycloadditions
The DielsAlder reaction is the most synthetically ecient method for the prepa-
ration of six-member rings and is a well-established reaction in combinatorial
chemistry. The typical [4 2] cycloaddition is a DielsAlder reaction with normal
electron demand between an electron-rich diene and an electron-poor dienophile.
There have been many examples of this reaction, especially in solid-phase synthe-
sis [2].
16.2.1
DielsAlder Reaction with Resin-bound Dienes
a Stille coupling with vinyl stannanes generates the diene for the DielsAlder
reaction with dierent dienophiles, which proceeds with complete facial selectivity
and endoselectivity. The high facial selectivity is rationalized by the sterically de-
manding ketal group which is eciently shielding the b-face of the molecule. In
sum, 16 cycloadducts are synthesized in high yields and purities.
Another example of an enantioselective cycloaddition is the reaction between
polymer-supported chiral amino furans and a variety of dienophiles, which can be
applied to the formation of sugar derivatives in natural product synthesis (Scheme
16.7) [10]. In order to generate the resin-bound furans, silylchloride resin [11] is
treated with the potassium enolate of the optically pure amino furanone. Reaction
with methyl acrylate provides the oxabicycloheptene adduct in a regio- and stereo-
selective way and, because of the steric demand of the pyrrolidine residue, the di-
enophile reacts at the less hampered a-face. Cleavage o the resin with tetrabutyl-
ammoniumuoride resumes the synthesis of a-substituted cyclohexenones.
16.2.2
DielsAlder Reaction with Resin-bound Dienophiles
with high regioselectivities. The observed ortho/meta ratio is in accordance with the
results obtained in solution phase. It is well documented that 1-substituted-1,3-
butadienes react with 1-substituted alkenes containing an electron-withdrawing
group, forming predominantly the ortho adduct.
Polymer-supported dehydroalanine derivatives are used as dienophiles in Diels
Alder reactions (Scheme 16.10) [14]. The dehydroalanines are generated by
coupling of N- and S-protected cysteines to the resin, oxidation of the suldes
to the sulfones, and elimination to the desired olens. After cycloaddition with
cyclopentadiene at 80 C and cleavage with 20% triuoroacetic acid in dichloro-
methane, mixtures of endo/exo isomers (1:2 to 1:4, 5181% yield) are detected
(determined by 1 H-NMR). The selectivities are similar to those reported in solu-
tion phase.
Besides Lewis acids, the application of high pressure also facilitates the cy-
cloaddition reaction. Especially highly substituted and unreactive starting materi-
als, which show no reaction under normal conditions, can be forced to react. In a
microwave-assisted Knoevenagel reaction between resin-bound nitroalkenes and
aldehydes, E/Z mixtures of trisubstituted dienophiles are generated (Scheme
16.12) [16]. Treatment with 2,3-dimethylbutadiene under high pressure conditions
(15 kbar, 25 C) yields the cycloadducts, whereas stereoselective reduction with
lithium aluminumhydride gives the cyclic amines via a traceless linker strategy.
The stereoselective formation of one diastereomer from a diastereomeric mixture
of two cycloadducts (from a E/Z mixture of nitroalkenes) is rationalized by an
aci-nitro intermediate.
16.2.3
Intramolecular DielsAlder Reaction on Solid Support
In order to increase the stereoselectivity, novel unsaturated amino acids are used
as dienophiles (Scheme 16.15) [19]. Phosphonoacetyl Wang resin [20] is hence
treated with an optically pure uorenylmethoxycarbonyl (Fmoc)-protected amino
acid aldehyde, forming the electron-decient dienophile with the intention of in-
16.2 [4 2] Cycloadditions 449
16.2.4
Hetero-DielsAlder Reaction on Solid Support
The aza-DielsAlder reaction allows one of the most convenient and versatile
approaches to nitrogen-containing six-member heterocycles. In a typical reaction
imines act as dienophiles and react with dienes to the precious cycloadducts. In
most cases the imines can be prepared in situ from amines and aldehydes, and
even simply mixing all three components (amine, aldehyde, and olen) together
16.2 [4 2] Cycloadditions 451
gives good results using Lewis acids for catalysis. A very short synthesis is the one-
pot reaction between amino-methylated polystyrene resin, aldehydes, and diene
catalyzed by ytterbium triate (Scheme 16.19) [23]. The cycloaddition proceeds
smoothly at room temperature over 1248 h and yields and purities are >90%.
The cycloadducts are released from the resin utilizing a tertiary amine N-dealkyla-
tion method which involves chloroethyl chloroformate treatment and methanolic
decomposition of the resulting carbamates [24]. Only the desired [4 2] products
are released from the solid support, enabling a clean resin cleavage.
16.2.5
DielsAlder Reaction in Solution Phase
It is true that transformations on solid support have some advantages over those in
solution, such as the use of an excess of reagents and the ease of removal of non-
resin-bound byproducts; in contrast, solution-phase chemistry often requires min-
imal investment of time during method development, has feasible scale-up, has
easy reaction monitoring, and no attachment points are required. A variety of imag-
452 16 Cycloadditions in Combinatorial and Solid-phase Synthesis
inative techniques have been developed to rapidly purify the many reactions of a
solution-phase library in a parallel fashion. Some of these techniques include acid/
base extraction [27], uorous-phase extraction [28], polymer-supported reagents
[29] and catalysts [30], solid-phase extraction [31], or polymer-supported quench/
scavenging reagents [32].
The reaction with imines and dienes is well suited to solution-phase combinato-
rial chemistry using the scavenger methodology. Scheme 16.21 outlines the cyclo-
addition to 2,3-dihydro-4-pyridones under Lewis acid catalysis [33]. An equimolar
mixture of aldehydes and primary amines in trimethylorthoformate reacts to form
the imines. After evaporation of the solvent the cycloadducts are formed using an
excess of Danishefsky diene under ytterbium triate catalysis, which nally hydro-
lyzes both the cycloadducts to the desired pyridones and the excess of the diene to
the corresponding ketone. The diene decomposition product and, if the reaction
does not go to completion, any unreacted imine are removed with a polyamine
resin. After simple ltration followed by an acidic aqueous work-up dihydropyri-
done products are obtained with good yields (up to 90%) and high purities (80
90%). A variety of dierent imines, derived from alkyl, alkylaryl, pyridyl amines,
and from substituted anilines, undergo the cyclization.
16.3
[3B2] Cycloadditions
16.3.1
Formation of Isoxazoles, Isoxazolines, and Isoxazolidines
nitrones and olens (Scheme 16.23). As nitrile oxides often suer from decom-
position and dimerization in solution [38], these transformations should be carried
out on solid phase on which either the dipolarophile or the 1,3-dipole can be
immobilized.
During the synthesis of a natural product-like library (see also Chapter 21), an
intramolecular cycloaddition is used as the key step in building up a polycyclic
template [43]. This product is formed by a Tamura tandem reaction [44] of a poly-
mer-bound epoxycyclohexanol and a set of nitrone carboxylic acids (Scheme 16.26).
After initial coupling of the 1,3-dipoles to a shikimic acid-derived alcohol, the sub-
sequent [3 2] cycloaddition proceeds with high stereo- and regioselectivity. A
variety of reagents and conditions have been screened for further manipula-
tions of the tetracyclic core thus formed and a split-and-mix library of more than 2
million compounds has been synthesized.
On the other hand, when nitrones are prepared in situ from 2-bromobenzalde-
hyde and methyl hydroxylamine and consequently reacted with polymer-bound
acrylates through a nontethered transition state (Scheme 16.27), they were found
to be less suitable for solid-phase combinatorial synthesis [45]. The yields recorded
are in the range of 2445% and even boosting the excess of reagent up to 40 equiv.
and extending the reaction times does not improve the results. In the latter case,
cleavage of the acids from the 2-chlorotrityl resin is observed owing to the pro-
longed exposure to heat.
The isoxazolidines thus obtained are then derivatized further and consequently
converted to their corresponding isoxazolines by oxidative cleavage using 2,3-
dichloro-5,6-dicyanobenzoquinone (DDQ). During the initial work-up, ascorbic
16.3 [3 2] Cycloadditions 457
acid is added in order to reduce the amount of remaining oxidant, but chromato-
graphy on silica gel is still found to be necessary.
Isoxazolidines can also be prepared from immobilized nitrile oxides, which are
easily generated through oxidation of polymer-bound aldoximes. In one case, N-
chlorosuccinimide (NCS) was used as the oxidant and the corresponding hydrox-
imoyl chlorides were converted to the reactive species by the dropwise addition of
triethylamine (Scheme 16.30) [47]. Immediate trapping with an excess of olens
gave the desired heterocycles with yields of 6080% and purities of >90%. Iso-
xazoles can also be prepared using this methodology.
In another example, the use of an additional amine base can be avoided when
a polymer-bound aldoxime is oxidized with commercially available household
bleach. After elimination of hydrogen chloride, the corresponding nitrile oxides are
obtained [48]. The generality and ease of this protocol is demonstrated when the
inverse strategy is pursued and resin-bound acrylates are successfully converted to
isoxazolines.
Solid-supported reagents can also be used for the in situ preparation of nitrones
with regard to the solution-phase synthesis of isoxazolidines (Scheme 16.31) [49].
This process has been carried out using polymer-supported perruthenate (PPS)
as the oxidant, but this procedure is limited to symmetrical hydroxylamines only.
In order to circumvent this limitation, aldehydes are normally condensed with
primary hydroxylamines in the presence of solid-supported acetate. After removal
of the polymer-bound reagent and transfer of the crude nitrone to a solution of
methyl acrylate, the desired cycloaddition product is isolated with an 81% yield.
16.3.2
Formation of Pyrrolidines
aldehyde, do not yield any products. There has been no success using silver(I)
acetate as an additive the catalyst most often used in imine cycloaddition reac-
tions [54].
Other examples that make use of silver salts include the silver acetate-induced
cycloaddition of tryptophan-derived imines to polymer-bound acrylates (Scheme
16.35) [55] and synthetic eorts toward a split-and-mix library of mercaptoacyl
proline-based inhibitors of angiotensin-converting enzyme (ACE) [56]. Silver ni-
trate has also been used in the synthesis of fully substituted prolines derived from
histidine precursors [57], while intramolecular cycloadditions have yielded poly-
cyclic cores when both the imine and the enone are immobilized on solid support
[58]. It is worth mentioning that, during the syntheses of hydantoin-containing
16.3.3
Formation of Furans
16.3.4
Formation of Imidazoles, Pyrroles, Pyrazoles, and Other Nitrogen-containing
Heterocycles
nchnone intermediates.
Scheme 16.38. Formation of imidazoles via mu
nchnone intermediates.
Scheme 16.39. Formation of pyrroles via mu
462 16 Cycloadditions in Combinatorial and Solid-phase Synthesis
acidic cleavage with TFA, a phenomenon also observed during the transfer of the
Tsuge reaction to solid-phase chemistry [71].
The formation of the open-chain pyridinium salts is suppressed through oxi-
dation of the bicyclic core with the bimetallic complex TPCD [Co(pyridine)4 -
(HCrO4 )2 ]. After treatment with TFA, aromatic indolizines are obtained and an
exploratory library of nine members has been prepared.
16.4
[2B2] Cycloadditions
The thus formed highly functionalized four-member ring heterocycles are also
valuable precursors for further chemical manipulations, particularly, when the b-
lactam strain is used to facilitate ring-opening reactions. A striking example of b-
lactams as versatile intermediates was given en route to a split-and-mix library of
4140 dihydroquinolinones (Scheme 16.43) [74]. Here, the nitro group of a [2 2]
cycloadduct is reduced and used as an internal nucleophile for the ring expansion
reactions.
Another method for the solid-phase preparation of b-lactams from imines in-
volves titanium ester enolates derived from 2-pyridinethiols (Scheme 16.44) [75].
16.5
[6B3] Cycloadditions on Solid Support
16.6
Rearrangements
Scheme 16.48. Claisen rearrangement on silica gel and mesoporous molecular sieves.
References 467
References
1 a) J. Sauer, R. Sustmann, Angew. 14 B. A. Burkett, C. L. L. Chai,
Chem. 1980, 92, 773801; b) W. C. Tetrahedron Lett. 1999, 40, 70357038.
Herndon, Chem. Rev. 1972, 72, 157 15 J. D. Winkler, W. McCoull,
179; c) K. N. Houk, A. P. Marchand, Tetrahedron Lett. 1998, 39, 49354936.
R. E. Lehr, Pericyclic Reactions, Vol. 2, 16 G. J. Kuster, H. W. Scheeren,
Academic Press, New York 1977; d) P. Tetrahedron Lett. 2000, 41, 515519.
V. Alston, R. M. Ottenbrite, J. Org. 17 J. D. Winkler, Y.-S. Kwak, J. Org.
Chem. 1975, 40, 11111116; e) B. M. Chem. 1998, 63, 86348635.
Trost, J. Ippen, W. C. Vladuchik, J. 18 K. Paulvannan, T. Chen, J. W.
Am. Chem. Soc. 1977, 99, 81168118. Jacobs, Synlett, 1999, 10, 16091611.
2 a) R. E. Sammelson, M. J. Kurth, 19 a) S. Sun, W. V. Murray, J. Org.
Chem. Rev. 2001, 101, 137202; b) S. Chem. 1999, 64, 59415945; b) S. Sun,
Booth, Tetrahedron, 1998, 54, 15385 I. J. Turchi, D. Xu, W. V. Murray, J.
15443; c) B. A. Bunin, The Org. Chem. 2000, 65, 25552559.
Combinatorial Index, Academic Press, 20 P. Wipf, T. C. Henninger, J. Org.
New York 1998. Chem. 1997, 62, 15861588.
3 E. M. Smith, Tetrahedron Lett. 1999, 21 L. F. Tietze, T. Hippe, A. Steinmetz,
40, 32853288. Synlett, 1996, 10431044.
4 M. Crawshaw, N. W. Hird, K. Irie, 22 S. Leconte, G. Dujardin, E. Brown,
K. Nagai, Tetrahedron Lett. 1997, 40, Eur. J. Org. Chem. 2000, 65, 639
71157118. 643.
5 J. Barluenga, F. Aznar, M.-P. Cabal, 23 W. Zhang, W. Xie, J. Fang, P. G.
C. Valdes, J. Chem. Soc. Perkin Trans. Wang, Tetraderon Lett. 1999, 40, 7929
1 1990, 633. 7933.
6 C. Chen, B. Munoz, Tetrahedron Lett. 24 P. Conti, D. Demont, J. Cals, C. J.
1999, 40, 34913494. Ottenheijm, D. Leysen, Tetrahedron
7 C. P. Ball, A. G. M. Barrett, A. Lett. 1997, 38, 2915.
Commercon, D. Compere, C. Kuhn, 25 C. O. Ogbu, M. N. Quabar, P. D.
R. S. Roberts, M. L. Smith, O. Boatman, J. Urban, J. P. Meara,
Venier, Chem. Commun. 1998, 2019 M. D. Ferguson, J. Tulinsky, C. Lum,
2020. S. Babu, M. A. Blaskovich, Bioorg.
8 a) D. A. Heerding, D. T. Takata, C. Med. Chem. Lett. 1998, 8, 23212326.
Kwon, W. F. Huffman, J. Samanen, 26 A. M. Boldi, C. R. Johnson, H. O.
Tetrahedron Lett. 1998, 39, 68156818; Eissa, Tetrahedron Lett. 1999, 40, 619
b) S. C. Schurer, S. Blechert, 622.
Synlett, 1999, 12, 18791882. 27 a) D. L. Boger, W. Chai, R. S. Ozer,
9 S. Wendeborn, A. de Mesmaeker, C. Andersson, Bioorg. Med. Chem.
W. K.-D. Brill, Synlett, 1998, 865868. Lett. 1997, 7, 463466; b) D. L. Boger,
10 R. H. Schlessinger, C. P. C. Tarby, P. L. Myers, L. H.
Bergstrom, Tetrahedron Lett. 1996, 37, Caporale, J. Am. Chem. Soc. 1996,
21332136. 118, 21092112; c) S. Cheng, D. D.
11 a) M. J. Farrall, J. M. J. Frechet, Comer, J. P. Williams, P. L. Myers,
J. Org. Chem. 1976, 41, 3877; D. L. Boger, J. Am. Chem. Soc. 1996,
b) J. T. Randolf, K. F. McClure, 118, 25672570.
S. J. Danishefsky, J. Am. Chem. Soc. 28 a) A. Studer, D. P. Curran,
1995, 117, 5712. Tetrahedron 1997, 53, 66816688;
12 L. Blanco, R. Bloch, E. Bugnet, b) A. Studer, S. Hadid, R. Ferritto,
S. Deloisy, Tetrahedron Lett. 2000, 41, S. Kim, P. Jeger, P. Wipf, D. P.
78757878. Curran, Science 1997, 275, 823830.
13 V. Yedida, C. C. Leznoff, Can. J. 29 a) J. J. Parlow, Tetrahedron Lett. 1996,
Chem. 1980, 58, 11441150. 37, 52575261; b) J. J. Parlow, J. E.
468 16 Cycloadditions in Combinatorial and Solid-phase Synthesis
17
Main Group Organometallics
Christopher Kallus
17.1
Introduction
solutions to this enormously challenging eld have been presented and will be re-
viewed in this chapter.
CC bonds are formed by the attack of metalated carbon nucleophiles onto
electrophiles such as alkyl and aryl halides, carbonyl groups such as aldehydes,
ketones, and their heteroanalogs, as well as carbonic acid derivatives. Among meta-
lated species, organomagnesium and -lithium reagents are most commonly used
in combinatorial chemistry, whereas cuprates have drawn less attention. Only oc-
casionally and in special cases have organoaluminum, -boron, or -indium species
been employed.
Since the chemistry of the dierent metalated species is quite often similar, the
structure of this chapter is not guided by the various main group elements. In-
stead, it is organized by the type of chemical transformation, thus avoiding double
citations of the literature examples, where dierent organometallics are used for
the same transformation. Rather than dividing the subject into a strictly mecha-
nistic sense, the order is completely practical, following the synthetic intention
focused on a diversiable substrate (organonucleophile or electrophile) which can
be modied in a combinatorial sense. For instance, metalated aromatics are used
to make structurally diverse aromatic compounds by reaction with dierent types
of electrophiles while resin-bound ketones give diverse alcohols when reacted with
organometallics.
17.2
Reactions of Metalated Aromatics
provides access to an attack on the position between the heteroatom and the hy-
droxymethyl group [10].
Solid-supported phenyl lithium and thiophenyl lithium serve as versatile pre-
cursors to the preparation of the corresponding resin-bound aryl isopropylsquarene
by reaction with diisopropyl squarate. The reactive intermediates provide a reaction
platform to generate several completely dierent cores, the so-called Multiple Core
Structure Libraries (MCSLs). From various possible compound shapes, quinones,
hydroquinones, and vinylogous amines derived from arylsquarenes have been re-
alized in small libraries. Further transformations at the squarene carbonyl group
are also possible (see below) [11].
In a similar fashion, the replacement of bromine with Grignard reagents leads
to magnesiated heterocycles as useful intermediates for further diversication.
This procedure has successfully been applied to the synthesis of several function-
alized thiophenes (Scheme 17.3). In contrast to reactions with the very polar orga-
nolithiums, the ester linkage and other functional groups are stable below 20 C
in the presence of an excess of Grignard reagent. Moreover, selective exchanges
on thiophene dibromides can be achieved at low temperatures. It should be men-
tioned that, in contrast to the alkoxymethyl thiophenyl lithium species, the rst
exchange and reaction with an electrophile takes place at the sterically more hin-
dered position between the heteroatom and the linking ester group. The direct-
ing eect of the anchoring group is based on a complexation between the magne-
sium and the ester group. The reaction is typically performed in the presence of
CuCN2LiCl in THF [12, 13].
Furthermore, organozinc bromides serve as building blocks for the solid-phase
synthesis of substituted aromatics. Since this reaction is catalyzed by Pd, it will be
discussed in Chapter 19.
474 17 Main Group Organometallics
17.3
1,2-Additions to C X Groups
17.3.1
Reactions with Aldehydes
13
by elemental analysis of sulfur and characteristic gel-phase C-NMR signals of the
dithiane methylenes (Scheme 17.8).
17.3.2
Reaction with Ketones
For the preparation of tertiary alcohols from aryl squarenes in MCSLs, two
strategies using phenyl lithium reagents have been considered [23]. One involves
the use of a polymer-bound aryl lithium species, generated from an aryl bromide
with n-BuLi, which was reacted with alkyl 4-isopropoxysquarenes to yield benzo-
fused quinones after thermal cyclization and cleavage from the resin (Scheme
17.13, see also Section 17.2). The other pathway starts with a resin-bound squarene
which is then converted to a tertiary alcohol with phenyl lithium. If the hydroxyl
17.3 1,2-Additions to CbX Groups 479
17.3.3
Reaction with Imines
at elevated temperatures. The scope of the solid-phase version of this reaction has
been demonstrated on aldimines prepared from the amino group of Rink resin
and aldehydes making use of the tea-bag method [25] (Scheme 17.15). In prin-
ciple, aldimines derived from aromatic aldehydes work well, independent of their
electronic nature. The polymer support functions as a suitable NH-protecting
group, while other masked imines such as sulfenimines, sulfonimines, and TMS-
protected imines were not applicable in solid-phase chemistry. The cleavage of the
products thus formed from the resin yielded primary amines. Similarly, Schi
bases from polymer-bound benzaldehydes were reacted with allyl magnesium bro-
mide, providing secondary a-allyl amines [26].
17.3.4
Reaction with Enolates
Scheme 17.18. a-Alkylated ketones by reaction of TMS enolates with alkynyl lithium reagent.
17.4
Conjugate Addition to a,b-Unsaturated Carbonyls and Related Systems
Because of their particularly soft character, cuprates are the reagents of choice for
conjugate additions, but only a few examples of this reaction type in solid-phase
syntheses or combinatorial chemistry have yet been described. An application has
been demonstrated by the 1,4-addition to resin-bound dihydropyridones (obtained
from a Grignard reaction, see above) [27]. Cuprates have been prepared in a clas-
sical manner by the reaction of copper(I) iodide and Grignard reagents in the
presence of borontriuoride etherate (Scheme 17.19).
this reaction can be prepared in situ from readily accessible terminal alkynes by
hydrozirconation or stannylation followed by transmetalation using Lipshutzs or
Babiaks protocol [35, 36].
The synthesis of b-substituted esters can be performed in a similar manner.
High enantioselectivities have been obtained in addition reactions of Me 2 CuLi or
Ph2 CuMgBr to chiral unsaturated esters immobilized on Wang Resin [37]. The
resulting trisubstituted hydroxy ester cyclizes upon cleavage from the resin form-
ing d-lactones in excellent yields and enantiopurities (Scheme 17.21).
17.5
Nucleophilic Substitutions
Grignard reagents can also serve as agents for functionalizing cleavage products
synthesized on the benzotriazole linker [39]. In an extremely short reaction se-
quence, consisting of resin capture of aldimines in a Mannich reaction and subse-
484 17 Main Group Organometallics
17.6
Reactions on Carboxylic Acid Derivatives and Related Systems
17.6.1
Reaction with Esters
In many examples, Grignard reagents have been applied for the release of prod-
ucts from the solid support. During the solid-phase synthesis of a PDE4 inhibitor,
MeMgBr was used to cleave the ester linkage leading to a dimethylphenyl alcohol
after treatment with ammonium chloride in ethyl acetate (Scheme 17.26) [42].
Scheme 17.26. Tertiary alcohols by cleavage of ester linkage with Grignard species.
17.6.2
Reactions with Weinreb Amides and Related Systems
For the synthesis of ketone or aldehyde libraries, the use of Weinreb amides as
anchoring groups is preferable. Although additional eorts during the work-up are
necessary, this strategy is advantageous because the reactive keto group is released
only at the stage of the cleavage step. In this context, the N-alkyoxyamide function
of b-alanin Weinreb amide on aminomethyl polystyrene has been used as a linker
group [43]. Since preformation of the linker conjugate with the substrate has to be
carried out in solution phase, this methodology appears to be somewhat laborious
(Scheme 17.28).
by a Weinreb amide linker [44]. The cleavage was carried out using EtMgBr to
provide ethyl ketones. Another successful method for the preparation of ketone
libraries on solid phase uses mercaptoacetylamide on polystyrene resin, which acts
in the same way as a Weinreb amide [45]. The starting carboxylic acid building
blocks were coupled as thioamides onto this linker. After treatment with Grignard
reagents, analytically pure products were released from the resin.
The combination of a Weinreb amide and an isonitrile functionality in one sin-
gle synthetic building block allows the preparation of a dipeptide methyl ketone
library by application of methyl Grignard reagent after the Ugi 4CC reaction
(Scheme 17.29). A 96-member library around a known anticonvulsant structure
has been synthesized according to this strategy [46].
hydrazones with SOCl2 . The tetrahedral intermediates from the Grignard reaction
were decomposed by addition of macroporous polystyrene-sulfonic acid resin (MP-
TsOH) and the ketones thus produced were captured by gentle heating with PS-
TsNHNH2 in acetic acid/THF (Scheme 17.32).
17.7
Aminolysis of Esters
Scheme 17.33. Aminolysis of ester bonds with organoaluminium reagents on solid support.
490 17 Main Group Organometallics
The ltrates must be carefully quenched and additional work-up was found to
be necessary. This reaction type has also been used for the simultaneous lactone
opening and amidation on solid support [37].
References
18
Enolates and Related Species in Combinatorial
and Solid-phase Synthesis
Jochen Kruger
18.1
Introduction
18.2
Aldol Reactions
18.2.1
General Aspects
The direct condensation between an aldehyde and a CaH acidic ketone was suc-
cessfully employed in solid-phase synthesis and library construction (see Section
18.2.2). In particular, Na and Zn enolates have played a dominant role in this eld.
Although the basic reaction conditions entail limitations arising from the base
sensitivity of some polymeric linkers or starting materials, this method presents a
powerful tool to perform aldol reactions on solid support.
For stereoselective aldol protocols, we discuss recent progress in the application
of boron enolates in solid-phase chemistry (see Section 18.2.3). In general, the
eciency of a boron aldol process on solid support is highly dependent on the
nature of the polymer, the polymeric linker, and the reactivity of the enolate. Thus,
for a given synthetic problem in solid-phase chemistry, the boron aldol protocol
has to be adjusted to the requirements of the desired transformation. To the best
of our knowledge, multicomponent library synthesis utilizing stereoselective boron
aldol reactions has not yet been reported.
In Section 18.2.4 we detail the latest achievements in the eld of liquid- and
solid-phase Mukaiyama aldol chemistry. This mild version of the aldol reaction can
serve as an attractive alternative for the direct aldol condensation employing eno-
lizable carbonyl compounds.
18.2.2
Li, Na, K, and Zn Enolates in Aldol Reactions
The synthesis of b-hydroxy carbonyls via a crossed aldol reaction between an alde-
hyde and an enolizable carbonyl compound has been adopted to solid phase,
whereas liquid-phase protocols in this eld are rather scarce.
Ruhland and Kunzer described the formation of aldol intermediates (2) upon
reaction of resin-bound aldehyde 1 with acetophenone in the presence of K2 CO3 at
elevated temperature (Scheme 18.1) [2]. Six aromatic and heteroaromatic methyl-
ketones were reported to be suitable substrates for this transformation, however,
yields and purities were not given. The intermediates (2) were used in the synthe-
sis of a diverse quinoline library of 12 products. Thereby, the eectiveness of the
aldol transformation as a key step in this sequence was demonstrated.
A related strategy was applied by Nicolaou et al. for the synthesis of an epothi-
lone library [4]. A key step in this route leading to diversity with concomitant con-
struction of the CaC skeleton was an aldol reaction mediated by in situ-generated
Zn enolates (see Scheme 21.14). Both contributions demonstrate that even
more sophisticated reaction conditions including inert gas techniques and low-
temperature protocols are routinely applied in solid-phase synthesis today. The
only drawback in terms of manipulative convenience arose from the fact that the
products were puried by preparative TLC after cleavage. This aspect might be
dicult to realize in an automated library synthesis process.
The synthesis of a,b-unsaturated ketones using a crossed aldol strategy on solid
support is well established and has successfully been applied to solid-phase or-
ganic synthesis. Either the aldehyde moiety or the CaH acidic ketone can be linked
to the polymer via an ether or an amide linkage, whereas linkage via an ester
proved to be incompatible with the strongly basic reaction conditions.
In general, a solution of NaOMe as a solution in methanol is routinely used as a
standard base to promote the desired condensation event. A typical example of this
reaction is illustrated in Scheme 18.3. Here, the immobilized acetophenone (5) was
treated with 12 equiv. of NaOMe (0.5 M solution in MeOH) and 12 equiv. of aro-
matic aldehydes at room temperature to yield enones (6) (ve examples) [5]. The
reaction was conducted in tetrahydrofuran (THF) to eect sucient swelling of
the polymer. Alternatively, trimethylorthoformate (TMOF), which facilitates the
dehydration step, was used as the solvent instead [6]. This type of aldol condensa-
18.2 Aldol Reactions 495
tion of ketones has also been carried out employing Li enolates which were gen-
erated in situ using an excess of LiOHH2 O (20 equiv.) in dioxane [7] or dime-
thoxyethane (DME) [8] as solvent.
In conclusion, reliable aldol protocols for the solid-phase synthesis of a,b-unsat-
urated ketones are available; however, these processes are primarily restricted to
the condensation of methyl-arylketones with aromatic aldehydes, thus limiting the
options for diversication in library synthesis.
18.2.3
Boron Enolates in Aldol Reactions
A). The resulting target molecules were cleaved from the resin using HF-pyridine
to yield the anti-congurated diols (12) after ash chromatography with high levels
of diastereoselectivity (ve examples: 9097% de) and good yields. Complementary
to this, the syn-congurated product (14) could be accessed using the (Z)-enolate
(13) (Scheme 18.5. path B). To achieve complete conversion in this reaction, alde-
hyde 10 was successively incubated three times, each time with an excess of eno-
late 13. The aldol product was cleaved from the resin and the resulting diol was
isolated after chromatography with 99% yield with 95% de in favor of the syn con-
guration.
Scheme 18.5. Stereoselective synthesis of anti and syn aldols using chiral B enolates.
formed upon cleavage from the resin was obtained in poor yield. Moreover, isola-
tion of the nal product required chromatography since the crude reaction mixture
was contaminated by oxazolidinone (19). Optimal results for the solid-supported
Evans aldol protocol were elaborated for soluble polyethylene glycol (PEG)-bound
aldehyde 20. This scenario comes closest to solution-like conditions having a
favorable impact on reaction kinetics (Scheme 18.6, path C) [13]. The polymeric
aldol adducts were precipitated from solution to obtain 21 (six examples, 5688%
yield) and 1 H-NMR measurements revealed that each product was formed as a
pure stereoisomer.
The three examples depicted in Scheme 18.6 clearly demonstrate that the choice
of the polymeric linker plays a crucial role in this solid-supported aldol process.
Furthermore, the scope of these reactions is limited to aldehydes bearing addi-
tional functional groups which allow the electrophile to be anchored to the poly-
mer. After cleavage, these functional groups reside as integral parts in the target
molecules.
An alternative strategy is based on resin-bound oxazolidinone 22, which was
synthesized on polymer support using l-tyrosine as the source of chirality [14]. As
depicted in Scheme 18.7 the (Z )-congurated boron enolate was generated by re-
action of 22 with an access of Bu2 BOTf in the presence of a tertiary amine such as
triethylamine using methylene chloride as the solvent of choice.
The excess of Bu2 BOTf had to be drained away prior to the addition of the alde-
hyde since deleterious eects with respect to the stereochemical outcome were re-
ported if additional Bu2 BOTf was present in the reaction mixture [15]. After com-
pletion of the aldol reaction, the resin was subjected to an aqueous LiOH solution
498 18 Enolates and Related Species in Combinatorial and Solid-phase Synthesis
in THF to eect cleavage of carboxylic acid 23. Alternatively, NaOMe was used to
form the corresponding methyl ester instead. Finally, product 23 was obtained as
the optically pure syn diastereomer with a chemical purity of 94%, as concluded
from high-performance liquid chromatography (HPLC).
In summary, a number of solution-phase protocols for the addition of boron
enolates to aldehydes have been successfully adapted to solid-phase synthesis.
These reactions proceed with similar eciency to the well-established solution-
phase reactions. However, a relatively small number of examples have been
reported to date, rendering an assessment of the scope of solid-supported boron
aldol reactions exceedingly dicult. Moreover, product isolation is routinely
achieved by aqueous work-up in combination with chromatographic methods. This
aspect could certainly hamper the application of such a protocol for automated
synthesis and the construction of multicomponent libraries still remains a chal-
lenging goal.
18.2.4
The Mukaiyama Aldol Reaction
Entry Silyl enol ether Electrophile Catalyst Product Conditions and comments Reference
R4 aromatic
and aliphatic
Entry Silyl enol ether Electrophile Catalyst Product Conditions and comments Reference
The Mukaiyama protocol has also been successfully conducted with immobi-
lized aldehydes. As depicted in Scheme 18.10, cyclic TMS enolether 33 was reacted
with aldehyde 32 mediated by Yb(OTf )3 to yield a,b-unsaturated ketone 34 after
cleavage in 91% yield [29]. The salient features of this process include the use of a
MeCN/H2 O mixture as the solvent for this reaction. Since conventional poly-
styrene resins were not compatible with these hydrophilic conditions, a resin
based on a polar polyoxyethylene/polyoxypropylene matrix was introduced for this
reaction.
18.3
1,4-Addition of Enolates to Michael Acceptors
Scheme 18.11. Enolate additions to Michael acceptors mediated by a basic exchange resin.
More recently, Ley and Massi reported a tandem Michael protocol for solution-
phase combinatorial synthesis leading to the bicyclo[2.2.2]octanone skeleton 40
that was further elaborated in subsequent amination reactions [33]. The bicyclic
scaold was constructed upon reaction of 2 equiv. of conjugated dianions (39) with
tert-butyl acrylates (38). As outlined in Scheme 18.12, the reaction mixture was
quenched with Amberlyst 15 (path A) and a simple ltration process provided
pure endo products (40) (with R2 OEt or OiBu), whereas the crude product
(with R2 methyl) was contaminated by hydrolyzed 3-methyl-2-cyclohexen-1-one.
Therefore, this mixture was treated with polymer-supported thiophenol and diiso-
propylethylamine (DIPEA) to scavenge the excess of hydrolyzed starting material
(39c) followed by neutralization of ammonium salts using MP carbonate (path B).
After ltration the desired products (40) were nally isolated in pure form. These
carefully orchestrated reaction and work-up conditions formed the basis for a rapid
synthesis of a larger library, since the reaction protocol can easily be automated
and avoids tedious chromatography.
The synthesis of the bicyclo[2.2.2]octanone framework via a tandem Michael re-
action has also been adapted to solid phase [34].
The 1,4-addition of enolates to Michael acceptors in solid-phase chemistry has
been examined by Domnguez et al. [35]. These authors investigated the addition
of resin-bound glycinates (41) to various a,b-unsaturated acceptors (Scheme 18.13).
The standard conditions for typical Michael acceptors such as acrylates, include 3
equiv. of Schwesinger-type base BEMP and 5 equiv. of acceptor 42 in NMP at room
temperature (rt) for 16 h to yield the Michael adducts 43 in high purity. For more
reactive acceptors (e.g. acrylonitriles) only 1.5 equiv. of 42 were used to avoid dou-
ble addition, while less reactive electrophiles 42 (e.g. substituted acrylates) required
additional base and acceptor 42 (each 10 equiv.). In total, 15 examples are given
composing a diverse set of enolate additions to a,b-unsaturated Michael acceptors.
Additionally, the resin-bound intermediates 42 were further modied on the poly-
mer, including deprotection of the imine, acylation of the amino group with qui-
naldic acid, and cleavage using triuoroacetic acid (TFA). The nal products were
obtained in good yields (6188%) and purities (7487%), indicating that the
Michael addition as depicted in Scheme 18.13 proceeded on the resin without
incident and with high eciency.
18.4
Alkylation of Enolates
18.4.1
a-Alkylation of Carbonyl Compounds
presence of ZnCl2 -doped acidic alumina to give ketone 52 in 72% yield (Scheme
18.16) [38]. This reaction worked well with allylic, benzylic, and tertiary halides,
leading to the alkylated products after a simple extraction step with CH2 Cl2 in
good yields (6072%).
Scheme 18.16. Alkylation of silyl enol ethers using ZnCl2 -doped alumina.
The elegant studies of Ellman and coworkers toward the synthesis of arylacetic
acid libraries represented classical solid-phase synthesis [39, 40]. This concept
called for a reliable enolate-alkylation protocol that proved to be feasible for the
construction of a diverse set of target molecules (Scheme 18.17). Accordingly,
Kenner et al.s sulfonamide linker [41] was chosen to bind the aryl acetic acid pre-
cursor to solid phase to yield polymer 53. The resin was subsequently treated with
15 equiv. of LDA to generate the corresponding trianion, which was trapped by
quenching with an excess of alkylhalide to selectively yield the monoalkylated
product 54. This alkylation reaction works equally well for a range of halides, in-
cluding benzylic and aliphatic halides (ve examples). The intermediates 54 were
further manipulated to give a diverse set of arylacetic acids which were isolated in
excellent yields (88100%) and high purities after cleavage. From these results, it
could be concluded that the alkylation event further upstream in this sequence
proceeded smoothly and with high eciency.
The alkylation of amide and ester enolates derived from deprotonation of the
corresponding a-acidic precursor has been reported in solid-phase synthesis using
LiHMDS [42] and KHMDS [43], respectively.
18.4.2
a- and g-Alkylation of 1,3-Dicarbonyl Compounds
The 1,3-dicarbonyl moiety represents a versatile template for the synthesis of het-
erocycle libraries, especially if alkylation protocols can provide an additional an-
508 18 Enolates and Related Species in Combinatorial and Solid-phase Synthesis
chor for diversity. Accordingly, the development of methods which selectively tar-
get the mono- or dialkylation of a,b-dicarbonyls has attracted recent attention.
In this context, a variety of solution-phase strategies has been elaborated. Most
strategies are based on the use of polymer-supported bases such as Amberlite
anion-exchange resins [44] or immobilized uoride sources [45]. A typical example
is illustrated in Scheme 18.18 [46]. Celite coated with KF cleanly eected the
monoalkylation of acetylacetone to yield 56 in 96% yield. In contrast, the system
Al2 O3 aKF predominantly delivered the corresponding disubstituted product under
otherwise identical conditions.
Z electron-withdrawing group.
18.4 Alkylation of Enolates
509
510 18 Enolates and Related Species in Combinatorial and Solid-phase Synthesis
chemistry has been adapted to solid phase and it has been shown that resin-bound
malonates could be cleanly a-dialkylated by this method [54].
An important modication of 1,3-dicarbonyls is the g-alkylation. This has
been accomplished on solid phase commencing from resin-bound b-ketoester (63),
which was deprotonated with an excess of LDA to generate the corresponding di-
enolate (Scheme 18.19) [55].
Subsequently, the excess LDA was drained away and 25 equiv. of alkyl halide
(R1X) were added to eect mono-g-alkylation. The resulting monoenolate could
serve as starting material for a second alkylation step. Thus, treatment of the eno-
late intermediate with n-BuLi regenerated the dianion which was quenched with
R2X to yield dialkylated products (64) in good yields. The g-alkylation worked well
with aliphatic, benzylic, and allylic halides (X Br or I; nine examples).
18.4.3
Stereoselective Alkylations of Enolates
drolysis (LiOH) to liberate the carboxylic acid or, alternatively, by reduction (LiBH4 )
leading to the corresponding alcohols.
For all examples depicted in Table 18.3, convenient recycling of the chiral auxil-
iary has been demonstrated and the polymers could be reused for further applica-
tions. However, the majority of protocols required rather drastic reaction con-
ditions for the cleavage of the nal product, which implies tedious work-up and
purication procedures rendering this chemistry less amenable for automation.
Diastereoselective alkylations have also been successfully applied in the synthe-
sis of more complex molecules on solid support. For instance, Hanessian et al.
showed that polymeric ester 73 could be deprotonated by the action of KHMDS
followed by the addition of reactive electrophiles such as allylic iodides (three ex-
amples) to yield the corresponding a-alkylated product on solid support (Scheme
18.20) [63].
This resin was treated with TFA to eect cleavage and lactonization leading to
enantiopure lactones (74) which were isolated in 75% yield (using two steps). The
stereochemical outcome of this reaction was rationalized via a Felkin-type transi-
tion state model.
In Chen and Jandas elegant studies toward a resin-supported synthesis of pros-
taglandin F2a , the a-alkylation of the central cyclopentanone core represented a key
step [64].
As depicted in Scheme 18.21, the requisite enolate was generated by lithiation of
the TMS enol ether (75) using excess methyl lithium. Upon addition of triate 76
the desired product 77 was isolated in pure form. For these studies a soluble non-
crosslinked polystyrene was selected as polymer support, since this choice guaran-
teed solution-like kinetics for each step of this sequence. Additionally, the soluble
Scheme 18.21. Enolate alkylation as a key step in solid-phase synthesis of prostaglandin F2a .
18.4 Alkylation of Enolates 513
18.4.4
Alkylation of Protected Glycines
Suitably protected glycine esters are readily alkylated to yield protected a-amino
acids. Accordingly, this route oers convenient access to tailor-made unnatural
amino acids which are valuable templates in organic synthesis.
Benzophenone imines derived from glycine esters serve as the starting materials
of choice for the alkylation of glycine. This is the result of the convenient protocols
that are available for imine formation as well as the high selectivity of these imines
toward monoalkylation.
Polymer-supported reagents have been developed to produce the above-
mentioned transformation. For instance, Palacios et al. reported the use of immo-
bilized P-ylides as strong non-nucleophilic bases [65]. Moreover, a stereoselective
alkylation procedure has been elaborated previously [66].
As depicted in Scheme 18.22, protected glycine 78 was reacted with an excess of
reactive halides, such as benzyl or allylic bromide under phase transfer conditions
(seven examples). As chiral catalyst, the immobilized tertiary ammonium salt 80
was used, and for a range of products medium to good levels of induction were
observed. The products 79 were isolated after ltration and extraction into the or-
ganic layer without the need for chromatographic purication.
The alkylation of glycine Schi bases has also been established in solid-phase
chemistry (Table 18.4). The most frequently reported conditions include the use of
Schwesinger-type bases in combination with an excess of alkyl halides in polar
solvents. For instance, imine 81 immobilized on Wang resin was treated with
BEMP and alkyl halide to selectively yield the monoalkylated products 82 (Table
18.4, entry 1). These products were further modied on solid support by cleavage
of the benzophenone imine followed by acylation of the amino group. Finally, the
resulting peptides were cleaved from the resin with TFA to deliver the corre-
Tab. 18.4. Alkylation of glycine derivatives on solid support.
89% ee
2. NH2 OH
3. TFA
18.5 Claisen-type Condensations 515
sponding products in excellent yields (77100%). These results indicate that the
alkylation event at the outset of the sequence must have proceeded with high e-
ciency.
A related process was described for PEG-bound glycine 83 (Table 18.4, entry 2).
The authors showed that the PEG polymer not only simplied product isolation,
but also accelerated the reaction signicantly. This catalytic eect was attributed
to the close relationship of the PEG polymers with crown ethers, which are well
known to be benecial in phase transfer reactions.
Dialkylated amino acids were accessed starting from natural a-amino acids
which were transformed to aldimines of type 85, and subsequent alkylation with
halides (Table 18.4, entry 3) led to the expected products 86. This concept has been
validated for a broad range of amino acids (36 examples).
Finally, a stereoselective alkylation of protected glycine on polymer support has
been reported by ODonnell et al. (Table 18.4, entry 4) [71]. Optimal results were
obtained if Coreys quaternary ammonium catalyst 89 was used as the source of
chirality. The immobilized intermediates were deprotected (NH4 OH) and cleaved
from the resin (TFA) to yield the free amino acids (88) in good chemical and opti-
cal purity.
18.5
Claisen-type Condensations
A more recent contribution in this eld reported a novel selenium linker which
was used to immobilize methylketone (96). Enolate generation was eected by
LiHMDS followed by addition of acylcyanides as electrophiles to produce 1,3-
diketones (97) in good yields (Scheme 18.25) [74]. Alternatively, the use of Weinreb
amides [75] or b-ketoester [76] as electrophiles in this Claisen-type reaction have
been described.
18.6
Dieckmann Condensations
Owing to the convenient reaction protocols both the liquid- and solid-phase syn-
thesis of tetramic acids appears to be an attractive tool for multicomponent library
synthesis.
18.7
Knoevenagel Condensations
In contrast, solid-phase protocols have been far more popular for Knoevenagel
reactions since the condensation products are often important intermediates in
multistep sequences which are best supported by solid-phase chemistry. Depend-
ing on the nature of the linker, the CaH acidity of the substrate, and the structure
of the aldehyde, dierent conditions have been employed. Piperidine or piperidine
acetate have been successfully used as catalysts for the Knoevenagel reaction and
18.7 Knoevenagel Condensations 519
seem to be the methods of choice. For instance, Gallop, Bhandari, and coworkers
reported that malonate (105) immobilized on TentaGel resin cleanly reacts with
aromatic aldehydes (eight examples) under the conditions depicted in Scheme
18.29 [90, 91]. The Knoevenagel products (106) were further transformed to pyrro-
lopyridines utilizing a Hantzsch protocol for construction of the pyridine hetero-
cycle. The nal products in this sequence were obtained in decent yields (2050%)
with excellent purities (> 90%), reecting almost quantitative conversion for each
step of the synthesis. Additionally, this methodology was extended to a library syn-
thesis of 4800 compounds, thereby demonstrating the broad scope of the Knoeve-
nagel reaction. Instead of TMOF, molecular sieves were also used to remove water
from the reaction mixture [92]. Under similar reaction conditions, salicylaldehydes
were eectively condensed with resin-bound malonates to yield the coumarin
skeleton on solid support [93].
If the basic piperidine is buered by acetic acid, even milder conditions for the
Knoevenagel condensation are obtained. As illustrated in Scheme 18.30, malonates
(107) loaded onto Wang resin were reacted with a range of aromatic and aliphatic
aldehydes to yield the corresponding Knoevenagel products on solid support. After
cleavage, the carboxylic acids (108) were obtained as mixtures of E and Z isomers
in good yields (4598%) and high purities (conversion 9298%). This protocol was
further applied in a library synthesis comprising 96 products of type 108 and the
majority of products were obtained with purities > 75% as determined by HPLC
[94]. Moreover, this procedure was especially suited for the use of enolizable ali-
phatic aldehydes in the Knoevenagel condensation on solid support [95]. In this
context, it was also reported that ethylenediamine diacetate (EDDA) very eectively
catalyzed the condensation of both aromatic and aliphatic aldehydes with resin-
bound malonates [5c].
18.8
Addition of Enolates to Imines
18.8.1
Synthesis of b-Amino Esters and Alcohols via Enolate Addition to Imines
Mannich-type reactions include the addition of enolates to imines and deliver val-
uable b-amino acid derivatives. Mild reaction conditions were reported employing
silyl enol ether or ketene silyl acetals which add to preformed or in situ-gen-
erated aldimines in the presence of catalytic amounts of Lewis acids. In this re-
gard, a variety of polymeric rare earth metal salts were reported to promote the
aforementioned process eciently [96]. For instance, Kobayashi et al. introduced a
microencapsulated scandium triate which exhibited high catalytic activity in the
Mannich reaction [22a]. As outlined in Scheme 18.31, the imino-aldol reaction be-
tween aldimine 109 and propiophenone-derived silyl enol ether 110 proceeded
smoothly to produce b-aminoketone 111 in excellent yield. Furthermore, the cata-
lyst was recycled and reused up to seven times in the Mannich process without
there being any deleterious eects on the yield. These reactions work best with al-
dimines originating from aromatic or heteroaromatic aldehydes.
The solid-phase synthesis of novel a-amino acid derivatives (119) has also been
elaborated for soluble silyl enol ethers in combination with immobilized imines
[102]. A recent report described the use of polymer-supported imino acetates (118)
as suitable substrates for the Mannich reaction (Scheme 18.34) [103]. Accordingly,
the Sc(OTf )3 -catalyzed process delivered a set of unnatural aspartic acid derivatives
(119) which were cleaved from the resin by transesterication using NaOMe.
Under similar conditions polymer-supported acylhydrazones reacted with ketene
silyl acetals and the corresponding adducts underwent cyclative cleavage upon
treatment with NaOMe to yield substituted pyrazolones [104].
522 18 Enolates and Related Species in Combinatorial and Solid-phase Synthesis
18.8.2
Solid-phase Synthesis of b-Lactams via Enolate Additions to Imines
18.9
Nitro-aldol Reactions
The nitro-aldol reaction (Henry reaction) represents a useful tool for the introduc-
tion of a nitrogen functionality onto an organic scaold. The base-promoted addi-
18.9 Nitro-aldol Reactions 523
tion of CaH acidic nitroalkanes to aldehydes has been realized in solution phase
by utilizing polymer-supported reagents. For instance, immobilized 1,5,7-triaza-
bizyclo[4.4.0]dec-1-ene (TBD) has been reported to catalyze the aforementioned
process [110]. The use of polymeric TBD dramatically simplied work-up, and
product isolation was reduced to a simple ltration process. However, only a lim-
ited number of examples were reported that lacked data on experimental details
and product purity.
Other polymeric bases which catalyze the Henry reaction are polymeric ylides
[65], ion-exchange resins such as Amberlyst A21 [111], or basic La 3 polymers
[112]. Again, the scope of these methods is rather limited to a small set of exam-
ples, and an assessment regarding the suitability of these reagents for library syn-
thesis is therefore exceedingly dicult.
Polymer-supported reagents were also successfully used to promote the 1,4-
addition of nitroalkanes to Michael acceptors. Accordingly, KF on basic alumina
was found to catalyze the addition of CaH acidic nitro-compounds to a range of
a,b-unsaturated ketones and esters (Scheme 18.36) [113, 114]. The reactions were
driven to completion by using a large excess of the nitroalkane. Finally, the adducts
124 were isolated after ltration and evaporation of the excess nitroalkane in good
yields (39100% for 14 examples). The best results were obtained if the Michael
acceptors were unsubstituted in the b-position.
and the resulting Henry adduct was cleaved from the resin to produce phenol
126 with a 96% yield and with 95% purity. Practically identical conditions were re-
ported for other polymeric substrates [118].
18.10
The BaylisHillman Reaction
The mixture was maintained at rt for 2 days and the procedure was repeated to
achieve complete conversion. After acidolysis with TFA, the BaylisHillman prod-
uct 128 was obtained with a crude purity of 97% and with an 85% yield after prepa-
rative HPLC. Similar protocols have been disclosed by other groups. Kunzer and
coworkers used dimethylformamide (DMF) as solvent and 3-QDL as base to pro-
mote the BaylisHillman addition [120], while Kulkarni and Ganesan reported
that a 3:1 DMF/MeCN solvent mixture in combination with DABCO and catalytic
amounts of La(OTf )3 were benecial to the process [121]. The major limitations
for the solid-phase versions of the BaylisHillman transformation arise from in-
complete reactions and a narrow range of aldehydes that are able to react smoothly
under the aforementioned conditions. Large excesses of the reagents or prolonged
reaction times with repeated incubations did not dramatically improve the situa-
tion.
Preferably, aromatic aldehydes with electron-withdrawing substituents work well
in these reactions while less reactive aromatic or aliphatic aldehydes gave sluggish
reactions. A slightly dierent protocol for the polymer-supported BaylisHillman
reaction was described by Reiser and coworkers [122]. Herein, acrolein immobi-
lized on a soluble PEG resin via an ester linkage was reacted with a range of alde-
18.11 Miscellaneous 525
hydes using ethanol as solvent and 3-QDL as base. The reaction with a diverse
set of aldehydes (six examples), including aliphatic aldehydes, could be driven to
completion since the soluble polymer guaranteed solution-like kinetics. However,
it was also reported that the ester linkage to the PEG polymer was partly cleaved
under the reaction conditions, leading to a lower yield of the desired product.
In conclusion, the BaylisHillman reaction on solid phase is an established pro-
cess; however, owing to the inherent limitations, a careful selection of the poly-
meric linker as well as the aldehyde set used in this reaction should be addressed
prior to multicomponent library synthesis.
Finally, an interesting three-component version of the BaylisHillman reaction
was reported on solid support. Thus, resin 127 was treated with aldehydes and
sulfonamides in the presence of DABCO in dioxane at 70 C. This cocktail gave
rise to the formation of functionalized sulfonamides 129 which represent useful
intermediates for further manipulations (Scheme 18.39) [123]. The versatility of
this chemistry has been shown for aromatic and heteroaromatic aldehydes in
combination with arylsulfonamides. The best results were obtained if electron-
decient aldehydes were employed in this process.
18.11
Miscellaneous
in THF at 50 C to yield the corresponding silyl esters which were cleaved from the
resin by transesterication. The reaction worked well for a range of allylic esters,
however, esters derived from acetic acid failed to give the desired rearrangement
product (134). This result is presumably due to a large extent to C-silylation in
the enolizationimmobilization event.
An elegant approach to substituted phenols via a cyclative cleavage strategy was
reported by Katritzky and coworkers (Scheme 18.41) [125]. Merrield resin loaded
with 3-hydroxypyridine was alkylated with bromoacetone to give pyridinium salt
(135). The latter was reacted with a,b-unsaturated ketones under basic conditions
to yield Robinson anellation product 136, which aromatized with concomitant
cleavage from the resin. A range of chalcones (12 examples) were successfully uti-
lized in this reaction and the phenols 137 were isolated after aqueous work-up and
column chromatography in good yields with high purities.
References
1 a) For solid phase, see: R. E. Gerlach, Angew. Chem. Int. Ed. 2000,
Sammelson, M. J. Kurth, Chem. Rev. 112, 34533457.
2001, 101, 137202; b) for polymer- 11 M. Reggelin, V. Brenig, Tetrahedron
supported reagents, see: S. V. Ley, Lett. 1996, 37, 68516852.
I. R. Baxendale, R. N. Bream, P. S. 12 M. Reggelin, V. Brenig, R.
Jackson, A. G. Leach, D. A. Welcker, Tetrahedron Lett. 1998, 39,
Longbottom, M. Nesi, J. S. Scott, 48014804.
R. I. Storer, S. Taylor, J. Chem. Soc., 13 M. Reggelin, V. Brenig, C. Zur,
Perkin Trnas. 1, 2000, 38154195. Org. Lett. 2000, 2, 531533.
2 T. Ruhland, H. Kunzer, Tetrahedron 14 a) A. V. Purandare, S. Natarajan,
Lett. 1996, 37, 27572760. Tetrahedron Lett. 1997, 38, 87778780;
3 M. J. Kurth, L. A. Ahlberg Randall, b) C. W. Phoon, C. Abell, Tetrahedon
C. Chen, C. Melander, R. B. Miller, Lett. 1998, 39, 26552658.
J. Org. Chem. 1994, 59, 5862. 15 H. Danda, M. M. Hansen, C. H.
4 a) K. C. Nicolaou, D. Vourloumis, Heatchcock, J. Org. Chem. 1990, 55,
T. Li, J. Pastor, N. Winssinger, Y. 173181.
He, S. Ninkovic, F. Sarabia, H. 16 T. Mukaiyama, H. Iwakiri, Chem.
Vallberg, F. Roschangar, N. P. Lett. 1985, 13631366.
King, M. R. V. Finlay, P. 17 a) S. Kobayashi, R. Akiyama, T.
Giannakakou, P. Verdier-Pinard, E. Furuta, M. Moriwaki, Molecules
Hamel, Angew. Chem. 1997, 109, Online 1998, 2, 3539; b) L. Yu, D.
21812187; b) K. C. Nicolaou, N. Chen, J. Li, P. G. Wang, J. Org.
Winssinger, J. Pastor, S. Ninkovic, Chem. 1997, 62, 35753581.
F. Sarabia, Y. He, D. Vourloumis, Z. 18 M. Kawai, M. Onaka, Y. Izumi, Bull.
Yang, T. Li, P. Giannakakou, E. Chem. Soc. Jpn. 1988, 61, 12371245.
Hamel, Nature, 1997, 387, 268272. 19 T. P. Loh, X.-R. Li, Tetrahedron 1999,
5 a) S. P. Hollinshead, Tetrahedron 55, 1078910802.
Lett. 1996, 37, 91579160; b) S. P. 20 B. C. Ranu, R. Chakraborty,
Hollinshead, WO 98/08813; c) for Tetrahedron 1993, 49, 53335338.
additional examples, see: P. Grosche, 21 B. C. Ranu, M. Saha, S. Bhar,
A. Holtzel, T. B. Walk, A. W. Synthetic Commun. 1997, 27, 3065
Trautwein, G. Jung, Synthesis 1999, 3077.
11, 19611970. 22 a) S. Kobayashi, S. Nagayama, J. Am.
6 C. Chiu, T. Tang, J. W. Ellingboe, J. Chem. Soc. 1998, 120, 29852986; b) S.
Comb. Chem. 1999, 1, 7377. Kobayashi, S. Nagayama, J. Am.
7 A. L. Marzinzik, WO 98/155532. Chem. Soc. 1998, 120, 4554; c) S.
8 A. L. Marzinzik, E. R. Felder, J. Org. Nagayama S. Kobayashi, Angew.
Chem. 1998, 63, 723727. Chem. Int. Ed. 2000, 39, 567569.
9 C. Gennari, S. Ceccarelli, U. 23 F. Haunert, M. H. Bolli, B. Hinzen,
Piarulli, K. Aboutayab, M. Donghi, S. V. Ley, J. Chem. Soc., Perkin Trans. 1
I. Paterson, Tetrahedron 1998, 54, 1998, 22352237.
1499915016. 24 S. Kiyooka, Y. Kido, Y. Kaneko,
10 I. Paterson, M. Donghi, K. Tetrahedron Lett. 1994, 35, 52435246.
528 18 Enolates and Related Species in Combinatorial and Solid-phase Synthesis
19
Solid-phase Palladium Catalysis for High-
throughput Organic Synthesis
Yasuhiro Uozumi and Tamio Hayashi
19.1
Introduction
19.2
CarbonCarbon and CarbonNitrogen Bond-forming Reactions of Aryl and Alkenyl
Halides
19.2.1
Cross-coupling Reactions
Ref. 10: yields are based on incorporation of halobenzoyl group on the resin.
542 19 Solid-phase Palladium Catalysis for High-throughput Organic Synthesis
Scheme 19.4. Synthesis of tetrasubstituted ethylenes on solid support via resin capture.
Scheme 19.7. Solid-phase synthesis of macrocyclic b-turn mimics with boronate linker.
The resin-supported arylstannane 45 was also examined for Stille coupling with
bromobenzenes (Table 19.6) [18]. Thus, the reaction of the PS resin-supported p-
tributylstannylbenzoate 45 with 3 equiv. of aryl bromide took place in 1-methyl-2-
pyrrolidinone (NMP) at 90 C to give the coupling products on solid support. A
palladacycle catalyst 46 was used for the coupling in the presence of LiCl to pro-
mote the reaction more eciently. Methanolysis of the resulting resin gave the
corresponding methyl p-arylbenzoates in 8095% yields.
The solid-phase synthesis of the library of the highly functionalized fused ring
system shown in Scheme 19.8 was achieved via the Stille coupling of an alkenyl
bromide with a series of alkenylstannanes, resulting in conjugate dienes and a
subsequent DielsAlder reaction (Scheme 19.8) [19].
Ref. 18
59 in 72% yield in one pot, thereby showing that the Sonogashira coupling and
subsequent intramolecular cyclization to form the indole ring proceeds on solid
phase (Scheme 19.9) [22].
The Rink resin-supported o-iodo(N-methanesulfonyl)anilide 60 gave the 2-
substituted indoles 61 which underwent, after deprotection of the N-mesyl group,
a solid-phase Mannich reaction to aord the (2-alkyl-3-aminomethyl)indole 5-
carboxamides 64 (Scheme 19.10) [23].
19.2 Carbon--Carbon and Carbon--Nitrogen Bond-forming Reactions of Aryl and Alkenyl Halides 549
A solid-supported terminal alkyne was also examined for the Sonogashira cou-
pling (Scheme 19.14) [27]. Thus, propiolic acid was connected to PS oxyamino
resin prepared from the phthalimide 75 via the Gabriel synthesis to give the solid-
supported propiolic amide 76. The palladium-catalyzed coupling of the supported
propiolic amide 76 with the halogenonucleoside 77 proceeded under the standard
Sonogashira conditions to give the nucleoside hydroxamic acid 79.
Ref. 28
19.2 Carbon--Carbon and Carbon--Nitrogen Bond-forming Reactions of Aryl and Alkenyl Halides 555
biaryl 2-carboxylic ester and the 3-carboxylic ester, respectively (Table 19.8, entries
9 and 10). Heteroaromatic biaryl derivatives were obtained from solid-supported
bromopyridine or thiophenylzinc (Table 19.8, entries 8 and 11).
It is noteworthy that zinc-mediated cross-coupling was used for the prepara-
tion of diarylmethanes (Scheme 19.15) [29]. Thus, the coupling of the Rink amide
of the o-, m-, and p-iodobenzoyl substrates with 10 equiv. of 4-cyanobenzylzinc
and 2,6-dichlorobenzylzinc reagents was catalyzed by a catalyst generated from
Pd2 (dba)3 and tri(2-furyl)phosphine to give the corresponding benzylphenylcar-
boxamides 80 and 81 in high yields. The synthetic sequences shown in Scheme
19.16, where the zinc-mediated arylaryl and arylbenzyl couplings were per-
formed on resin support, gave AraAraCH2 aAr products in high purity.
19.2.2
Palladium-catalyzed Arylation and Alkenylation of Olens
The Heck reaction is generally thought to proceed via the reaction pathway
shown in Scheme 19.17. Oxidative addition of RX to palladium(0) gives LnPd(R)X.
Coordination of an olen substrate to LnPd(R)X gives Pd(II) (h 2 -olen)R. Insertion
of the h 2 -olen ligand into the PdR bond gives a s-alkylpalladium intermediate
which subsequently undergoes b-hydride elimination to give the Heck product and
LnPd(H)X.
The intermolecular Heck reaction of resin-supported aryl iodides with olen
substrates was examined on various polymer supports. Representative results are
19.2 Carbon--Carbon and Carbon--Nitrogen Bond-forming Reactions of Aryl and Alkenyl Halides 557
summarized in Table 19.9, where conjugated olens were used to exhibit high re-
activity [31, 32]. The Heck reaction of the Wang resin-supported 4-iodobenzoate
with ethyl acrylate was catalyzed by Pd(OAc)2 in DMF to give the ethyl cinnamate
product in 91% yield (Table 19.9, entry 1). The unsymmetrical stilbene was ob-
tained from methoxystyrene in 90% yield (Table 19.9, entry 2). Alkenylation of io-
dobenzoates supported on PSpolyethylene glycol (PEG) resin and peptide amide
linker (PAL) resin took place at 37 C in the presence of Pd(OAc)2 -PPh3 in aque-
ous DMF (Table 19.9, entries 47).
The Heck reaction of a supported styrene with solution-phase aryl iodides was
also examined under essentially the same conditions (Table 19.10). Thus, the Wang
resin-supported styrene carboxylate reacted with iodobenzene, bromonaphthalene,
bromothiophene, and bromopyridine in DMF on heating to give the corresponding
b-arylstyrene-4-carboxylic acids upon release from the resin support. A palladium
complex generated in situ by mixing Pd2 (dba)3 and tri-(2-furyl)phosphine (Table
19.9, conditions B) exhibited high catalytic activity for the reaction with aryl bro-
mides (Table 19.9, entries 2, 4, and 5).
Aryl halides having allylamine or acrylamide substituents at their ortho posi-
tions have been well documented to undergo an intramolecular Heck reaction to
form indole ring systems. The intramolecular indole ring construction has been
applied to the solid-phase synthesis of libraries of indole derivatives. Thus, the 2-
bromoaniline 85, which was connected at its 5-position to PSPEG resin by a
Wang-type linker, was converted to the N-acyl-N-allylanilines 86 using the standard
uorenylmethoxycarbonyl (Fmoc) method. The supported substrates 86 were sub-
558 19 Solid-phase Palladium Catalysis for High-throughput Organic Synthesis
Entries 13, ref. 31; entries 47, ref. 32. Conditions: entries 1 and
3, Pd(OAc)2 , Et3 N, Bu 4 NCl, DMF, 8090 C, 16 h; entry 2,
Pd2 (dba)3 , (o-tol)3 P, DMF, 100 C, 20 h; entries 47, Pd(OAc)2 ,
PPh3 , Et3 N, Bu 4 NCl, DMF/H2 O (9:1), 37 C, 4 h.
jected to the palladium-catalyzed Heck reaction. The resulting resin was treated
with CF3 COOH to give N-acyl-3-alkyl-6-hydroxyindoles (87) in good to excellent
yields and purities (Scheme 19.18) [33]. An acrylamide group showed good reac-
tivity for the intramolecular Heck reaction (Scheme 19.19) [34]. Thus, cyclization
of Rink resin-supported (2-iodo)acrylanilides (90) was catalyzed by Pd(OAc)2 -PPh3
19.2 Carbon--Carbon and Carbon--Nitrogen Bond-forming Reactions of Aryl and Alkenyl Halides 559
Tab. 19.10. Solid-phase Heck reaction of supported styrene with aryl halides.
in DMF to give the oxindoles 91. N-Allyl-2-iodoanilines (94) which were supported
on Rink amide resin by their N-alkyl side-chains also underwent Heck cyclization
in the presence of Pd-PPh3 species in aqueous DMF at 80 C to aord 3-(resin-
connected alkyl)indoles (95) (Scheme 19.20) [35]. Cleavage of the resin support
under acidic conditions gave high yields of the indoles 96, which bear 3-(amino-
carbonyl)methyl and N-benzyl groups. Preparation of an oligopeptide library at the
5-position of the indole skeleton was achieved by use of resin-supported 5-carboxy
indole 95 as a scaold (Scheme 19.20, bottom).
A macrocyclic peptide was synthesized on solid support via Heck cyclization
(Scheme 19.21) [36]. An oligopeptide chain (98) bearing aryl iodide and acrylamide
groups was prepared by the standard Fmoc method on PS resin. Heck reaction of
the oligopeptide 98 took place at 37 C in aqueous DMF in the presence of the
Pd(OAc)2 -PPh3 catalyst to give, after deprotection and release from resin support,
19.2 Carbon--Carbon and Carbon--Nitrogen Bond-forming Reactions of Aryl and Alkenyl Halides 561
the macrocyclic peptide 100 in 30% overall yield based on the loading of the start-
ing resin. A benzazepine skeleton was constructed on solid support by intramo-
lecular Heck reaction (Scheme 19.22) [37]. The Wang resin-supported (N-butenyl)-
2-iodobenzamides 101 (R CH3 or CH2 Ph) underwent Heck cyclization with
Pd(OAc)2 -PPh3 to give 102. Acidic cleavage of the Wang ester, followed by treat-
ment with diazomethane, gave the benzazepines 103 in high yields.
19.2.3
Amination of Aryl Halides
Entries 15, ref. 39; entries 610, ref. 40. Conditions: A, amine
(3 equiv.), Pd2 (dba)3 (5 mol%), (o-tol)3 P, NaOBu-t (1020 equiv.),
toluene, 100 C; B, amine (10 equiv.), Pd2 (dba)3 (20 mol%),
(o-tol)3 P (80 mol%), NaOt-Bu (1020 equiv.), toluene, 100 C; C,
amine (10 equiv.), Pd2 (dba)3 (20 mol%), BINAP (80 mol% P),
NaOBu-t (1020 equiv.), toluene, 100 C
nations were carried out with the Pd2 (dba)3 /tri-(o-tolyl)phosphine catalyst system
and NaOt-Bu in toluene at 100 C to give complete conversion of the substrates.
Resin-bound o-bromides showed little activity, presumably owing to their steric
hindrance (Table 19.11, entry 3). PSPEG Rink amide (TG RAM) resin-bound p-
bromobenzamide was also examined for coupling with piperidine and pyrrolidine
to give the N-arylpiperidine and N-arylpyrrolidine in 81% and 49% yields, respec-
tively, under essentially the same conditions (Table 19.11, entries 6 and 7) [40]. It
has been documented that primary and secondary aliphatic amines result in sig-
nicant reduction of the bromide using (o-tol)3 P and that the improved conditions
with 2,2 0 -bis(diphenylphosphino)-1,1 0 -binaphthyl (BINAP) (Table 19.11, conditions
C) decrease this side reaction. The yield of N-arylpyrrolidine (49%) increased to
93% with BINAP (Table 19.11, compare entry 7 with 8). The use of BINAP as a
ligand also allowed for the successful coupling of primary amines. Thus, benzyl-
amine reacted with the TG RAM-supported p- and m-bromobenzamide with
the Pd/BINAP catalyst to give 99% and 89% yields, respectively, of the p- and m-
(benzylamino)benzamide (Table 19.11, entries 9 and 10).
19.2 Carbon--Carbon and Carbon--Nitrogen Bond-forming Reactions of Aryl and Alkenyl Halides 565
19.2.4
Miscellaneous Reactions [41]
19.2.4.1 Heteroannulation
It has been reported that annulation of a 2-iodoaniline with an internal alkyne
takes place in the presence of a palladium catalyst to give a 2,3-disubstituted indole
in one step (Larock annulation) [42]. The annulation of 4-carboxamide-2-iodoani-
lines (104) supported on Rink resin with an excess amount of disubstituted al-
kynes was catalyzed by Pd(OAc)2 -PPh3 to give the indoles 105 (Table 19.12, route
A) [43]. Cleavage of the resin moiety from 105 by triuoroacetic acid gave the
2,3,5-trisubstituted indoles 108 in excellent yields (Table 19.12, entries 15). The 2-
iodoaniline 106 bound to the resin support at its N1 position by the THP linker
reacted with alkynes under palladium-catalyzed conditions to give N-resin-bound
indole 107 (Table 19.12, route B). Acidic cleavage of the NTHP linkage gave high
1 A i CONH2 H Pr Pr 91
2 A i CONH2 H Me t-Bu 87
3 A i CONH2 H Me Ph 86
4 A ii CONH2 COCH3 Pr Pr 95
5 A i CONH2 COCH(CH3 )2 Me t-Bu 75
6 B iii H H Ph SiMe3 73
7 B iii H H Me t-Bu 55
Entries 15, ref. 41; entries 67, ref. 42. Condition i, alkyne
(1015 equiv.), Pd(OAc)2 (10 mol%), Ph3 P (20 mol%), LiCl
(1 equiv.), K2 CO3 (5 equiv.), DMF, 80 C; ii, alkyne (1015 equiv.),
Pd(OAc)2 (10 mol%), Ph3 P (20 mol%), Bu 4 NCl (1 equiv.), KOAc
(5 equiv.), DMF, 80 C; iii, alkyne (excess), PdCl2 (Ph3 P)2
(20 mol%), tetramethylguanidine (10 equiv.), DMF, 110 C.
566 19 Solid-phase Palladium Catalysis for High-throughput Organic Synthesis
Polystyrene C 67
ArgoPore A 60
TentaGel Substrate (6 equiv.) C 50
PdCl2 (Ph3 P)2 (10 mol%)
Polystyrene Cul (20 mol%) C 92
Dioxane, Et3 N, rt, 24 h
ArgoPore C 86
TentaGel Substrate (5 equiv.) C 90
Pd2 (dba)3 (10 mol%)
Polystyrene Ph3 As (40 mol%) C 79
Dioxane, 60 C, 24 h
ArgoPore C 80
iodide was connected to PSPEG resin (TentaGel), standard PS resin, and macro-
porous PS resin (ArgoPore) with a hydrazine linker. The supported aryl iodides
(115) were subjected to the coupling reactions under the conditions listed in Table
19.13 to give 116. The resulting resins (116) were subsequently subjected to linker
cleavage conditions A, B, or C (see Table 19.13) to give the substituted aromatics in
a traceless fashion.
19.3
Solid-phase Reactions by Way of p-Allylpalladium Intermediates
19.3.1
Cleavage of Allyl Ester Linkers
A carboxylic acid moiety connected to a polymer resin by an allyl ester linker was
released under palladium-catalyzed allylic substitution conditions. Thus, an allyl
ester group of the PS resin-supported tripeptide 117 was cleaved reductively by tin
hydride in the presence of a palladium-PPh3 catalyst to release the peptide in high
yield. Carbonoxygen bonds of supported allyl esters 118120 were also readily
cleaved by morpholine by way of p-allylpalladium intermediates (Scheme 19.25) [47].
19.3 Solid-phase Reactions by Way of p-Allylpalladium Intermediates 569
Tab. 19.14. Preparation of conjugated dienes via Ru-catalyzed cross-metathesis and Pd-
catalyzed allylic substitution on solid support.
Ref. 48.
19.3.2
N-Allylation via p-Allylpalladium Intermediates
The reaction of an allyl ester with a nitrogen nucleophile bound to the PSPEG
resin gave the N-allylation product (Scheme 19.27) [50]. Thus, the reaction of 2-
methoxycarbonylmethyl-2-propen-1-ol (128) with the TentaGel-bound benzylamine
127 in the presence of Pd(PPh3 )4 gave the N-allylation product 129. After ester-
ication of the allylic alcohol of 129, the resulting allyl acetate 130 was subjected
to palladium-catalyzed allylic substitution, again with various nitrogen nucleo-
philes. A resin-supported p-allylpalladium intermediate generated in situ under-
went nucleophilic attack by primary, secondary, tertiary, and cyclic amines to give
the corresponding allylic amines (131) on solid support. The N-(2-aminomethyl-2-
propenyl)-N-benzylglycine derivatives 132 were released from the resin 131 by al-
kaline hydrolysis in moderate to high yields.
19.3.3
Insertionp-Allylic Substitution System
dium 137 via the alkylpalladium 136. The p-allylpalladium should readily undergo
intramolecular nucleophilic attack of nitrogen atom at the ortho position to form
the 2-(alkenyl)indoline 134.
Three-component coupling of an aryl halide, 1,5-hexadiene, and the Rink-sup-
ported piperidine 138 was catalyzed by palladium to give the N-(6-aryl-2-hexenyl)-
piperidine 141 via the insertionp-allylic substitution pathway (Scheme 19.29) [52].
The alkylpalladium intermediate 140 generated in solution phase underwent a b-
eliminationinsertion process which was terminated by the formation of thermo-
dynamically stable p-allylpalladium 143. The resulting p-allylpalladium complex
143 reacted with piperidine on the resin supports to give the N-alkylated piper-
idines 139 in high yield.
19.4
Palladium Catalysis with Solid-supported Complexes
Homogeneous transition metal catalysts are widely used for a variety of organic
transformations. High-throughput synthesis by solution-phase catalysis has also
574 19 Solid-phase Palladium Catalysis for High-throughput Organic Synthesis
19.4.1
Preparation of Solid-supported Palladium Complexes and Their Use in Palladium
Catalysis
The biarylphosphines 152 also reacted with the chloromethylated PS resin under
basic conditions to give the PS-supported biarylphosphines 153 (Scheme 19.32)
[60]. The resin-bound biaryl-(dialkyl)phosphines 153 were the ligands designed for
use in the palladium-catalyzed amination and SuzukiMiyaura coupling of aryl
halides, especially those of aryl chlorides, whereas the use of electron-rich phos-
phine ligands allowed for an increase in the scope of the aryl halide substrate [61].
turnover numbers (TON) up to 5000. The supported carbene complexes 155 were
air-stable and recyclable catalysts.
The polypyrrole-bound mono- and bisphosphines 162 and 163 were prepared
as their P-borane complexes from the corresponding monomers 160 and 161 via
FeCl3 -induced or electrochemical polymerization conditions (Scheme 19.36) [64].
These phosphineborane complexes reacted with palladium(II) without prede-
complexation to give the polypyrrole-bound palladium(0)phosphine complexes,
where the borane on the phosphorus atom served as a reducing agent of palladiu-
m(II). The resulting immobilized polypyrrole palladium(0)phosphine complexes
catalyzed the Heck reaction and the p-allylic substitution of allyl acetates.
plex 167 generated by treatment of 166 with H2 PdCl 4 catalyzed the Heck reaction
of aryl bromides and even aryl chlorides. Thus, the reaction of chlorobenzene with
styrene in the presence of 0.003 mol% of the palladium species 167 and tetrabuty-
lammonium bromide in dimethylacetamide at 140 C gave an 89% isolated yield
of trans-stilbene where the TON observed reached 23,600.
19.4.2
Solid-supported Chiral Palladium Catalysts
amount of research has been reported to date on asymmetric reactions using homo-
geneous catalyst systems in which activity and stereoselectivity can be tuned by vary-
ing the ligand structure. Recently, immobilization of the enantioselective catalysts
has been recognized as one of the most promising strategies for achieving highly
stereoselective catalysis under heterogeneous conditions [72]. Several examples of
chiral ligands supported on polymer resin, which have found utility in asymmetric
palladium catalysis, are shown in Scheme 19.41. Palladium complexes of the resin-
supported 2-diphenylphosphino-2 0 -substituted-1,1 0 -binaphthyl (MOP) ligand 176
[73] and pyridinoxazoline 179 [74] catalyzed allylic substitution with good to high
stereoselectivity. The PS-supported BINAP 177 [75] was applied to a palladium(II)-
catalyzed aldol reaction of a silyl enolate [76]. A novel chiral ligand, (3R,9aS)-(2-
aryl-3-(2-diphenylphosphino)-phenyl)-tetrahydro-1H-imidazo[1,5-a]indole-1-one was
designed, prepared, and immobilized on an amphiphilic polystyrene-poly(ethylene
References 581
glycol) graft copolymer (PSPEG) resin (178) [77]. A palladium complex of the PS
PEG resin-supported ligand 178 catalyzed the allylic substitution of both cyclic and
acyclic allylic esters in water with high enantioselectivity (up to 98% ee). The PS
PEG-supported Pd complex was readily recovered by simple ltration and reused
without loss of catalytic activity or enantioselectivity.
References
20
Olen Metathesis and Related Processes for CC
Multiple Bond Formation
Florencio Zaragoza
20.1
Introduction
Olen metathesis refers to a reaction in which two alkenes exchange their alkyli-
dene fragments (Scheme 20.1) [15]. This reaction has been applied to the prepa-
ration of compound libraries in solution and has also been used for solid-phase
synthesis, thus enabling its application to automated parallel synthesis. Examples
of metathesis on solid phase include the chemical transformation of resin-bound
intermediates as well as the cleavage of nal products from the support.
Most of the currently used catalysts (e.g. 1, 2, and 3) are ruthenium benzylidene
complexes. One of the reasons for choosing the benzylidene ligand is that styrene
does not generally undergo cross-metathesis eciently when using ruthenium
carbene complexes as catalysts [28]. For this reason, only small amounts of prod-
uct resulting from a cross-metathesis with styrene will result, even if large amounts
of catalyst are used (see, for example, Scheme 20.20).
Olen metathesis reactions have been grouped into dierent categories (Scheme
20.2). These include self-metathesis (reaction of one alkene with itself ), cross-
metathesis (reaction of two dierent alkenes with each other), ring-opening meta-
thesis polymerization (ROMP, polymerization of a strained, cyclic alkene), and
ring-closing metathesis (RCM, cyclization of a diene). All these types of olen
metathesis can be conducted on insoluble supports, and can thus be adapted to
automated, parallel solid-phase synthesis. For combinatorial chemistry in solution,
mainly cross-metathesis and ROMP have been used.
Alkynes are also suitable substrates for catalytic olen metathesis (Scheme 20.3).
Treatment of a mixture of an alkene and an alkyne with a metathesis catalyst can
lead to the clean formation of dienes [29, 30]. Because alkynes usually react faster
than alkenes with carbene complexes, the formation of dienes from alkenes and
588 20 Olefin Metathesis and Related Processes for CC Multiple Bond Formation
alkynes is assumed to proceed via the initial addition of the catalyst to the alkyne,
followed by cycloreversion and reaction of the resulting vinylcarbene complex with
the alkene (Scheme 20.3). Treatment of alkynes with carbyne complexes can bring
about alkyne metathesis [1, 31], which presumably also proceeds via reversible
2 2 cycloadditions (Scheme 20.3). Some of these intriguing transformations have
also been performed on insoluble supports, and might be suitable for parallel syn-
thesis.
20.2
Olen Metathesis in Solution
Scheme 20.4. Strategies for the preparation of compound mixtures using olen metathesis.
20.2.1
Scope and Limitations of Olen Metathesis in Solution
[14, 28, 3234]). Usually, mixtures of the products of cross-metathesis and of self-
metathesis are obtained, each of them as mixtures of E- and Z-isomers. Unfor-
tunately, some alkenes show a high tendency to undergo self-metathesis (to form
symmetric dimers; see, for example, [35, 36]), whereas other alkenes (acryloni-
trile, styrenes) undergo self-metathesis only slowly or not at all. For this reason,
cross-metathesis of mixtures of dierent olens will not always yield the statisti-
cally expected amounts of internal olens. This feature can cause problems during
the deconvolution of such compound libraries because potent ligands formed only
in low quantities are usually dicult to identify by deconvolution.
A further problem of cross-metathesis in solution is that a purication step will
usually be required to remove the catalyst. With the recent development of immo-
bilized catalysts (e.g. 46) [3741], however, this problem has been reduced. Un-
fortunately, all of the immobilized ruthenium carbene complexes described so far
(Scheme 20.5) lose activity rather quickly; this fact might be due to the inherent
instability of these complexes and to the fact that during catalysis detachment of
the metal from the support can readily occur.
20.2.2
Examples of Library Preparation by Cross-metathesis in Solution
20.2.3
Examples of Library Preparation by Ring-closing Metathesis in Solution
Some new strategies for the preparation of cyclic peptides by ring-closing meta-
thesis are presented below to illustrate the scope of these cyclizations.
Liskamp and coworkers have investigated the cyclization of N-alkenylated pep-
tides (12) by ring-closing metathesis (Scheme 20.11) [4850]. The peptides were
prepared by standard solid-phase synthesis, and the N-alkenylation was eected
during the assembly of the peptide by N-sulfonylation with 2-nitrobenzenesulfonyl
chloride, followed by N-alkenylation under Mitsunobu conditions and sulfona-
mide cleavage by treatment with mercaptoethanol/1,8-diazabicyclo[5.4.0]undec-7-
ene (DBU). Ring-closing metathesis could be performed either in solution or on
solid phase, but in solution higher yields were usually obtained [50]. Cyclization
experiments showed that the length of the N-alkenyl group was crucial for ring
594 20 Olefin Metathesis and Related Processes for CC Multiple Bond Formation
closure. N-Allyl peptides (12) could only be cyclized to yield eight-member rings
(13). Larger ring sizes required the use of N-homoallyl or N-(4-penten-1-yl) pep-
tides. The cyclization of tripeptides (to form a 15-member ring, e.g. (14)) was
particularly dicult, and only proceeded satisfactorily with N-(4-penten-1-yl) sub-
stitution (Scheme 20.11).
20.2.4
Examples of Library Preparation by Ring-opening Metathesis Polymerization in
Solution
20.3
Olen Metathesis on Solid Phase
In solid-phase synthesis, the metathesis of alkenes has been used both for the
chemical transformation of support-bound intermediates as well as for the cleav-
age of products from the support. Although these techniques have not yet been
596 20 Olefin Metathesis and Related Processes for CC Multiple Bond Formation
extensively used for the preparation of large libraries by parallel synthesis, solid-
phase chemistry is generally well suited for this purpose, and some of the reac-
tions described below can probably be used for the preparation of compound li-
braries.
20.3 Olefin Metathesis on Solid Phase 597
20.3.1
Cleavage from the Support by Olen Metathesis
Scheme 20.16. Mechanism of the cleavage of dienes from supports by RCM [70].
olen metathesis will only be of limited use for the preparation of such libraries
unless more stable metathesis catalysts or selective scavengers for metal-containing
byproducts become available.
A similar strategy has been described by Peters and Blechert [74], in which RCM
of a support-bound diene was used to release styrenes from a polystyrene-based,
insoluble support. Linkers of this type can also be cleaved by cross-metathesis with
ethylene [68].
Several groups have investigated the preparation of cyclic compounds by RCM
with simultaneous cleavage from the support [71, 75, 76]. One recent example, re-
ported by Piscopio et al. [77], is shown in Scheme 20.19. The substrate (23) for
olen metathesis was prepared in one step by an Ugi reaction. The product (24), a
Freidinger lactam, was designed to mimic b-turns, which play a pivotal role in the
20.3.2
Ring-closing Metathesis on Solid Phase
Scheme 20.23. Preparation of cyclic peptides by RCM on solid phase [55]. TG, Tentagel.
allylation, strained triene (30) was generated, which upon treatment with a meta-
thesis catalysts underwent a ring-opening/ring-closing cascade to yield highly sub-
stituted, tetracyclic compounds (31). A valuable feature of this synthesis is the
ready availability of some of the four building blocks. One drawback of this reac-
20.3 Olefin Metathesis on Solid Phase 603
tion sequence is the low selectivity of the allylation reaction, which necessitates the
protection of all nucleophilic functional groups.
20.3.3
Cross- and Self-metathesis on Solid Phase
nates (39) with o-alkenols (40) in a solid-phase synthesis of muscone analogs [88]
(Scheme 20.26). Unsaturated a-amino acid derivatives (42) have also been prepared
on solid phase by cross-metathesis [35] (Scheme 20.26). The low loading of the
starting resin (0.07 mmol g1 ) was required to suppress self-metathesis.
the alkene), high yields of dienes can be obtained. This reaction can also be con-
ducted on insoluble supports, with either the alkene or the alkyne attached to the
resin (Scheme 20.28). The resulting dienes can be further transformed by 2 4
cycloaddition with suitable dienophiles to yield substituted cyclohexenes [94].
20.4
Conclusion
With the development of highly ecient and selective catalysts for olen meta-
thesis in recent years, this reaction has become a valuable tool for organic chem-
ists. Cross-metathesis and ROMP in solution can now be performed with func-
tionalized alkenes, and can oer interesting new possibilities for the preparation of
compound libraries. In particular, the selective cross-metathesis of dierent al-
kenes, for which the underlying principles are now slowly emerging, has huge
potential and could become a process with an impact similar to those of the Wittig
or the DielsAlder reactions if its scope and limitations are clearly understood.
Solid-phase synthesis has also greatly beneted from these new catalysts, and new
cleavage strategies and other methodologies based on carbene complex-mediated
olen metathesis on solid phase have been developed successfully.
The properties of the currently available catalysts are, however, far from ideal.
Because of their limited stability, large amounts of catalyst are often required
to drive reactions to completion. This feature can lead to signicant amounts of
metal-derived impurities in the crude products. Moreover, most carbene complexes
suitable for olen metathesis are also highly sensitive toward amines, azoles, and
other nucleophiles, which severely limits the choice of functional groups tolerated
in the starting materials. This facet is particularly problematic for the preparation
of libraries of biologically active compounds because nucleophilic functional groups
are often of crucial importance for biological activity. Future research should aim
to overcome these limitations of current catalysts by further enhancing their selec-
tivity and stability.
References
Part III
Special Synthetic Topics
21
Solid-phase Synthesis of Natural Products and
Natural Product-like Libraries
K. C. Nicolaou and Jerey A. Pfeerkorn
21.1
Introduction
Natural products have always played, and continue to play, important roles in
both drug discovery and chemical biology. In fact, from 1989 to 1995, 60% of all
approved drugs and new drug application (NDA) candidates were derived from
natural sources [1]. Furthermore, natural products have been used extensively for
studying complex biological processes such as signal transduction and cell cycle
regulation [2]. Yet, in spite of their importance to both biology and medicine, it
was only recently that solid-phase chemistry has been applied to the synthesis of
natural products and their analogs. Initial attempts were hampered by the limited
scope of early solid-phase reactions for the construction of nonoligomeric struc-
tures as well as the lack of appropriate solid-phase analytical techniques [3]. How-
ever, given the remarkable advances in solid-phase chemistry over the last decade,
it is not surprising that numerous research groups are now actively engaged in
the construction of structurally complex, natural product-based libraries using
solid-phase combinatorial synthesis techniques [3, 4]. Figure 21.1 shows a diverse
collection of naturally occurring structures which have been synthesized using
solid-phase technologies. Within this chapter, we seek to highlight these accom-
plishments, focusing particular attention on the novel strategies developed to solve
each problem and on how the synthetically demanding complexity of natural
products has spurred the development of new linking and cleavage devices and
protocols, as well as on the chemical biology contributions made possible by these
technologies. The discussion will be divided into three sections paralleling, chro-
nologically, the trends which have emerged in the solid-phase synthesis of natural
products. These include: (1) combinatorial derivatization of immobilized natural
product skeletons; (2) solid-phase target-oriented total synthesis of natural prod-
ucts; and (3) construction of diversity-oriented natural product-like libraries for
chemical biology and drug discovery purposes.
21.2
Solid-phase Derivatization of Natural Product Scaolds Combinatorial
Semisynthesis
21.2.1
Solid-phase Semisynthesis of Rauwolfa Alkaloids
was deprotected with piperidine to furnish free amine 15 which was then coupled
to a series of N-uorenylmethoxycarbonyl (Fmoc)-protected amino acids (16) via
HATU activation. These 36 amino acids represented the rst element of diversity
in the library. After piperidine-mediated deprotection of the newly coupled amino
acid 17, the liberated amine was coupled to the free C1 carboxylic acid of yo-
himbinic acid (14) to eect loading of the natural product skeleton onto the solid
support. A second element of diversity was then installed at the C2 hydroxyl posi-
tion of yohimbinic acid through reaction with the symmetrical anhydrides of a
variety of carboxylic acids [(R2CO)2 O] to aord esters of type 20. After acidic cleav-
age, 22 pools each containing 36 compounds were obtained with the structures of
selected pools conrmed by ow injection mass spectroscopy (MS) analysis. While
validating the overall strategy, this rst approach required extensive deconvolution
eorts to identify the active component(s) of any of the pools. As such, the authors
then elaborated on the above approach by cosynthesizing surrogate analytes, or
tags, onto each bead which could be cleaved under conditions orthogonal to the
chemistry of the actual library member on the same bead. In this way, if during
biological assay the subanalytical amount of cleavage product from a given bead
demonstrated activity, the tag on that bead could be subsequently cleaved and
identied by MS, thereby revealing the structure of the active compound. The suc-
cess of this strategy hinged upon construction of orthogonally dierentiated beads;
this was accomplished by treatment of TentaGel-S-NH2 with a 9:1 mixture of Fmoc-
Cl and allyl chloroformate, giving a resin with a @5:1 ratio of Fmoc/Alloc sites.
The N-Fmoc-protecting group was then cleaved with piperidine and the generated
amine was acetylated with the N-Fmoc Rink amide (RAM) linker to give resin 22.
The spacer for the amine tag was installed by removal of the Alloc protecting
group [Pd(PPh3 )4 /TMSN3/TBAF] and reaction with N-Alloc iminodiacetic anhy-
dride, leading to carboxylic acid 25. The free carboxylic acid of 25 was subsequently
coupled with 36 secondary amines ( n Ry NH) through the intermediacy of an acti-
vated pentauorophenolate ester. These 36 amines (each with a distinct molecular
weight) served as the tags, and each of them encoded for a dierent amino acid
to be coupled in the next step. After deprotection of 26 using piperidine, each of the
36 N-Fmoc-protected amino acids (16) were coupled to their respective resins us-
ing HATU activation to provide amides of type 28. Upon completion of this opera-
tion, the resins were pooled and then split into 22 new reaction vessels where the
C2 hydroxyl group was acylated using a symmetrical acid anhydride as described
above. This nal acylation did not require tagging since the beads were already
spatially segregated for the reactions. The products of this nal reaction were then
maintained in the 22 separate pools for screening. Since the synthesis was en-
coded, if any of the products released from a particular bead proved biologically
active the surrogate amine tag from that bead could then be cleaved to provide
identication, as illustrated at the bottom of Scheme 21.2. Treatment of the resid-
ual resin with HCl followed by Li 2 CO3 and subsequent labeling of the released
amine with dansyl chloride would provide conjugate 32, which could be readily
analyzed by MS to determine the structure of the active library member.
21.2 Solid-phase Derivatization of Natural Product Scaffolds -- Combinatorial Semisynthesis 617
21.2.2
Solid-phase Synthesis of Purine Derivatives
21.2.3
Solid-phase Semisynthesis of a Taxoid Library
21.2.4
Solid-phase Semisynthesis of Sarcodictyns A and B and Libraries Thereof
With the linker-cleavage protocol validated, the avenue toward library construc-
tion was then open, as illustrated in Scheme 21.6. Inspection of the sarcodictyn
structure suggested the following three sites for diversication; (1) the side-chain
appended to C(8), (2) the hemiketal at C(4) and (3) the C(15) ester moiety. As out-
lined in Scheme 21.7, each of these sites was modied in a sequential fashion.
Thus, the previously linked scaold 61 was rst esteried at the C(8) hydroxyl
moiety employing ve distinct (R1COX) acyl donors to give 68. After desilylation of
the C(15) hydroxy group, intermediate 69 was split along three pathways. The rst
entailed acylation (R2COX) of the C(15) hydroxyl group followed by transketaliza-
tion release with a series of alkyl alcohols (R3OH), ultimately yielding structures of
the general type 75. The second pathway involved oxidation of the C(15) hydroxyl
group of 69 to the corresponding carboxylic acid, which was converted to either an
amide or an ester (74) by coupling with a series of amines (R4NH2 ) or alcohols
(R4OH) respectively. These derivatives were then released by transketalization as
before, giving rise to 76. The nal path involved conversion of alcohol 69 to the cor-
responding amine (71), which was subsequently acylated (R5COX) giving structures
of the general type 74. Release of the latter structures by transketalization provided
77. This approach yielded a 66-member library which was evaluated using the col-
orimetric tubulin polymerization assay. Active compounds were also tested in cyto-
toxicity assays employing ovarian cancer cells (1A9) and two Taxol TM -resistant cell
lines (1A9PTX10 and 1A9PTX22), leading to a number of promising antitumor
agents with higher potencies than those exhibited by the naturally occurring sub-
stances themselves [14].
21.2.5
Solid-phase Semisynthesis of Vancomycin
Scheme 21.8. Use of selenoether 84 as a safety-catch linker for carboxylic acids [17].
to provide polymer-bound phenol 92, which was then glycosylated with the tri-
chloroacetimidate 93 to aord monosaccharide 94. Removal of the C2 Alloc group
from glucose using palladium(0)-mediated allyl transfer conditions aorded com-
pound 95 in excellent yield. The polymer-bound conjugate 95 was then subjected
626 21 Solid-phase Synthesis of Natural Products and Natural Product-like Libraries
21.3
Solid-phase Total Synthesis of Natural Products Combinatorial Total Synthesis
The excitement generated in the 1990s by combinatorial chemistry did not escape
the attention of synthetic chemists involved in the total synthesis of natural prod-
ucts. The attractiveness of solid-phase chemistry and the lure of the possibility of
using the split-and-pool strategy for library construction prompted several studies
directed toward diversity-oriented total synthesis on solid support. In a typical case,
the natural product itself is rst targeted for synthesis on solid phase an en-
deavor that provides the opportunity to develop the necessary chemistry for load-
ing, elaborating, and cleaving the nal target. Once the technology is developed,
the practitioner then employs a pool of building blocks to construct a library of
small to medium size, depending on the complexity of the target molecules in-
volved. Below, we discuss several such solid-phase syntheses, paying particular at-
tention to the novel linking and cleavage strategies developed during these pro-
grams and as a consequence of the challenges posed by each natural product.
21.3.1
Solid-phase Synthesis of Prostaglandins and Libraries Thereof
One of the early examples of total synthesis on polymeric support was reported by
Chen and Janda in which the construction of prostaglandin E2 (PGE2 ) methyl ester
was carried out using a soluble polystyrene polymer; although, strictly speaking,
this is not a solid-phase synthesis, it is included here for its elegance and for his-
torical reasons [21]. The practical features of soluble polymeric supports have been
well known for some time and include solubility in most organic solvents [tetra-
hydrofuran (THF), CH2 Cl2 , CHCl3 ], thus allowing reactions to be carried out
under standard solution-phase conditions and, more importantly, allowing stan-
dard analytical techniques, especially NMR, to be used to monitor reaction prog-
ress [22]. When necessary, however, the polymer support can be precipitated out of
the reaction mixture since it is insoluble in solvents such as methanol and water
so that the benets of easy purication typically associated with solid-phase chem-
istry can still be realized [22].
21.3 Solid-phase Total Synthesis of Natural Products -- Combinatorial Total Synthesis 627
four dierent a-chains to furnish structures of the general type 110. After reduc-
tion of the alkyne of the a-chain to the alkene moiety of the natural prostaglandins,
the pools were cleaved with aqueous HF to aord four new pools of prostanoids,
each consisting of four compounds. These mixtures were then assayed for inhibi-
tion of cytomegalovirus (CMV) a member of the herpesvirus family and de-
convolution of an active pool led to the identication of a lead compound.
A similar solid-phase synthesis of a related series of prostaglandins was reported
by Ellman and coworkers, who employed a traditional, crosslinked polystyrene-
based resin [24]. The Ellman approach relied on tethering the cyclopentene build-
ing block 113 (Scheme 21.13) through its hydroxyl group to a suitably derivatized
resin. Thus, treatment of the polystyrene-derived di-n-butylsilyl chloride resin with
113 in the presence of imidazole aorded conjugate 114. The tri-(4-methox-
yphenyl)methyl (TMT) protecting group of 114 was then removed from the mole-
cule by treatment with formic acid and the a-side-chain was installed through a
Suzuki coupling with a borane derivative to provide structure 116. DessMartin
oxidation of 116 followed by attachment of the lower side-chain via a conjugate
addition provided 118. The sulfonamide moiety of the a-side-chain was then acti-
vated with bromoacetonitrile and coupled with various alcohols (R1OH) and
amines (R1NH2 ). Cleavage of the silyl ether bridge with HFpyridine resulted in
the formation of 120, whereas l-selectride reduction of 119 followed by a similar
cleavage resulted in the formation of prostaglandins of type 122. As with the Janda
approach, the strategy developed by Ellman allowed the construction of a small li-
brary demonstrating its usefulness in combinatorial chemistry.
21.3.2
Solid-phase Synthesis of Epothilone A and Libraries Thereof
21.3.3
Solid-phase Synthesis of (S)-Zearalenone
Continuing the theme of cyclorelease (see Scheme 21.14a), the Nicolaou group re-
ported the solid-phase synthesis of (S)-zearalenone (10) using a Stille coupling re-
action to form and simultaneously release the macrocyclic product from the resin
21.3 Solid-phase Total Synthesis of Natural Products -- Combinatorial Total Synthesis 631
21.3.4
Solid-phase Synthesis of (DL)-Muscone and Libraries Thereof
In an eort to develop a resin suitable for the solid-phase version of the powerful
HornerWadsworthEmmons reaction, we targeted (dl)-muscone and a library
thereof [30]. The adopted strategy represents yet another example of the cyclo-
release strategy (see Scheme 21.15, top) in which the cyclic product is simulta-
21.3 Solid-phase Total Synthesis of Natural Products -- Combinatorial Total Synthesis 633
21.3.5
Solid-phase Synthesis of the Vitamin D3 System
Among the most recent examples of the solid-phase synthesis of natural products
and their analogs is one reported by Takahashi and coworkers around the vitamin
D3 skeleton [31]. As shown in Scheme 21.18, the Takahashi group employed
the trialkylsilane linker 168, which was alkylated with the presynthesized bicyclic
upper unit of vitamin D3 to furnish intermediate 169. The ketone of 169 was then
subjected to a HornerWittig condensation with the presynthesized lower unit in
the presence of n-BuLi to aord triene 170. The side-chain was then installed by a
Grignard reaction in which the tosylate moiety of 170 was displaced to furnish
compound 171, which was ultimately cleaved to provide the 1a,25-(OH)2 -vitamin
D3 analog 172.
21.3.6
Solid-phase Synthesis of Carpanone-like Molecules
The Shair group has described the synthesis of a small library of carpanone-like
molecules employing a solid-phase variant of the biomimetic carpanone synthe-
sis rst reported by Chapman et al. [33]. The strategy, as outlined in Scheme
21.19, relied on initial tethering of a series of electron-rich o-hydroxystyrenes to a
solid support to generate adducts of type 173. These styrenes were then treated
with a series of electron-poor o-hydroxystyrenes in the presence of PhI(OAc)2 to
eect an oxidative coupling furnishing 174, which underwent a spontaneous in-
verse electron-demand DielsAlder reaction to aord a small library of benzox-
anthenones (175) representing the carpanone skeleton. Cleavage with HF led to
the release of carpanone-like compounds 176. The researchers minimized oxida-
tive homocoupling of the resin-bound styrenes species through electronic consid-
erations, low resin loading, and linker optimization.
21.4
Combinatorial Solid-phase Synthesis of Natural Product-like Libraries
Over the past centuries, nature has served as a most generous source supplier of
cures and, in more recent times, leads for drug discovery purposes. Indeed, many
of the clinically used drugs of the twentieth century have their origins in nature,
whose library of compounds exceeds our wildest imagination in terms of both
numbers and molecular diversity. However, the tremendous strides made by biol-
ogy and high-throughput screening technologies in recent times demand a much
faster pace of isolation and characterization of new natural products than currently
possible [34]. This problematic phenomenon is compounded by the pressure to
discover new drugs and to fulll the promise of the human genome project, the
fruits of which will soon be available for further biomedical advances [35]. The
advent of combinatorial chemistry provided a powerful alternative to nature, for
now chemists can design and rapidly synthesize large libraries of natural product
analogs or even libraries of natural product-like molecules. The value of such en-
deavors should be measured by various factors, such as novelty of the chemistry
developed, the molecular diversity generated, the practicality of library construc-
tion, and the relevance of the designed structures to biology.
The rst examples of combinatorial synthesis of natural product analog libraries
are those already discussed above. More recently, several research groups have
21.4 Combinatorial Solid-phase Synthesis of Natural Product-like Libraries 635
21.5
Conclusion
21.6
Addendum
References
22
Solid-phase Synthesis of Heterocyclic Systems
(Heterocycles Containing One Heteroatom)
Roland E. Dolle
22.1
Introduction
22.2
Solid-phase Synthesis of Heterocycles Containing One Nitrogen Atom
22.2.1
Aziridines
Tab. 22.1. Generic structures of heterocyclic libraries containing a single heteroatom: 1992
2000. *Solid-phase attachment point.
22.2 Solid-phase Synthesis of Heterocycles Containing One Nitrogen Atom 645
[85] [84]
646 22 Solid-phase Synthesis of Heterocyclic Systems (Heterocycles Containing One Heteroatom)
of the aziridine carboxylic acids 6 with triuoroacetic acid (TFA) aords products
in >70% yield and in high purity. A second complementary method for azir-
idine synthesis was described in which substrates 1 are acylated with activated 2,3-
dibromopropane 7 to yield resin-bound a-bromoacrylamides 8. The use of 7
required as an alternative three-step acylation of 1 with acrylic acid chloride, bro-
mine addition, and HBr elimination gave a low yield of 8 owing to premature
cleavage. Treatment of Michael substrates 8 with a range of primary amines and
amino acid esters, followed by TFA cleavage, aords aziridines 10 in high purity.
22.2.2
b-Lactams
22.2.3
Pyrrolidines and Derivatives
Another early example of heterocyclic synthesis on solid support is the 1,3-dipolar
cycloaddition reaction of resin-bound azomethine ylides with olens to furnish
functionalized pyrrolidines (Scheme 22.3A). This work was carried out by pioneers
at Aymax in 1995 [8, 9]. In situ ylide formation occurs upon treatment of resin-
652 22 Solid-phase Synthesis of Heterocyclic Systems (Heterocycles Containing One Heteroatom)
bound amino acid imines (11) with AgNO3 and triethylamine in acetonitrile. Re-
action of these silver azomethine ylides with electron-poor olens (28) furnishes
proline derivatives after resin cleavage. The synthetic scheme was optimized using
gel-phase 13 C-NMR (nuclear magnetic resonance) analysis. Product yields range
from 50% to 80% and diastereoselectivities from 2.5:1 to >10:1. Because the pyr-
22.2 Solid-phase Synthesis of Heterocycles Containing One Nitrogen Atom 653
22.2.4
Tetramic Acids
relies on some variation of the following three-step sequence: (1) reductive alkyla-
tion of resin-bound a-amino acids with aldehydes (44); (2) acylation of the second-
ary amine (44 to 4850); and (3) base-promoted Claisen-type intracyclative cleavage
(4850 to 5153; Scheme 22.5). Acylating reagents include Meldrumss acid, ma-
lonic half esters, aryl acetic acids, and acetic acids bearing electron-withdrawing
groups, with the last two reagents providing convenient access to diversity at posi-
tion C3 on the heterocycle.
22.2.5
Pyrroles
Tetra- and penta-substituted pyrroles were rst prepared by Mjalli and co-
workers via the 1,3-dipolar cycloaddition of alkynes to polymer-bound munch-
nones (Scheme 22.6A) [22]. Resin-bound amino acids 54 are subjected to a four-
component condensation with aldehydes, carboxylic acids, and either phenyl or
pyridinylisocyanates to yield Ugi products (55, 56). Hydrolysis of the terminal
amide bond in 55 and 56 to 57 is carried out after Boc anhydride activation. The
selection of either phenyl or pyridinylisocyanate stems from the necessity to hy-
drolyze the amide bond in 55 and 56 under conditions compatible with resin link-
age. Originally, benzylisocyanate was used in this process, but the rate of hydroly-
sis of the corresponding amide is too slow (t1=2 > 7 days) to be of practical utility.
The reaction of 57 with electron-decient alkynes either in neat acetic anhydride as
solvent or in a toluene solution of Et3 N and isobutyl chloroformate at 100 C (24
48 h) leads to sequential in situ formation of munchnones 58 and 1,3-dipolar cy-
cloaddition. Some ten pyrroles were obtained in 2676% overall yield using this
protocol (TFA cleavage). The purity of all the products was exceptional.
A slight variant of this reaction was reported by Armstrong wherein resin-bound
succinate 61 is used in the Ugi condensation employing 1-isocyanocyclohexene as
a convertible isocyanide (Scheme 22.6B) [23]. Munchnone precursors 62 are su-
ciently reactive such that heating 62 in toluene containing 10 equiv. HCl and 25
equiv. dimethyl acetylenedicarboxylate (DMAD) aords the pyrrole ring, negating
the earlier requirement for intermediate amide hydrolysis (55, 56 to 57; Scheme
22.6A).
Resin-bound enaminones 69 are particularly versatile intermediates with respect
to pyrrole synthesis (Scheme 22.7). They are prepared from the condensation of
resin derivative 68 and primary amines [trimethylorthoformate (TMOF)] as the
dehydrating reagent in DMF. The intramolecular net reaction of 69 with a-alkyl-
substituted nitroalkenes (neat or prepared in situ from nitroalkanes and aldehydes)
aords pyrroles in 4690% yields [24]. Enaminones 69 readily react with a-bromo-
ketones in the classical Hantzch pyrrole synthesis [25]. The key to the success of
this reaction is the use of the non-nucleophilic Lewis base 2,6,-di-t-butyl-pyridine
and DMF as solvent.
22.2.6
Piperidine and Derivatives
nchnones.
Scheme 22.6. Pyrroles via immobilized mu
71, an aldehyde (72), a diene (73), and ytterbium triate (10 mol%) in methylene
chloride is shaken at room temperature for @24 h (Scheme 22.8). Traceless cleav-
age of adducts 75 is performed using ACE chloride (N-debenzylation) to give col-
lections of 76. Electron-rich dienes in combination with ethyl glyoxylate, phenyl-
glyoxal, and 37% aqueous formaldehyde are the preferred reagents.
22.2 Solid-phase Synthesis of Heterocycles Containing One Nitrogen Atom 657
Michael addition with 3-butenone in EtOH; (4) oxidation to the sulde to sulfone
81; (5) bromination; and (6) reaction with triphenylphosphine. Phosphonium salt
82 undergoes Wittig condensation to 83 in high yield with a variety of aldehydes
using NaOMe as base. Treatment of vinyl ketones 83 with an excess of benzyl-
22.2 Solid-phase Synthesis of Heterocycles Containing One Nitrogen Atom 659
22.2.7
Dihydropyridines
The dihydropyridine (DHP) pharmacophore, well known for its anity for cal-
cium channels, is regarded as a privileged scaold possessing broad-based biologi-
cal activity. The synthesis of DHPs has been largely explored by Gordeev and co-
workers at Aymax [30, 31]. Their synthesis is based on the multicomponent
cyclocondensation of resin-bound enamino esters 92 with either 2-benzylidine b-
keto esters 93 or b-keto esters and aldehydes (94, 95) followed by TFA cleavage
from solid support (Scheme 22.12A). The success of the cyclocondensation relies
upon the use of pyridine as solvent, which facilitates the formation of the thermo-
dynamically favored enamine 97 from imine 96 (isomerization of the p-bond).
An elegant application of 13 C-labeled substrates and 13 C-NMR was used to dene
the reaction mechanism. Acyclic adducts 97 are cleaved from resin prior to cycli-
zation. The solid-phase protocol is suciently robust for the construction of a 300-
member DHP library (Scheme 22.12B). Rink amine resin was split into ten por-
tions and condensed with one of ten b-keto esters. The enamines were then com-
bined and portioned into 30 reaction vessels. Each enamine lot was treated with
one of three 1,3-dicarbonyl building blocks and ten aldehydes in pyridine at 45 C.
Each resin lot was cleaved with 3% TFA in CH2 Cl2 to yield 30 pools of ten DHPs
per pool. All 30 pools were evaluated for Ca 2 channel binding (rat brain mem-
branes). Two of the pools displayed binding anities at @10 nM. All 20 com-
660 22 Solid-phase Synthesis of Heterocyclic Systems (Heterocycles Containing One Heteroatom)
pounds from the two pools were resynthesized as discrete analogs from which 3,5-
dicarboxymethyl-2,6-dimethyl-4-(2-uorophenyl) DHP 105 was identied with an
IC50 14 nM.
22.2.8
Pyridines
Ellingboe and coworkers [35] generated 1,5-diketones (114) on resin from im-
mobilized hydroxyacetophenones via ClaisenSchmidt reaction with an aromatic
aldehyde followed by Michael addition of trimethylsilyl enol ether (112 ! 113 !
114; Scheme 22.14). Heterocycle formation 116 occurs upon heating 114 with
NH4 OAc in HOAc/DMF at 100 C in an atmosphere of air. Presumably, DHPs 115
initially form and spontaneously oxidize to the corresponding aromatic pyridines.
The yield and purity of ten examples (117) range from 19% to 62% and 21% to
70%, respectively.
Jung [34] condensed the same enone substrates 113 with 1-(2-oxo-2-arylethyl) pyr-
idinium salts (118) and ammonium acetate in a Krohnke-type pyridine synthesis
662 22 Solid-phase Synthesis of Heterocyclic Systems (Heterocycles Containing One Heteroatom)
(Scheme 22.14). DMF containing glacial acetic acid (1:1) is the preferred solvent.
Conducting the reaction in the absence of acetic acid gives a side product arising
from the addition of pyridine to 113.
22.2.9
Azepanes, Benzazepines, and Derivatives
To date, the solid-phase synthesis of seven- and eight-member ring nitrogen het-
erocycles has relied primarily on the use of the ring-closing metathesis (RCM) re-
action (Schemes 22.1522.17) [3638]. Metathesis reactions generally proceed in
good yield with minimal intermolecular cross-linking. The challenge is devising a
synthetic scheme that permits the introduction of diversity elements in the requi-
site bis-olen substrates. The traceless linker strategies reported by Piscopio intro-
duce up to three points of diversity in azepinone structures (Schemes 22.15 and
22.16). In one example, bis-olen substrates 123 with two points of diversity are
prepared by sequential FukuyamaMitsunobu and acylation reactions [39]. Ugi
four-component coupling of resin-bound cinnamyl amine 125, aldehyde 128, iso-
cyanate 127, and olen-bearing carboxylic acid 126 conveniently introduces three
diversity parameters in a single reaction (Scheme 22.16) [36]. Elevated temperature
is generally required for successful RCM. For example, treatment of substrate 129
with Grubbs catalyst (510 mol%) in CH2 Cl2 at 25 C fails to generate metathesis
products. However, conducting the RCM reaction in dichloroethane at 80 C for 24
h gives the corresponding lactams as a 1:1 mixture of diastereomers in good yield.
22.2 Solid-phase Synthesis of Heterocycles Containing One Nitrogen Atom 663
Immobilized diene substrates for azepane synthesis may also be prepared by the
addition of allyl lithium to tethered imine ethers (132 to 133 to 134; Scheme 22.17)
[37].
Polyhydroxylated azepane scaold 137 was obtained upon the reaction of Rink
linker 71 with chiral l-iditol bis-epoxide 135 (Scheme 22.18) [40]. The intramo-
lecular ring closure required heating for 5 days at 80 C in DMF. Azepanes 137 are
obtained following acylation of diol 136 and resin cleavage.
Bolten and Hodges [41] at Parke-Davis Pharmaceuticals (now Pzer) developed
a solid-phase synthesis of substituted benzazepines (147) via intramolecular Heck
reaction of allyl or propargyl glycines bearing a 2-iodophenyl group (140, 144)
664 22 Solid-phase Synthesis of Heterocyclic Systems (Heterocycles Containing One Heteroatom)
22.2.10
Indoles
The synthesis of the indole system has received more attention from combinatorial
chemists than any other heterocycle. This is a reection of the broad-based biolog-
ical activity ascribed to indole-containing compounds. Some 11 reports from eight
laboratories describe the solid-phase methodology used to synthesize substituted
666 22 Solid-phase Synthesis of Heterocyclic Systems (Heterocycles Containing One Heteroatom)
indoles [4352; see also 141, 144]. With the exception of the classical Fischer in-
dole synthesis by Merck [51], all of the examples employ Pd catalysis as a central
theme. The requisite o-iodoanilines are generally tethered to solid support through
an appended carboxylate function. The immobilized substrates are then subjected
to the Sonogashira protocol for terminal alkyne coupling followed by Pd-catalyzed
cyclization. The overall yield and eciency of the reaction sequence are highly de-
pendent on the structure of the alkyne and the base used in the heteroannulation
step (Schemes 22.21 and 22.22) [44]. Because of its ability to provide homogeneous
reaction conditions, tetramethylguanidine (TMG) is the preferred base in many of
the reported protocols.
Indoles with three points of diversity were generated by Collini and Ellingboe
(Scheme 22.23) [45] by coupling triuoracetylated alkyne intermediates 168 with
vinyl triates 170 using catalytic Pd(PPh3 )4 and potassium carbonate as base. Re-
moval of the triuoroacetyl protecting group followed by N-alkylation gave a set of
functionalized indoles (173).
Using an intramolecular Heck reaction, o-iodoanilines 176, bearing a N-allyl sub-
stituent, may be cyclized to give indoles 177 (Scheme 22.24) [43, 46]. Several varia-
tions on this theme have been published, including the synthesis of 2-oxindoles [52]
(178 to 179; 180 to 181).
Solid-phase reaction conditions have been optimized for Fischer indole synthesis
(Scheme 22.25) [51]. In a single step, resin-bound arylketones 182 are treated with
a solution of an arylhydrazine (183) and ZnCl2 in glacial acetic acid at 70 C for
1820 h. Methanolysis of the resin generates indole esters 186. Because acid catal-
22.2 Solid-phase Synthesis of Heterocycles Containing One Nitrogen Atom 667
ysis is required for the reaction, the acid-stable hydroxymethylbenzoic acid (HMB)
is the preferred linker (ester linkage). A survey of resins revealed that polystyrene
was superior to polyethylene glycol (PEG) polystyrene (PEG-PS), minimizing
resin-based impurities in the indole products. Electron-rich, electron-decient, and
even sterically demanding 2,5-disubstituted hydrazines can be used in this solid-
phase synthesis.
668 22 Solid-phase Synthesis of Heterocyclic Systems (Heterocycles Containing One Heteroatom)
22.2.11
Tetrahydroquinolines
22.2.12
Quinolinones
22.2.13
Quinolines
The rst step in both sequences is an aldol condensation of 224 (225) with
an aryl ketone. Optimal reaction conditions in solution employ K2 CO3 in CH3 CN
at room temperature. Optimal reaction conditions for the solid-phase synthesis
require CH2 Cl2 /THF (1:1) and K2 CO3 at reux temperature for 2 days. Dichloro-
methane was used as a cosolvent to ensure resin swelling (polystyrene hydroxy-
674 22 Solid-phase Synthesis of Heterocyclic Systems (Heterocycles Containing One Heteroatom)
methyl resin). In solution, the nitro reduction to quinoline N-oxide (226 to 227)
was eected with SnCl2 in reuxing EtOH for 2 h. This same transformation on
solid phase requires a 4-h reaction time, and again preswelling the resin with
CH2 Cl2 is essential. The solution- and solid-phase reaction conditions for the re-
maining two-steps, deoxygenation (TiCl3 ) and aminolysis (amine, AlMe3 ), are vir-
tually identical. Solid-phase reaction monitoring was conveniently achieved via
fourier transform infrared FT-IR spectroscopy. A small parallel array of 12 quino-
lines was synthesized and the products were puried by silica gel chromatography
(yields 2065%).
The solid-phase synthesis of quinolines based on the Doebner quinoline synthe-
sis (multicomponent condensation) has been described by Gopalsamy and Pallai
[60].
22.2.14
Tetrahydroisoquinolines
22.3
Solid-phase Synthesis of Heterocycles Containing One Oxygen Atom
22.3.1
Tetrahydrofurans and g-Butyrolactams
the benzylic position of the polymeric backbone. The polyethylene spacing units in
TentaGel minimize the formation of backbone radicals. No byproducts are pro-
duced, and tin residues are conveniently washed out of the resin to give pure cy-
clized products. This chemistry is also useful for constructing dihydrobenzofurans.
As further conrmation of the observations made by Balasubramanians group,
Toru and coworkers [68] found that linear spacing groups on polystyrene support
were necessary to enable high-yielding radical cyclization of substrates 245 en route
to g-butyrolactones 247 (Scheme 22.34). The traceless linker solid-phase synthesis
begins with the preparation of b-bromoacetals (245). Intramolecular radical cycli-
zation furnishes cyclic actetals (246), whereupon g-butyrolactones are isolated after
Jones oxidation. Six g-butyrolactones were prepared using this methodology, with
yields ranging from 47% to 93%. The same six examples based on Merrield resin
with linear spacing groups gave reduced yields on the order of 2343%.
22.3.2
Furans
spectroscopy. Furans 252 are obtained in 70% yield and in 98% purity upon ther-
molytic cleavage of 251 in benzene at 70 C.
The solid-phase chemistry was applied to a split-and-pool synthesis of a small
furan library (18 members). Both aryl and alkyl groups may be incorporated into
the furan ring using a rhodium(II) peruorobutyramidate catalyst. The chemistry
appears to be limited to acetylenes possessing electron-withdrawing groups.
22.3.3
Benzofurans
Given the structural similarity of benzopyrans and indoles, many of the solid-
phase approaches developed for indoles have been applied to benzofuran syn-
thesis. Solid-phase strategies to this class of compounds utilize Pd-catalyzed hetero-
annulation of terminal acetylenes in the presence of o-hydroxyaryl iodides [77,
78]. Fancelli and coworkers prepared 254 by esterifying TentaGel S-OH resin (Mit-
sunobu reaction) with acetate-protected 3-iodo-4-hydroxy benzoic acid (Scheme
22.36) [77]. Heteroannulation with terminal acetylenes is carried out using Pd(0)
22.3.4
Pyrans, Benzopyrans, and Derivatives
In contrast to furans, there are few reports on the synthesis of pyrans and their
derivatives. Solid-phase chemistries for the construction of dihydropyrans (241)
have been reported that utilize a hetero-DielsAlder reaction of oxabutadienes
(239) with enol ethers (Scheme 22.37) [82]. Baldwin and coworkers at Pharmaco-
peia, have described the solid-phase synthesis of a dihydrobenzopyran library using
molecular tagging technology [83]. In this chemistry, resin-bound o-hydroxyaryl
methylketones are attached to resin and condensed with a variety of cyclic and
acyclic ketones. When bifunctional ketones were used, four points of diversity
were achieved (Scheme 22.38).
22.4
Solid-phase Synthesis of Thiophenes
22.5
Summary
The solid-phase syntheses of all the major classes of heterocycles containing a sin-
gle heteroatom have been described in the literature. Much of the solid-phase het-
erocyclic chemistry published thus far is synthetic methodology that is typically
680 22 Solid-phase Synthesis of Heterocyclic Systems (Heterocycles Containing One Heteroatom)
References
23
Multicomponent Reactions
Arounarith Tuch and Stefan Walle
23.1
Introduction
23.2
Mannich Reaction
because they led to better swelling of the polymer. The aldehydes tested by Ka-
tritzky were of diverse reactivity HCHO, PhCHO, and iso-BuCHO and only
secondary amines were used, whereas Showalter also used primary amines. In the
case of the less reactive benzaldehyde, harsher conditions were necessary to carry
out the reaction: toluene, reux, and a DeanStark trap for 12 h [2]. These Man-
nich adducts (1, 3) were then cleaved from the polymer-supported benzotriazole
with Grignard (4 mol equiv., THF, reux, 4 h) [2], (5 mol equiv., toluene, reux,
overnight) [3], or organozinc reagents (5 mol equiv., THF, 60 C, 12 h) [2]. After
work-up, the amines (2, 3) were isolated in good overall yields [2, 3].
In their synthesis of tertiary methylamines (8), Blaney et al. [4] (Scheme 23.3)
also used benzotriazole, but this component was not polymer supported. In this
alternative strategy, the reaction conditions are milder than those in the previous
two strategies [2, 3]. The resin-bound secondary hydroxylamine 5 reacted with an
excess of aliphatic or aromatic aldehyde (10 mol equiv.) in the presence of an ex-
cess of benzotriazole (10 mol equiv.) in CH2 Cl2 at room temperature (rt) for 18 h
to form the resin-bound benzotriazole Mannich adduct 6. The latter was then dis-
placed with Grignard reagents (10 mol equiv., THF, rt, 16 h). Quaternization of 7
(5 mol equiv. MeOTf, CH2 Cl2 , rt, 16 h) followed by base-induced cleavage (5 mol
equiv. NEt3 , CH2 Cl2 , rt, 16 h) gave very pure amines (8).
3-Aminomethylindoles [5] (Scheme 23.4), which are interesting because of their
biological activity, were prepared from polymer-bound indoles (10) by Mannich re-
action. The resin-supported 2-substituted indoles 10 obtained in three steps from
Rink amide resin and 4-amino-3-iodobenzoic acid [5b] were subjected to Mannich
reaction conditions with an excess of formaldehyde (28 mol equiv.) and secondary
amines (28 mol equiv.) at rt for 1.5 h in 1,4-dioxane/HOAc (4:1). The reactions
were reported to proceed more eectively in this mixture than in only HOAc. After
acidic cleavage, the 2-substituted 3-aminomethylindoles 11 were isolated in good
yields and purities.
[7] eciently applied this Mannich-type reaction to solid phase to prepare suc-
cessfully a 48-amino alcohol library [7b] (Scheme 23.8). Aldehyde (1.2 mol equiv.),
primary amine (1.2 mol equiv.), Drierite, and scandium triate (0.1 mol equiv.)
were stirred in CH2 Cl2 at rt for 1 h prior to the addition of the polymer-supported
silyl enol ether 22. The mixture was then stirred at rt for 20 h and, after comple-
tion of the reaction, the desired amino alcohols (24) were reductively cleaved from
the support in good overall yields.
23.3
Hantzsch Reaction
b-acetoacetate (29) was formed upon treatment of diverse amino alcohol linkers
with 2,2,6-trimethyl-1,3-dioxanone. The Hantzsch reaction was then carried out
with aminocrotonate derivatives and aryl aldehydes in DMF at 80 C. The desired
4-aryl-1,4,-dihydropyridines (30) were obtained along with the byproduct (31)
(15%). This Knoevenagel product (31) was selectively removed by washing the
resin with hydrazine. After acidic cleavage and supported liquid extraction, the 4-
aryl-1,4,-dihydropyridines 30 were obtained in modest to good overall yields (44
88%) and in good purities (7790%).
23.4
BaylisHillman Reaction
23.5
Grieco Three-component Reaction
23.6
Biginelli Reaction
A major drawback of this so-called Biginelli reaction is the moderate yield asso-
ciated with it only 2060% of the expected product is usually obtained. Since 4-
aryldihydropyrimidinones have emerged as an integral backbone of several drugs
such as calcium channel blockers or antihypertensive agents during the past few
23.7 Multicomponent Reactions with Isocyanides 693
decades, some improved procedures for their preparation have been reported, in-
cluding various solid-phase modications suitable for combinatorial chemistry
[13].
Recently, the synthesis of Biginelli compounds involving the use of BF3 OEt2 /
CuCl and polyphosphate ester leading to high yields of dihydropyrimidinones was
published [14]. The use of ferric chloride hexahydrate as a mild catalyst in ethanol
as the solvent consistently produced yields of 8090% of the desired dihydropyr-
imidinones [15].
23.7
Multicomponent Reactions with Isocyanides
23.7.1
History of Isocyanides
Xanthocillin
Leptocillin
4-((E )-2-carboxyvinyl)-3,4-epoxy-
1-cyclopentenyl isocyanide
Acenthellin 1
Aerocyanidine
these compounds were known; by 1971, more than 300 isocyanides had been de-
scribed in the literature.
The synthesis of tetrazoles by Olivieri-Mandala and Malagna in 1910 was an
important and remarkable event that took place during the period when isocy-
anide chemistry was not very popular [23]. In the 1920s Passerini worked on the
synthesis of a-acyloxy carbonamides [22, 24, 25], inventing a new type of three-
component reaction utilizing carbonyl compounds, carboxylic acids, and isocy-
anides. Again, the lack of a general method to synthesize isocyanides in good
yields postponed the general use of the Passerini reaction by another 30 years. In
1958, a breakthrough in isocyanide chemistry occurred: a high number of various
isocyanides could be synthesized in good yields by the dehydration of N-formyl
amines. Even though dozens of methods for the preparation have been described
[26], the reaction of N-formyl amines with phosgene, triphosgene, or other inor-
ganic dehydratants and matching bases is still the method of choice regarding
costs and yields of the desired products [27, 28]. Various organic bases such as
triethylamine, pyridine, quinoline, diisopropylamine, or DABCO have been used.
Depending on further functionalities, many alternative isocyanide synthesis meth-
ods have been developed, some of which are summarized in Table 23.2 [19].
23.7.2
Isocyanide Chemistry
Reference
20
29
30
31
32
33
34
the reaction of the isocyanide carbon with electrophiles and nucleophiles. (Most
other functional groups in organic chemistry react with nucleophiles and electro-
philes at dierent centers, only carbenes and carbon monoxide share this behavior
with isocyanides.) A very important eld of chemistry to which isocyanides con-
tribute is the synthesis of heterocyclic compounds. Among others, useful imida-
zoline, oxazoline [35], thiazoline, pyrrole [36], imidazole [37], oxazole [38], and
thiazole [39] syntheses have been described [40].
23.7.3
Isocyanides on Solid Phase [41]
23.7.4
Passerini Reaction
In 1921, the reaction between carboxylic acids, oxo components, and isocyanides
was described for the rst time by the Italian chemist Passerini. The reaction leads
to a-acyloxycarboxamides, which are produced in one step [24] (Scheme 23.17).
Dierent mechanisms have been discussed for the Passerini reaction [44, 45];
since the reaction is accelerated and usually carried out in aprotic solvents, a non-
ionic mechanism is indicated [46]. Most likely, the rst step of the reaction leads to
a hydrogen-bound adduct of the carbonyl compound and the carboxylic acid 39; in
the second step, the a-addition of the electrophilic carbonyl carbon and the nucle-
ophilic oxygen atom of the carboxylic acid to the isocyanide carbon forms the cyclic
transition state (40) involving all three parent compounds. The a-adduct cannot be
isolated and rearranges to the stable a-acyloxycarboxamide (41) (Scheme 23.18).
The solvent of choice for the Passerini reaction is CH2 Cl2 , and the reaction is
usually carried out at rt with a high concentration of the three components. There
are almost no limitations in the choice of the oxo component used: both aldehydes
and ketones can be utilized. Some sterically hindered and a,b-unsaturated ketones
as well as bicyclic extremely hindered ketones such as camphor do not react
with isocyanides in a Passerini reaction [47]. Besides normal C-isocyanides,
23.7 Multicomponent Reactions with Isocyanides 697
Me3 SiCN, which is in an equilibrium with Me3 SiNC, also reacts with mineral
acids in the presence of ZnI2 to give 2-hydroxy carboxylic acids and amides [48].
The treatment of ketones such as 42 with two equivalents of the isocyanide in
the presence of BF3 OEt3 in nonpolar solvents leads to the b,g-unsaturated a-
oxocarboxylic amide 43 (Scheme 23.19).
23.7.5
The Ugi Reaction
In a simplied mechanism, the oxo component and the amine react in the rst
step to the imine (Scheme 23.21). Then, the acid component protonates the nitro-
gen atom of the imine, increasing the electrophilicity of the CbN bond. In the
third step, the electrophilic iminium cation as well as the nucleophilic acid anion
add to the isocyanide carbon atom. An intramolecular acylation of the nitrogen
atom of the former amine and the subsequent rearrangement to an a-acylamino-
amide follow, forming the stable Ugi product. All the steps of the reaction sequence
are equilibria; however, the sequence for the last step lies completely on the
product side. In this respect, this type of MCR is dierent from other MCRs. The
698 23 Multicomponent Reactions
driving force for the reaction sequence is the oxidation of the isocyanide C(II)
atom to the amide C(IV) atom. Changes in the nucleophiles and electrophiles of
the components of the Ugi condensation can be observed during the reaction se-
quence: the reactive centers of the acid component and the imine change the sign
of their reactivity several times. First, the CbN bond of the imine behaves like a
base toward the acid component. Then, the protonated Schi base functions as the
electrophilic and the acid anion as the nucleophilic component of the a-addition.
In the cycloaddition and elimination that follows the addition, the reactive centers
change their signs once again. In the course of the Ugi reaction, one CaC bond
and several heteroatomC bonds are newly formed.
Reactivity tables for isocyanides have been generated; the formation of the prod-
uct with dierent isocyanides has been examined as a function of solvent and
concentration [53]. The reactivity is mainly inuenced by inductive, mesomeric,
and steric eects the last of which play the least important role; in addition, the
concentration of the reactants has a signicant inuence on the reactivity and the
products of the reaction!
The Ugi reaction can be carried out in solution as well as on solid phase. In so-
lution phase, low-molecular-weight alcohols, such as methanol, ethanol, or tri-
uoroethanol, are used as typical solvents. Aprotic polar solvents such as DMF,
chloroform, dichloromethane, THF, or dioxane have also been described in many
cases. However, the choice of the solvent has a decisive inuence on the product of
the reaction: benzoic acid, benzyl amine, dichloroacetic aldehyde, and tert-butyl
isocyanide react in methanol at 0 C almost quantitatively to form the Ugi product;
in methylene chloride the Passerini product of benzoic acid, dichloroacetic aldehyde
and tert-butyl isocyanide is formed [54].
The Ugi condensation is an exothermic reaction and usually proceeds at a fast
rate at room temperature or below; for large batches, an external cooling of the re-
action mixture is recommended. If the reactants are present in high concen-
23.7 Multicomponent Reactions with Isocyanides 699
trations, the reaction proceeds better this is typical behavior for many multi-
component reactions.
23.7.6
The Ugi Reaction on Solid Phase
The Ugi reaction has also been described on solid phase. Basically, each of the re-
actions components can be attached to resin; which component is best attached to
the solid phase certainly depends on the structure and the electronic properties of
each component and varies from case to case.
In 1997, the synthesis of small-molecule combinatorial libraries on solid phase
by the Ugi reaction was published [56]. Besides pyrroles [57, 58] and imidazoles
[58], the solid-phase preparations of small ring lactams [59], a-(dialkylamino)-
amides, hydantoin-4-imides, 2-thiohydantoin-4-imides, and 5-(1-aminoalkyl)tetra-
zoles have been described. The last are prepared by replacing the carboxylic acid
with HN3 (Scheme 23.22).
The reaction on solid phase has been carried out using NaN3 ; thus, the exposure
of various isocyanides to a range of amines, aldehydes, and sodium azide with
pyridine hydrochloride in a methanol/dichloromethane/water (3:3:1) mixture for
4 days led to the Ugi product [60]. The resin was washed with methanol and
dichloromethane (DCM), then agitated with 20% TFA in DCM. After removing
the solvent, the residues were puried by preparative thin layer chromatography
(TLC).
With the isocyanides immobilized on Wang resin, 45 was obtained by stirring
the isocyanides with aldehydes, amines, and in situ-generated HOCN [HXCN can
be conveniently generated in situ by the addition of pyridineHCl to form KXCN
(X O, S)] for 24 h in methanol/chloroform/water (5:5:1). The TFA/CH2 Cl2
cleavage provided the desired hydantoin-4-imides 46 in yields up to 80% (Scheme
23.23).
The synthesis of the a-(acylamino)amides 47 was also carried out by an Ugi re-
action, starting from pyrimidine carboxylic acids [44] (Scheme 23.24).
700 23 Multicomponent Reactions
23.7.7
Other Multicomponent Reactions with Isocyanides
During the last few years, several new multicomponent reactions have been
discovered, mostly involving isocyanides. In 1998, Bienayme and Bouzid [61],
Blackburn et al. [62], and Groebke et al. [63] described a synthesis of fused 3-
aminoimidazoles 48. Heteroaromatic amidines, such as 2-aminopyridine or pyr-
imidine, react with isocyanides and aldehydes in the presence of a catalytic
amount of protic acids in a very ecient three-component reaction to form the
corresponding imidazo-pyridines and -pyrimidines respectively (Scheme 23.25).
Many heteroaromatic amidines have been tested in this reaction with good results;
only electron-poor amidines tend to react slowly, with the frequent appearance of
side-products.
Scheme 23.29. Synthesis of the imidazole 51 on solid phase via four-component condensation.
the imidazole nucleus using the above-described methodology increases the overall
diversity and size of an imidazole library.
References
24
Strategies for Creating the Diversity of
Oligosaccharides
Pamela Sears and Chi-Huey Wong
24.1
Introduction
Carbohydrates are essential for many important biological functions [14]. When
conjugated to protein to form glycoproteins, they can alter protein structure and
function. As components of glycolipids, they can play pivotal roles in intercellular
recognition and signaling. The extracellular matrix contains proteoglycans that not
only modify the physicochemical properties of the matrix, but also are involved in
a variety of recognition processes such as cell adhesion in response to inammatory
factors and cancer metastasis. Although numerous carbohydrate structures occur
in nature, in general the role of saccharide structure in function at the molecular
level has been minimally studied. This longstanding problem can be attributed
mainly to the diculty of synthesizing saccharides, especially when compared
to proteins and nucleic. Nucleic acids can now be easily made via chemical and
biological synthetic techniques, and proteins, which are encoded by DNA, can
therefore be easily produced and manipulated through recombinant DNA technol-
ogy. In addition, automatic synthesizers are available for the synthesis of poly-
peptides and oligonucleotides. Saccharides, however, are made in nature with a
diverse set of enzymes competing to produce very diverse products [1]. There is no
information carrier that encodes a particular saccharide structure, and so creat-
ing libraries of saccharides with methods analogous to protein mutagenesis is not
possible. Furthermore, unlike proteins and nucleic acids, saccharides are more dif-
cult to synthesize chemically because (1) oligosaccharides are typically branched
rather than linear; (2) the monosaccharide units can be connected by a- or b-
linkages; and (3) oligosaccharide synthesis requires multiple selective protection
and deprotection steps, a process called protecting group manipulation.
This last requirement is quite formidable, and currently there is no general
route for combinatorial saccharide synthesis. In a glycosidation reaction, both do-
nors (monosaccharides activated for reaction) and acceptors (which receive the ac-
tivated monosaccharide) contain many hydroxyl and other functional groups that
must be dierentially and selectively protected. The product must then be se-
lectively deprotected for the next round of reactions. The complexity of protecting
24.2
Chemical Synthesis of Oligosaccharides
Several practical approaches have been taken with success for the chemical syn-
thesis of oligosaccharides (Fig. 24.1) [524]. Most involve the activation of the
anomeric leaving group with a Lewis acid and then displacement of that leaving
group by the free hydroxyl of the acceptor sugar. The KoenigsKnorr method of
coupling glycosyl halides, one of the rst techniques to gain widespread use, is still
employed [5] and most other glycosidation reagents used to date proceed by the
same basic mechanism. The relative instability of the sugar halide necessitates
the construction of the saccharide from the reducing end, and in fact, many of the
most successful approaches are those that minimize side-reactions of the activated
sugar. New leaving groups have been further developed to improve the stability of
the glycosyl donors and their reactivity. Trichloroacetimidates [6], prepared by the
reaction of free sugars with trichloroacetonitrile and base, are used most fre-
quently for coupling, as are glycosyl sulfoxides [7], phosphites [8, 9], phosphates
[10], thio-glycosides [11], and pentenyl-glycosides [12]. Another scheme for glyco-
side synthesis is to build the saccharide from the nonreducing to the reducing
end using glycals [13], which can be activated through epoxidation for either direct
attack of the epoxide with the aglycon or intermediate formation of, for example,
the thioacetal or phosphate (Fig. 24.2).
The control of anomeric conguration of the product can be complicated, espe-
cially because the reaction can occur readily via either an SN 1- or an SN 2-type pro-
cess. The anomeric conguration of the activated sugar, therefore, does not ensure
708 24 Strategies for Creating the Diversity of Oligosaccharides
Fig. 24.1. (A) Common mechanisms for phosphites will not activate thioglycosides or
glycosidation. (B) Commonly used pentenyl glycosides. Tf, triate; TMSOTf,
glycosidation reagents and their activators (in trimethylsilyl triate; TfOH, triic acid; NIS,
parentheses). Some of these glycosidation N-iodosuccinimide; Et, ethyl; DMDO, 3,3-
reagents can be used orthogonally. For dimethyldioxirane; DMTST, dimethylthiosul-
example, the activator for glycosyl uorides or fonium triate.
the anomeric conguration of the product. Furthermore, which products form can
be heavily inuenced by the protecting groups used. Acyl protecting groups at C2
can strongly direct the trans conguration at C1 by forming an intermediate diox-
ocarbenium ion (Fig. 24.1A). In general, a-1,2-cis-glycosides such as a-d-glucosides
and a-d-galactosides can be formed either by taking advantage of the kinetic
anomeric eect [14] in the displacement of glycosyl halides and thioglycosides or
by direct displacement of b-trichloroacetimidates under conditions that favor in-
version (with no participating substituent at C2 and a nonpolar solvent) [15]. b-1,2-
trans-Glycosides such as b-d-glucosides and -galactosides can be obtained by using
neighboring group eects mediated by the 2-O-acyl protecting group or polar me-
dia to favor SN 1 displacement and formation of the dioxocarbenium species. Glu-
24.2 Chemical Synthesis of Oligosaccharides 709
cosyl and galactosyl phosphates have, in all cases explored, produced the b-1,2-
trans-glycoside, regardless of the anomeric conguration of the phosphate [10],
and glycal chemistry also produces mainly the b-anomer. a-1,2-trans-Glycosides,
such as a-d-mannosides, are simple to obtain as they are favored both by the kinetic
anomeric eect and by the presence of participating groups at C2, but b-1,2-cis-
glycosides are still quite dicult to construct. Preparation of the b-d-glucoside
followed by inversion at C2 has been one common method, and recent attempts to
direct the attack of the incoming sugar by tethering it in a position that allowed
only b-attack have met with success [1619].
In general, control of anomeric stereochemistry is still a problem, especially
when neighboring group participation is lacking. Also, there are certain chem-
istries that do not work well with some sugars. In nature, only a-sialic acid link-
ages are observed, but sulfoxide and trichloroacetimidate chemistries only give the
b-anomer, a problem that can be solved by using other activating groups such as
phosphites [8, 9], thioglycosides [20], and 2-xanthates [21].
In automating oligosaccharide synthesis, it is convenient for the reactions to be
performed on solid phase. This approach allows rapid removal of reactants, rela-
tively easy purication, and (in the case of library construction) the encoding of the
product either by position (as in a two-dimensional array chip format) or, for
mix-and-split type library construction, by an accessory encoding reaction [7] in
which labels are added to the solid support as the chain is extended or by radio-
frequency-encoded combinatorial chemistry technology [22]. Most of the saccha-
ride synthetic techniques outlined above have been applied to solid-phase synthetic
strategies on a variety of supports [7, 13, 2226]. Polystyrene-based resins such as
the Merrield resin are commonly used [6, 24], although these do not necessarily
have the optimal characteristics for synthesis of sugars with regard to swelling
properties and reactant accessibility, particularly in hydrophilic media [26]. More
hydrophilic supports such as polyethylene glycol-based resins have been used with
good success [26], as have hybrid resins such as TentaGel that have a polysty-
710 24 Strategies for Creating the Diversity of Oligosaccharides
rene core coated in polyethylene. To a lesser extent, soluble supports such as poly-
ethylene glycols and derivatives as well as thermoresponsive polyacrylamide de-
rivatives [27] have been used in oligosaccharide synthesis.
However, there are many disadvantages to using a solid support earlier. As
mentioned, protecting group manipulation on a solid support is extremely di-
cult, and with the protecting group chemistry known to date, it is impossible to
create true diversity based on this stepwise solid-phase method. In addition, oligo-
saccharides and glycopeptides are sterically hindered compounds. Blocking one
side of the molecule further with a solid support is likely to drop yields dramati-
cally. Long, exible linkers can be used to alleviate this problem somewhat, but
such linkers must be both cleavable and yet still compatible with the coupling
and protectiondeprotection reactions (e.g. photo- or enzyme-sensitive linkers or
linkers which can be cleaved by Pd(0) or by olen cross-metathesis). Monitoring
reaction progress on solid phase is also not trivial. In addition, protecting group
manipulation on resins is extremely dicult, as non-soluble reagents are generally
not amenable to solid-phase synthesis. Palladium nanoparticles, however, have
been found to be useful in the debenzylation of sugars attached to a polyethylene
glycolacrylamide (PEGA) resin [28].
The most challenging task, however, is the selection of orthogonal protecting
groups and their selective manipulation during synthesis. Commonly used pro-
tecting groups include benzyl or silyl ethers and derivatives, as well as acid- or
base-sensitive protecting groups [15, 23, 29] (Fig. 24.3). Although conditions have
been developed for their selective deprotection, in general their application to the
synthesis of oligosaccharide libraries with great diversity has not been demon-
strated. To date, the largest oligosaccharide made by solid-phase synthesis is that
reported by Nicolaou et al. (Fig. 24.4) [25] and Seeberger and coworkers (Fig. 24.5)
[30]. Both groups synthesized the same branched dodecasaccharide on solid phase
using phenyl thioglycosides [25] or glycosyl phosphates and imidates [30], and the
products were released from the support with photolysis [25] or olen cross-meta-
thesis [30]. In the Nicolaou groups synthesis, trisaccharide blocks were coupled
successively, with typically 5060% yields on the coupling steps, while Seebergers
group alternated mono- and disaccharide couplings to obtain the repeating tri-
saccharide unit of the phytoalexin elicitor, and the process has been automated
using a modied peptide synthesizer. Although the individual coupling yields were
not tabulated, making direct comparison of the two strategies dicult, the overall
yield of Seebergers synthesis was very good, in excess of 50%. The approaches are
similar in principle. The diculty of these approaches is the generality of the
methods. In both cases, building blocks were tailor-made to t the synthesis of this
particular compound. The blocks used by the Seeberger group are more general
only in that they are less complex, but the pattern of protecting and activating
groups still pigeonholes them into the synthesis of a certain class of compounds,
namely b1,2/6-linked polymers. The use of either type of scheme for the general
synthesis of many dierent polysaccharides will require the maintenance of a very
large stock of building blocks that are appropriate for the construction of dierent
types of links.
24.3 Enzymatic Synthesis of Oligosaccharides 711
Fig. 24.3. Commonly used protecting groups toluenesulfonic acid; MsOH, methanesulfonic
and their removal conditions (in parentheses). acid; TFA, triuoroacetic acid; Tr, trityl; All,
See references 523 and citations therein. allyl; DMAP, 4,-N,N-dimethylaminopyridine;
Ar, aryl; DDQ, 2,3-dichloro-5,6-dicyano-1,4- Lev, levulinoyl; Piv, pivaloyl; Bn, benzyl; Phth,
benzoquinone; TBDMS, tert-butyl-dimethylsilyl; phthalimidyl; Ph, phenyl; Troc, trichloro-
TBDPS, tert-butyl-diphenylsilyl; TsOH, p- ethoxycarbonyl.
24.3
Enzymatic Synthesis of Oligosaccharides
Over the past few decades, enzymatic approaches have been gaining popularity for
the synthesis of saccharides and glycopeptides [31, 32]. Enzymes feature exquisite
stereo- and regioselectivity and catalyze the reaction under very mild conditions.
Extensive protectiondeprotection schemes are thus unnecessary, and the control
of anomeric conguration is simple. Both glycosyltransferases, the enzymes which
are naturally used to synthesize saccharides, and glycosidases, enzymes normally
used to hydrolyze glycosidic bonds, have been used. Drawbacks to an enzymatic
approach are the availability and cost of the catalysts and substrates, which can be
high. The enzymes themselves are in many cases only just becoming available,
particularly in the case of glycosyltransferases. The substrates, which for glycosyl-
712 24 Strategies for Creating the Diversity of Oligosaccharides
transferases are the nucleotide-activated sugars, are relatively expensive, but can be
prepared from sugars or sugar phosphates through enzymatic or biological meth-
ods that have been worked out [31, 33]. Glycosidases, which use cheaper substrates
such as sugar halides and p-nitrophenyl glycosides, can be used but the yields
have typically been lower. However, the Withers group recently found that muta-
genesis of glycosidases to remove one of the two catalytic carboxylates in the active
site produces an enzyme, coined a glycosynthase, that can catalyze the synthesis
of a saccharide from a uorosugar donor but cannot catalyze hydrolysis of the re-
24.3 Enzymatic Synthesis of Oligosaccharides 713
sulting product [34] (Fig. 24.6). Whether this approach will be applicable to other
exo-glycosidases remains to be investigated.
Another drawback of the enzymatic approach is that while enzymes are excellent
at catalyzing the synthesis of natural products, their ability to accept novel saccha-
rides with unusual or unnatural sugars as substrates may be poor; at best, it will
be unknown. Models for the substrate preferences of glycosyltransferases are cur-
rently unavailable, and alteration of their specicity using protein engineering has
experienced limited success. Prediction of reaction products with novel substrates
will become easier as the enzymes begin to enjoy more widespread use and their
substrate specicities become better characterized. Since the preparative scale en-
zymatic synthesis of N-acetyllactosamine involving sugar nucleotide regeneration
in the 1980s [35], enzymatic and chemoenzymatic approaches have been used in
the synthesis of a great number of oligosaccharides and glycoconjugates [32]. For
example, the synthesis of sialyl trimeric Lewis X [36] was accomplished through
714 24 Strategies for Creating the Diversity of Oligosaccharides
the transfer of sialic acid and fucose to a chemically synthesized trimeric LacNAc
acceptor. Further improvement in the area with the multiple enzymes required
for sugar nucleotide regeneration immobilized on beads has been developed (Fig.
24.7) [37]. The four enzymes required for the (re)generation of UDP-galactose
from uridine diphosphate (UDP), galactose, and phosphoenol pyruvate [with cata-
lytic amounts of glucose-1-phosphate and adenosine diphosphate (ADP)] are
coimmobilized on a bead, which can be added to the reaction medium to allow in
situ generation of the galactosyltransferase substrate, UDP-galactose. This process
both facilitates the reaction by removing the product, UDP (a galactosyltransferase
inhibitor), and can reduce the cost of the reaction by allowing the use of cheaper
substrates, assuming that the immobilized enzymes are stable enough to be reused
multiple times. However, if one or more of the immobilized enzymes is in-
activated, replacement of that enzyme will be dicult.
Application of enzymes to an automated scheme is possible. The logic of such a
reaction scheme is conceptually simple, as it is determined by the enzymes pre-
ferred reaction: the saccharide must be built stepwise, in a linear fashion, from the
reducing end (Fig. 24.8). Conducting the reaction on solid phase requires sup-
plying the enzymes in solution, from which they must be either recovered for re-
cycling or discarded. Recovery can be achieved via a variety of techniques such as
anity-based capture (of anity-tagged enzymes), passage through a microlter,
or enzyme precipitation. Enzymes are large molecules, and thus care must be
taken in choosing the support for solid-phase synthesis. The support, if porous,
should have pores large enough to accommodate these macromolecules and
24.3 Enzymatic Synthesis of Oligosaccharides 715
Fig. 24.9. Enzymatic synthesis of the ganglioside from the support. Polymer-bound
ganglioside pseudo-GM3 on a polymeric versions of GM3 have also been shown to be
support. Ceramide glycanase-catalyzed useful as picomolar inhibitors of inuenza
transglycosidation can be used to remove the hemagglutinin [48].
24.4
Programmable One-pot Synthesis
A recent approach that shows promise for automation is the use of one-pot reac-
tion schemes that use the reactivity prole of dierent protected sugars [42, 43] to
determine the outcome. The reactivity of a sugar is highly dependent on the pro-
tecting groups and the anomeric activating group used. By adding substrates in
sequence from the most to least reactive, one can assure the predominance of a
desired target compound (Fig. 24.10). Compared with the stepwise solid-phase
method, the one-pot method involves no protecting group manipulation in the
iterative process. The key to this approach is to have extensive quantitative data re-
garding the relative reactivities of dierentially protected sugar building blocks. A
large amount of reactivity data for more than 100 protected p-methylphenyl thio-
glycosides (Fig. 24.11) were recently generated and used as the basis of a computer
program, termed OptiMer, which selects the best reactants for the one-pot synthesis
718 24 Strategies for Creating the Diversity of Oligosaccharides
Fig. 24.10. (A) Traditional step-wise solid- database of relative reactivities. This approach
phase synthesis requires on-resin protecting requires preparation of a number of building
group manipulation, which can become very blocks with their glycosidation reactivities
complicated as the number of glycosidic quantitatively measured. A reactivity dierence
linkages increases. (B) OptiMers one-pot greater than 1000 between the building blocks
approach. OptiMer is a program which will give a high-yield in coupling. No protecting
predicts the optimal type and order of addition group manipulation and intermediate isolation
of partially protected sugars, based on a are required during the one-pot synthesis.
Fig. 24.12. Synthesis of the cancer antigen reactive one, in the presence of an activator.
Globo-H using OptiMer technology. In brief, The product obtained is then puried and
the sequence of Globo-H is entered into the deprotected to give the target. Ac, acetyl;
computer, which predicts the best building Bn, benzyl; Bz, benzoyl; Tol, tolyl; Troc,
blocks to be used. These building blocks are trichloroethoxycarbonyl.
then mixed in sequence, starting with the most
scheme. These reactions are typically performed in solution, but, in order to facili-
tate removal of reactants at the end, the nal acceptor may be attached to solid
support or other aglycons.
Future development in this approach is to expand the building-block repertoire
and to ensure its applicability in programmable one-pot synthesis. Compared with
stepwise solid-phase synthesis, the one-pot approach requires protecting group
manipulation only at the stage of building-block synthesis, and thus holds greater
potential for automation and for a greater diversity of oligosaccharide structures.
Other solution-phase syntheses of oligosaccharide libraries have been reported,
especially the method for the synthesis of mixtures [47], but the methods have
not been demonstrated to create great diversity. In addition, characterization of the
mixture represents a dicult problem.
24.5
Conclusions
for the synthesis of oligosaccharides. Future eorts to reach this goal include,
for example, the development of new methods and strategies for rapid assembly of
oligosaccharides, the design of new orthogonal protecting groups, and the simpli-
cation of protecting group manipulation. Development of such automated sys-
tems easily accessible to both biologists and chemists will have a signicant im-
pact on our understanding of carbohydrates in biological systems and on the
development of new therapeutics to modulate or to control carbohydrate functions.
References
1 P. Sears, C.-H. Wong, Cell. Mol. Life Large, D. G., Warren, C. D. (eds),
Sci. 1998, 54, 223252. Dekker, New York 1997, pp. 207
2 A. Varki, Glycobiology 1993, 3, 97 243.
130. 16 D. Crich, S. Sun, J. Am. Chem. Soc.
3 H. Lis, N. Sharon, Eur. J. Biochem. 1998, 120, 435436.
1993, 218, 127. 17 F. Barresi, O. Hindsgaul, J. Am.
4 T. Kolter, K. Sandhoff, Angew. Chem. Soc. 1991, 113, 83769377.
Chem. Int. Ed. Engl. 1999, 38, 1532 18 G. Stork, J. J. La Clair, J. Am. Chem.
1568. Soc. 1996, 118, 247248.
5 H. Paulsen, Angew. Chem. Int. Ed. 19 Y. Ito, T. Ogawa, J. Am. Chem. Soc.
Engl. 1982, 21, 155173. 1997, 119, 55625566.
6 J. Rademann, A. Geyer, R. R. 20 P. J. Garegg, Adv. Carbohyr. Chem.
Schmidt, Angew. Chem. Int. Ed. Engl. 1997, 52.
1998, 37, 12411245. 21 H. Lonn, K. Stenvall, Tetrahedron
7 R. Liang, L. Yan, J. Loebach, M. Ge, Lett. 1992, 33, 115116.
Y. Uozomi, K. Sekanina, N. Horan, 22 K. C. Nicolaou, X. Y. Xiao, Z.
J. Gildersleeve, C. Thompson, C. Parandoosh, A. Senyei, M. P. Nova,
Smith, K. Biswas, W. C. Still, D. Angew. Chem. Int. Ed. Engl. 1995, 34,
Kahne, Science 1996, 274, 15201522. 22892291.
8 T. J. Martin, R. R. Schmidt, 23 P. H. Seeberger, W.-C. Haase, Chem.
Tetrahedron Lett. 1992, 33, 61236126. Rev. 2000, 100, 43494393.
9 H. Kondo, Y. Ichikawa, C.-H. Wong, 24 J. T. Randolph, K. F. McClure, S. J.
J. Am. Chem. Soc. 1992, 114, 8748 Danishefsky, J. Am. Chem. Soc. 1995,
8750. 117, 57125719.
10 S. Hashimoto, T. Honda, S. 25 K. C. Nicolaou, N. Watanabe, J. Li,
Ikegami, J. Chem. Soc. Chem. J. Pastor, N. Winssinger, Angew.
Commun. 1989, 685. Chem. Int. Ed. Engl. 1998, 37, 1559
11 R. J. Ferrier, R. W. Hay, N. 1561.
Vethaviyasar, Carbohydr. Res. 1973, 26 P. M. St. Hilaire, M. Meldal, Angew.
27, 5561. Chem. Int. Ed. Engl. 2000, 39, 1162
12 B. Fraser-Reid, J. R. Merritt, A. L. 1179.
Handlon, C. W. Andrews, Pure Appl. 27 X. Huang, K. Witte, D. E.
Chem. 1993, 65, 779786. Bergbreiter, C.-H. Wong, Adv.
13 S. J. Danishefsky, K. F. McClure, J. Synth. Catal. 2001, 343, 17.
T. Randolph, R. B. Ruggeri, Science 28 O. Kanie, G. Grotenberg, C.-H.
1993, 260, 13071309. Wong, Angew. Chem. Int. Ed. Engl.
14 R. U. Lemieux, K. B. Hendriks, R. V. 2000, 39, 45454546.
Stick, K. James, J. Am. Chem. Soc. 29 B. Ernst, G. W. Hart, P. Sinay,
1975, 97, 40564062. Carbohydrates in Chemistry and
15 S. L. Flitsch, G. M. Watt in: Glyco- Biology, Wiley-VCH-Verlag, Weinheim
peptides and Related Compounds. 2000.
722 24 Strategies for Creating the Diversity of Oligosaccharides
Part IV
Molecular Design and Combinatorial
Compound Libraries
25
Design Criteria
Josef Pernerstorfer
25.1
Introduction
25.2
Properties of Combinatorial Libraries for Drug Development
ity. These authors identied: (1) molecular weight, (2) lipophilicity, (3) number of
hydrogen bond donor groups (i.e. number of NH and OH bonds), and (4) number
of hydrogen bond acceptors (i.e. number of N O) as the key properties which
have an essential eect on the permeation through lipid bilayers and therefore on
the intestinal absorption process. Properties 1, 3, and 4 are obviously easily calcu-
lable from the structural formula. As a measure of lipophilicity, Lipinski and co-
workers used the partition coecient of the substance between n-octanol and
water, P, which can be approximated by increment systems. The authors used two
such systems. The rst was the Pomona College Medicinal Chemistry program
which calculated log P values (C log P) from structural fragments and gave very
accurate results, but failed in many instances because of fragments in the mole-
cule which were not assigned in the program. The other system they used had
been described by Moriguchi et al. [2]. This system gives less but still reasonably
accurate results and allows the calculation of M log P directly from the structural
formula [2].
Lipinski et al. determined the distribution of these four properties among the
drug database and found cut-o levels for each parameter such that @90% of the
drugs were within these levels. From these, they stated that poor absorption and
permeation are more likely when:
Tab. 25.1. Distribution of M log P and molecular weight in the pre- and post-HTS era [1].
Percentile M log P MW
more lipophilic and heavier. Both these factors led to the discovery of a number of
highly active compounds in vitro but caused severe problems in developing orally
bioavailable drugs owing to insucient solubility and permeability in vivo. Ob-
viously, the strategy in combinatorial chemistry of taking a multifunctional central
building block and decorating it with organic residues runs the risk of generating
very high MW compounds which are furthermore very lipophilic since all polar
groups (acids, amines, alcohols, and phenols) would be masked by combinatorial
variation (acid amides, ureas, ethers, esters). Although this can be useful during
the late stages of the optimization process, it is not a good procedure for generat-
ing lead compound libraries.
In accordance with these results, Teague et al. [3] studied the drug optimization
process at Astra Zeneca. They found that only very few lead compounds have such
desirable properties and that it would be very unlikely for one of them to pass on
to clinical development directly. Generally, potency and pharmacokinetic prole
have to be further optimized this goal is accomplished by the addition of suitable
groups and side-chains and causes an increase in MW and lipophilicity of the drug
compared with the initial lead. Teague et al. divided the properties of lead struc-
tures identied by HTS into three groups with respect to anity, MW, and log P:
1 Intermediate anity (> 0.1 mM), low MW (< 350), and low C log P (< 3). These
are classical lead structures which can be optimized by introducing lipophilic
groups to increase potency and improve pharmacokinetic properties.
2 High anity (f 0.1 mM), high MW (g 350), and low C log P (< 3). These leads
are often derived from natural products and can be optimized by derivatization
to improve pharmacokinetics while potency is retained.
3 Intermediate anity (> 0.1 mM), high MW (g 350), and high C log P (> 3).
These leads are often generated in HTS assays of combinatorial libraries. An
optimization of these drug-like leads normally proves dicult, since the anity
results from many nonoptimized interactions between lead and target and a
further optimization produces very lipophilic, poorly soluble compounds.
Teague et al. concluded that combinatorial libraries for lead identication should
be fundamentally dierent from those for lead optimization. Lead structure li-
braries have to produce compounds with low MW (100350) and low C log P (1.0
3.0). Focused libraries based on such leads can quickly improve drug properties.
728 25 Design Criteria
25.3
Dierentiation of Drug-like and Nondrug-like Compounds
Since an accessible virtual library is much larger than one that is actually syn-
thesized, the rst criterion when choosing compounds for synthesis should be
that they have reasonable drug-like properties. For example, toxic and reactive
substructures which can react with proteins can cause false-positive read-outs in
screening assays and should, therefore, be eliminated (Fig. 25.2) [5].
Several authors have tried to dierentiate drugs from nondrugs by looking at the
25.3 Differentiation of Drug-like and Nondrug-like Compounds 729
Fig. 25.2. Reactive functional groups leading to in vitro false positives [5].
25.4
Diversity in Combinatorial Chemistry for Drug Development
25.4.1
Introduction
25.4.2
Descriptors
very high. Therefore, sometimes rigid descriptors are used, which only consider
one or a few minimum energy conformations of the molecule.
25.4.3
Selection Algorithms
After diversity has been assessed using a descriptor system, compounds can be
subdivided into classes which cover the available chemical space with representa-
tive molecules. A widely used system to evaluate the similarity between the bit-
strings of two compounds is the Tanimoto coecient. If two molecules have A and
B bits set in their fragment bitstrings and C of these in common [27], then:
Tanimoto C=A B C
The Tanimoto coecient can adopt values between 1 (identical substances with
respect to the descriptor system) and 0 (no similarity).
Using this similarity metric, the set of compounds can be subdivided into clus-
ters of similar compounds which are dierent from the members of other clusters.
There are two categories of clustering: hierarchical and nonhierarchical. Hierar-
chical clustering can be divided into agglomerative and divisive approaches. The
agglomerative approach begins with all the molecules as singleton clusters and
combines these gradually into larger clusters by successively joining the most
similar compounds. Divisive clustering begins from one large cluster containing
all entities and splits this cluster successively into smaller clusters of dissimi-
lar compounds. Both methods continue until the desired number of clusters
is reached. An example of a nonhierarchical clustering algorithm is the Jarvis
Patrick method. For this approach, in the rst stage, the K nearest neighbors for
each member are calculated, according to the Tanimoto coecient. In the next
step, clustering is performed. Two structures are assigned to the same cluster,
when the following three conditions are met:
25.4.4
Diversity Assessment
There is much discussion in the literature about the relevance of dierent de-
scriptors and diversity selection processes [2835]. Although 3D descriptors can
encode more information on molecules, it was shown that 2D descriptors perform
better in separating biologically active molecules from inactive ones. Hierarchical
clustering in combination with MACCS 2D descriptors performed best in several
dierent datasets of enzyme assays, high-throughput screening results, and activ-
ities in an monoamine oxidase assay. JarvisPatrick clustering was shown to be
rather ineective since it generated either singleton clusters of actives or very large
clusters containing the actives together with many inactives. Using hierarchical
clustering, it was possible to separate active clusters containing 3050% actives on
average whereas only 0.82% of the whole dataset was active. It appeared that a
reasonable cluster size is around ve molecules per cluster, which means that
about 20% of a dataset still has to be sampled in order to obtain valid results [21].
The MACCS structural keys also performed best in predicting hydrophobicity,
electrostatics, sterics, dispersion interaction, and hydrogen bonding of molecules.
These factors are relevant for ligandreceptor interaction and molecular recogni-
tion. This result is the reason why this descriptor system performs well in biologi-
cal clustering [36].
Another statistical evaluation of random selection methods and of designed
compound subsets for biological screening systems proved that the latter were, in-
deed, superior. About 3.5 times more compounds have to be sampled randomly to
obtain results equivalent to a rational design approach. For initial screening, a Ta-
nimoto similarity radius of 0.7 between similar compounds in the subset is ex-
pected to be sucient to cover relevant biological eects, whereas, for lead rene-
ment, a Tanimoto similarity radius of 0.85 is recommended [37].
The question of which level of diversity of combinatorial libraries should be
assessed has also been widely discussed in the literature. Three main strategies
have been proposed [38]: reactant-based design means that the diversity of the re-
agents is evaluated and these diverse reactants are assumed to give a diverse prod-
uct library. The number of reagents is orders of magnitude lower than the number
of products, therefore this design process is very fast. However, it is not guaran-
teed that a maximum diverse set of reagents also provides a maximum diverse set
of products. Therefore, product-based design uses fully enumerated combinatorial
libraries and picks the most diverse products thereof. During this process, the com-
binatorial nature of the products gets lost and, therefore, the process loses e-
ciency. A compromise between these two approaches is a reactant-biased/product-
based algorithm. This means maintaining the combinatorial nature of chemistry
at the reactant level, but optimizing the diversity at the product level by use of
mathematical models such as genetic algorithms and simulated annealing [39
41]. A general problem with these product-based approaches is that diversity has to
be calculated for the whole virtual library, which needs very large computational
25.5 Privileged Structures 735
25.5
Privileged Structures
25.5.1
Introduction
When Evans et al. [44] were looking for new cholecystokinin (CCK) antagonists in
1988, they introduced the concept of privileged structures for structural motifs
which are capable of providing useful ligands for more than one single target. As
shown in Fig. 25.5, the benzodiazepine skeleton is a recurring structural motif ad-
dressing not only CCK but the structurally similar Triuadom 2 shows high an-
ity to the opiate receptor [45] and 3 is an NK-1 antagonist [46].
This nding contradicts the common principles of individuality and selectivity of
biological receptors. According to Evans et al., a possible explanation for this phe-
nomenon might be that these G-protein-coupled receptors descend from the same
ancestral genes and that they still remain structurally related. Similarly, Wiley and
Rich [47] have reasoned that the similar structure of G-protein-coupled receptors
which are characterized by seven transmembrane a-helices might favor a recurring
binding element. These hydrophobic pockets might be complementary to the dif-
ferent conformations of privileged structures [47]. Therefore, libraries of such
privileged structures can give rise to a number of lead compounds in new screen-
ing assays [26, 48, 49].
25.5.2
Further Examples of Privileged Structural Motifs
25.5.3
Substructure Analysis of Drugs
In a paper by Bemis and Murcko [61], a database of 5120 known drugs [Compre-
hensive Medicinal Chemistry (CMC)] was analyzed for the molecular frameworks
of the drugs. These authors dened the framework of a drug as the union of all
ring systems and linker atoms on the direct path between the ring systems. On the
basis of the two-dimensional representations of the drugs (ignoring atom type, hy-
bridization, and bond order), they found 1179 dierent frameworks, but only 32 of
these accounted for half of all frameworks. A few representative samples of these
are outlined in Fig. 25.11. The numbers show the frequency of each framework
within the database.
Although the trivial six-member ring appears most often, the other frameworks
indicate a clear preference for certain ring systems. The aforementioned 5-phenyl-
1,4-benzodiazepine also belongs to these privileged frameworks (45 occurrences).
With regard to the atom type, hybridization, and bond order, Bemis and Murcko
found more than 2500 dierent frameworks in the database, but only 42 of these
accounted for 25% of the total. In another paper [62], the same authors studied
the frequencies of side-chains in the CMC database. The side-chains are all parts
of the drug which do not belong to the framework. They found an average number
of four side-chains per molecule, including the trivial oxygen side-chain of the car-
bonyl groups. When ignoring the carbonyl group (3545 occurrences) as a side-
chain, only 20 dierent side-chains account for 75% of the total. Examples of these
are given in Fig. 25.12. The number gives the frequency of occurrence among the
5090 drugs studied.
These data clearly show that certain substructures within a drug appear more
often than others, a nding which can aect the design of new drug-like libraries.
An interesting application of these ndings is the Retrosynthetic Combinatorial
Analysis Procedure (RECAP) [63]. In this approach to the design of biased com-
binatorial libraries, all compounds in the World Drug Index (WDI) were broken
down into fragments by dissection of 11 classes of bonds which are easily formed
by combinatorial chemistry until structurally nontrivial building blocks were ob-
tained (Fig. 25.13). Trivial residues such as methyl ethers were not broken down
further so that pharmacophoric information was not lost.
Clusters of compounds which exhibit potency toward a given biological target
were examined for common building blocks. In the next step, the frequency of
these building blocks in the specic cluster was compared with the frequency in
740 25 Design Criteria
the total WDI in order to dierentiate pharmacophoric subunits from trivial com-
mon substructures. Combinatorial libraries using these building blocks are ex-
pected to lead to rapid identication of novel hits and lead compounds for a given
biological target. A possible drawback is the similarity of the compounds synthe-
sized in this way to known drugs. Many leads produced according to this approach
might therefore be under patent protection. Furthermore, it is unlikely that new
bioisosteric pharmacophores will be produced using this method.
25.6
Conclusion
References
26
Estimation of Physicochemical and ADME
Parameters
Michael W. Harter, Jorg Keldenich, and Walter Schmitt
26.1
Introduction
The advent of combinatorial chemistry has added a new momentum to the early
stages of drug discovery and drug development. Nowadays, combinatorial libraries
are produced that serve dierent purposes in the drug discovery process. Random
libraries are generated with the aim of identifying one or more hits, i.e. identi-
fying one or more members of a library which bind to a given target in a screening
approach. Here, the term random is not precisely dened: complete randomness
is never realized. The preference for a certain template structure and the limi-
tations set by the chemistry that can be performed with the given template as well
as practical reasons confer restrictions even on a random library. Once a hit (or an
island of activity) has been identied, so-called focused libraries might be syn-
thesized in order to explore the chemical space close to that hit. These focused li-
braries are usually smaller in size and are used to gain structureactivity relation-
ship (SAR) information. The properties of the initial hit are usually improved and
a member of the focused library might serve as a lead structure; eventually after
repeated cycles of optimization a compound can be obtained that fullls the re-
quirements of a drug candidate and enters the stages of preclinical and clinical
development.
In the early days of combinatorial chemistry, library design centered exclusively
on the aim of nding molecules which bind to a given target or optimize binding
anity, as mentioned above. There has been debate in the pharmaceutical indus-
try as to why the number of successful development candidates has not grown at
the same or a similar pace as the number of molecules which are synthesized and
tested for pharmacological eects. A possible answer to that question might be the
limited attention that has been paid during library design to the overall properties
of the intended test compounds. This response is surprising since it has been
known for a long time that it takes more than just one pharmacophore to produce
a potential drug molecule. Assessing the reasons why a drug candidate fails in
preclinical development (i.e. the development stage that follows the discovery
stage) might be helpful in realizing any additional properties that might be neces-
sary for a molecule to become a successful drug candidate [1, 2]. Beside commer-
cial and strategic reasons, which consitute about one-third of all discontinued
projects at that stage, toxicity and safety issues cause a further 30% of all drop
outs. Another 17% of all discontinued development projects are due to poor phar-
macokinetics or an unfavorable physicochemical compound prole. Depending on
the source of information, this gure rises up to 3550% when all stages of pre-
clinical and clinical drug development are considered. It is this last fraction of
molecules that we will discuss in this chapter.
26.2
ADME/PK Considerations in Combinatorial Library Design
It is only very recently that the need to design libraries of drug-like molecules
has become a central topic among the practitioners of combinatorial or parallel
synthesis in the life science arena [3, 4]. A number of rules have been estab-
lished from the analysis of various databases in order to aid chemists in library
design with respect to drug-likeness [59], some of which were discussed in
Chapter 25. In some companies, rules of that kind have been implemented in
software tools that are used as guides in library design. But what exactly does
drug-likeness mean?
Most potential drugs are intended for oral application. In order for a compound
to reach its site of action, it has to cross several barriers within the body. After in-
take, it has to be suciently water soluble in the lumen of the gastrointestinal (GI)
tract for it to be absorbed. It also has to be suciently stable in this environment,
which has, for instance, marked dierences in pH values. On the other hand, the
compound should not be too hydrophilic or too bulky because then it might re-
main in the gut lumen and eventually be excreted rather than being absorbed.
Hence, some degree of lipophilicity is required for a compound to be absorbed
through the lipid membrane of the gut wall, but compounds with a molecular
volume greater than a certain threshold range are restricted from being (passively)
absorbed. However, if the compound is too lipophilic, it might remain in the
membrane or stick to other biological material in the gut and might not reach the
portal vein. In the following sections, the factors which govern absorption, for ex-
ample permeability and solubility, will be discussed in greater depth. After being
absorbed, a compound has to pass through the liver, where it might be metabo-
lized, it might be transformed into active, inactive or even toxic metabolites, or it
might be excreted via the bile. Having passed through the liver, the compound is
part of systemic circulation. As such, it has to be stable against plasma components,
e.g. certain enzymes. It has to be distributed within the body and must penetrate
the respective target tissue. For that purpose, it is likely that the compound again
has to cross one or even several lipid membranes. Simultaneously, the compound
is subject to continued hepatic and renal clearance.
From this brief description, it is obvious that a drug is more than just a phar-
macophore. It must possess properties that allow a sucient proportion of the ad-
26.3 Estimation of ADME/PK from Physicochemical Parameters 745
ministered amount of drug molecules to reach their site of action in order to trig-
ger the desired pharmacological eect. In recent literature, these properties are re-
ferred to as ADME/PK properties, where ADME stands for absorption, distribu-
tion, metabolism, and excretion and PK stands for pharmacokinetics.
Once the importance of ADME/PK properties has been recognized, it is desir-
able to incorporate ADME/PK considerations early into the hit- or lead-nding
process because the likelihood of successful optimization is higher and the time it
would take to nd a suitable drug candidate is shortened when the starting point
already possesses some quality features [10]. Especially in library design, it would
be advantageous to lter virtual libraries in order to select a sublibrary containing
discrete members with some minimum likelihood of being orally bioavailable and
to synthesize and screen just the subset [3]. For that purpose, tools are needed
which allow an estimation of the ADME/PK properties from the chemical struc-
ture [11]. The physicochemical properties of a compound that underlie ADME/PK
behavior and how these can be predicted are discussed below. A review that sum-
marizes the current status of the entire eld of ADME/PK optimization was pub-
lished recently [12].
26.3
Estimation of ADME/PK from Physicochemical Parameters
culated from the structure. For library design, these calculations have to be fast
enough to calculate the property space even for very large virtual libraries.
Most of the ADME/PK properties at least those which are driven by passive
processes can be explained by appropriate physiological and physicochemical
models having physicochemical parameters as their input. Some ADME/PK pa-
rameters are partly or exclusively dependent on active processes with partially un-
clear or insucient scientic background and, therefore, are not covered in this
chapter. Physiological and physicochemical models have the advantage that they
can be generalized much more easily than QSAR models and are less dependent
on the molecular space of the compounds they are derived from.
The aim of this chapter is to describe briey some of the models that explore the
physicochemical parameter space in which drug-like molecules are found.
26.3.1
Models for Permeation Through Membranes (Absorption)
Biological membranes are built up either by complex cell layers as they are found
in the gut wall, the skin, and the endothelia cell layers dividing organs from the
blood cycle (e.g. bloodbrain barrier), or by lipids and proteins dividing the cell
plasma from its environment. Overviews of in vitro models for measuring perme-
ation through such membranes either in vitro or in vivo are given in the literature
[1619].
The physicochemical model used most frequently today is one where the per-
meation of compounds through membranes is described as follows [20, 21]. The
model consists of a layer of thickness X with area A (the membrane) dividing
aqueous compartments from each other. Adjacent to both sides of this layer, un-
stirred water layers (thickness Y,Y 0 ) are located. The transport of a compound
through such a model layer is then described by two phenomena:
. diusion inside the water layers and inside the membrane, and
. distribution into and out of the membrane.
The permeation constant Pe for this process depends on the diusion coe-
cient (Df ) of the compound in the water layer (Dfa ) and in the membrane (Dfm ) as
well as on the distribution coecient of the compounds to the membrane Km . The
mathematics have been well described in the literature and are beyond the scope
of this chapter [21].
For the chemist, real biological barriers have the disadvantage that they are not
well dened with respect to such a model. The necessary parameters for exam-
ple thickness of the individual layers, or area of the membrane are unknown or
can only be roughly estimated. A practical approach toward this problem is the t
of Eq. (1) to permeation coecients from in vivo or in vitro measurements [20].
Dfa; m Km
Pe A cm s1 1
Dfa BDf m Km
26.3 Estimation of ADME/PK from Physicochemical Parameters 747
MVab DpH
Pe A cm s1 2
MVa B MVb DpH
The parameters MV, which can be substituted by MW, and DpH can be calcu-
lated from structure (see Section 26.3) and are therefore key parameters in library
design for assessing the permeation of compounds through biological barriers.
Examples of the use of such a model for gut wall permeation (absorption) [20], in-
testinal cell layer [22], bloodbrain barrier [23], and skin [24] are given in the lit-
erature.
In addition to absorption, a term characterizing paracellular (pore) transport for
very small compounds (MW < 250 g mol1 ) has to be included (Eq. 3).
As a proof, Eq. (3), Fig. 26.1 shows a comparison of 133 compounds with mea-
sured fraction dose absorbed in humans [2529], and predictions have been made
by tting Eq. (3) to these data. As input, only measured lipophilicity (membrane
anity, see Section 26.4.1) and MW were used. The fraction dose absorbed is cal-
culated from Pe (using Eq. (9) in [30]) for compounds exhibiting complete solu-
bility while being absorbed.
The t parameters of Eq. (3) used for Fig. 26.1 are:
A B C D a b g
3940 1:1 10 7 2:5 107 202 0.6 4.25 16
Some authors go beyond such a model and further divide the distribution coef-
cient into volume-dependent and polarity terms, which in most cases are related
748 26 Estimation of Physicochemical and ADME Parameters
An oral drug transport and absorption model has been described in the literature
[44].
Solubility itself is dicult to calculate directly from structure because the crystal
and dissolution properties of a real compound depend also on the presence of
varying amounts of impurities which cannot be modeled well with simple and fast
calculation algorithms (see Section 26.3). Moreover, high-quality datasets with
measured solubility over a wide range of chemical space are not yet available. Last,
but not least, the solubility in the intestinal environment is dicult to dene, be-
cause compounds are administered in tablets with dierent vehicles. Even when
solutions are administered, it is not possible to know whether the compounds stay
in solution or whether they precipitate; if they do precipitate, it is not possible to
know to what extent this takes place (e.g. in the stomach) nor to know which crys-
tal form or the crystal diameter. Therefore, until now, in most cases it has only
been appropriate to apply solubility in a qualitative way.
Nevertheless, it is a very important parameter, and it is common sense that a
compound with low solubility will have absorption problems especially when
higher doses are needed. The limit of when the solubility is suciently high de-
pends strongly on the vehicle used, the animal species, the dose, the volume ad-
ministered with the compound, and the permeability of the gut wall.
26.3.2
Models for Distribution in the Body
A prerequisite for the action of a drug in vivo is that it must not only enter the
systemic circulation, because the site of action is only rarely found in the blood it-
self, but the drug must also be distributed into the target organs. A drugs concen-
tration at its target is important for the strength of the pharmacological response.
Depending on the mode of action, this requirement can be a mean tissue concen-
tration or an unbound concentration, which is the concentration of free drug that
is not bound to lipids or proteins. The last case is of interest if, for example, the
binding site for the drug is available directly from the aqueous phase of the tissue
(cytoplasm or interstitial uid). While the unbound concentration can in the rst
instance for all organs be regarded as being equal to the free concentration in
plasma (which is precisely true for steady-state conditions), the mean concentra-
tion is dependent on the composition of the organs. For an estimation of the con-
centration at the site of action, the steady-state distribution coecient between tis-
sue and plasma K tp C tissue =Cplasma is often very helpful.
As is evident from this discussion, no general rule can be drawn for optimum
distribution behavior. This fact is also true for binding to lipids and proteins. The
optimum properties depend very much on the purpose for which a particular
active compound shall be optimized. A very simple and obvious example is that of
a membrane-bound protein with a lipophilic binding site. It is desirable to design
compounds with a certain degree of lipophilicity in order to achieve high concen-
trations at the binding site. In contrast, for cytosolic proteins (in solution in the
cytoplasm) the opposite is true. In this case, a lower lipophilicity leads to a higher
750 26 Estimation of Physicochemical and ADME Parameters
unbound concentration, which is then available at the target. As a result, the dis-
cussion in this section is more important if a library is to be designed for optimiz-
ing a lead structure for a known target and is less important for random libraries.
Nevertheless, it is very desirable to estimate the unbound fraction in plasma
( fup ), which determines the free concentration (Cfree plasma fup Cplasma ), and the
distribution coecients as early as possible. Both fup and the K tp can be deter-
mined experimentally. The rst by separating the protein fraction of plasma sam-
ples and determining the concentration in the water phase, and the latter from
in vivo animal experiments or in vitro, e.g. by microdialysis of the substance
from plasma into homogenized tissue. However, these experiments are rather
time-consuming because in all cases quantitative analyses of concentrations in
complex biological matrices have to be carried out. Thus, they cannot be carried
through for large numbers of compounds in the early stages of research. This
problem is especially true for animal studies. Methods of estimating this distribu-
tion behavior from other properties or even from the chemical structure directly
are therefore very desirable.
The tissue/plasma distribution coecients are determined by the anity of a
drug to the dierent constituents of the tissues and the blood. In some cases, spe-
cic binding to biological macromolecules may play a signicant role. However,
generally, distribution between lipids and water as well as unspecic adsorption
to proteins are the main factors determining K tp . It is generally accepted that
these eects are related to other distribution coecients that describe a drugs
lipophilicity, therefore it may be considered that K tp can be correlated with such
lipophilicity descriptors. Several attempts to estimate the tissueplasma distribu-
tion coecients depend on this relationship and can be found in the literature (see
below).
The simplest approach is to nd a correlation function between K tp and, for ex-
ample, the octanolwater distribution coecient using tting [4547]. The core of
the correlation functions is a potential function in both cases. Because this
approach alone is not sucient to describe the correlation correctly, additional pa-
rameters have to be incorporated. These are the fraction of nonionized compound
[45] or some information about the tissue composition, either in the form of water
content and a so-called binding capacity factor [46], or as water and lipid content
[47]. Several other approaches take the composition of the organ tissues as the
starting point for the calculation of K tp . In most cases, only the fractions of water
and dierent lipids (neutral and phospholipids) are taken into consideration [48
53]. The binding to proteins in the organs and plasma is explicitly taken into ac-
count only in [54], whereas in [51, 52] protein binding is incorporated only in the
form of a free fraction in plasma. Generally, the calculation of K tp relies on linear
functions in which the dierent fractions of the tissue components are multiplied
by a parameter describing the distribution between the respective component and
water. The general form is shown in Eq. (4),
with
26.3.3
Models for Clearance and Metabolism
Clearance is dened as the irreversible removal of a drug substance from the body,
the rate constant that correlates the elimination rate of a drug with its plasma
concentration. In principle, a compound can be eliminated in two dierent ways:
excreted unchanged in its original chemical form, or modied chemically, i.e. me-
tabolized, with the metabolite(s) still being part of the systemic circulation. The
main metabolizing organ in the human body is the liver, but it should be noted
26.4 Estimation of Physicochemical Parameters 753
that metabolism also takes place in other parts of the body, such as the gut wall
[61, 62]. Hepatic metabolism is characterized by so-called phase I and phase II
transformations. Oxidation, reduction, and hydrolyses are all phase I reactions that
are employed by the liver to introduce functional groups into (mostly) lipophilic
drug molecules. These reactions are catalyzed by various enzyme families. Among
the oxygenases the cytochrome P450-dependent monooxygenases are the most
important ones [63]. Phase II reactions are conjugation reactions. Drug molecules
or their phase I metabolites are coupled with endogenous, highly hydrophilic mole-
cules such as glucuronic acid or sulfate anion in order to enhance their solubility
in water. As a result, urinary excretion is accelerated. Thus, it is not surprising that
lipophilicity is a useful physicochemical parameter for the estimation of renal
clearance. Renal clearance happens by ltration of blood serum at the glomerulus
into the kidney tubules from which reabsorption of most of the water content and
solutes occurs. If the logarithm of the distribution coecient log D at pH 7.4 (i.e.
log D7:4 ) between octanol and aqueous buer of a drug molecule (or its metabolite)
is <0, reabsorption is almost completely prevented and the molecule will be ex-
creted in the urine. However, if the molecule is strongly bound to plasma proteins,
it will escape rapid urinary excretion even if its lipophilicity is low. This feature
occurs because only the unbound fraction of a drug molecule in the plasma is l-
tered into the kidney tubules. Most drug molecules, however, are rather lipophilic,
rendering them subject to metabolic (hepatic) clearance. Thus, predicting the
metabolic stability of a molecule could be of great value in compound design.
Of all the physicochemical parameters, lipophilicity was linked rst to metabo-
lism by generating QSAR models for cytochrome P450 data [64, 65]. Unfortu-
nately, these and other examples of correlating physicochemical parameters with
metabolic stability [66, 67] do not seem to be widely applicable [67]. They work
quite well within a series of highly analogous compounds, but lose almost any
predictive power when dierent chemical entities are considered [11]. This dichot-
omy might reect the complex nature of these multienzyme-catalyzed biotrans-
formations.
Further approaches have been taken to predict metabolism, including three-
dimensional QSAR, pharmacophoric, and protein-modeling and articial neural
networks [6871]. All of these dierent approaches share a rather limited reliabil-
ity when they are used to predict the clearance of molecules that are not very sim-
ilar to the training set from which the models were derived. But, as metabolic
databases become larger, combinations of the above methods and perhaps also
new tools will arise and make in silico metabolism screening a part of an integrated
approach toward lead structure identication and optimization.
26.4
Estimation of Physicochemical Parameters
The estimation of physicochemical parameters has a long tradition [72]. The esti-
mation methods range from those using physicochemical theory, for example the
754 26 Estimation of Physicochemical and ADME Parameters
theory of ideal solutions, to those using molecular structural elements and corre-
lation equations with experimentally available parameters, e.g. log P.
The aim of this chapter is not the detailed description of how to measure and
predict the relevant physicochemical parameters. As such, the reader is referred to
handbooks and overviews [56, 73]. Instead, this chapter gives an overview of cur-
rently available tools for the estimation of these parameters from the structure of
compounds. These methods have to be fast and valid in order to help the combi-
natorial chemist prole even very large libraries.
26.4.1
Lipophilicity
Descriptors of lipophilicity are all distribution coecients between a buer (at dif-
ferent pH values) and a lipid-like phase, such as octanol, reversed phase (RP) high-
performance liquid chromatography (HPLC) column stationary phase, lipid mem-
branes, and so forth, as stated in Section 26.3. The most prominent member of
this distribution coecient family is the coecient between octanol and water. For
the neutral form of the desired molecule, this coecient is called the partition coef-
cient (K ow ). More commonly, its log P value is used. Also available are logarithms
for the distribution between octanol and buer at various pH values, log DpH (dis-
tribution coecient), or log K, which is the logarithm of the capacity factor from RP-
HPLC measurements. Some recently developed methods are distribution coe-
cients that have stationary phases on RP-HPLC columns built up by covalently
bound single lipid molecules log IAM (immobilized articial membranes) [74] or
those using liposomes [75], or complete lipid bilayers immobilized on silica beads
log MA (membrane anity) [76]. Only for log P and log D do commercially avail-
able calculation methods exist.
In general, one can divide the prediction methods for log P (log D) into three
groups:
1 Atom contribution methods (A log P) [ 77]. With this approach, methods are gath-
ered that fragment a molecule into its heavy atoms. Once the contribution of an
atom fragment (dened before analysis) to the molecules log P value is eval-
uated by a correlation analysis with measured log P values, the log P of a new
structure is calculated by summing up the contribution of all atoms. In some
cases, the contributions are somehow weighted when an atom is an exposed pe-
ripheral one compared with more shielded interior atoms.
2 Molecular property methods (B log P) [ 77]. These methods use structural features,
such as number of nitrogen and oxygen atoms or ring structures available, or
semiempirical molecular descriptors derived from quantum chemical calcu-
lations or solvatochromic parameters, for example hydrogen-bonding acceptor/
donor ability, and correlate these to known log P values of experimental datasets
of varying size.
3 Fragmental contribution methods (C log P) [ 77]. These methods are the most es-
tablished in the literature and are based on the contribution of molecular frag-
26.4 Estimation of Physicochemical Parameters 755
All of these methods have their own advantages and disadvantages, e.g. the
fragmental approaches give higher correlation coecients than the atomic contri-
bution methods but sometimes in the fragmental approaches a parameterization
is not possible because a fragment is missing from the training set. A benchmark
can be found in the literature [77].
It is our belief that a method using computer-automated structure evaluation
[78], which is now commercially available [79], is a special case. This method en-
codes the structural fragments into binary bitstrings, known as ngerprints. Two
ways of ngerprinting have been described:
The latter has the advantage that the ngerprints generated are characterized by
the nature of the chemical structures in the experimental dataset. Various con-
ditions can be selected to determine the uniqueness of a fragment. The fragment
length can be predened or can be kept exible in a length region, for example
between 1 and 5 atom units, and the program itself nds the most relevant frag-
ments by a multilinear regression analysis with measured properties. Alternatively,
neuronal networks can be trained with these ngerprints as input patterns. These
methods are very useful if a large dataset of measured properties, such as log P,
log K, log D, and log MA values, is available. The advantage is that for distribution
coecients such as log K and log MA, where neither commercial programs nor
public databases are available, a calculation prior to synthesis is possible even for
very large libraries.
26.4.2
Solubility Including pKa Eects
As discussed earlier, the solubility of a drug may play a crucial role in the process
of absorption from the gut. Although the solvent medium in the gut is not pure
water, aqueous solubility (Sa ) is usually taken as the surrogate parameter. Several
methods of calculating water solubility from chemical structure have been de-
scribed in the literature (see below). However, their accuracy is signicantly lower
than that of distribution coecients, particularly if they are applied to very chemi-
cally diverse compounds. The reason for this fault is that, at least for substances
which form solid crystals under normal conditions, the solubility is mainly deter-
mined by two energy contributions. One of them describes the energy required to
place the solute molecules in the solvent and is given mainly by the chemical po-
tential. This part is very much related to distribution coecients, such as log P,
and can be estimated from the two-dimensional chemical structure suciently ac-
756 26 Estimation of Physicochemical and ADME Parameters
curately. The second energy contribution arises from the need for the molecules to
leave the energetically favorable crystal structure. This energy relies very much on
the interactions of the drug molecules with each other and on their ability to be
densely packed, and is therefore strongly dependent on the three-dimensional
structure of the molecules. This feature is the principal reason why the accuracy of
prediction on the basis of two-dimensional structures is limited.
A broad survey of the dierent prediction methods is given in [80], including
examples of applications. A review of various methods with a critical discussion of
their quality can be found in [81]. Below, we give a short overview of the dierent
approaches. Basically, three classes of methods can be distinguished: fragment-
based methods, quantitative correlations with log P or with molecular descriptors,
and the universal quasi-chemical functional group activity coecient (UNIFAC)
method.
The fragment-based methods [82, 83] rely on the assumption that Sa is an addi-
tive property which can be calculated by summing up the contributions of the dif-
ferent chemical groups of a molecule. These methods have the advantage of being
easily applicable. However, their applicability is restricted to structures for which
all groups contained within the structure are parameterized in terms of their con-
tribution to solubility.
Because methods for calculating log P are commonly available (see Section
26.4.1), using a log Sa /log P correlation is a very convenient method to estimate
Sa . A quite simple physicochemical consideration [80] leads to the result that, for
nonionic liquid solutes, a relationship of the form
log Sa a b log P 6
should exist. This relation is a quantication of the qualitative rule that water is a
good solvent for polar and a poor solvent for nonpolar, lipophilic solutes. Depend-
ing on the dataset, the parameters a and b have values of 0.251.5 and 0.951.2
respectively [81]. In its simplest form, the correlation gives good estimations only
for liquid compounds and can therefore generally only be used to make rough es-
timations. For compounds that are solid, a term regarding the energy required to
remove a molecule from the crystal has to be included, which can be accomplished
by considering the melting temperature of the crystal [84, 85]. However, because
the value of this property is generally not known for virtual structures and sub-
stances produced by combinatorial chemistry, these approaches cannot be helpful
for the evaluation of combinatorial libraries and are therefore not discussed in
greater detail here.
Several relationships between solubility and molecular connectivity have been
described [8689]. The advantage of these methods is that their application is quite
simple because the connectivity indices needed can easily be calculated from the
structure. On the other hand, their applicability is somewhat restricted because the
single relationships are in each case restricted to one single or only a few classes of
chemical structures.
26.4 Estimation of Physicochemical Parameters 757
26.4.3
Plasma Protein Binding
As stated in Section 26.3.2, plasma protein binding is one of the key factors de-
termining the distribution of a compound within the body. This parameter can be
measured using plasma itself or by using single plasma proteins such as albumin
or acidic a1 -glycoprotein to estimate plasma binding. The latter has the advantage
that the measurements can be carried out in a relatively straightforward way and
with a high throughput [92]. If a dataset of sucient size and enough chemical
space is available, this parameter can also be predicted from the structure, using
the same methods as those described in Section 26.4.1.
References
27
Virtual Compound Libraries and Molecular
Modeling
Roger M. Brunne, Gerhard Hessler, and Ingo Muegge
27.1
Introduction
27.2
Lead-nding Libraries
The need for large diverse screening compounds for HTS is a major reason for the
rapid technological advances in combinatorial chemistry. A combinatorial library
that is used for HTS against many cell-based or enzymatic assays with dierent
targets requires a wider range of structural and physicochemical properties than
a focused library screened against a particular target. Therefore, the assessment
of the diversity of the library compounds is an important step during the design
phase of a lead-nding combinatorial library [9, 10].
27.2.1
Diversity Assessment of Library Compounds
descriptor space that was eciently and generally applicable for the assessment of
the diversity of chemical structures in a combinatorial library.
Pharmacophoric or three-dimensional (3D) ngerprint keys include informa-
tion from the 3D representation of the molecules. The dierent pharmacophoric
groups of the molecule, such as hydrogen acceptor and hydrogen donor groups,
and lipophilic and aromatic centers are dened. After generation of (evenly sam-
pled) conformational models the two-, three-, and four-center distance combina-
tions between the pharmacophoric groups for each conformer are determined and
encoded in a bitstring [14, 15]. The computational eort for large libraries is con-
siderable owing to the conformational sampling step.
For a quantitative assessment of structural diversity a metre of the dissimilarity
or similarity of all pairs of structures is needed. Using ngerprint keys as the de-
scriptor this is achieved by a bit-by-bit comparison of the corresponding ngerprint
keys, as in similarity searching [16]. A widely used similarity metre is the Tani-
moto coecient, which is the quotient t of the number NAB of all common set bits
in molecules A and B divided by the number of all noncommon set bits in mole-
cules A and B (Eq. 1).
NAB
b
t 1
NA^ NB^ NAB
b
The values of the Tanimoto coecient range from 0 for most dissimilar struc-
ture pairs to 1 for pairs with all identical structure elements. Since t only refers to
set bits, it scales somewhat with the size of the molecules involved. However, the
Tanimoto coecient lacks the properties of a metre and the coecient values
do not resemble true distances [16, 17]. Nevertheless, t is computed very fast and
has even been used to assess properties such as biological activity between similar
structures [18, 19].
The diversity of a library can be computed by the pair-wise calculation of the
similarity of all structure pairs (see also Section 27.4.2.2). A representative mea-
sure is the self-similarity of a library that is determined by nding for each com-
pound the most similar but not identical structure and averaging the correspond-
ing Tanimoto coecients [13c]. The diversity of a library increases with decreasing
self-similarity.
27.2.2
Drug-likeness of Library Compounds
Scheme 27.1.
1 Pharmacophore points are fused and counted as one when their heteroatoms
are not separated by more than one carbon atom.
2 Compounds with fewer than two or more than seven pharmacophore points are
dismissed.
3 Primary, secondary, and tertiary amines are considered to be pharmacophore
points but not pyrrole, indole, thiazole, isoxazole, other azoles or diazines.
4 Compounds with more than one carboxylic acid are dismissed.
27.3 Focused Libraries 765
Based on these simple rules it has been shown that about two-thirds of drug-like
compounds as dened by their presence in drug-like databases such as Compre-
hensive Medicinal Chemistry (CMC) or MACCS-II Drug Data Report (MDDR) are
classied as drug-like while only one-third of the compounds in the Available
Chemicals Directory (ACD) pass the lter. Together with statistical tools that prune
distributions of such important parameters as MW or clog P to be commensurate
with those of drug-like molecules the above described pharmacophore lter pro-
vides a simple guiding tool in the design of virtual libraries that can easily be im-
plemented.
27.3
Focused Libraries
Diverse libraries and random libraries are large screening libraries synthesized
without a particular target in mind. With the advent of HTS a decade ago it was
generally assumed that lead structures could be obtained with high success rates
by screening large diverse compound libraries. However, the history of HTS that
runs throughout the pharmaceutical industry tells a dierent story. Only one in
ten HTS runs produces a viable lead compound that is subsequently optimized in
a drug discovery eort [33]. One reason for this somewhat disappointing track
record is the fact that the available chemical space is so overwhelmingly large.
Estimates for accessible small organic molecules range from 10 60 to 10 100 com-
pounds. Even if the space of drug-like compounds is considerably smaller
(maybe only 10 12 compounds) it is still six orders of magnitude larger than the
largest screening libraries processed today. From this obvious dilemma it is clear
that additional knowledge about the drug target is needed to optimize the output
of a library screen. Several strategies are available to design focused libraries that
have an increased chance of hitting a drug target:
27.3.1
Targeting Protein Families
pounds that target other classes such as enzymes and nuclear receptors. The com-
pounds have been identied by dividing the MDDR into two structural groups:
@20,000 GPCR-like and @55,000 non-GPCR-like compounds by searching the
activity-class eld of the database for entries such as 5-hydroxytryptamine (5-
HT), leukotriene, PAF, and then randomly choosing 5000 compounds from each
class. Training of the neural net is performed for 2000 cycles with a learning rate
of 0.2 and a momentum term of 0.1 reshuing the training dataset before each
cycle. The resulting neural net classies GPCR-like compounds correctly with 80%
certainty. An independent test of compounds in our proprietary database that were
found to hit GPCRs or other targets showed a correct prediction of GPCR-like
compounds in 70% of the cases.
An important question for the design of combinatorial libraries is whether
the neural net is able to attribute the GPCR-likeness directly to possible building
blocks or whether the GPCR-likeness is an attribute of the enumerated molecule
only. Table 27.1 shows a list of possible GPCR-like primary amines taken from the
ACD [39] as putative building blocks for a virtual triazine library of 160,000 com-
pounds. Ten percent of the enumerated library was identied as GPCR-like by the
neural net. Table 27.1 shows the fraction of building blocks occurring most fre-
quently in the enumerated GPCR-like portion of the library. A high percentage
suggests that the GPCR-likeness of an enumerated compound can be directly at-
tributed to its building block (100% means that all enumerated compounds that
contain the building block have been recognized as GPCR-like a strong indicator
that the GPCR-likeness of the enumerated compound is solely due to the GPCR-
likeness of the building block). Table 27.1 shows the expected functional groups
such as piperazines, morpholines, and piperidines to be carriers of high GPCR-
likeness. This table also shows additional functional groups, such as pyrrolidines,
aliphatic tertiary amines, and acridine, that may not have been obvious to be very
GPCR-like. Other functional groups, such as imidazoles, pyrazoles, triuoro-
methyl, and benzylethers, show somewhat weaker GPCR-likeness. Table 27.1 sug-
gests that the GPCR-likeness in most cases can be attributed to the building blocks
alone. This is an enormous advantage in the design of combinatorial libraries
since a virtual library does not have to be enumerated to optimize its GPCR-
likeness. Expensive and sometimes inconclusive optimization routines such as ge-
netic algorithms [40, 41] that act on the enumerated library can be avoided. Rather,
the choice of core and building blocks becomes sucient for the library design. As
a result, there are two easy principles of designing GPCR-like libraries:
27.3.2
Privileged Structures
The term privileged structures was introduced by Evans and coworkers for struc-
tural types of small molecules that are able to bind with high anity to multiple
classes of receptors [42]. Benzazepine analogs were found to be eective ligands for
768 27 Virtual Compound Libraries and Molecular Modeling
100 73 22
98 68 18
97 63 18
95 57 18
95 54 17
94 34 16
94 30 15
93 29 12
92 27 11
87 24 11
84 23 10
27.3 Focused Libraries 769
82 22 10
an enzyme that cleaves the peptide angiotensin I while others are eective CCK-A
receptor ligands. Cyproheptadiene derivatives were found to have peripheral anti-
cholinergic, antiserotonin, antihistaminic, and orexigenic activity. Hydroxamate and
benzamidine derivatives have been shown to be privileged structures for metal-
oproteases and serine proteases respectively. For the class of 7-transmembrane G-
protein-coupled receptors a large number of privileged structures have been found,
including diphenylmethane, diazepine, benzazepine, biphenyltetrazole, spiropiper-
idine, indole, and benzylpiperidine [43]. Recently, the concept of privileged struc-
tures has also been applied to the design of natural product-like libraries [4446].
The wide use of privileged structures in the design of combinatorial libraries is evi-
dent from the large number of libraries being made for various target classes [47].
Many of the privileged structures are rigid, polycyclic heteroatomic moieties that
allow for varied substitutions in the three-dimensional space [43]. If one is inter-
ested in privileged structures that do not occur ubiquitously as structural motifs of
active compounds (such as biphenyls [48]) but are specic for a target class, it is
instructive to assess the target class likeness of the privileged structure. Privileged
structures are thought to interact directly with a specic binding site in a particular
class of proteins. However, for prominent drug target classes such as GPCRs this
assumption is not easy to validate without crystallographic evidence. An analysis of
some privileged structures that are thought to be specic for GPCRs reveals that
only some of them show high GPCR-likeness according to the above-described
neural net. For instance, benzylpiperidines, spiropiperidines, and biphenyltetra-
zoles exhibit high GPCR-likeness according to the neural net. On the other hand,
diphenylmethane, benzodiazepines, indoles, and benzamidines show mediocre to
no GPCR-likeness. Interestingly, the GPCR-likeness of ligands that bear privileged
structures is sometimes very high, e.g. PD 132,177 (an AT-2 ligand) which pos-
sesses a diphenylmethane moiety. This nding suggests that the reason why a
770 27 Virtual Compound Libraries and Molecular Modeling
privileged structure hits a certain target class may not always be due to a direct and
specic interaction with the protein. Sometimes it may rather be attributed to its
ability to provide substituents with optimal vectors to interact within the protein
pocket. Not all privileged structures may therefore be optimal starting points for
the design of target class-directed libraries following principle 1 mentioned in
Section 27.3.1. They may rather be ubiquitously favored substructures that provide
an increased chance to hit almost any protein. Some of those ubiquitously privi-
leged structures have recently been identied [48]. They include; carboxylic acids,
biphenyls, diphenylmethane, and, to a lesser extent, naphthyl, phenyl, cyclohexyl,
bibenzyl, benzimidazole, quinoline, etc. This observation suggests that those sub-
structures ubiquitously enriched in active compounds among various protein
classes may be used as privileged building blocks rather than as privileged core
structures in combinatorial libraries. A more rigorous analysis of privileged drug-
like structural motifs has been published by Bemis and Murcko [49, 50].
27.3.3
Similarity
one of the three- or four-point pharmacophore patterns a bit in the binary nger-
print is set, signaling the ability of the molecule to assume this pharmocophoric
pattern. The examination of all possible conformations and all pharmacophore
patterns results in a bitstring for each molecule that can then be used to calculate
the similarity between pairs of molecules (Fig. 27.1).
In an earlier study, 3D pharmacophoric ngerprints were found to be inferior
to standard 2D ngerprints in retrieving active compounds from databases [58].
Nevertheless, recent evidence was presented demonstrating their potential in clas-
sifying kinase inhibitors [59]. Furthermore, Pickett and coworkers could iden-
tify known brinogen antagonists hidden within a 100,000-compound subset of
Rhone-Poulenc Rorers repository using the four-point pharmacophoric ngerprint
of RGD peptide. All known inhibitors were found among only 3% of the database
[60]. They also report the application of pharmacophoric ngerprints in conjunc-
tion with a genetic algorithm-based compound selection, which maximizes the av-
erage pharmacophoric similarity of the library to a lead structure, but no practical
application is demonstrated. Thus, the use of 3D pharmacophoric ngerprints in
library design appears to be promising, although applications are restricted to vir-
tual libraries containing fewer than 100,000 compounds owing to high computa-
tional expense.
In principle, descriptors, which do not rely on chemical fragments alone but
give a more general picture of the molecular properties, may be more suitable for
library design based on the similarity principle. Especially the adequate calculation
of similarity between compounds with dierent scaolds needs to be improved
over chemical fragment similarity measures. Examples of such descriptors are fea-
ture trees [61], topological pharmacophoric ngerprints [62], and BCUTs [63] that
have the ability to cope better with changes in the scaold between lead structure
and combinatorial library because of graph theoretical and 3D pharmacophore ap-
proaches.
QSAR models have also been used for designing a focused library [64], but it
should be kept in mind that the validity of QSAR models usually is constrained to
the close chemical neighborhood of the molecules used as the training set. There-
fore, QSAR models are probably not well suited for designing libraries around a
new scaold.
772 27 Virtual Compound Libraries and Molecular Modeling
27.3.4
Docking
The 3D protein structure presents valuable information for drug design, since de-
tails of the binding site of the target protein are available at atomic resolution.
Molecular docking routines use this information to match molecules to the protein
binding site and to predict the binding anity of the putative ligand [65]. There-
fore, docking is used for virtual screening to identify potential ligands from large
databases [66, 67] or from virtual combinatorial libraries. Docking can be applied
for cherry-picking only the most promising few compounds out of a virtual com-
binatorial library. On the other hand, the predicted binding anity also provides
reactant selection criteria to give the best library in terms of a large number of
highly potent molecules.
Various molecular docking programs such as DOCK [6870], FlexX [7173], or
GOLD [74, 75] are available. These programs dock only whole molecules. There-
fore, each molecule of the library has to be docked and scored individually. This is
a time-consuming process, which limits the number of virtual compounds to be
docked to about 1000 per day and processor.
In the following section, we will describe some examples of successful applica-
tions of structure-based design to combinatorial chemistry and compare structure-
based design with diversity-based library design.
At Roche a virtual library of roughly 4500 diaminopyrimidines has been chosen
to yield dihydrofolate reductase (DHFR) activity [76]. The diaminopyrimidine core
moiety of the library matches pharmacophoric key features in the folate binding
site and is part of many high-anity DHFR ligands. Therefore, a diaminopyr-
imidine library is expected to contain a fairly large fraction of DHFR active com-
pounds. The virtual library has been docked into Staphylococcus aureus DHFR [77]
with FlexX by xing the position of the diaminopyrimidine core at its key interac-
tion points. Of the 150 top-scoring ligands, 104 have been synthesized by parallel
chemistry, from which 16 have shown activity in antibacterial assays a hit rate of
about 15%. The same diaminopyrimidine virtual library has been clustered and
375 diverse molecules have been selected to span adequately its chemical space
and have been tested for antibacterial activity. Thirty-two of these compounds were
selected giving a hit rate of 8.5%. Although antibacterial activity rather than the
activity against the actual target has been monitored, the structure-based selection
has given a twofold higher hit rate than diversity selection. It has been suggested
that this encouraging result is due to the ability of the virtual screen to focus the
selection of the compounds on molecules that are most likely to interact with the
molecular target DHFR.
For library design, docking of whole compounds is only useful in a few cases
because of the limited throughput of docking programs and the size of virtual
combinatorial libraries, which usually comprise more than several hundred thou-
sand compounds. Virtual combinatorial libraries are built by combining every re-
actant from one site with N variations with any other reactant from another site
27.3 Focused Libraries 773
with M variations, thus giving N*M products for a library with two points of vari-
ation.
In structure-based library design, this combinatorial problem is often solved by
docking individual building blocks. Thus, instead of docking N*M products only
N M fragments have to be investigated. Experimental evidence for the validity of
this approach is given by the structureactivity relationship by nuclear magnetic
resonance (SAR by NMR) technique [78]. Here, small fragment-like molecules
are screened and the best molecules are linked to the complete drug later in the
design process.
One such approach was presented by Bohm and coworkers in a search for new
thrombin inhibitors [79]. The molecular design was divided into two steps. First,
high-anity fragments for the recognition pocket of thrombin were identied by
docking 5300 commercially available primary amines in the recognition pocket of
thrombin. Amines were chosen because they are good starting materials for a va-
riety of chemical reactions. In the second step, the complete compounds were built
up within the thrombin binding site by LUDI [80, 81], a de novo design program,
which joined small fragments to whole molecules within the protein binding
pocket. The top-scoring amine fragment, p-amino-benzamidine, was used as an
anchor to add 540 aldehydes by reductive amination with an extension of LUDI
that could dierentiate between types of chemical links. Ten of the 100 top-scoring
molecules were synthesized, producing a hit with an inhibition constant of 95 nM.
The design of a combinatorial library against cathepsin D, an aspartyl protease,
represented pioneering work in the eld of structure-based library design by Ell-
mans and Kuntzs groups [82]. The library was designed upon the (hydroxyethyl)-
amine isostere, since statins and analogs were well-known mechanism-based in-
hibitors for aspartyl proteases. The central (hydroxyethyl)amine core was modeled
in the active site with a binding mode similar to that of pepstatin. Substituents of
the central core were introduced as acids and amines selected from ACD with a
molecular weight limit of 270. The three sites of variation were independently in-
vestigated within the protein-binding site using CombiBuild [83], which employed
an exhaustive conformational search in the corresponding binding pockets. Next,
for each site of variation, ten compounds were selected from the 50 top-scoring
reactants by cost and diversity criteria. A 10*10*10 library was obtained containing
23 compounds with 50% inhibition at a concentration of 0.33 mM.
For comparison, a second library of the same size was synthesized using diver-
sity criteria alone for picking reactants. In this library, only three compounds had
inhibitory activity at a concentration of 0.33 mM. The focused library not only
showed a sevenfold increased hit rate compared with the diverse library, but also
contained compounds with higher activity.
The focused library was also assayed against plasmepsin II, another aspartyl
protease; it produced 13 hits at an inhibitor concentration of 1 mM [84]. The
lead compounds could be further optimized by selecting reactants for each site of
variation independent of diversity criteria and of structure-based design with
DOCK.
774 27 Virtual Compound Libraries and Molecular Modeling
The examples presented above show that it is possible to use 3D protein struc-
tures to design combinatorial libraries with hit rates signicantly higher than
those obtained by diversity screening (twofold for the DHFR example and seven-
fold for the cathepsin D example). The examples presented above also exemplify
the usefulness of privileged structures. In both examples the designed library was
based on structures that are known to have some intrinsic anity to the corre-
sponding target [diaminopyrimidines for DHFR and (hydroxyethyl)amine for as-
partyl protease]. When these fragments are decorated with a diverse set of sub-
stituents, an impressive hit rate is obtained. Decorating a privileged core with a
diverse set of substituents appears to be especially useful when no 3D protein
structure is available.
27.4
Methods for Library Optimization
The denition of a template and the selection of available reactants easily gives
rise to virtual combinatorial libraries with millions of compounds. Only a subset of
those compounds can actually be synthesized. The reagents for the actual library
could be chosen based upon the properties of the individual reagents or on the
properties of the combinatorial products. Reagent-based selection is easy to ac-
complish and does not require an elaborate selection procedure. On the other
hand, reagent-based selection neglects the properties of the whole molecule, which
govern the interaction with the protein target.
Furthermore, a product-based approach called cherry-picking selects single
enumerated compounds from a large virtual library based on one or more favor-
able properties. Cherry-picking is unlikely to provide a set of reagents that can ef-
ciently be applied to parallel or combinatorial synthesis. Usually, multititer plates
(MTPs) are used for parallel synthesis. Whenever one reactant is selected, it is
used for the entire row or column of the MTP, generating a complete matrix of
products. If one reactant is changed to obtain another product in one well, the
products in all other wells of that row or column are changed too.
Libraries obtained from product-based selection were shown to be more diverse
than libraries that are obtained from reagent-based diversity selection [85, 86].
Furthermore, many design objectives rely on whole-molecule properties, e.g. the
similarity to a lead structure or drug-likeness. Therefore, only product-based selec-
tion procedures will be discussed further here.
Traditionally, anity has been considered as the key point in the initial phase of
lead optimization. This is still true, however, for reducing pharmacokinetic and
toxicological problems in the clinic [87]; the consideration of ADME-related prop-
erties becomes more and more important in the early phase of lead optimization.
Thus, dierent aspects have to be reconciled in the library design.
In the following section, some methods are described that are used to select an
optimal set of reagents with respect to the various design goals outlined above. The
selection procedure consists of two independent components: a search method,
27.4 Methods for Library Optimization 775
which selects the subset of the virtual library space, and a tness function, which
evaluates how well the actual selection ts the desired properties. As a search
method only a genetic algorithm will be discussed, although other methods have
also been successfully applied [88, 89].
27.4.1
Genetic Algorithm
27.4.2
Fitness Function
The tness function measures how well the actual selection of compounds ts the
desired properties. It can consist of only one term, if one property is to be opti-
mized, or can comprise several terms, if various properties have to be optimized
simultaneously. Examples of how the individual terms of the tness function can
be constructed are given below.
27.4.2.1 Potency
Various methods have been described above that can be used to predict the po-
tency of virtual compounds, e.g. similarity to lead structures and structure-based
776 27 Virtual Compound Libraries and Molecular Modeling
design methods. These techniques can give an estimate of the activity of each
member of the virtual library. From this information the suitability of the actual
designed library can be obtained. According to the target function in Eq. (2) that
needs to be maximized, libraries are preferred that contain a large fraction of po-
tent molecules. Nactive is the number of compounds better than a threshold and
Ntotal is the total number of compounds selected.
Nactive
tness 2
Ntotal
1XN
tness property 3
N i
27.4.2.2 Diversity
The diversity of a library in product space can be calculated as the sum of pair-wise
dissimilarities over all selected molecules [92], as shown in Eq. (4), where N is the
number of compounds in the selection. Diversity is given by (1 similarity). The
averaged pair-wise similarity can be eciently calculated using the cosine coef-
cient and the centroid algorithm [92]. This function (Eq. 4) also allows for the
ecient comparison of the relative diversity of two datasets, simultaneously max-
imizing the diversity within the actual selection and the diversity to in-house
screening pools.
X
N X
N
similarityi; j
i j
D1 4
N2
Diversity can be measured by cell-based methods [93]. The chemical space, de-
scribed by a set of descriptors, spans a multidimensional grid. This grid is divided
into several bins along the dierent axes [94]. Now, diversity can be measured by
the number of bins that are populated by the actual selection. Similar molecules in
this context belong to the same or to neighboring bins.
Fig. 27.2. The bisarylether library used in the each reactant, three sections were reserved in
analysis had three sites of variation: R1 and R2 the chromosome. The full library is obtained by
were at xed positions, while the position of R3 combining each reactant with every other
at the second aromatic ring was varied. For reactant from the other sections.
. rule of ve
. diversity
. both diversity and the rule of ve
The library was coded in one chromosome divided into three sections for the
three sites of variation, each section containing ten reactants. The enumeration of
778 27 Virtual Compound Libraries and Molecular Modeling
the combination of these reactants gives the 1000 products of the library. The se-
lection is based on the product properties.
The tness function for the rule of ve-based selection is the sum of individ-
ual scores. Whenever a molecule is within the limits of the rule of ve criteria, it
obtains a score of 1. Thus, a molecule which meets all criteria obtains a score of 4.
The average of this score was optimized over the selected set of 1000 molecules,
resulting in a library design in which all molecules stayed well within the limits
given by each rule of ve. However, the averaged pair-wise diversity, calculated
from HQSAR ngerprints [96] with the cosine coecient, of this library is fairly
low at only 0.09.
In contrast, the diversity-based selection gives a signicantly higher averaged di-
versity of 0.25, however, a signicant fraction of the selected molecules violate the
rule of ve limits.
The combination of both design objectives in one scoring function as the sum
over the diversity score and the above-mentioned rule of ve score results in the
design of a diverse library with an averaged pair-wise diversity of 0.23 that is in
agreement with the rule of ve criteria.
Thus, the combination of both tness functions allows for the combined opti-
mization of dierent parameters, which results in an improved property prole for
the whole library (Fig. 27.3).
The above example illustrates the benets of combining dierent objective func-
tions in the optimization procedure. A genetic algorithm provides a simple scheme
to perform this combined optimization by adding various terms to the tness func-
tion, as shown in Eq. (6):
The rst term represents the internal diversity of the selection. Other terms repre-
sent the distribution of any optimizable property. The weights wprop for these pa-
rameters have to be adjusted according to the problem.
Thus, powerful tools are available that allow the selection of a good set of com-
pounds out of a large virtual library. According to the intended library design, vari-
ous terms can be combined within the tness function:
. For lead nding, a diverse library, which consists of lead-like or drug-like mole-
cules, appears to be desirable. Diversity could also be dened with respect to the
existing in-house screening pool.
. For lead optimization, potency-related aspects such as similarity to HTS hits may
be optimized in conjunction with drug-likeness.
27.5
Conclusion
Over the past years combinatorial chemistry has become an established method
within drug discovery. In parallel, computational methods have played an increas-
Fig. 27.3. a) Distribution of molecular weight, library selected by a combination of the rule
27.5 Conclusion
log P, number of donors and the number of of ve and diversity. Compounds with
acceptors for the whole virtual library. (b) For properties above the rule of ve limits are
779
the library selected by the rule of ve; (c) for represented by dark grey bars.
the library selected by diversity; and (d) the
780 27 Virtual Compound Libraries and Molecular Modeling
ing role in supporting this process. With relatively little computational eort vir-
tual library compounds can be optimized with respect to maximum diversity, high
similarity with an already known screening hit or lead structure, or improved drug-
likeness and ADME properties. Even the eciency of virtual screening can be in-
creased by applying appropriate lters for functional groups, molecular weight, or
other physicochemical properties that reduce the number of compounds for the
actual docking process. Depending on the intended purpose for which the library
is designed and synthesized, a combination of dierent computational methods is,
in most cases, appropriate.
A screening hit from a library that has been optimized using computational
methods can be optimized more eciently based on the design principles or rules
used for the original library. The new library can be focused on compounds similar
in structure to the screening hit while areas of the property space that have not led
to bioactive compounds can more easily be avoided.
Starting from calculated physicochemical properties and standard 2D ngerprint
keys, as available with chemical structure databases for structure searching, new
descriptors have been developed that can address more complex properties (e.g. 3D
pharmacophore keys) or encode properties more eciently (e.g. BCUT) for library
evaluation. In general, ngerprint keys or bitstrings provide an ecient way to
encode properties and can be used for the diversity assessment of combinatorial
libraries. They can be tailored according to the properties suitable for a particular
purpose.
Given the wealth of available descriptors and methods it is already becoming
more and more challenging to nd the most suitable descriptors for a particular
library design task and the best optimization method. This is reminiscent of the
situation in classical QSAR. However, the situation is much more promising with
virtual library design than with QSAR. With combinatorial chemistry a whole set
of new test compounds can be synthesized and tested, whereas in classical QSAR
projects typically only single compounds (usually only the estimated most active
ones) become available. The educated shot gun approach with a combinatorial
library has a much higher probability for success during screening. The overall
risk of failure due to insucient information on the true nature of the interaction
of bioactive compounds with their biological target or system decreases when
whole libraries of potentially active compounds (rather than single compounds)
are synthesized and screened.
While combinatorial chemistry can eciently be supported by computational
methods, support of computer-assisted drug design can be evaluated by library
synthesis and screening. As a result of the combined eorts of computer-assisted
library design and combinatorial chemistry, a new quality in the overall drug dis-
covery process has been achieved. While its usefulness has been proven for several
cases, library design is an active research eld that is expected to gain more im-
portance and applicability in the future.
References 781
References
28
Erythropoietin Sensitizer A Case Study
Berthold Hinzen, Gabriele Braunlich, Christoph Gerdes, Thomas
Kramer, Klemens Lustig, Ulrich Nielsch, Michael Sperzel, Josef
Pernerstorfer
28.1
Introduction
28.2
Results
28.2.1
High-throughput Screening and Biological Evaluations
Fig. 28.1. Dihydopyridazinone 1 was the lead treatment with 10 mU mL1 Epo without
structure of the Epo-sensitizer project sensitizer; 4.5% point increase in the
[EC50 0.23 mM (luciferase assay); 6.5- to hematocrit after p.o. application (10 mg kg1 );
tenfold stimulation of erythropoiesis over the IC50 (PDE3) 0.2 mM].
Fig. 28.2. The lead and some of the options of the chemical optimization program.
28.2 Results 787
model used human primary CD34-positive stem cells. The stimulation of the pro-
liferation of these erythroid cells was determined by counting benzidine-positive
cells after 14 days culture in methylcellulose medium. In vivo results were ob-
tained from mice, for which the increase in the hematocrit was measured after 5
days of treatment.
28.2.2
Concept for Chemical Optimization
28.2.3
4-Fluoro-3-nitroaniline as Central Core
To test a broad variety of heterocyclic head groups, the initial strategy was to use a
specic single intermediate as a central core and to modify this core with a variety
of heterocyclic nucleophiles as an approximate guide to which of these hetero-
cycles would be worth following up. 4-Fluoro-3-nitroaniline (2), because of its func-
tional group pattern, is a well-known building block in combinatorial chemistry
788 28 Erythropoietin Sensitizer -- A Case Study
(Scheme 28.1). Using formyl-functionalized resin (3), this core was immobilized
by reductive alkylation. Owing to the low nucleophilicity of the aniline moiety, the
standard conditions for the imine formation [trimethylorthoformate (TMOF),
room temperature, 2 h] had to be enforced slightly: the mixture of starting mate-
rials was heated to 40 C overnight using TMOF as the dehydrating agent and sol-
vent [7]. Reduction was subsequently performed with tetrabutylammonium boro-
hydride in a mixture of acetic acid and dimethylformamide (DMF) (5:95) [8].
Acylation of the core was then performed under standard conditions to give the
left-hand part of the desired analog (4) of lead structure 1 [9].
In the next step, one of the heterocycles (510) was attached to the core by a
nucleophilic aromatic substitution reaction to give intermediates 11a. The reaction
worked well with reactive nucleophiles. Piperazine, for example, was coupled
quantitatively after stirring the resin with the amine (15 equiv.) in dimethyl sulf-
oxide (DMSO) at room temperature overnight. However, with poor nucleophiles
such as imidazole the alkylation reaction did not proceed in good yields at room
temperature and gave unsatisfactory results at elevated temperatures. However,
repeating the coupling reaction twice at room temperature gave products of su-
cient purity [> 85% by high-performance liquid chromatography/ultraviolet
(HPLC/UV) detection] for biological testing.
Following this route, the rst library of compounds with the general structure
11a was prepared using the indicated heterocycles 510 and aromatic carboxylic
28.2 Results 789
28.2.4
Libraries Around Single Heterocycles
28.2.4.1 Hydantoins
The targeted library of compounds 14 (Fig. 28.3) can be prepared retrosynthetically
from three types of building blocks: the hydantoin moiety can be built up from
an amino acid 15 and isocyanate 16, which are coupled and then cyclized under
basic conditions. The amide bond can be formed from the corresponding carbox-
ylic acid derivative 17 and an amine functionality can be prepared from the nitro
group in 16. Based on literature precedent [13], the ease of cyclo cleavage from the
support, and the fact that immobilized amino acids are commercially available,
the latter was chosen as the attachment point to the support, as indicated in Fig.
28.3.
790 28 Erythropoietin Sensitizer -- A Case Study
In the rst series of experiments, the preparation of 18 and 19 started from im-
mobilized N-uorenylmethoxycarbonyl (Fmoc)-protected amino acids 20, 4-nitro-
phenylisocyanate 16, and carboxylic acid chlorides 21 (Scheme 28.2). The amino
groups of the amino acids were liberated using piperidine in DMF and then re-
acted with 16 to give urea 22. Subsequently, the nitro group was reduced to 23 us-
ing SnCl2 2H2 O in DMF at room temperature.
The acylation of the resulting aniline 23 was rst investigated using carboxylic
acid chlorides (e.g. benzoyl chloride 21), although ultimately the synthesis of the
target amide 18 was envisaged to start from carboxylic acids since important
building blocks were not readily available as acid chlorides (e.g. nicotinoyl chlor-
ides). Benzoyl chloride (21) in dichloromethane and triethylamine (TEA) as base
gave the amide functionality in good yield and purity. The replacement of the
highly reactive acid chlorides by carboxylic acids, e.g. benzoic acid 24, proved to be
possible using TBTU as the activating agent [14]. The corresponding urea func-
28.2 Results 791
Fig. 28.4. Compound 25 was one of the few hydantoins displaying a weak biological activity.
tionality in 19 was prepared using isocyanates in ethyl acetate. This coupling reac-
tion gave signicantly better yields and purities in ethylacetate than in the more
commonly used solvents CH2 Cl2 [15] or DMF [16].
The cyclo cleavage to hydantoins 18 and 19 was highly dependent on the reac-
tion conditions. Based on literature precedent [13], the reaction temperature was
chosen to be 6070 C. Using TEA as base and a 1:1 mixture of methanol and
ethanol as solvent, only moderate yields of the cyclized products 18 and 19 were
obtained. The yield of the reaction was increased by using pure methanol as sol-
vent and a reaction time of 23 days. In accordance with the concept of cyclo
cleavage only the desired heterocyclic compounds are cleaved from the support
the purity of the products was good (>85% by HPLC) and approximately 600
compounds were tested for biological activity. Some of these compounds showed
weak activity, e.g. 25 (Fig. 28.4), which stimulated erythropoiesis approximately
1.5-fold. However, these hits were not further pursued since the activity was too
weak and, based on the already existing structural information, there was no po-
tential for further improvement.
28.2.4.2 Pyrazoles
Following the concepts outlined above, the replacement and modication of the
polar heterocyclic head group in lead structure 1 by a pyrazolyl group was the next
aim. A solid-phase synthesis of pyrazoles 26 based on 1,3-dicarbonyl compounds
27 as the key intermediates has been described by Marzinzik and Felder (Fig. 28.5)
[17]. The diketone 27 is prepared in a Claisen reaction from acetophenone 28 and
ester building block 29 [18]. Intermediate 27 is then condensed with a hydrazine
building block to give the corresponding pyrazole [19].
Following this retrosynthetic concept, two routes to the pyrazole analogs (30) of
lead 1 were investigated (Fig. 28.6). The rst route (a) proceeds via immobilized
Scheme 28.3. The Claisen condensation of substituted acetophenones. This result might
resin-bound acetophenone derivatives was be due to the fact that the immobilized enolate
reported for secondary amide-substituted is a much better nucleophile when bound to
acetophenone derivatives. However, the the support as secondary amide than as
reaction failed with resin-bound tertiary amide- tertiary amide.
28.2 Results 793
The failure of route (b) (Fig. 28.6) prompted further investigation of the second
route (a): TentaGel resin TM 41 [22] was employed as support instead of Rink
amide resin [23] as used by Marzinzik and Felder since better and more reliable
results were obtained with the more solution-phase-like resin 41. Immobilization
of 4-acetyl benzoic acid (28) was performed under standard conditions. The subse-
quent Claisen condensation to 42 was achieved using benzoic acid esters in DMA
with NaH as base at 90 C (1 h). However, the reaction remained unreliable and no
general trends describing the reactivity of the ester components were observed.
Furthermore, reaction conditions could not be improved by using other bases (so-
dium methoxide [24], NaHMDS [25], sodium amide, potassium tert-butoxide [26]),
temperatures, reaction times, or solvents.
In solution phase, closely related reactions gave improved yields when 18-crown-
6 (18-C-6) was added [27]. This modication was also tested on solid support. In-
terestingly, in some cases the addition of 18-C-6 gave improved yields, but in many
794 28 Erythropoietin Sensitizer -- A Case Study
other cases the reaction still did not work. The Claisen condensation remained a
critical step in the sequence, and for library production each ester building block
had to be tested prior to use. Furthermore, only esters which had no (acidic) a-
proton could be used. For the rst library, only aromatic esters were employed
(Scheme 28.5).
Scheme 28.5. Solid-phase synthesis of intermediate with hydrazines gave in all cases
primary and secondary carboxamides near equimolar mixtures of the two possible
containing the pyrazole moiety as the key regioisomers. For clarity, only one of the
fragment. Cyclization of the dicarbonyl regioisomers is shown.
were obtained. This low success rate was due to the last modication step, which
required rigorous exclusion of any traces of water. This requirement is not readily
achieved using fully automated synthesizers.
The library was strongly inuenced by the SAR of the original lead structure 1,
and only a limited diversity was achieved. In vitro activity was not observed for this
library. However, since only a limited area of the available and accessible diversity
space was exploited, the biological activity of the entire virtual library was by far
not veried by this rst set of products. Therefore, a second library of 96 com-
pounds was prepared from 288 theoretically accessible compounds (Fig. 28.7). To
exploit further the chemical space accessible by the described chemistry, not only
the secondary amides 46 but also the primary amides 45 were submitted to testing.
Furthermore, a more diverse set of hydrazine building blocks was employed. In-
stead of using a fully automated synthesizer, the radiofrequency-encoding technol-
ogy from Irori TM (see Chapter 6) was used. This technique allows the preparation
of larger libraries as it is not restricted to spatial encoding of the products. Fur-
thermore, reaction conditions can be much better controlled and inert conditions
can be ensured more easily.
Within this second library, the rst biologically highly active compounds were
identied and a major breakthrough in the project was realized. Interestingly, it
was not the secondary amides (46) that were the active compounds but rather the
corresponding synthetic precursors (45). Furthermore, a rst SAR of the hydrazine
building block (R2) was identied: the N-methyl-substituted pyrazoles were the
796 28 Erythropoietin Sensitizer -- A Case Study
1 R H (50) 156
2 R Me (47) 410
3 R Et (48) 219
4 R i-Pr (51) 100
5 R t-Bu (52) 100
6 R Ph (49) 100
7 R Bn (53) 100
8 Isoxazole 54 314
most active compounds, e.g. 47, but already N-Et (48) and N-Bn (49) were signif-
icantly less active. The unsubstituted pyrazole 50 displayed some weak activity
(Table 28.1).
These initial results were conrmed by follow-up libraries and isoxazoles, e.g. 54
(Table 28.1), were identied as being almost as potent as N-methyl pyrazoles. Re-
sults for the R1 group were less clear and a variety of libraries were prepared using
a large set of ester building blocks. Based on the already established SAR the
number of hydrazine building blocks was kept small within these libraries. Bio-
logical testing nally established an SAR for the R1 building block (Table 28.2).
This residue was the most variable part of the molecule and many para-substituted
phenyl derivatives showed good activity; meta-Substituted phenyl derivatives were
less active and ortho-substituted compounds were inactive. Furthermore, a num-
ber of heterocyclic ester building blocks led to active compounds (Table 28.2).
Until this stage, three out of the four building blocks employed [acetophenone,
ester (R1), hydrazine (R2), and alkylating agent (R3)] had been modied. From this
optimization process involving around 600 compounds, pyrazole 55 was one of the
most active molecules in vitro (entry 1, Table 28.2). Compound 55 displayed a 4.5-
fold stimulation of erythropoiesis at rather high concentrations (0.1 mM). Based on
this result, in vivo experiments were initiated, but, disappointingly, no increase in
the hematocrit was observed. This result was most probably due to the poor phar-
macokinetic prole of 55: Cmax 0.16 kg L1 and area under the curve
(AUC) 0.11 kg h L1 after being given intraperitoneally to mice (Table 28.3).
Mono- and dialkylation of the primary amide with small alkyl groups increased
the metabolic stability [dimethylamide analog of 55 and 71 (for the synthesis, see
below): Cmax 0.193 kg L1 and AUC 0.416 kg h L1 after intraperitoneal ap-
plication to mice; Table 28.3], but was also associated with a loss of in vitro activity
[3.4-fold stimulation of erythropoiesis at a high concentration (10 mM)]. The rst
result indicated the primary amide functionality as an important point in the met-
28.2 Results 797
abolic degradation of 55 and probably also for the other hits within this class. To
address this problem and to obtain a thorough picture of the structureactivity
space around 55 and of the pyrazoles in general, the following issues were next
examined:
1 How can the primary amide functionality in hits 5670 be replaced to improve
metabolic stability?
Fig. 28.8. Modications of the primary amide functionality in 55 to increase metabolic stability.
Answering the rst question was a chemically demanding task since it required
a change in the link to the solid support and, as mentioned above, the Claisen
condensation was sensitive to substitution at the nitrogen. As potential replace-
ments for the primary amide, ve functionalities were considered: (1) carboxylic
acids prepared by hydrolysis of the primary amides, (2) (benzylic) amines as prod-
ucts of reduction of the corresponding amides, and (3) ureas, (4) sulfonamides,
and (5) inverse amides as derivatives of the benzylic amines obtained by reduction
(Fig. 28.8).
Some methods for the reduction of polymer-bound amides to the corresponding
benzylic amines have been described in the literature [32, 33]. However, we have
not obtained positive results for the reduction and subsequent acylation of pyr-
azoles on solid support, e.g. the sequence of reduction of 43 to 72, subsequent
acylation, and cleavage to give 73 failed (Scheme 28.6).
Since literature precedent is only given for easily accessible amide function-
alities, this failure was thought be due to steric hindrance around the amide func-
tionality. This problem was solved by solution-phase parallel synthesis: after cleav-
age of the products from solid phase the primary amides 45 were reduced using
borane in THF at room temperature overnight to give 74 [34]. The borane reduc-
tion proved to be superior to reductions with lithium aluminum hydride since the
work-up of the latter reaction in a parallel fashion was tedious [35]. The resulting
amines (74) were submitted to testing and, in addition, derivatized using acylating
agents and isocyanates to give analogs 73a and 73b (Scheme 28.6). However, the
activity of these libraries was very low. Therefore, these classes were abandoned.
To prepare the carboxylic acid analogs of 55 and other active pyrazoles, the linker
to the resin was redesigned. Instead of the amine-functionalized TentaGel SAM
resin TM 39 employed so far, it was thought that the corresponding Wang resin [36]
28.2 Results 799
Fig. 28.9. Carboxylic acid 79 the compound selected for preclinical development.
Tab. 28.4.
Entry Maximum
stimulation
(%) [ mM]
28.2.5
The Candidate for Preclinical Development
Carboxylic acid 79 has no PDE3 activity and dose-dependently increases the eects
of EPO (10 mU mL1 ). The EC50 is approximately 30 nM and the maximal re-
sponse is an approximately fourfold stimulation of the eects of 10 mU mL1
EPO. Thus, the sensitizing eects are displayed at the relevant EPO concentration
encountered in renal anemia. Even at high EPO concentrations, EPO-mediated ef-
fects were increased. In a mouse model, 79 (3 and 10 mg kg1 day1 orally, once
daily, 5 days of treatment) increased the hematocrit on day 8 by 1.9% points and
2.8% points respectively. However, the eects of 79 appear to be rather modest
compared with treatment with rhEPO, which gives an increased hematocrit of at
least 10% points after 5 days of treatment with 400 U kg day1 following the same
protocol.
The lack of a more pronounced increase, a fast return to baseline, and the total
absence of activity in rats were surprising. To exclude metabolic instability as a
potential explanation for these disappointing results, the kinetic parameters of 79
were determined in several species. In all species tested (mouse, rat, dog), no
pharmocokinetic problems were seen after peroral administration and very high
expositions could be achieved in mice and dogs: Cmax 0.847 kg L1 , AUC 9.9
kg h L1 and a 100% bioavailability after p.o. application to mice. These data ex-
cluded unfavorable pharmacokinetics as the reason for the modest in vivo eects.
However, a possible explanation for the limitation of the eects of EPO-sensitizing
agents can be given by the very tight regulation of the red blood cell mass in mice
and rats. Further studies concerning the biological activity of 79 are ongoing.
802 28 Erythropoietin Sensitizer -- A Case Study
28.3
Combinatorial Chemistry in Drug Discovery
Over a period of 7 months approximately 700 pyrazoles and isoxazoles were pre-
pared within nine libraries with 2.5 Full-time-employees (FTEs). The libraries
varied according to size, the synthetic techniques employed (automated synthesizer,
radiofrequency tagged, manual synthesis), and the diversity of the building blocks
used. The rst libraries were mostly a verication of the method development and
the selection of building blocks was mostly ruled by the pre-existing SAR of the
lead structure dihydropyridazinone 1 and its close analogs. Still, these libraries
helped to identify the primary and not the secondary amides as relevant com-
pounds. The sizes and diversities of the next libraries were increased. These li-
braries further validated the SAR for the primary amides but did not help to solve
the lack of in vivo potency. However, based on the SAR, small and highly focused
libraries were prepared, in most cases manually, to verify the medicinal chemistry
hypotheses. By using parallel methods, this objective was carried out more quickly
and more thoroughly than would have been possible by conventional methods.
Combinatorial chemistry in solution and on solid support has been shown to be
a very successful tool in supporting drug discovery. To realize this potential, some
strategic considerations have been important:
Acknowledgements
References
1 a) A. J. Erslev, Blood 1974, 44, 77; b) Alewood, Tetrahedron Lett. 1995, 36,
A. J. Erslev, N. Engl. J. Med. 1991, 7709; c) R. Ramage, S. L. Irving, C.
324, 1339; c) L. O. Jacobson, E. McInnes, Tetrahedron Lett. 1993, 34,
Goldwasser, W. Fried, L. Plzak, 6599; d) J.-U. Peters, S. Blechert,
Nature 1997, 179, 633. Synlett 1997, 348; e) S. Tomasi, M. Le
2 a) E. Goldwasser, Blood Cells 1984, Roch, J. Renault, J.-C. Corbel, P.
10, 147; b) W. Jelkmann, Physiol. Rev. Uriac, B. Carboni, D. Moncoq, B.
1992, 72, 449. Martin, J.-G. Delcros, Bioorg. Med.
3 R. T. Means, S. B. Krantz, Blood Chem. Lett. 1998, 8, 635; f ) S. H. Lee,
1992, 80, 1639. S.-H. Chung, Y.-S. Lee, Tetrahedron
4 W. Fried, C. Morley, Steroids 1985, Lett. 1998, 39, 9469; g) M. B. Francis,
46, 799. E. N. Jacobsen, Angew. Chem. Int. Ed.
5 J. W. Esbach, J. C. Egrie, M. R. 1999, 38, 937; h) R. C. D. Brown, M.
Downing, J. K. Browne, J. W. Fisher, Chem. Commun. 1999, 1547;
Adamson, N. Engl. J. Med. 1987, 316, i) C. Blackburn, B. Guan,
73. Tetrahedron Lett. 2000, 41, 1495.
6 C. D. Demoliou-Mason, Exp. Opin. 9 a) S. M. Dankwardt, T. M. Phan, J.
Ther. Patents 1995, 5, 417. L. Krstenansky, Mol. Diversity 1996,
7 a) G. C. Look, M. M. Murphy, D. A. 1, 113; b) Y. Pei, R. A. Houghten, J.
Campbell, M. A. Gallop, Tetrahedron S. Kiely, Tetrahedron Lett. 1997, 38,
Lett. 1995, 36, 2937; b) M. M. 3349; c) S. Curtet, M. Langlois,
Murphy, J. R. Schullek, E. M. Tetrahedron Lett. 1999, 40, 8563; d) D.
Gordon, M. A. Gallop, J. Am.Chem. G. Hall, J. Tailor, M. Gravel, Angew.
Soc. 1995, 117, 7029; c) B. Ruhland, Chem. Int. Ed. 1999, 38, 3064.
A. Bhandari, E. M. Gordon, M. A. 10 a) D. A. Goff, R. N. Zuckermann, J.
Gallop, J. Am. Chem. Soc. 1996, 118, Org. Chem. 1995, 60, 5744; b) H. V.
253; d) G. C. Look, J. R. Schullek, C. Meyers, G. J. Dilley, T. L. Durgin,
P. Holmes, J. P. Chinn, E. M. T. S. Powers, N. A. Winssinger, H.
Gordon, M. A. Gallop, Bioorg. Med. Zhu, M. R. Pavia, Mol. Diversity 1995,
Chem. Lett., 1996, 6, 707; e) E. A. 1, 13; c) P. Grosche, A. Holtzel, T.
Boyd, W. C. Chan, V. M. Loh Jr, B. Walk, A. W. Trautwein, G. Jung,
Tetrahedron Lett. 1996, 37, 1647; f ) M. Synthesis 1999, 1961.
Schuster, J. Pernerstorfer, S. 11 a) S.-M. Yeh, C.-M. Sun, Tetrahedron
Blechert, Angew. Chem. Int. Ed. Engl. Lett. 1999, 40, 7247; b) P.-C. Pan,
1996, 35, 1979; g) Y. Wang, S. R. C.-M. Sun, Tetrahedron Lett. 1999, 40,
Wilson, Tetrahedron Lett, 1997, 38, 6443; c) C.-M. Yeh, C.-M. Sun, Synlett
4021; h) D. Maclean, J. R. Schullek, 1999, 810.
M. M. Murphy, Z.-J. Ni, E. M. 12 H. Perrier, M. Labelle, J. Org. Chem.
Gordon, M. A. Gallop, Proc. Natl. 1999, 64, 2110.
Acad. Sci. USA 1997, 94, 2805. 13 a) S. W. Kim, S. Y. Ahn, J. S. Koh, J.
8 a) B. Chenera, J. A. Finkelstein, D. H. Lee, S. Ro, H. Y. Cho, Tetrahedron
F. Veber, J. Am. Chem. Soc. 1995, 117, Lett. 1997, 38, 4603; b) J. Matthews,
11999; b) W. D. F. Meutermans, P. F. R. A. Rivero, J. Org. Chem. 1997, 62,
804 28 Erythropoietin Sensitizer -- A Case Study
29
Estimation of Stability and Shelf Life for
Compounds, Libraries, and Repositories in
Combination with Systematic Discovery of New
Rearrangement Pathways
Ferenc Darvas, Gyorgy Dorman, Tamas Karancsi, Tamas Nagy,
and Istvan Bagyi
29.1
Introduction
29.1.1
Stability and Shelf Life Characterization: the Need
didates that fail to exhibit sucient stability should be stopped before investing
millions of dollars [5].
29.1.2
Stability Characterization: Empirical Studies
Giving an estimation about the purity degradation of a stored library over a time
period, such as a year, is principally a simple procedure: quality sampling before
and periodically during the storage can lead to a quantitative description about the
purity deterioration. These data can be utilized to develop a qualitative picture of
the library or repository stability. Such empirical studies of the library stability are
easy to perform and might be useful as a rst guess regarding the expected stabil-
ity issues.
To the best of our knowledge, the rst open account of an empirical stability
study was presented by Ortner in 1995 [6], in which she stated that approximately
20% of a collection of several thousand organic compounds (combinatorial library
members and individually synthesized samples) stored in ethanol degraded over a
1-year period.
At Merck [7], @10,000 compounds were stored at 20 C, 0 C and at room
temperature in DMSO for a year. The decomposition ratio, which was measured
by high-performance liquid chromatography/mass spectrometry (HPLC/MS), sug-
gested that compounds fall into two distinct classes: stable and unstable samples.
Compounds belonging to the unstable class had already decomposed to a large ex-
tent within 6 months. Storing the samples at 20 C did not signicantly improve
the stability of the compounds compared with storage at 0 C. The European
Union is also sponsoring an empirical study involving regular periodic purity
checking of 10,000 compounds by LC/MS and nuclear magnetic resonance [1 H-
nuclear magnetic resonance (NMR)] for a period of 3 years [8].
29.1.3
Stability and Shelf Life Estimation: Model-based Approaches
Empirical studies unfortunately suer from the problem that they do not deliver
an exact forecast about the future quality of the samples studied. Also, it is dicult
and subjective to extrapolate conclusions drawn from these studies to other sam-
ples. They generally do not allow conclusions to be drawn on a number of impor-
tant stability issues, such as structurestability relationships, the inuence of dif-
ferent storage conditions on the stability, or the implications of the quantity and
nature of the impurities for the stability of the compounds.
Three years ago, the authors concluded that a more qualitative stability descrip-
tion was needed in the eld of combinatorial libraries and repositories, and, as a
result, developed a model that extends beyond empirical description. As a starting
point, models introduced at stability expiry date determination of drug substances
are applied (see Section 29.7).
808 29 Estimation of Stability
29.2
Methods and Tools for Combinatorial Stability Assessment
29.2.1
Modeling Intrinsic and Extrinsic Factors Inuencing the Stability of Individual
Compounds
The major factors of combinatorial library stability can be classied into two
groups: intrinsic and extrinsic factors [12]. Intrinsic factors are the properties that
are inherent to a compound with absolute purity and depend solely on the chemi-
cal structure. From our point of view, the decomposition rate of the pure com-
pound at dierent temperatures is the most important property. Less relevant fac-
tors of the pure compound are sensitivity to light, sensitivity to exposure to air or
oxygen, and sensitivity to humidity.
Extrinsic factors of compound stability are the conditions that are beyond the
chemical structure of the compounds being studied. Such factors are the purity of
the library members, the chemical nature and the distribution of the impurities,
such as the accumulated side products and reagent residues, and the nature of the
29.2 Methods and Tools for Combinatorial Stability Assessment 809
solvent(s) used for storage of the compounds. External factors, such as the tem-
perature of storage, the speed of the repeated freeze/thaw cycles, or the chemical
nature of the container are also included in this group.
The authors have proposed that the eect of the intrinsic factors should be
modeled by the Arrhenius equation [11], as described in Section 29.8. This model
enables a qualitative shelf life time to be predicted (and, therefore, the expiry date)
for solid compounds in a pure state or in solution for a single solvent system
under xed storage conditions, if we assume that no structure-dependent interac-
tion exists between the compounds and the impurities or the solvents respectively.
These assumptions can be investigated by statistical signicance testing calculations.
However, the model cannot be used to give a forecast of how the shelf life will be
changed if essential storage conditions are altered, for instance whether the sub-
stances are stored in DMSO instead of as a solid or in plastic instead of glass vials,
since the listed extrinsic factors are not part of the Arrhenius model. For treating
such eects, a dimensional analysis approach has been suggested by the authors
[10], but it is not detailed here.
29.2.2
Modeling: from Compounds to Libraries
Theoretically, the best stability estimation can be gained when both Arrhenius pa-
rameters (a and b) are determined for each individual compound in the library.
But, this would require vast numbers of tests to be conducted. For fast estima-
tion of the entire compound library composed of structurally similar compounds,
a single slope of the Arrhenius line (b) can be used in theory. A good approxima-
tion is to apply an average b, which is calculated from the previously determined b
values of the individual compounds or cluster representatives. According to this
assumption, a slope transposition can be used within clusters to determine the
interception of the Arrhenius line (a) for the compounds within the rest of the
library cluster.
Stability estimation of combinatorial libraries by experimental determination of
the Arrhenius parameters for individual compounds is time-consuming work. It
would be advantageous to reduce the number of experiments by identifying corre-
lation between structural similarity and stability. The correlation can be deter-
mined by performing a homogeneity test, which is a statistical parallelism test [13]
of the Arrhenius lines for library members (Fig. 29.1). If no signicant dierence
between the slopes of the regression equation can be proven, a joint regression
line can be calculated for the entire library, which will be used to estimate a shelf
life for each individual compound in the library.
29.2.3
Modeling: from Libraries to Repositories
entire repository. The stability test includes a homogeneity test for each library in
the repository followed by a decrease in the number of homogeneous libraries
using cluster analysis [14], and an experimental stability test for individual com-
pounds that represent the reduced homogeneous library clusters.
There is another way to model stability: if the collection of compounds can be
divided into several, discrete structure-related stability classes, the number of ex-
periments can be reduced by using simplied experimental techniques, plus sta-
tistical calculation, which ensures a robust classication of the products into the
stability classes.
29.2.4
Realization of Shelf Life Estimation for Individual Compounds and Combinatorial
Libraries
Fig. 29.3. Calculation module of the shelf life for (a) individual
compounds and (b) combinatorial libraries.
base management system at ComGenex, Inc. are utilized. Part of the information
system is an expert system, which is still under development. It is intended to pro-
vide a virtual stability screening during the evaluation phase of the library design
and to establish structurestability relationships.) The decomposition rate con-
stants for each elevated temperature are calculated from the measured degradation
812 29 Estimation of Stability
according to the decomposition kinetic equation (Section 29.8, Eq. 1). The rate
constants are used to generate Arrhenius parameters by linear regression (Section
29.8, Eq. 2). If the Arrhenius line is extrapolated to lower temperatures (e.g. room
temperature), k25 and the shelf life can be calculated according to Eq. 2 (now T
refers to 25 C) and Eq. 3 (Section 29.8) respectively. The minimum acceptable
purity limit or percentage of acceptable purity loss should be dened before cal-
culation.
As noted above, in the case of structurally unrelated individual compounds, each
of them should be tested at multiple temperatures. However, for structurally clus-
tered compounds the shelf life estimation for the rest of the library (series 2) is
based on the degradation at one, optimally selected temperature, assuming that
their decomposition kinetics (the temperature dependence of the rate constants)
are similar.
29.2.5
Instrumentation
For thermal decomposition analysis, test compounds are distributed to wells and
incubated in various media at dierent temperatures in time- and temperature-
controlled multiwell heating blocks, e.g. MultiAger TM (see Fig. 29.12, [15]). The
decomposition of compounds is monitored by a HT HPLC system equipped with
ultraviolet (UV) and integrated with a two-arm robotic sample processor that in-
jects samples and collects fractions. For identication of degradation products on-
line LC/MS experiments are performed. Data acquisition and overall system oper-
ation software are integrated to a central database management system.
29.3
Validation Studies for Combinatorial Stability Assessment
29.3.1
General Experimental Conditions
29.3.2
Stability Study for an Indole Library
For validation studies, from an indole library [11] of 1385 members (Fig. 29.4), ten
compounds were selected randomly as a learning set (to determine the model
29.3 Validation Studies for Combinatorial Stability Assessment 813
Fig. 29.4. The indole core-based library investigated in the stability validation study.
parameters). Another set of 16 compounds was chosen from the same library for
the validation of the stability assessment model.
In order to check the linearity and parallelism of the correlation between log kT
and 1=T, the learning set of ten compounds was stored at ten dierent temper-
atures (115190 C) for two dierent durations (in time intervals of 1 h and 17 h).
Two rate constants for each temperature were calculated from the measured deg-
radation (Section 29.8, Eq. 1). The average values of the rate constants were used
for the generation of the Arrhenius parameters (Section 29.8, Eq. 2).
The model was validated at two levels. First, using the Arrhenius equation, the
rate constants of the compounds were calculated and extrapolated to 75 C and 100
C. The rate constants at 75 C and 100 C were used to estimate the decreased
purity of the compounds after storage for a predetermined time. The decomposi-
tion was monitored experimentally by HPLC. The estimated and the measured
purity values were compared for six compounds (shown in Table 29.1; the re-
maining four compounds did not show signicant degradation).
Two examples of linear regression calculated for the learning set are shown in Fig.
29.5. Since they are structurally related, their b values are similar. They conrm the
expected correlation between the log kT and 1=T values. Based on the temperature
dependence, kT was estimated at lower temperatures in order to estimate the shelf
life of the compounds under storage conditions.
To further validate the model, b values determined experimentally from the
learning set were used to predict the purity of the structurally similar set of 16 com-
Tab. 29.1. Comparison of the predicted and detected values of purity for the learning set of
compounds.
1 99 82 77 5 96 81 15
2 92 80 85 5 90 88 2
3 97 78 86 8 95 90 5
4 93 92 92 0 91 92 1
5 97 87 90 3 92 93 1
6 98 93 92 1 97 96 1
814 29 Estimation of Stability
pounds. The a and kT values for the validation set were determined from a single
thermal decomposition experiment at 150 C according to the Arrhenius law (Sec-
tion 29.8, Eq. 2). Then, a and b were used to calculate the decomposition rate con-
stants at two temperatures: 105 C and 115 C. The purity loss, estimated for a par-
ticular length of time at 105 C and 115 C, was compared with the experimental
results for the ten compounds as shown in Table 29.2 (the remaining six com-
pounds did not show signicant degradation).
The majority of the predictions (80%) t the experimental data well. However, in
some cases, the decomposition was overestimated or underestimated, possibly as a
result of dierent degradation pathways (Table 29.1, entry 1, experiment 2; Table
29.2, entries 7, 9, and 11). At these high temperatures, the overall degradation
mechanism may be dierent from that at storage temperature. The predicted HPLC
29.3 Validation Studies for Combinatorial Stability Assessment 815
Tab. 29.2. Comparison of the predicted and detected values of purity for the validation set of
compounds.
7 97 55 71 16 72 88 16
8 95 95 95 0 95 95 0
9 95 80 66 14 85 85 0
10 99 97 97 0 98 99 1
11 97 75 10 65 96 47 49
12 97 93 91 2 95 95 0
13 95 95 88 7 92 93 1
14 99 99 99 0 99 99 0
15 98 97 98 1 98 98 0
16 97 82 85 3 95 93 2
results of the remaining test compounds were in good agreement with the mea-
sured ones.
In some cases (Table 29.2, entries 8 and 14), the initial purity and the purity after
the thermal decomposition test were identical, i.e. these compounds did not show
thermal decomposition they had a long shelf life.
Although monitoring thermal decomposition is faster at higher temperatures,
this results in a loss of reliability of the predictions because of the changes in the
overall kinetics along with the increased temperature.
29.3.3
Combinatorial Stability Investigation for a Small Repository
29.4
Stability Investigations in Combinatorial Drug Discovery
29.4.1
Pilot Design Phase
hydroxyl groups were protected with bulky pivaloyl groups, sixfold higher stability
was detected. Proton NMR studies based on the downeld eect of the chemical
shift of the pyrrole proton conrmed that in the rst case the electron density was
signicantly reduced while in the second case no changes were observed, indicat-
ing that the stabilizing factors are purely steric eects.
To avoid similar problems, a two-stage stability checking procedure is advisable.
First, empirical knowledge in combination with an expert system should be used
to determine structural features which may lead to unstable library structures that
are still in the design phase. Second, during the pilot chemistry feasibility studies,
the pilot matrix representing the major reactivity classes of substituents can be
used in a combinatorial stability study. The matrix can show how the substituents
aect the chemical stability and which substituents should preferably be eliminated
from the full library. Initial stage stability investigations should be extended to
the key intermediates. Since the last intermediates can often be the main source
of impurities of the nal library, knowledge about their stability is a key factor in
ensuring that library products not only show high quality immediately after the
production but also retain their quality for a longer period of time.
Poor chemical stability often prevents the synthetic realization of a promising
new library. On the other hand, changing the substitution pattern in the library
design phase can avoid the stability problems in a number of cases.
29.4.2
Libraries
All library members reect their synthetic history through their impurity prole.
Properly designed combinatorial stability assessment of the library members could
reveal interactions between the impurities and the major product library member,
which result in a deterioration of its long-term stability. Poor stability often prevents
the optimization of a lead structure or extra eort is required to stabilize the mol-
ecule at later stages of drug development. (In one of the HIV entry inhibitor hit
818 29 Estimation of Stability
29.5
A Way Towards Systematic Discovery of New Rearrangement Pathways
Selected library members have already been used for the discovery and charac-
terization of thermally induced transformations by combining MS fragmentation
with standard structural analytical methods [21, 22]. To detect thermal rearrange-
ment, routine process-specic MS fragmentation patterns (rearrangement nger-
prints) need to be identied and subsequently recognized by the appropriate
software tools. The MS ngerprint species a unique fragmentation pattern that
29.7 Appendix I: Stability Testing of Drug Substances 823
indicates the formation of a new sublibrary via thermal rearrangement and will
rapidly speed up the identication of such transformations.
Such rearrangements can be transferred to a synthetic preparative protocol. This
approach can be considered as an equivalent of the Houghten library-from-library
approach [23] via thermal rearrangements.
29.6
Summary
The approach presented in this chapter for predicting shelf life and expiry date
for compounds, libraries, and repositories is based on modeling reaction kinetics,
modeling, completed with thermal decomposition experiments in which a series of
purity measurements of the compounds are involved. To save experimental time
and substances, extrapolation techniques of the experimental values for sub-
libraries, libraries, and repositories are proposed. Algorithms, software and expert
system solutions, instrumentation for thermal decomposition, and high through-
put analyses are integrated with a chemical database management system to a
uniform, high-performance stability estimation solution the Stabex TM system.
Stabex TM is capable of estimating the stability of several tens of thousands of
library members per month in the library design and synthesis phases, or, later,
during the systematic, long-term stability investigation of the whole repository.
Stability investigations of the key intermediates promise the synthesis of com-
pounds which may remain pure for a long time, thus contributing to a future total
quality discovery management system.
As a side product of the high throughput analytical procedures associated with
the stability investigation, a systematic search and discovery of new thermal re-
arrangements is proposed and has been demonstrated in this chapter.
29.7
Appendix I: Stability Testing of Drug Substances
Stability expiry date determination of drug substances is well regulated and con-
trolled by standardized international guidelines. Besides the photochemical- and
media-initiated structural changes, the most important factor that inuences chem-
ical degradation is the temperature.
Guidelines for stability testing are determined by the International Conference
on Harmonization (ICH), which dene three levels of stability testing: stress, accel-
erated, and long-term (real) stability testing. All three have dierent roles in stability
analysis [24, 25].
29.7.1
Stress Stability Testing
29.7.2
Accelerated Stability Testing
29.7.3
Long-term Stability Testing
Long-term stability testing is carried out to evaluate the stability of the physical,
chemical, biological, and microbiological characteristics of a drug product and a
drug substance, covering the expected duration of the shelf life and retest period,
which are claimed in the submission and will appear on the labeling.
The shelf life, according to the ICH guidelines, is a time interval that a drug
product is expected to remain within an approved specication provided that it is
stored under the conditions dened on the label in the proposed containers and
with the correct closure.
A related term is the expiry/expiration date, which is a date placed on the
container/labels of a drug product designating the time during which a batch of
the product is expected to remain within the approved shelf life specication if
stored under dened conditions and after which it must not be used [24, 25].
29.8
Appendix II: The Arrhenius Model
ln C0 =C
kT 1
t
where C0 is the initial purity (v/v, %); C is the measured concentration of the com-
pound of interest in the untreated sample (v/v, %); Ccrit is the minimum purity
requirement of compounds at a given storage temperature (v/v, %); and t is the
duration of the thermal burden treatment (h).
Having measured the rate constant at dierent elevated temperatures, a (inter-
cept of the Arrhenius line) and b (slope of the Arrhenius line) parameters can be
determined according to the Arrhenius law (Eq. 2). Using a and b, the rate con-
stant can be extrapolated to lower (including ambient) temperatures:
b
log kT a 2
T
ln Ccrit =C0
t shl 3
kT
29.9
Appendix III: Model Realization the Stabex TM System
1 Evaluating the shelf life of libraries under initial study before production.
2 Evaluating the shelf life of library members after production.
3 Retrospective evaluation of the shelf life of the repository already available.
4 Development and methodological studies of combinatorial stability issues.
Stabex TM has specic links to the elements of the CMT library technology [28],
which integrates library design, synthetic plan evaluation, synthesis and purica-
tion, HT analyses, storage, and data management. The overall owchart for
Stabex TM is shown in Fig. 29.12.
826
29 Estimation of Stability
Acknowledgements
Thanks to Peter Slegel (EGIS Pharmaceuticals, Budapest Hungary) for his contri-
bution to this work and helpful discussions. The help given by Robert Ferenczi
and Gabor Pocze (ComGenex) is much appreciated.
References
Part V
Novel Applications of Combinatorial Chemistry
30
Concepts of Combinatorial Chemistry in
Process Development
Markus Eckert and Ulrich Notheis
30.1
Introduction
30.1.1
General
30.1.2
Subject of this Chapter
In this chapter, we want to give a description of the possibilities and challenges for
combinatorial chemistry in the dierent phases of process development with the
focus on ne chemicals. Therefore, we review some principal issues of process
development which have had an inuence on the use of combinatorial methods in
process development. Furthermore, strategies and equipment for the implementa-
tion of parallel screening methods in the laboratory will be discussed.
30.1.3
Literature
The pioneering work on spot test plates was performed in the 1920s and 1930s
by F. Feigl [2], who began preparing materials on spot test plates to analyze spe-
cic properties afterwards. Impressive examples of fast quantitative analysis of
heavy metals are given in the journals of that time [3].
Also, parallel catalyst screening was used by Mittasch in the development of the
ammonia process during the rst 20 years of the twentieth century [4]. The use
of parallel reactors for the testing of heterogeneous catalysts was described in the
1950s [5]. However, parallelization was limited, as, for example, Dowden and
Bridger stated in 1957: a test apparatus . . . can advantageously be multiplied to
give a greater rate of testing, providing the temptation to overtax both observer and
supervisor is avoided [6].
Obviously, this limit could not be overcome until the era of computers and pro-
cess automation. Therefore, just recently, numerous articles and books have been
published covering the topic of automated synthesis in process development. The
main emphasis has been on chemical process development in the pharmaceutical
industry.
An excellent review was published in 1999 by Harre et al. describing automated
chemistry in organic process research and development [7]. Describing the pio-
neering work that started during the 1980s, they revealed that most published ex-
amples were based on a Zymark robot system. Second, owing to the lack of other
commercially available solutions some companies launched programs with in-
house developments such as the DART (development automated reaction toolkit)
system of Glaxo Wellcome [8], or the ATLAS (assessing technologies for laboratory
automation in synthesis) initiative by SmithKline Beecham [9].
Nowadays, the user has the choice between several commercially available sys-
tems with dierent degrees of automation and features. Harre et al. presented a
table with a dozen systems available as at summer 1999 [7]. In mid-2001, the boom
for parallelization in process development is continuing, leading to a greater vari-
ety of apparatus available on the market (see Tables 30.1 and 30.2).
Since 1996, numerous examples of the optimization of organic reactions by par-
allel process development have covered almost the whole area of organic chemistry.
Recommended literature in this respect is by Hird [10], Orita et al. [11, 12], Zhang
et al. [13], Gooding et al. [14], Kirchho et al. [15], and Emiabata-Smith et al. [8].
In addition, there are two books covering special aspects of parallel process devel-
opment. The book Automated Synthetic Methods for Specialty Chemicals, edited by
William Hoyle, includes a number of papers presented at a Royal Society of Chem-
istry symposium in 1999 [16]. For process development especially, the contribu-
tions by Evens [17] and by Armitage and Smith [18] are relevant, describing their
experience with the Anachem SK233 workstation and the HEL AutoMate system.
Owen and Dewitt wrote about Laboratory automation in chemical development
in Process Chemistry in the Pharmaceutical Industry, edited by Gadamasetti [19].
They predict a shift of paradigm, where simply the emulation of manual methods
with automation is not enough, but the chemists must develop techniques that are
suited to automation, e.g. statistical experimental strategies.
834 30 Concepts of Combinatorial Chemistry in Process Development
30.2
Process Development
30.2.1
Overview and Denition
For basic chemicals, process development is carried out along similar lines, how-
ever the route scouting, the choice of dierent starting materials and routes, is
usually very limited and much more emphasis is put on process design, engineer-
ing, and plant design.
30.2.2
Combinatorial Chemistry in Process Development
30.2.3
Demands on Process Development
For our discussion, it may be helpful to understand the special requirements that
process development has to meet in the whole business workow (Fig. 30.3).
As already mentioned, one of the main goals of process development is to en-
able the manufacture of the desired product in the desired quantity to the required
specication. But, additionally, often equally important goals include shortening
the time to market, and achieving the least expensive manufacturing process. For
the production department, robustness and simplicity of the process are the two
guiding principles in this endeavor.
The process, moreover, has to be environmentally benign and should be the best
long-term route. Last, but not least, the synthetic route must be unimpeded by ex-
ternal patents and, in order to strengthen ones intellectual property position, it
should be patentable.
836 30 Concepts of Combinatorial Chemistry in Process Development
ket shares will be lost. This result can easily exceed the additional costs of a more
expensive alternative process. All these associations have to be borne in mind when
strategies for parallelization in process development are discussed.
30.2.4
Process Development for Dierent Applications
Process development is a complex and diverse eld because the targets of process
development, its strategies, and approaches dier depending on the kind of product
and the production scale. These can range from millions of tons of basic chemicals,
where an improvement in the reaction selectivity of less than 1% saves several mil-
lion Euros, to sensitive pharmaceuticals, where world production varies between
100 kg and several tons per year under strict quality requirements (GMP). For any
product in between these extremes, an adequate process needs to be developed,
taking into account a number of specic basic conditions:
Given that in the ne chemical business dedicated plants will only be built for
runaway successes, the production of most ne chemicals requires processes
with a high degree of simplicity which t a standard apparatus.
. Topical stage of development
Improving an existing industrial process usually includes minimizing invest-
ments and changing as little as possible in the plants equipment. Often, process
parameters are optimized within narrow margins determined by the already ex-
isting plant reactors, which can be changed only at extraordinary costs. In this
case, it is not necessary to set up a primary screening for a totally new synthesis.
If the goal is to nd a new, revolutionary route to a product using new feedstocks
or totally dierent chemistry (dream reactions), the situation is reversed.
All of these factors have to be taken into account when discussing parallelization
in process development more precisely. Table 30.1 gives a short summary of the
key issues.
30.3
Parallelization in Process Development
30.3.1
Number of Experiments Compared with Scale of Experiment
On the one hand, it is obvious that by carrying out more experiments a higher de-
gree of optimization will be achieved and, thereby, a better process. Here, paral-
lelization can help either by reducing the time for a given number of experiments
or by performing more experiments in a given time.
On the other hand, the process has to be scaled up: thermochemistry, speed of
addition, and work-up (e.g. ltration, precipitation, extraction, or distillation) have
to be adjusted to the available apparatus, or a suitable apparatus has to be con-
structed. Therefore, the reaction is run on dierent scales, and, if applicable, in dif-
30.3 Parallelization in Process Development 839
30.3.2
Requirements and Equipment for Parallelization in Dierent Development Phases
In those cases, the rst-shot experiments should involve several parallel runs
over the chiral ligand database (e.g. 3 96 experiments) to obtain a true estimate
of the potential of this route.
What are the requirements for parallel process screening equipment? In this devel-
opment phase, most chemists use standard apparatus such as 50- to 100 mL asks
with magnetic stirrers. If sucient starting material is available, batches of 2550
mmol are preferred. The more sophisticated the intermediates and the more valu-
able the 23 g of step 8 material are, the more opportune are microscale reactions:
however, the handling of small amounts of substances becomes more dicult when
changing from the few grams scale to the few milligrams scale.
To transfer a standard ask set-up into a parallel screening set, it is desirable
to keep the most common features constant, including working under an inert
atmosphere (degas/nitrogen purge), mixing (preferably stirring), cooling/heating,
reux, adding liquids through a dropping funnel, adding solids altogether through
an opening under counternitrogen ow, collecting small samples for analysis dur-
ing the reaction, and no cross-contamination by neighboring cells. Some special
features such as distilling a reaction component out of the reaction mixture for
equilibration shifting are sometimes necessary.
The preferred vessel size is around 2050 mL, it should not be smaller than
5 mL and not larger than 100 mL. The temperature range should be 50 C to
200 C. Sometimes, systems operate at up to 120140 C, which is often not high
enough. An independent temperature control for each reaction vessel would be the
best case, but often this increases apparatus costs exponentially. We regard 612
simultaneous reactions per set as a good throughput, especially when the temper-
842 30 Concepts of Combinatorial Chemistry in Process Development
ature for all vessels per set is the same. In the case of automated workstations in a
dedicated laboratory, a higher throughput can be realized.
Tab. 30.2. Workstations for route scouting and parallel process screening.
Workstationb Number of simultaneous Vessel size (mL), agitation Temperature control, Price Internet;
reactions individual (yes/no) category a http://www
tion, the option of collecting samples at reux conditions, and the exibility and
connectability to other glassware and equipment need to be taken into account.
Sometimes a system appears to be the best despite oering only a few features;
often, these systems provide a high degree of freedom and exibility for arranging
a personalized and chemistry-oriented set-up. In our laboratories, we have used
one set-up over ten times: a simple block set-up based on a H P block for six or
12 standard 100-mL side-cap bottles. These bottles are available with one or two
additional screw cap arms originally made for enzyme chemistry. These arms are
suitable for liquid and solid addition, sample collection, etc. Instead of a closed
screw cap, we use three-bore Teon caps, which are available with standard screw
threads for Bola ttings. A stop cock for argon/vacuum, a reux condenser, Teon
tubing, for example to connect a dosimeter pump, or an internal thermometer can
be easily attached (Fig. 30.7). This set-up is more like a small vessel than a test
tube. Each system is independent and built from mass-produced inexpensive
equipment. This multiple reaction set-up has exhibited good properties in numer-
ous standard reactions for ne chemical processes, including Grignard and diazo-
nium salt chemistry. Of course, it is not automated, but it enables the bench
chemist to carry out up to 12 experiments in parallel.
For high pressure reactions dierent pieces of apparatus are required. There are
only a few systems available on the market that are suitable for higher pressure
hydrogenation or carbonylation chemistry. Argonaut Technologies oers a work-
station called Endeavor for reactions up to 33 bar. It has individual temperature
control (restricted to 20 C temperature dierence between vessels) from room
temperature to 200 C and mechanical stirring for eight vessels each containing
58 mL. For reactions at these low pressures we regard the Endeavor workstation
as a good system. Parr instruments oers a multiple autoclave system for high-
pressure applications up to 200 bar and 300 C. Six autoclaves, each containing
75 mL, can be run with magnetic stirring at the same temperature. Besides variants
working with inlets for a single autoclave, such as similar systems developed by
SmithKline Beecham, Parr, and others, this multiple reaction station is suitable for
process screening of, for example, asymmetric hydrogenations and reactions that
need very high temperatures (200 C and 300 C). A pressure range up to 100 bar
is oered by the HEL High Pressure Chem-Scan. Here, eight 10-mL reactors for
temperatures up to 200 C can be congured in one row; the stirring is accom-
plished by a specially xed magnetic anchor agitator.
What are the requirements for the parallel optimization equipment? The vessel
should have a volume of 50250 mL, a mechanical stirrer, an independent heating/
cooling aggregate as well as an accurate temperature control. Calorimetric analysis
is advantageous at this stage; reux condensers and metering pumps for the re-
actants are necessary as peripheral equipment.
150,000 euros. At the time of writing, the Auto-mate from HEL has been on the
market for about 2 years and has users in most pharmaceutical and ne chemical
companies. It has four mechanically agitated vessels per unit (volume 50100 mL),
the opportunity to monitor calorimetric data, and a whole set of extension equip-
ment at an additional charge. From the chemists view, the transparent glass set-
up has the advantage that one can recognize color changes or precipitations in the
reaction mixture. Recently, the Automate has become available in combination with
an x, y,z robot for automated sample collection and handling named DUET.
The four FlexyLab reactors are of larger size (volume 250 mL) and can be run
completely independently. They come with two mass-controlled liquid dosage units
with ow inducers for each reactor. The reactor head has two standard NS 14/NS
19 glass joints, convenient for attaching standard laboratory equipment. The tem-
perature dierence between the two reactors can be up to 200 K, depending on the
cooling aggregate.
The MultiMax from Mettler-Toledo is a compact system with four 50-mL reactors
and a dispenser box for one reagent. Three standard glass joints in the reactor head
allow the attachment of additional equipment, and agitation is carried out with a
stirring bar. Additional features such as mechanical stirring and additional dosage
units may be added; the STAVEX program for statistical experimental design and
the possibility of equipping the system with online in situ FT-IR analysis are useful
additional features.
The Surveyor has been developed by Argonaut Technologies for process devel-
opment chemists. It has ten transparent Teon reaction vessels with a volume of
75 mL and an integrated x, y,z robot. The typical magnetic Argonaut agitation sys-
tem provides good mixing even of viscous media and suspensions (although the
transfer quality to common plant stirrers has not been examined). Online sampling
and the integrated HPLC analysis (not included in the basic station price) are in-
teresting for determining reaction kinetics, an important issue during the process
optimization. With ten parallel reactors and a working volume of 45 mL, this sys-
tem could also be useful as a process-screening workstation.
30.3.3
Requirements for Analytical Instruments
. Installing an online analysis. The analytical system needs to be fast enough, i.e.
the run time for one sample has to be the 1/nth part of the required time reso-
lution for all reactors. Usually, this requires more than one analytical instrument
(e.g. 24 GLCs) for one multiple reactor, or simplied analytical methods, or both.
Online analysis may give fast data, but it is more dicult to set up and has to
run reliably and stably to oer a real advantage.
. Analyzing the samples o line. This feature provides more exibility and makes the
system less complex to run. Sample preparation may be performed by a robotic
system which additionally transfers the samples to a standard format that may be
directly introduced into the chromatograph.
For every project a new decision has to be made as to whether time and money
should be invested in a better performing high-throughput method or whether
slow standard methods should be applied.
Quality and type of analytical method depends, on one hand, on the analytical
problem, but, on the other hand, they also depend on the stage of development
and the type of information required. Whereas in a process screening a simple
yes/no answer determined by TLC may be all that is required to eliminate 7080%
of inactive catalysts, the requirements will rise with the progress of the develop-
ment until, nally, elaborate, high-precision quantitative methods will be required
to allow the setting up of mass and mole balances and the specication of trace
amounts of side products.
Important contributions in the development of new methods for high-throughput
testing and analysis in the primary screening have been made recently by re-
searchers working in combinatorial catalysis. Maier and coworkers and Reetz and
coworkers showed infrared (IR) thermography to be a feasible screening tool for
analyzing the catalytic activity of heterogeneous catalysts in hydrogenations or
oxidations [25], homogeneous catalysts in enantioselective reactions, ring-closing
metathesis, and even enzyme-catalyzed reactions [26, 27].
Another issue of increasing importance is the high-throughput determination of
enantiomeric excess; here, the adaptation of parallelization techniques from other
disciplines is a tremendously fruitful approach. By parallel capillary electrophoresis,
Reetz et al. were able to analyze the ee values of 7000 samples per day [28]. Mi-
kami et al. combined a synthesis robot with an HPLC circular dichroism system to
30.4 Planning of Parallel Process Development 851
screen their library of chiral ligands in the asymmetric addition of diethyl zinc
to aldehydes [29]. In the future, one can expect new methods in this eld to be
developed.
30.4
Planning of Parallel Process Development
30.4.1
Screening Strategies
parallelization, meaning that the more you wish to know about a single experi-
ment, the lower the degree of parallelization and the larger the scale. The goal is
that the additional information gained by performing more experiments over-
compensates for the loss of information from each single experiment. But, never-
theless, it may be more ecient to have an approximate look at a huge experi-
mental parameter space with highly parallelized, small-scale reactions with a yes/
no answer and then selecting the most interesting areas for closer investigation on
a larger scale.
For every loop, and for dierent phases of process development, questions have
to be answered, such as: How precise must the answer from the experiment be?
How many experiments have to be analyzed for the next set of experiments?
The determination of some reaction parameters may cause diculties in small-
scale experiments, e.g. the enthalpies of reactions and parameters related to mass
transfer eects (inuence of mixing, inuence of particle size in heterogeneous
catalysis, mass transfer limitations at interfaces, etc.). In critical cases, it has to be
checked in advance whether or not relative statements under standard conditions
can be translated into reliable information on a larger scale. For example, if experi-
ments are performed at very low conversions, data may be extrapolated to higher
conversions if the formation of the desired compound takes place mainly within a
system of parallel reactions; if the desired compound is an intermediate in a con-
secutive reaction, this may not be the case.
are copied from nature. For example, in the rst step of optimizing a heteroge-
neous catalyst, the number of components is dened, then a rst generation of
catalysts is initialized and tested in parallel [34]. The results are evaluated, leading
to a selection of catalysts from the population to create the next generation. The
next generation is created by application of evolutionary operators such as crossover,
or qualitative or quantitative mutation. The loop of testing, selection, and mutation
is repeated until the convergence criterion is satised. In heterogeneous catalyst
development, sometimes the question of whether a quaternary will show activity
if all binary and ternary combinations of a given set of elements are completely
inactive must be considered. Here, the application of genetic algorithms seems to
be a promising strategy [34]. A similar approach in which a genetic algorithm is
used increasingly is in directed evolution for the creation of new enzymes display-
ing improved enantioselectivity in a given reaction [35].
30.4.2
Strategies for the Selection of the Equipment
The ultimate goal would be to have each laboratory equipped for parallel working
at all screening stages. Therefore, a complete cascade of screening apparatus (for
process screening, process optimization, process validation) would have to be pur-
chased or developed, which would require signicant investment. Hence, some
moderating decisions have to be made and preliminary questions have to be
answered:
ments is still the best approach, and the eort and expense incurred in setting up
an automated workstation will pay dividends.
In the case of process optimization and validation, the number of experiments
drops to <20, but these experiments require huge data collections and highly accu-
rate processing. The adaptation to dierent chemistry in independently controlled
vessels is easier and faster than in the high-throughput workstations since these
vessels are designed to enable multiple dosing, to have heat ramps over a wide tem-
perature range or reux, and to have joints for special equipment, making their use
worthwhile even for chemists with rapidly changing projects en route to the pilot
plant.
A question arising in this context is the percentage of chemistry covered by the
systems: a 100% perfect solution compared with a 70% standard solution. Sys-
tems covering the total range of possible parameters (temperature range, pressure
range, materials, special safety requirements) are much more expensive and more
dicult to construct. Therefore, we recommend an approximate 70% coverage,
which can be achieved with a number of commercial systems. A broadly applicable
workhorse is better than one sophisticated wonder machine. Supplementary to
these workhorses, we have set up dedicated special equipment for our core tech-
nologies, such as high-pressure reactions and phosgenations.
30.4.2.2 What is the Right Distribution and Number of Pieces of Apparatus for
Dierent Stages?
For a group working in ne chemical synthesis with rather fast-changing organic
chemistry demands, we have had positive experiences in starting with nonauto-
mated multiple reaction blocks for process screening. The advantages are the rather
low costs, the ready-to-go installation, the exibility in chemistry, and the ability to
equip every laboratory (broad introduction) with multiple reaction blocks without
any special training or environment.
The decision of where to expand next with parallel techniques is mostly depen-
dent on the problems and the orientation of the process development group (Fig.
30.10). Either priority is given to a primary screening, which means expanding to
automated workstations and robotics, or priority is given to process optimization
and validation, which means expanding to parallel, independently controlled vessel
stations for scaling up experiments.
In general, it is advantageous to set up and use sophisticated, highly parallelized
equipment in long-running basic chemical projects [4], whereas for the ne chemi-
cal area a more general apparatus with a lower degree of parallelization seems to
be appropriate, reecting the fact that large-scale productions take place in dedi-
cated plants and ne chemicals are produced in multipurpose plants (without tak-
ing into account the quality requirements for dedicated apparatus).
An often underestimated point in the parallelization debate is the generally low
acceptance of new equipment by technicians: expensive workstations often remain
unused (investment ruins). As a consequence, aversion to new developments in the
eld of automation and parallel working increases. Therefore, a stepwise approach
to the introduction and use of new equipment is advisable, e.g. beginning with a
856 30 Concepts of Combinatorial Chemistry in Process Development
30.5
Case Studies
Acid chloride addition to bis-imine Solvents, bases, equivalents, order of 112 Yes PSg DARTb 23
addition
Alcohol oxidation Catalysts, solvents pressure, oxidants 5000 ? PS In-house platforms 38, 39
Aldol condensation Base, solvent, concentration, temperature, 60 ? PS Bohdan development 41
order of addition workstation
Bredereck reaction, imidazole Temperature, formamide equivalent, 26 Yes POh Gilson 215 Bohdan 19
synthesis workstation
Condensation reaction Cocatalysts, concentration 284 Yes PS/PO In-house workstation 42
Cyclization of 3-chloroamide Water addition, temperature, equiv. NaOH ? ? PS Anachem SK 233 22
Phase transfer catalyst
Deprotection of silyl ether Temperature, concentration equivalents 6 Yes PO DART/PROSPERc 33, 23
Deprotection of TFA groupd Base, equivalent, concentration, volume, 116 Yes PS/PO 43
solvent
FriedelCrafts acylation Lewis acids, equivalent of Lewis acid, 52 Yes PS/PO Bohdan workstation 17
solvents, temperature
30 Concepts of Combinatorial Chemistry in Process Development
Heck reaction Catalyst, Pd/ligand ratio, loading 96 No PS/PO Anachem SK 233, ReactArray 13
Hydrogenation Catalyst, loading, pressure, temperature, 44 Yes PS Argonaut Endeavor 44
equivalent MsOH
Hydrogenation Catalyst, loading, solvent, stoichiometry of 77 No PS Multihydrogenation vessel 22
base, quality of starting material
Imidazolidin-2-one synthesis Reagent, solvent, temperature, concentration, A60 No RSf/PS Argonaut Surveyor 18
stoichiometry
Liquidliquid extraction Solvent, temperature 30 No PS Anachem SK 233 21
Michael Addition Solvent 10 No PS Anachem SK 233, 45
REACTarray
Mitsunobu coupling Temperature, concentration, addition rate, 36 Yes PS/PO DART 12
stoichiometry
Nucleophilic substitution Bases, solvents, temperature prole ? ? PS STEM Block 22
Reductive amination Catalyst, additives, stoichiometry Several Yes PS Multi vial reactor 46
hundred
Silyl enol ether addition Lewis acids, solvents 140 No PS Anachem SK 233, ReactArray 45
Suzuki reaction Concentration, water level, catalyst load, 10 Yes PS Anachem SK 233, 45
boronic acid stoichiometry REACTarray
a Numerous names of companies and equipment used in this table
are registered trademarks.
b DART development automated reaction toolkit.
c PROSPER process research optimization screening parallel
experimentation robot.
d TFA triuoroacetic acid.
e DOE design of experiments.
f RS route scouting.
g PS process screening.
h PO process optimization.
30.5 Case Studies
859
860 30 Concepts of Combinatorial Chemistry in Process Development
sively how a combination of dierent screens on dierent scales can lead eec-
tively to new catalytic systems in a very short period of time.
Scheme 30.1.
Scheme 30.2.
ratio of the two solvents aected the product quality. The second example included
two prolonged liquid additions and two strictly controlled cooling ramps. In both
cases, excellent product quality and a robust process were reached by using this
automated robotic system together with an experimental design approach.
Summing up the examples, it is evident that most of them involve the combi-
nation of statistical experimental design (DOE) and automation to screen rapidly
the parameter space of interest for the process.
30.6
Summary
Combinatorial chemistry has now moved into process development in the form of
parallelization. Although the degree of parallelization is still moderate compared
with biological assay high-throughput screening, over the last few years this num-
ber has signicantly increased. There are several examples in the literature, includ-
ing some hundreds of experiments during a single development. The probation in
everyday use and the signicant acceleration of process development by parallel-
ization still has to be proven.
In the past, several companies developed individual solutions to transfer parallel
techniques into their process development department, but recently various pieces
of apparatus and workstations for all development phases have become commer-
cially available, making it easier for newcomers to enter the eld of parallelization.
However, signicant nancial outlay and eort are still required for successful im-
plementation. Especially when moving to highly parallel solutions, the expenditure
for automation and software becomes the determining factor and makes interdis-
ciplinary teams necessary.
High-throughput experiments in themselves have no value, thus deliberate ex-
perimental design has to be included to reach the true objective a highly ecient
workow. A well thought out tuning of the design steps, testing, and analysis, as
well as good coordination between the development phases, are necessary to avoid
limiting bottlenecks.
Yet, process development is still carried out in several phases. Even hundreds of
experiments on a small scale cannot replace a pilot run. It will be interesting to see
during the next few years if and how combinatorial chemistry can have an impact
on this problem and is able to shortcut the route from the test tube to the plant.
It is our belief that combinatorial chemistry is not a substitute for experience
and creativity in process development, but an additional powerful tool in the chem-
ists hand.
References
Surf. Sci. Catal. Vol 130, Coruia, A., 42 R. W. Wagner, F. Li, H. Du, J. S.
Mel, F. V., Meuioroz, F. J. L. G. Lindsey, Org. Process Res. Dev. 1999, 3,
(eds.), Elsevier, Amsterdam 2000. 2837.
39 P. Desrosiers, A. Guram, A. 43 V. W. Rosso, J. L. Pazdan, J. J. Venit,
Hagemeyer, B. Jandeleit, D. M. Org. Process Res. Dev. 2001, 5, 294
Poojary, H. Turner, H. Weinberg, 298.
Catalysis Today 2001, 67, 397. 44 J. M. Hawkins, T. W. Makowski, Org.
40 P. D. Higginson, N. W. Sach, Org. Process Res. Dev. 2001, 5, 328330.
Process Res. Dev. 2001, 5, 331. 45 M. A. Armitage, G. E. Smith, K. T.
41 M. Harre, H. Neh, C. Schulz, U. Veal, Org. Process Res. Dev. 1999, 3,
Tilstam, T. Wessa, H. Weinmann, 189195.
Org. Process Res. Dev. 2001, 5, 335 46 M. A. Huffman, P. J. Reider,
339. Tetrahedron Lett. 1999, 40, 831834.
864
31
High-Throughput Screening Applied to Process
Development
Oliver Brummer, Bernd Jandeleit, Tetsuo Uno, and
W. Henry Weinberg
31.1
Introduction
31.1.1
General
three phases: (1) route scouting, (2) screening and optimization, and (3) character-
ization and validation. In addition to these phases, the real combinatorial discovery
activity will be necessary to incorporate a totally new process where a novel catalyst
or reaction is required. New process discovery is an important activity that can
provide for the introduction of more environmentally benign processes, the use of
more cost-eective starting materials, or provide a higher purity of an intermediate
or a product. Process hazard evaluation must also be carried out for all new pro-
cesses. Hazard analysis should be considered as early as possible in the process
development to insure that a safely scalable process is being pursued.
Finally, the authors want to point out that they do not comment on any of the
described automation devices and tools. This review only describes the latest devel-
opments in high-throughput techniques for process automation but does not give
any recommendations.
31.1.2
Automation and Experimental Design
. The time necessary for optimization of the process is shortened, leading to the
faster development of the drug candidate.
. A much wider range of variables can be selected, giving an improved chance of
success in process screening.
. The quality of data will improve by eliminating human operational errors.
. Better yields can be obtained owing to the more precise control of reaction con-
ditions.
. Large amounts of experimental data are eciently collected with minimum op-
erator involvement.
. The sta is released from tedious repetitive tasks, allowing more time for cre-
ative work.
. Researchers will have more exibility so that operations involving monitoring
reactions over extended time periods are possible.
866 31 High-Throughput Screening Applied to Process Development
31.1.3
High-throughput Process Development
Cawse has reviewed these strategies developed in the eld of combinatorial chem-
istry for drug discovery and materials science [10].
Process optimization utilizes parallel reactors of 550 mL volume. The process
parameters, such as reaction temperatures, stirring eciency, and rates of reagent
addition, should be strictly controlled for high-quality data. At this stage, online
(real time) reaction monitoring is essential to obtain information regarding reac-
tion proles. Even when the number of reactions is small, the amount of data can
be large owing to multiple sampling from the same reaction over time. Currently,
high-throughput automated process optimization attracts most developmental ef-
forts in this eld.
Reactors for process characterization and validation should be a good model for
the real pilot plant. Throughput is low for this stage, but these developments have
also been included in this review since automation can accelerate it, thus higher
throughput can be achieved than in the conventional operation. Calorimetry has
emerged as a useful technique because it can provide information on both reaction
kinetics and reaction hazards. Real time in situ infrared technology (FT-IR) has
also been used to assess reaction hazards [11].
In high-throughput process development, the use of software can signicantly
reduce the time required to design combinatorial experiments, can manage data
collection and monitoring during runs, and can analyze data to reach conclusive
insight into the process. Instruments should be controlled by user-friendly inter-
faces. Because high-throughput experiments generate data orders of magnitude
faster than conventional methods, powerful database and data handling will be an
integral part of the system. The resulting database, in eect, becomes an extension
of the laboratory notebook, and, in conjunction with an appropriate database
structure and access tools, this extension becomes a powerful intellectual asset.
31.2
Case Studies
31.2.1
High-throughput Combinatorial Process Discovery
reaction entity (spatially addressable format), and automatically stores the process
data in a central database, thus allowing for analysis of the results at the end of the
screening experiments. The technology also uses a variety of 96-well reactor for-
mats equipped with a liquid-handling robot to manage exibly both homogeneous
and heterogeneous liquid-phase reactions under high pressures or reux conditions.
In one example, 2000 high-throughput screening experiments were performed
to discover a novel catalyst system for a homogeneous palladium-catalyzed Suzuki
coupling reaction of 2-chlorobenzonitrile (1) with 4-(methylphenyl)boronic acid (2)
to yield ortho-tolylbenzonitrile (OTBN) (3), an important key intermediate of Du-
Ponts antihypertensive compound Losartan (6) (Scheme 31.1). All 2000 screening
experiments were performed in a custom-made 96-well reactor format within 4
weeks. Multidimensional combination of archived ligands, palladium metal pre-
cursors, various bases, and solvents were screened, leading to the discovery of a
lead catalyst structure. The lead was optimized into a new catalyst system that ex-
hibited 5000 turnovers (TO) with >99.5% selectivity of the desired heterocoupling
product over the undesired homocoupling impurity [13].
The same group also developed appropriate technology for high-throughput syn-
thesis and screening of metal-doped polyoxometalate (POM) libraries in 96-well
format. Libraries of the heterogeneous POM catalysts were screened to discover se-
lective aerobic oxidation catalyst systems. In this case, the imidazole alcohol 4 was
oxidized to aldehyde 5, which also represents an important key intermediate for the
Losartan (6) synthesis (Scheme 31.1). Within 4 weeks, 5000 experiments had been
performed to screen dierent catalyst compositions, eects of solvents, and addi-
tives under variable reaction conditions. Two classes of catalyst systems that give
more than 90% yield were identied. These results were conrmed and validated
in subsequent scale-up experiments [14].
Symyx Technologies has also extended its high-throughput capabilities to asym-
metric reactions. As an example, hundreds of catalyst compositions were screened
consisting of transition metals and an array of chiral phosphine ligands under a
variety of conditions for asymmetric hydrogenation of a particular substrate, iden-
tifying the optimum composition that provided high enantiomeric excess and cat-
alyst productivity. The hits were then scaled up in a secondary screening cycle
using a system that is capable of running 96 5-mL pressure reactions in parallel
(Fig. 31.1) [15].
31.2.2
Route Scouting, Screening, Optimization, and Validation
Fig. 31.1. Illustration of hard- and software array of 96 pressure reactors; top left, table
package developed by Symyx Technologies for with conversion data for 48 parallel reactions;
scaled up secondary screening of hits top right, graphical representation of
identied in primary screening. Bottom left, conversion data from the table.
This study showed that it was possible to carry out reactions in a reaction station
located within the working envelope of an autosampler. Although adequate for this
study, the Gilson 231XL and the Peltier rack had severe limitations in the feasibil-
ity of other chemistries. The study also showed that having minimized the bottle-
neck of analyzing the samples, reaction preparation and data manipulation became
the labor-intensive task. For all of these reasons, researchers at Glaxo Wellcome
designed the development automated reaction toolkit (DART) and the process re-
search optimization screening parallel experimentation robot (PROSPER) systems,
which are specically designed to accelerate process optimization using automa-
tion and experimental design.
The DART system became commercially available from Anachem as SK233 TM .
The SK233 TM consists of one or two STEM reaction blocks and a Gilson 233XL
Autosampler. Each STEM reaction block allows up to ten reactions at a tempera-
ture of 30 C to 150 C. The SK233 TM was designed to deliver liquids to each re-
action and perform online automated HPLC analysis without attendance [16]. Re-
searchers at SmithKline Beecham extended the ability of the SK233 TM workstation
by improving the performance of the SK233 TM under reux conditions. New glass
reaction vessels with a quick-t female joint and a cold nger-type reux con-
denser with a hollow center were implemented in the SK233 TM [17]. This multiple
31.2 Case Studies 871
Glaxo Wellcomes new PROSPER system has over 50 reactors that are individu-
ally controlled with features to enable rapid process development [19]. One example
872 31 High-Throughput Screening Applied to Process Development
ditions from this initial screen gave 17 in 85% yield, 18 in 90% yield, and 19 in
96% yield using 5% Pd/CaCO3/Pb with 0 equiv. of MsOH, 10% Pd/C with 0 equiv.
of MsOH, and 10% Pd/C with 1.1 equiv. of MsOH, respectively. These three con-
ditions were then optimized for the two catalysts involved by examining low and
high points for catalyst loading (5% and 20%), pressure (25 and 100 psi), temper-
ature (25 C and 60 C), and, for Pd/C, the amount of MsOH (0 and 1.5 equiv.).
The best conditions from this full factorial screen gave 17 in 97% yield, 18 in 95%
yield, and 19 in 99% yield. The measurement of hydrogen uptake helped to set re-
action times and parameter ranges for the full factorial analysis, allowed for quickly
spotting under- and overreduction, aided predicting robust reaction endpoints, and
provided data for analyzing kinetic behavior. These techniques can be applied to
the rapid optimization of reactivity and selectivity for the hydrogenation of a vari-
ety of polyfunctional molecules.
Researchers at Argonaut have also developed the automated chemical develop-
ment platform Surveyor TM (http://www.argotech.com) especially for conducting
process screening and optimization with a consortium of six companies: Pzer,
Agouron Pharmaceuticals, Aventis Pharmaceuticals, Eli Lilly & Company, and
Rohm & Haas. Surveyor has ten individually controllable reaction vessels with au-
tomated reagent/solvent addition. The system is also capable of sample extraction/
quenching/injection for online HPLC analysis. It uses simple graphical software to
set up experiments and to control temperature, temperature ramp rates, reagent
addition, agitation, sampling, and analysis.
Researchers at Argonaut have demonstrated the automated optimization for the
multistep synthesis of chiral imidazolidinones, useful auxiliaries for the asymmet-
ric synthesis and key intermediates for some human immunodeciency virus (HIV)
protease inhibitors [24]. This study included the initial route scouting, the reaction
condition screening, and the optimization, followed by a scale-up run using the
conventional laboratory set-up. Two routes were compared and variables such as
choice of reagents, time, temperature, and concentration were examined using
Surveyor (Scheme 31.6). Each route was evaluated and compared from a scale-up
standpoint, and the optimized conditions for the selected route were subsequently
applied on a 6-g scale, aording the imidazolidin-2-one 20 in 90% isolated yield.
The discovery phase of this study was a combinatorial screening of ten bases
and seven 50% aqueous solvents. The reactions were performed by dispensing 100
mL of a stock solution of 21 in methylene chloride into each reactor. The solvent
was evaporated in a Savant SpeedVac, and the appropriate reaction solvent (200 mL)
and the appropriate base (200 mL of 0.5 M aqueous base) were dispensed into each
reactor. The reactors were incubated at 45 C for 3 h on a J-Kem shaker plate. After
the reaction aliquots were taken for HPLC analysis, this combinatorial screening
of the 66 reactions revealed that the best overall combination is lithium hydroxide
and water.
With the selection of aqueous alkoxide solutions for the screening DOE, the fol-
lowing factors were then subjected to the second-phase experiment (a screening
DOE): choice of hydroxide (lithium, sodium, or potassium hydroxide), concentra-
tion, temperature, and the equivalents of base. Thirty reactions were performed
including a full factorial 2 3 experiment with each base, including an additional six
center points distributed among the three bases to measure experimental error. A
Zymark Robot performed the experiments by dispensing the appropriate amount
of water and dilute aqueous hydroxide solution to each reaction. Of the four factors
evaluated, only the equivalent of base had a statistically signicant impact on con-
version of starting material to product. After optimization of the equivalents and
concentration of base, a two-factor central composite design with four center points
was employed. Twenty reactions were prepared on a Gilson 215 liquid handler,
heated for 3 h at 45 C in a reactor block, and sampled for HPLC analysis. A scale-
up of the optimized conditions resulted in an isolated yield of 95% of product. In
conclusion, 116 reactions were performed within 4 days, improving the yield from
63% to more than 95%.
Outsourcing of chemical development is becoming increasingly important. Com-
panies such as MediChem, Johnson Matthey, and Symyx Technologies oer auto-
31.2 Case Studies 877
. ALRs control multiple automated laboratory reactors, on many scales from 100
mL to 10 L, for optimization, validation, DOE, and kilo-lab work.
. Calorimetry: addition of an internal heater and controllable power supply gives
calorimetric capabilities.
. Plant simulation: multiple independently controlled reactors, hold/feed vessels,
and crystallizers can be used to simulate full-sized plant operation in the laboratory.
. Crystallization studies: using optical sensors to detect the onset of crystallization
and control crystallization processes.
. Optimization, DOE, and kinetics: using an HPLC and autosampler directly from
three small reactors, a fully automated parallel system has been produced.
. Route scouting: the autosampler system combined with a modied Stem-type
heating block allows the study of the inuence of solvent/catalyst/base, etc. of
ten parallel reactions.
Data logged from the system are generally handled in Excel, a suitable graphing
package, or a statistical package.
The latest and most exible version of the CAR system runs a set of three custom-
built jacketed reactors attached to a Gilson analytical autosampler and a high-
throughput HPLC system. Even though the author states that the system has been
used considerably for DOE optimization work, it appears to be most suitable for
process characterization or validation or both. The throughput is rather slow but it
oers a precise reaction control under realistic conditions as well as a high degree
of analytical capability. Each reactor is equipped with an overhead stirrer, reux
condenser, thermopocket and sample port, and two nondedicated ports for addi-
tions, pH probe, etc. The reactor volume is 50150 mL, and each reactor has an
independent heater/circulator to heat the jacket up to 200 C. Any other traditional
laboratory equipment can also be utilized balances, pumps, pH meters, etc. The
31.2 Case Studies 879
31.2.3
Miscellaneous
ined in 284 reactions over a 10-week period, yielding 1704 data points. Three e-
cient cocatalysts were identied and the rate and reactor volume productivity were
optimized. The authors concluded that the comprehensive set of data accumulated
from the automated experiments establishes the scope of BF3/ethanol cocatalysis
in the synthesis of TMP and should be useful for planning syntheses as well as for
studying the mechanism(s) of catalysis.
Otera and coworkers have developed a new type of automated synthesizer with
the ability to conduct a variety of synthetic reactions [33]. Although not intended
for the process development, their results suggest that the system will be useful for
automated process development with some modication. The system consists of
a control unit [automatic reaction system (ARS) and a sequencer], a jacketed glass
reactor (50 or 130 mL), reservoirs, volumetric ceramic valveless piston pumps, a
syringe for quenching the reaction, and a cooling unit. Reactions can be run under
inert atmosphere at reaction temperatures from 78 C to elevated temperatures.
Using the synthesizer, air-sensitive organolithium and Grignard reagents as well
as transition metal catalysts could be handled. Also, the dependence of chemical
yields on the reaction temperature for Peterson alkene synthesis and on the addi-
tion rates of the aldehyde for aldol reactions were examined. Because the order of
reagent addition is programmed and the reaction temperature is quickly tunable,
sequential reactions can be conducted smoothly. An advanced control system was
882 31 High-Throughput Screening Applied to Process Development
incorporated that allows a task to start immediately after the preceding one has
nished, minimizing the time for completing the multistep process.
In the current environment of intense market competition, batch process in-
dustries stand to benet from faster process development. Two batch process areas,
operating procedure synthesis (OPS) and process hazards analysis (PHA), are time-
consuming because they are often performed manually. Recently, a Purdue Uni-
versity Group developed two intelligent systems, iTOPS and Batch HAZOPExpert
(BHE), to automate OPS and PHA. Two applications from the specialty chemical
industry are presented to demonstrate the utility of the integrated system [34].
Resolution by forming diastereomeric salts is still an important method for ob-
taining enantiomerically pure chiral compounds. Generally, tedious trial-and-error
experiments are required to identify the satisfactory combination of resolving
agents. Researchers at Roche Discovery Welwyn described the use of dierential
scanning calorimetry (DSC) as a means to identify diastereomeric salts with a clear
eutectic composition that is needed for eective resolution and for forming the
basis of a resolving agent screening process. This work also included automated
salt synthesis using the ACT robot in 96-well microtiter plate format. (ACT refers
to the Advanced ChemTech synthesis robot, which is primarily used for solid-phase
synthesis.) Automation showed good correlation with the nonautomated experi-
ment, and is therefore suitable for future screening of resolving agents. Rapid data
analysis was facilitated using the in-house software package Resolution Compan-
ion, which also enabled identication of the optimum crystallization conditions
following a trial crystallization experiment. This software package enables (1) the
construction of binary-phase diagrams using the Schroedervan Laar equation, (2)
the rapid analysis of data from DSC thermograms, and (3) the construction of
ternary-phase diagrams for evaluation of optimal solution concentrations. The au-
thors also point out that the method has some limitations: (1) a failure to crystal-
lize under multiwell evaporation does not imply crystallization will not occur under
other conditions, (2) polymorphism, degradation, and signal overlap can compli-
cate DSC analysis, and (3) solvate formation can markedly alter the phase diagram.
Nevertheless, the use of DSC to aid in the selection of a resolving agent has been
demonstrated and forms the basis of an automated screening procedure [35].
31.3
Summary and Outlook
In the case that an entirely novel process has to be developed, where only a mini-
mal level of chemical knowledge exists, the true combinatorial high-throughput
methodology (hundred-at-a-time or thousand-at-a-time approach) needs to be im-
plemented. There will be, however, signicant technical hurdles between the ten-
at-a-time technology and the hundred-at-a-time technology, and it will take con-
siderable eorts to develop these systems [36]. Not only will the close collaboration
between process chemists and combinatorial chemists be necessary, but the involve-
ment of engineers and computer scientists will be likewise indispensable for full
implementation. Nevertheless, the truly high-throughput systems will fundamen-
tally change the way chemical process development is carried out in the future.
Acknowledgments
The authors wish to thank Dr Michael Cannarsa (Synthetech) for his contribution
to the original ideas of this manuscript, Ms Cate Larsen (Symyx Technologies) for
useful help, and Ms Silvia Lee (Symyx Technologies) and Ms Kathryn Boykin
(XenoPort) for their invaluable support in reference and patent searches. The
authors are grateful for Ron Krasnow (Symyx Technologies) for checking the
manuscript.
References
32
Combinatorial Methods in Catalysis
Bill Archibald, Oliver Brummer, Martin Devenney, Sasha Gorer,
Bernd Jandeleit, Tetsuo Uno, W. Henry Weinberg, and
Thomas Weskamp
32.1
Introduction
32.1.1
Combinatorial Catalysis
The conventional process used to search for new materials in important areas such
as biopharmaceuticals, materials science, and catalysis has traditionally been long
and expensive, relying on a large number of iterative steps to produce new or op-
timized materials. Increasing competition in the chemical industry demands faster
and more ecient identication and development of improved materials, includ-
ing catalysts. This drive toward increased research productivity was rst encoun-
tered in the pharmaceutical industry, where long development times and high re-
search costs necessitated the introduction of new research and development (R&D)
approaches designed to accelerate the drug discovery process. One approach in-
tended to speed up the discovery process is to develop high-throughput methods
for rapidly synthesizing and screening large numbers of diverse chemical com-
pounds, searching for trends in desired parameters. Virtually every major drug
manufacturer now utilizes this new research technology, called combinatorial
chemistry, as an integral part of its research and development program [110].
The combinatorial process involves the design and synthesis of discovery libraries
aimed at eciently exploring large numbers of structurally or compositionally di-
verse compounds thought to be of interest as a result of a review of their chemical,
biological, physical, and/or structural properties. Rapid, sensitive measurements of
one or more relevant chemical or physical properties of each library member result
in the identication of a family of lead compounds with a desired property. These
compounds are then optimized by continuously varying the stoichiometries or
structures of a more focused set of precursors in a focus library. Compounds with
optimum compositions are then synthesized in quantities sucient for detailed
characterization and evaluation. Finally, with methods of combinatorial chemical
synthesis evolving together with rapid analysis and high-throughput screening, data
management is becoming challenging and requires appropriate database technol-
ogy.
The biopharmaceutical industry has levered development of combinatorial tools
from the broad availability of low-cost, high-performance computers, robotics, mo-
lecular modeling, database software tools, and tools created for the clinical diag-
nostics, optics, and semiconductor industries. Adoption of combinatorial methods
has also created novel research areas such as bio- and chemoinformatics to man-
age and mine the large amounts of structural and functional information obtained
from combinatorial investigations [11].
Application of combinatorial methods to nonbiological materials and catalyst
development is still in its infancy. Work initiated by Schultz and coworkers in 1995
has initiated tremendous eorts in this eld [12], although the concepts and their
experimental realizations can be traced back to Hanak in the early 1970s [13]. The
combinatorial paradigm, however, is beginning to be explored increasingly in in-
dustrial and academic laboratories around the world. Combinatorial methods have
been applied to optical and electronic materials, magnets, polymers, and catalysts
for commodity, specialty, and ne chemical applications. Traditional materials de-
sign and screening approaches, which are often based on an a priori mechanistic
approach, typically only lead to structures that were known or expected by the
chemist. This applies even more so to catalysis, where knowledge of the underly-
ing mechanism of the catalytic reaction often drastically limits the number of prob-
able metal/ligand combinations. Combinatorial catalysis methods allow for both
the probable and improbable metal/ligand combinations to be investigated. Fur-
thermore, combinatorial catalysis is rapidly taking advantage of the numerous solid-
and solution-phase synthetic methodologies that exist, including polymer-supported
reagents [1418].
In general, three dierent approaches exist for preparing and testing libraries of
compounds (Fig. 32.1). Conventional research, performed in a one-at-a-time or
serial fashion, provides thorough quality control over samples entering the screen-
ing process. Combinatorial methods that involve, for example, split-and-pool
synthesis are much faster and make the preparation of relatively large numbers of
compounds feasible; however, compounds entering the screening equipment often
lack phase purity. Additional methods that are intermediate to the two extremes
outlined above are based on the parallel synthesis of compounds in a spatially ad-
dressable format with usually one composition per site, which are then coupled to
automated screens. This allows multidimensional problems in the discovery and
optimization of catalysts to be eciently addressed by coupling parallel synthesis
with automated or high-throughput screens.
A SciFinder= reference search performed in February 2001 using the key word
combinatorial resulted in more than 9000 hits. An analysis of the literature
search reveals a tremendous growth in scientic publications and patent applica-
tions in the combinatorial eld. Most contributions deal with combinatorial appli-
cations to pharmaceutical, biological, and medical disciplines; however, signicant
advances in the development of combinatorial approaches to the discovery and
32.1 Introduction 887
optimization of new materials and catalysts has occurred in recent years. An in-
creasing number of review articles have been written that address new methods in
combinatorial chemistry and high-throughput screening for chemical process de-
velopment [19] and, of course, combinatorial catalysis [20].
This chapter summarizes applications of combinatorial methodologies to the
discovery and optimization of new catalysts published in the scientic literature
between 1995 and August 2001; patents, patent applications, and conference pro-
ceedings have only been included if they are of the utmost importance.
After this introductory section, the following section introduces combinatorial
approaches to novel metal binders, combinatorially functionalized polymers, and
combinatorially discovered metal complexes as enzyme mimetics. Section 32.3
discusses combinatorial catalysis in asymmetric synthesis. Section 32.4 describes
examples of multidimensional screening in combinatorial catalysis. Section 32.5
summarizes recent eorts in one-pot multisubstrate screening approaches, while
Section 32.6 summarizes some applications of combinatorial methods in the search
for novel polymerization catalysts. Section 32.7 describes combinatorial inorganic
homogeneous catalysis, and Section 32.8 deals with combinatorial approaches in
heterogeneous catalysis. Section 32.9 emphasizes combinatorial electrocatalysis.
Novel high-throughput screening tools are discussed in Section 32.10. Section 32.11
includes a summary and an outlook concerning the future of combinatorial catal-
ysis science.
32.1.2
Combinatorial Organic and Organometallic Catalysis
32.2
Metal-binding Ligands, Polymeric Enzyme Mimetics, and Metal Complexes as
Enzyme Mimetics
32.2.1
Combinatorial Approaches to Metal-binding Ligands
ferent cyclen libraries were synthesized, diering in both length and nature of the
peptidic side arms (Fig. 32.2).
The side-chain-protected and -deprotected resins were agitated with diluted
aqueous Cu(II) and Co(II) ion solutions. Selective metal binding was screened
visually in a pooled assay identifying binding to Cu(II) and Co(II) by characteris-
tic bright blue and red colors. Deconvolution of the chemically encoded polymer
beads allowed the identication of ligand structures that were responsible for the
selective binding of the metal ions [25]. The peptidic sequences generally diered
for Cu(II) and Co(II) ions, and the length and chemical nature of the peptidic ap-
pendages in both protected and unprotected amino acid forms signicantly inu-
enced the anity of the new peptidic functionalized tetraamines for Cu(II) and
Co(II) ions relative to the corresponding unfunctionalized or alkylated cyclen core.
In a related combinatorial procedure, the development of uorescent sensors for
nanomolar aqueous copper was reported by Sames and coworkers [26]. A library of
ionophoric ligands was generated based on three dierent N-containing macro-
cyclic or tridentate scaold using a split-and-pool strategy and an orthogonal tert-
butyloxycarbonyl (Boc)/Fmoc protecting group protocol. The library was assayed
visually by incubating the solid support-bound ligands with aqueous Cu(II) solu-
tion followed by reaction with a copper-selective staining reagent (blue color). All
identied ionophores contained a carboxylic group as well as an aromatic nitrogen
heterocycle. A pyrazine-containing ligand was selected as a lead structure for fur-
ther development of a uorescent sensor bearing a covalently linked dansyl uo-
rophore (Fig. 32.3).
890 32 Combinatorial Methods in Catalysis
rise to colored metal complexes from the pooled assay, and catalytic activity of the
new metal complexes in the ring closure of the Arnstein tripeptide (ACV) to iso-
penicilline N was not addressed.
Metalloproteins, synthetically or biosynthetically appended to proteins or other
biomolecules, are nding increasing utility in the biochemical analysis of non-
covalent proteinnucleic acid and proteinprotein interactions. The amino termi-
nal Cu(II)- or Ni(II)-binding (ATCUN) motif is a structural feature of several natu-
rally occurring proteins such as certain types of albumins (e.g. human serum
albumin (HSA), bovine serum albumin (BSA), and rabbit serum albumin (RSA)
and neuromedins C and K, among several others [30]. Generally, this domain is
described as Ni(II)H2 NaAA1 aAA2 aHis. Long and coworkers employed a posi-
tional scanning combinatorial protocol to optimize the desoxyribose-based cleav-
age of B-form DNA by Ni(II)H2 NaAA1 aAA2 aHis metallopeptides [31]. Using a
standard Boc protocol and split-and-mix technique on methylbenzhydrylamine
(mBHA) resin, two libraries were generated from a selection of l-amino acids in
which the rst (AA1 ) and the second position (AA2 ) of the peptide ligand were
varied within the H2 NaAA1 aAA2 aHis sequence. The libraries were assayed after
cleavage from the solid support for increased direct DNA cleavage relative to
Ni(II)H2 NaGlyaGlyaHis after incubation with a Ni(II) source and oxidative acti-
vation of the metal complex with KHSO5 or magnesium monoperoxophthalate
(MMPP). Increased catalytic activity was found when the amino-terminal peptide
position contained a hydrophobic amino acid and the second peptide position con-
tained an ionic or polar amino acid. The optimized and resynthesized metal-
lotripeptide domain Ni(II)HNaProaLysaHis was found to oxidatively cleave DNA
892 32 Combinatorial Methods in Catalysis
anolic solutions of Ni(II) and Fe(III) salts. After staining with the appropriate indi-
cators (dimethyglyoxime (DMG) in the case of Ni(II) or KSCN in the case of Fe(III)),
beads containing high concentrations of Ni(II) or Fe(III) were visually identied by
their red or orangered color, respectively, using a light microscope. The chemical
tags on the selected beads were cleaved from the solid support, allowing deconvo-
lution of the chemical synthesis history of an individual selective metal-binding
ligand. Several new ligand structures were disclosed that selectively bind Ni(II) or
Fe(III), each of which consisted of histidine residues, and, in most cases, a set of
two turning elements and/or terminating caps.
Technetium-99m complexed to organic compounds, proteins, peptides, or anti-
bodies plays an important role in the radioimmunodetection (RIAD) of organs and
in tumor imaging because of its low cost, ideal physical properties, and broad avail-
ability. Schneider-Mergener and his group synthesized and screened a cellulose-
bound hexapeptide combinatorial library for the identication of technetium-99m
( 99m Tc)-binding peptides [34]. A combinatorial library of approximately 8000 hex-
apeptides of the general structure B1 aXaB2 aXaB3 aX was synthesized using auto-
mated synthesis on a cellulose platform (spot synthesis) yielding the hexapep-
tide library in a spatially addressable format. The amino acids Bi i 13) were
positioned on the dened positions 1, 3, and 5 within the hexapeptide sequence
and each was one of the 20 naturally occurring l-amino acids. The amino acids Xi
in positions 2, 4, and 6 were systematically randomized. After incubation with an
aqueous pertechnate solution as a 99m Tc source, the library was screened for 99m Tc
anity by phosphorimaging. Besides a variety of cysteine-containing hexapeptides,
known for their strong complexing capability toward 99m Tc, peptidic ligands con-
taining the amino acids His, Lys, Arg, and Met showed signicant anity toward
the radioisotope. In subsequent focused libraries, the three X-positions were dened
by positional scanning and revealed the noncysteine-containing primary se-
quences LysaGlyaHisaSeraHisaVal and LysaAlaaMetaTyraHisaGly as superior
99m
Tc ligands.
Pirrung and Park disclosed the discovery of selective metal-binding peptoids
using 19 F-encoded combinatorial libraries [35]. A method for encoding solid-phase
split-and-mix combinatorial libraries was developed taking advantage of the large
chemical shift dispersion of synthetic uoroarenes (F-tags). Nine uoroarenes ( 19 F-
tags) bearing linkers for attachment to solid support through a photocleavable linker
(Holmes photolinker) were prepared [36]. A 90-member library of N-alkylglycines
bearing substituted succinamides was prepared on solid support from a set of nine
amines, in which the amine was encoded by the uorinated tag, and ten anhydrides
(Fig. 32.6). Bead pools were visually assayed for metal binding by treatment with
a Cu(II) ion source (blue color). Selection of ten blue beads was followed by ir-
radiation to release the tag. Sensitive 19 F-nuclear magnetic resonance (NMR)
was used for decoding. Two compounds, obtained from solution-phase synthesis,
showed high anity either for copper(II) (CuOTf2 ) (K d 44 mM), or iron(III)
[Fe(2-ethylhexanoate)3 ] (K d 31 mM).
Walt and coworkers reported the solution-phase synthesis and characterization
of metal-binding indicators with diverse optical responses on exposure to various
894 32 Combinatorial Methods in Catalysis
heavy metal ions [37]. A combinatorial approach, based on azo coupling with sub-
sets of either phenolic compounds or aromatic amines, generated a library of azo
dyes. Each reaction mixture containing the product(s) of the azo coupling was in-
cubated with a series of a heavy metal ion sources. Physical parameters such as
absorbance and uorescence spectra of the resulting complexes were measured. Of
the azo dyes prepared, terdentate dyes were particularly useful, providing distinct
spectral responses to three or more metal ions.
Metal-binding agents that are immobilized on appropriate inorganic supports
are important in environmental chemistry, in sensing chemistry, and in bio-
chemistry. Bergbreiter and coworkers prepared a library of surface-bound metal-
complexing hydroxamic acids on silica-coated glass plates as selective and reversible
metal sensors [38]. Aminopropylated silica was reacted with a set of 12 dicarboxylic
acid derivatives as acylation agents to modify the aminopropylated surface into
a carboxylic acid-containing surface. Subsequent attachment of three dierent hy-
droxylamines furnished the immobilized hydroxamic acids (Fig. 32.7). Binding of
the hydroxamic acids to Fe(III) ions resulted in a reddish coloration which was as-
sayed qualitatively or semiquantitatively by either visual inspection or using diuse
reection spectrometry, respectively. The highest yields in the formation of an
Fe(III) complex were obtained when glutaric anhydride or maleic anhydride was
used as an acylation agent. The results were conrmed on bulk samples.
32.2 Metal-binding Ligands, Polymeric Enzyme Mimetics, and Metal Complexes as Enzyme Mimetics 895
32.2.1.3 Miscellaneous
Besides the potential utility of a particular class of ligands for a particular catalytic
application, several features must be considered in the selection of ligand libraries.
Reactions that use more than two dierent starting materials are called multi-
component reactions (MCRs) [39]. In the light of chemical productivity and gen-
eration of molecular diversity, an ideal MCR should comprise more than two com-
ponents. High atomic economy (ideally the incorporation of all of the atoms that
build the starting materials into the nal product) is an asset [40], and, further-
more, ligand synthesis and purication procedures must be readily amenable to
parallel synthesis techniques. With respect to their productivity, yield, conversion,
and facile execution, MCRs occupy an outstanding position among other reactions
making them especially interesting for the concept of combinatorial chemistry.
Attempts are now being made to discover new multicomponent reactions by means
of combinatorial technology [39].
In this context, Lapointe reported the parallel synthesis of substituted amino-
methylphosphines of the general structure R1 R 2 PCHR 3 NR 4 R 5 in 1999 [41]. By
combining two secondary phosphines (R1 R 2 PH), a subset of six substituted alde-
hydes (R 3 CHO), and eight secondary amines (HNR 4 R 5 ) in a facile Mannich-type
three-component condensation reaction, a 96-member library of substituted ami-
nomethylphosphines was synthesized (Scheme 32.1). According to their backbone-
building atoms, these ligands were termed PCN ligands. The relatively mild reac-
tion conditions [tetrahydrofuran (THF), room temperature], the large number of
substructures available, and the lack of protection/deprotection steps for the phos-
phine moiety made this protocol an attractive target for parallel synthesis. The
delivery of the reagents and the mixing were performed by an automated liquid
dispenser. The crude condensation products were of sucient purity (@ 95% by
1
H- and 31 P-NMR spectroscopy) for their direct use as ligands. The parallel synthe-
sis of libraries of metal complexes of these PCN ligands and their use as ethylene
polymerization catalysts, for example, is currently being investigated [42]. This ex-
ample is one of the rst reports about parallel syntheses of organophosphines,
and, owing to the broad availability of the appropriate building blocks, large num-
bers of diverse functionalized PCN ligands are accessible.
896 32 Combinatorial Methods in Catalysis
Fig. 32.8. Library of dipeptides each bearing two bidentate pyridine ligands.
tion of a small library of 25 dipeptides, each bearing two bidentate ligands. These
ligands can serve as building blocks for the construction of disparate metal-binding
molecules.
Schmalz and coworkers reported a modular approach to a new class of structur-
ally diverse bidentate P/N, P/P, P/S, and P/Se chelate ligands for transition metal
catalysis [46]. Bis-protected bromohydroquinones that are accessible from hydro-
chinone were used as the central building blocks (Scheme 32.4). The rst donor
functionality (L 1 ) is either introduced by Pd-catalyzed amination, by Suzuki cross-
coupling, or by lithiation and subsequent treatment with electrophiles (e.g. chlor-
ophosphanes, disuldes, or carbamoyl chlorides). In the case of the Suzuki reac-
tion, the bromohydroquinones were rst transformed into the corresponding bor-
onic acid via lithiation and reaction with triisopropylborate and then reacted with
various heteroaryl bromides. After selective deprotection, the second ligand tooth
is introduced via reaction of the phenolic OH moiety with a chlorophosphane, a
chlorophosphite, or a related reagent. Some of the resulting ligands were converted
into the corresponding PdCl2 complexes that were characterized by X-ray crys-
tallography.
32.2 Metal-binding Ligands, Polymeric Enzyme Mimetics, and Metal Complexes as Enzyme Mimetics 899
32.2.2
Combinatorial Libraries of Polymeric Catalysts as Enzyme Mimetics
be generally of great interest. The obvious ease and practicability of this approach
may allow for a rapid evolution and optimization of polymeric catalysts and may
be useful, especially if reproducibility issues are carefully addressed, to identify
active polymeric catalysts for many other important chemical reactions.
32.2.3
Combinatorial Synthesis Enzyme Mimetics
The researchers extended their methodology towards the synthesis of the corre-
sponding azacrown etherlanthanide complexes that were capable of hydrolyzing
phosphate esters, a crucial chemical transformation in areas such as nucleic acid
chemistry [60]. The lanthanide complexes synthesized catalyzed the cleavage of
904 32 Combinatorial Methods in Catalysis
phospho di- and triesters and double-stranded DNA as well, which opens up the
possibility of using this class of chemicals as therapeutics. Another important po-
tential application is the hydrolytic decontamination of toxic organophosphates,
phosphonates, and uorophosphates which are widely used as insecticides and are
stockpiled as chemical warfare agents.
A 625-member library of undecapeptides has recently been assayed for cata-
lytic activity in phosphate hydrolysis, using test substances that in the presence of
oxygen form insoluble indigo dyes. Using a split-and-mix approach, Berkessel and
Herault synthesized on solid support sequences PheaXaGlyGlyaXaGlyGlyaXa
GlyGlyaX, where X is one of Arg, His, Tyr, Trp, or Ser [61]. About 2500 beads were
incubated with Cu(II), Zn(II), Fe(III), Co(III), Eu(III), Ce(IV), and Zr(IV) solu-
tions; only in the presence of Zr(IV) did the observation of the blue indigo color
indicate catalytic hydrolysis activity. The most promising candidates were isolated,
identied, and synthesized on solid support and in solution to conrm the library
screening results.
Scheme 32.6. Evolution of diamide ligands for Mn(III)-catalyzed oxygenation on solid support.
32.2 Metal-binding Ligands, Polymeric Enzyme Mimetics, and Metal Complexes as Enzyme Mimetics 905
changes due to the exothermic catalytic acetylation of ethanol were monitored di-
rectly on the oating beads by infrared (IR) thermography. The hottest 23 beads
were then manually selected, and decoding revealed three structures as eective
catalysts (Scheme 32.7, bottom). Most of them have N-4-pyridylproline incorpo-
rated as the capping agent and contained either (R)- or (S)-3-aminoquinuclidine as
the primary amine component.
32.3
Combinatorial Catalysis in Asymmetric Synthesis
32.3.1
Combinatorial Catalyst Libraries in Enantioselective Additions of Dialkyl Zinc
Reagents
By using this methodology, the same group also synthesized a library of 60 chiral
Schi base ligands that contains a set of dierent metal-binding sites (a phenol, an
imine, and a secondary sulfonamine) [74]. By means of a high-throughput screen-
ing approach, a ligand was identied that catalyzed the copper-mediated, enantio-
selective conjugate addition of diethyl zinc to cyclic enones with up to 90% ee (Fig.
32.12).
In a highly related approach, Liskamp and coworkers developed a solid-phase
synthesis of bis-(sulfonamide) ligands [75]. A library of these ligands was screened
in a simultaneous substrate-screening procedure for the ability to enantioselectively
catalyze the Ti(OiPr)4 -mediated addition of diethyl zinc to aldehydes. The best solid-
phase candidate was resynthesized in solution and gave enantioselectivities up to
66% (Fig. 32.12).
32.3.2
Ligands for the Lewis Acid-catalyzed Asymmetric Aza-DielsAlder Reaction
32.3.3
Divergent Ligand Synthesis for Enantioselective Alkylations
In 1998, Burgess and coworkers highlighted in several examples the value of di-
vergent ligand synthesis strategies for the preparation of focused ligand libraries
910 32 Combinatorial Methods in Catalysis
and the use of automated ligand screening [82, 83]. By denition, divergent syn-
theses of chiral ligands are characterized by an optically pure material being pro-
duced and then subsequently being used to prepare many ligands of a structurally
similar class of ligands.
Based on l-serine, a chiral building block (chiral synthon chiron) was pre-
pared as a key intermediate that was then converted into a family of 13 novel chiral
phosphine oxazoline ligands using several dierent routes and employing conven-
tional organic solution synthesis [82]. It was shown that the divergent ligand syn-
thesis approach was very suitable for eciently accessing this ligand class with
good diversity in terms of steric bulk and/or electronic pertubation (Fig. 32.13).
The ability of these ligands to asymmetrically catalyze the addition of dimethyl
malonate to 1,3-diphenylprop-2-enyl acetate was investigated [84]. The best of these
ligands gave a palladium complex that catalyzes the allylic alkylation with up to
94% ee.
In a dierent modular approach to generate chiral catalysts for the same trans-
32.3 Combinatorial Catalysis in Asymmetric Synthesis 911
amide moieties were reduced using BH3 THF to give a library of 20 chiral penta-
mines. From this set of ligands, only moderate selectivities were obtained. Through
modication of the N-methylpiperazine moiety, one complex was identied that
catalyzes the alkylation of piperidinepropionamide with good yields and up to
84% ee (Scheme 32.12).
32.3.4
Chiral Phosphine Ligands for Asymmetric Hydrogenation
up to 18% were observed, but some trends and correlation between peptide se-
quence and selectivity were established. In a more recent publication, the same
group synthesized larger libraries, giving access to ligands that were used for the
catalytic asymmetric hydrogenation of methyl-2-acetamidoacrylate in the presence
of rhodium. Enantiomeric excesses up to 38% were obtained [92].
32.3.5
Asymmetric Reactions Catalyzed by Schi Base-type Ligands the Positional
Scanning Approach
automated screening, not only oers unique opportunities for the identication of
substrate-specic catalysts (ne tuning) but also permits the discovery of ligands
that possess unusual properties and which might otherwise elude detection.
In 1992, Hayashi and coworkers reported the addition of trimethylsilyl cyanide
(TMSCN) to epoxides catalyzed by titanium Schi base complexes to yield racemic
b-cyanohydrins [95]. Based on this work, Snapper, Hoyveda, and coworkers dis-
closed a diastereo- and enantioselective version of this reaction in 1996 (Scheme
32.14) [93]. With combinatorially optimized chiral titanium Schi base complexes,
nonracemic b-cyanohydrins were obtained in good chemical and optical yields. Chi-
ral dipeptidic hemisalen ligands of the general structure SBaAA2 aAA1 aGlyOMe
(SB Schi base or hemisalen; AA1 amino acid) were synthesized in a modu-
lar fashion from various amino acid and aldehyde building blocks using parallel
solid-phase synthesis techniques. After cleavage from the solid support, complex-
ation with a titanium alkoxide precursor, and parallel screening of the chiral Lewis
acids for eciency and enantioselectivity, a potent catalytic system for ring open-
ing of cyclohexene oxide in 89% enantiomeric excess was discovered (Scheme
32.14).
Subsequent work by the same group demonstrated that screening did not require
cleaving the Schi base ligand from the solid support, which signicantly accel-
erated the optimization process [96].
Using a closely related approach for their ligand synthesis, Sigman and Jacobsen
disclosed polymer-supported Schi base catalysts of the general structure linker1
amino acidlinker2 diaminephenylidene in 1998 [94]. In this example, the on-
bead iterative optimization of ligands for the asymmetric hydrocyanation of imines
(Strecker reaction) is described. The Strecker reaction constitutes one of the most
32.3 Combinatorial Catalysis in Asymmetric Synthesis 915
direct and viable strategies for the asymmetric synthesis of a-amino acid deriva-
tives [97]. Ligand synthesis was made amenable to solid-phase synthesis techniques
by choosing readily accessible building blocks and reaction conditions (Scheme
32.15). Systematic but non-obvious permutation cycles of their building blocks
(iterative optimization), such as sequential variation of the amino acid compo-
nents, the nature of the linker, or the diamino bridging moiety, revealed a transi-
tion metal-free catalyst after the synthesis of only three ligand generations that
mediated the reaction of both aromatic and aliphatic aldimines in more than 70%
and 83% enantiomeric excess, respectively. Screening was performed sequentially
by gas chromatography using a chiral stationary phase. The best catalyst was re-
synthesized in solution and showed further improved enantioselectivity of 91% ee
for an aromatic aldimine. Meanwhile, independent investigations by the same
group disclosed a very active chiral salen Al(III) complex for the same reaction, re-
sulting in imine hydrocyanation products of enantiomeric excesses as high as 95%
[98].
Hoveyda, Snapper, and coworkers applied their modular approach to Schi base-
type ligands for identifying catalysts for a variety of asymmetric transformations.
Scheme 32.16 summarizes the novel catalytic systems for enantioselective addi-
tions of cyanide to imines (A) [99], for Cu-catalyzed conjugate addition of alkylzinc
reagents to cyclic enones (B) [100], for Zr-catalyzed addition of diethylzinc to
imines (C) [101], and for Cu-catalyzed allylic substitutions (D) [102].
916 32 Combinatorial Methods in Catalysis
32.3.6
Identication of Novel Catalysts for the Asymmetric Epoxidations via the Positional
Scanning Approach
In 1999 Francis and Jacobsen disclosed the discovery of a novel catalyst for alkene
epoxidation [103]. Using a solid-phase synthesis protocol on an aminomethyl poly-
styrene resin, ligands bearing potential metal-binding moieties were prepared. The
ligands comprised ve amino acids with donor side arms (Asp, Cys, His, Met, and
Ser), three dierent chiral linking elements each diering in rigidity (1-amino-2-
indanol, trans-1,2-diaminocyclohexane, and Ser), and 12 dierent capping agents
32.4 Multidimensional Combinatorial Screening 917
32.4
Multidimensional Combinatorial Screening
pounds entering the assay screen. Methods intermediate between the two extremes
outlined above are based on parallel or array syntheses or screening, in a spatially
addressable format with usually one compound per well, coupled to automated
screens. This approach facilitates a relatively high level of quality control and a sig-
nicantly enhanced throughput compared with sequential methods. Thus, multi-
dimensional problems in the discovery and in the optimization of catalysts may be
eciently addressed by coupling parallel ligand synthesis with automated or high-
throughput multidimensional screens (e.g. metal precursors, ligands, solvents, and
reactant ratios). An integrated instrumental set-up for parallel high-throughput
catalyst screening should encourage researchers to investigate a broader variety of
systems, some of which may be regarded as too unusual to warrant testing in a
sequential fashion.
32.4 Multidimensional Combinatorial Screening 919
32.4.1
Catalyst Discovery and Optimization Using Catalyst Arrays
Allylic alkylation products with an enantiomeric excess as high as 94% were ob-
served. In a closely related experiment, using the same instrumental set-up, ligands,
metal sources, and additives, but a more dicult to control methyl-substituted
allylic substrate, the best enantiomeric excess was determined to be 74% [83]. The
related divergent ligand synthesis strategy and an identical instrumental set-up for
the high-throughput catalyst screening as described above revealed novel so-called
propeller-shaped, C3 -symmetric triarylphosphines as chiral ligands [108, 109].
The ligands were screened in an allylic amination reaction with phthalimide as the
nucleophilic component, resulting in N-Substituted phthalimide derivatives were
obtained with up to 82% enantiomeric excess.
32.4 Multidimensional Combinatorial Screening 921
Taylor and Morken uncovered an eective catalyst for the diastereoselective re-
ductive aldol reaction with the aid of an arrayed catalyst evaluation protocol [110].
An array of 192 experiments in glass 96-well microtiter plates was used to evaluate
the interdependence of a variety of reaction variables such as the metal precursor,
the ligand, and the hydride source in a parallel fashion. The parent test reaction
was the reductive coupling of methyl acrylate and benzaldehyde (Scheme 32.20).
Each reaction was analyzed for relative conversion and stereoisomer ratios by
conventional gas chromatography employing chiral stationary phases (GCCSP ). Sev-
eral relationships between reaction condition and yield could be rapidly concluded,
including that catechol borane was a superior reducing agent with the largest
number of catalysts under the reaction conditions. The novel catalyst derived from
[(COD)RhCl]2 , Me-DuPhos, and Cl2 PhSiH provided generally moderate yields (15
69%) for a variety of other substrates but with a high syn stereoselection of up to
23:1 syn/anti. The study demonstrated a rapid assessment of the interdependence
of reaction parameters showing that traditional empirical catalyst development
approaches, where reaction variables are independently optimized, may not have
revealed all active catalyst formulations found within the array.
The metal-catalyzed addition of amines to carbon double bonds, especially of
acrylic acid derivatives yielding b-amino acid derivatives, which are useful in peptide
analogs or as precursors of optically active amino alcohols, diamines, and lactams,
is of utmost importance. Kawatsura and Hartwig uncovered several late transition
metal complexes that catalyze the addition of amines to acrylic acid derivatives using
a novel parallel colorimetric assay to analyze the addition of primary and second-
ary alkyl amines to substrates with CbC bonds [111]. All possible combinations of
seven metal precursors and 11 phosphine ligands were evaluated in a glass 96-well
microtiter plate format for catalytic activity in the addition of piperidine or octyl-
amine to methacrylonitrile as representative experiments (Scheme 32.21).
Using multiple-well plates, discrete homochiral Lewis acid complexes were indi-
vidually generated in solution from four dierent metal salts, three dierent com-
mon enantiopure ligands, three dierent solvents, and two dierent additives.
Screening was performed sequentially in about 1 week by measuring the enantio-
meric excesses and conversions by means of an automated HPLC equipped with a
chiral stationary phase to result in 144 sets of approximate yields and enantiomeric
excesses. Reproducibility was conrmed and the most ecient chiral Lewis acid
was a combination of MgI2 and (R,R)-1,2-diphenylethylenediamine in acetonitrile
in the presence of 2,6-lutidine to aord the N-arylpiperidinone in 97% enantio-
meric excess.
Hydrosilylation of substrates such as acetophenones with ruthenium catalysts
requires mixed P/N ligands for activity and selectivity; neither pure P/P-chelating
nor pure N/N-chelating ligands show activity and selectivity alone [114]. Based
32.4 Multidimensional Combinatorial Screening 923
on that knowledge and adapting Noyori and coworkers concept of mixed ligand
ruthenium complexes as precatalysts [115], Frost and coworkers developed an e-
cient ruthenium-catalyzed asymmetric hydrosilylation of ketones using high-
throughput parallel screening to optimize ligand combinations for the target
reaction [116]. A small 50-member mixed ligand library of ruthenium diamine/
diphosphine complexes (precatalysts) was prepared in situ by addition of enantio-
merically pure tol-BINAP ligands to a ruthenium precursor [RuCl2 (C6 H6 )2 ], fol-
lowed by addition of various monoimine or diamine ligands. Diphenylsilane and
the precatalysts were added to a solution of acetophenone as the substrate to syn-
thesize the corresponding 1-phenyl ethanols (Scheme 32.23).
32.4.2
Parallel Array Screening for Catalyst Optimization Using Discovery and Focused
Ligand Libraries
Product analysis was performed by serial gas chromatography (GC). The results
of the initial 96-member library screening allowed for assessment of structure
activity proles. 8-Hydroxyquinoline derivatives proved to be outstanding ligands
for the copper-catalyzed phenoxylation of the test substrate, yielding the desired aryl
ether in @80% yield. Among the best ligands for methoxylation in terms of yield
and low arene-reduction byproduct were pyridines bearing amino groups in posi-
tion 2 (60% to >90%). Subsequent smaller focused ligand libraries including 8-
quinolinols (30 member) and 3-hydroxypyridine (23 member) for the phenoxylation
reaction and 2-aminopyridine-type ligands (41 member) for the methoxylation con-
rmed the results obtained. The most eective ligands were found to be generally
useful for alkoxylation of a variety of substrates, including intramolecular ether-
forming reactions.
Parallel solution-phase synthesis techniques and parallel array assaying facili-
tated the identication and optimization of new nonpeptidic chiral dirhodium(II)
carboxylates useful in the enantioselective SiaH insertion reaction of diazoesters
with silanes [121123]. A systematic search through the Available Chemical Direc-
tory (ACD) resulted in a set of 2000 useful alternative chiral carboxylic acid ligands
which were then clustered and the centroid of each of 80 clusters selected as a rep-
resentative acid [124]. The representative chiral carboxylic acid ligand from each
of the 80 clusters then underwent a 20-at-a-time parallel synthesis format ligand
exchange with a dirhodium(II) carbonate precursor to yield the desired chiral dir-
hodium(II) carboxylate catalyst. A total of 69 chiral dirhodium(II) catalysts were
assayed in the targeted enantioselective carbenoid SiaH insertion of methyl 2-
(diazo)phenylethanoate into silanes with varying degrees of steric bulk to yield in-
sertion product (Scheme 32.27).
32.5 One-pot, Multisubstrate Screening 927
32.5
One-pot, Multisubstrate Screening
enantiomeric ratios for aromatic and aliphatic aldehydes in favor of the (R)-
enantiomer [(R)/(S) 9798:32] were obtained. A ligand derived from trans-(S,S)-
diaminocyclohexane and the N-protected b-amino sulfonyl chloride derived from
l-Phe performed the title reaction best (Scheme 32.29). The puried ligand was
subjected to characterization, and the screening results were conrmed by reaction
with the four separate aldehydes.
32.6
Combinatorial Approaches to Olen Polymerization Catalysts
respect, Symyx Technologies has developed general methodologies for the combi-
natorial synthesis, high-throughput screening, and characterization of libraries of
supported and unsupported organometallic compounds and catalysts [139a, 140].
For the discovery of novel polymerization catalysts, libraries of ligands in combi-
nation with various metals are screened in the presence of dierent monomers. In
order to optimize the yield and selectivity of a given organometallic complex in a
polymerization reaction, a variety of factors are tested in a high-throughput fashion,
including the form of the ancillary ligand precursor, the choice of the metal pre-
cursor, the reaction conditions (e.g. solvent, temperature, time), and the stability of
the desired product.
Mullen and coworkers tagged silica- or polymer-supported heterogeneous cata-
lysts for industrial olen polymerization with uorescent dyes [141]. Here, direct
detection of the dierent product beads obtained by dierent catalysts is available
through uorescent dyes that exhibit dierent emission wavelengths. The approach
starts with producing the tagged catalysts by supporting various metallocenes with
silica, activating them with MAO, and labeling them with dierent rylene dyes (Fig.
32.16). The dyes were chosen because of their high chemical stability, their high
tendency to physisorb on silica, their high uorescence yield, and because of the
fact that they do not inuence the polymerization. Also, a great variety of rylene
dyes with dierent emission wavelengths covering the entire visible spectrum is
available. These labeled catalysts are then mixed and introduced in a single poly-
merization vessel. During the olen polymerization each catalyst particle forms
only one product granulate through a particular growth process and can be con-
sidered as a microreactor. To assign the dierent compounds of the product mix-
ture to the employed catalyst, the polymer products are exposed to UV light, and
can be directly identied, manually separated, and characterized because of the
dierent emission wavelengths of the labels. The authors could demonstrate the
feasibility of this concept for the polymerization of ethylene as well as for the co-
polymerization of ethylene/hexene.
Fig. 32.16. Catalysts and dyes for the synthesis of tagged, supported catalysts.
934 32 Combinatorial Methods in Catalysis
BuLi, and reacted with 0.5 equiv. of TiCl 4 to give a library of 78 possible titanium
species. This complex library was then activated with MAO ([Al]/[Ti] 100] in tol-
uene and the resulting catalyst solution was exposed to propylene (2.7 atm.). Even
though 90% of the formed polymer could be washed away in reuxing diethyl
ether (atactic polypropylene), the remaining 10% of polymer was found to be
syndiotactic polypropylene ( 13 C-NMR microstructural analysis). Deconvolution
methods, i.e. the synthesis and testing of sublibraries, were used to identify suc-
cessfully the most active catalyst.
Hinderling and Chen reported the use of electrospray ionization tandem mass
spectrometry (ESI-MS/MS) and gas-phase ion molecule reactions for the rapid
screening of Brookhart-type Pd(II) olen polymerization catalysts [143]. A test
library of eight catalysts was prepared by reacting eight diimine ligands with
[(cod)Pd(CH3 )(Cl)], washing and drying of the complex, and activation with AgOTf
(Scheme 32.32). An electrospray mass spectrum of the mixture of complexes
showed that all eight catalysts are present in similar concentrations. The mixture
Scheme 32.32. Rapid screening of Brookhart- desolvation conditions to give polymeric ions
type Pd(II) olen polymerization catalysts by (and loss of DMSO); (iv) reject all ions below
ESI-MS/MS. (i) [(cod)Pd(CH3 )(Cl)], then a certain mass and subject the remaining high-
AgSO3 CF3 ; (ii) excess ethylene, then quenching mass ions to collision with Xe to induce b-
with DMSO; (iii) electrospray under mild hydride elimination.
936 32 Combinatorial Methods in Catalysis
32.7
Combinatorial Inorganic Catalysis
32.7.1
Combinatorial Libraries of Homogeneous Polyoxometalate-based Catalysts
PV2 W10 O40 ] 5 as the most eective among them. All the samples evaluated, in-
cluding the starting materials, exhibited catalytic eciency to a certain degree in the
conversion of THT to THTO, suggesting that Keggin-type polymolybdato- or tung-
stophosphates have the ability to oxidize selectively THT to THTO regardless of
their vanadium content. It was assumed that the catalytically most active species
were primarily Keggin polyoxometalates rather than lower nuclearity polyoxometa-
lates. Several other combinatorial libraries containing boron, silicon, and arsenic
were evaluated under a variety of conditions, but the results were not quantied.
Selective catalytic oxidations that proceed at satisfactory rates at ambient condi-
tions and use oxygen or air as terminal oxidant are of utmost interest [146]. In a
more recent contribution, Hill and his group reported a diversity-based discovery
and mechanistic investigation of selective homogeneous thioether oxidation by
the Au(III)Cl2 NO3 (thioether)/oxygen system [147]. A 150-member library of inor-
ganic complexes was constructed by combinatorially combining polyoxometalate
anions and redox-active cations. The library design comprised POMs with reversible
redox chemistry and redox-active d-block ions or appropriate precursors including
HAuCl 4 , as well as s-block and p-block cations as counterions for the POMs. The
catalyst library included also POM-free control formulations such as the chloride,
nitrate, or perchlorate salts of the appropriate redox-active d-block cations.
Each library member was evaluated for its ability to catalyze the oxidation of the
thioether mustard analog 2-chloroethyl ethyl sulde (CEES) to the corresponding
sulfoxide (CEESO), using only oxygen as the oxidant (Scheme 32.34).
Product analysis and distribution was assessed by conventional serial GC. Most
catalyst preparations showed little or no activity in the target reaction but three cat-
alyst compositions exhibited considerable catalytic activity. The catalytically active
compositions included the POMs [CuPW11 O39 ] 5 and [MnPW11 O39 ] 5 , and the
POM-free system AgNO3 , each of them with ve equivalents of HAuCl 4 . The most
optimized combination of HAuCl 4 and AgNO3 formed a catalyst that exhibited
orders of magnitude higher reaction rates and higher turnover numbers (TONs) at
ambient temperature and at 1 atm. of air or oxygen than previously reported cata-
lysts. Extensive kinetic and mechanistic studies for the oxygen-based oxidation of
thioethers catalyzed by the parent system Au/Ag/NO3 revealed two gold complexes
as the catalytically active species (Fig. 32.17).
938 32 Combinatorial Methods in Catalysis
32.7.2
Combinatorial Libraries and High-throughput Screening of Heterogeneous
Polyoxometalate Catalysts
Two classes of catalyst systems that gave more than 90% yield were identied. These
results were conrmed and validated in subsequent scale-up experiments, but the
proprietary catalyst composition was not disclosed.
32.8
Combinatorial Heterogeneous Catalysis
32.8.1
Introduction
0.5% Ag, Bi, Co, Cr, Cu, Er, Fe, Gd, Ir, Ni, Solution-based impregnation H2 O2 Infrared thermography 164
Pd, Pt, Rh, Ti, V, Zn, on Al2 O3 pellets
37, 110% Co, Cr, Cu, Fe, Ir, Mn, Ni, Pd, Solution-based sol-gel method Hyrogenation of 1-hexyne Infrared thermography 165
Pt, Rh, Ru, V, Zn on Si, Ti oxides
100 Na2 O/Al2 O3/SiO2 zeolites with Li Solution-based hydrothermal a a 166
and Cs zeolite synthesis
37 TiO2/Al2 O3/SiO2/TiO2/ZrO2 zeolites Solution-based hydrothermal a a 167
zeolite synthesis
645 clusters of Pt, Ru, Os, Ir on carbon Solution-based Methanol direct fuel cell Fluorescence acidbase indicator 157
coprecipitation
120 ternary thin lm clusters of Pt, Pd, Thin lm deposition: CO oxidation, CO NO Scanning quadruple mass 168
Rh, and Pd, Rh, Cu sputtering spectrometry
66 ternary combinations of 1% Pt, Pd and Solution-based impregnation Cyclohexane dehydro- REMPI 161
In on Al2 O3 genation to benzene
16 Au/Co3 O4 and Au/TiO2 powders Solution-based CO oxidation Quadruple mass spectrometry 162
coprecipitation
33 16% Ag, Au, Bi, Co, In, Cr, Cu, Fe, Solution-based sol-gel method Propylene oxidation Spatially resolved analysis of an 169
Mo, Ni, Re, Rh, Sb, Ta, Te, V, Y on Si, array of batch microreactors
Ti, Zr oxides
50 ternary and quaternary oxides of Co, Solution-based Ethane and propane Six-parallel gas chromatography 170
Cd, Fe, Ga, Ge, In, Mn, Mo, Ni, Nb, V, coprecipitation dehydrogenation
W, Zn
32.8 Combinatorial Heterogeneous Catalysis
941
942
30 binary combinations of Na2 WO4 and Solution-based impregnation, Oxidative coupling of Quadruple mass spectrometry 171
Mn on SiO2 ; Au and In on ZrO2 , TiO2 , deposition, and methane and CO
SiO2 , MgO, ZnO, Nd2 O3 , Y2 O3 , CeO2 , precipitation oxidation
Mn2 O3
66 ternary combinations of oxides of Mo, Solution-based sol-gel Ethane oxidative Scanning quadruple mass 172
V and Nb deposition dehydrogenation spectrometry, photothermal
deection, gas chroma-
tography
32 Combinatorial Methods in Catalysis
144 ternary combinations of oxides of Solution-based sol-gel Ethane oxidative Scanning quadruple mass 173
V/Al/Nb, and Cr/Al/Nb deposition dehydrogenation spectrometry, photothermal
deection
66 ternary combinations of 1% Pt, Pd, Solution-based impregnation Cyclohexane Quadruple mass spectrometry 174
and In on Al2 O3 dehydrogenation to
benzene
45 3- to 5-element combinations of Pt, Solution-based impregnation C3 H8 total oxidation Quadruple mass spectrometry 175
Pd, Rh, Ru, Au, Cu, Ag, and Mn on
TiO2 and Fe 2 O3
56 quaternary combinations of Pt, Pd, In, Solution-based impregnation NO reduction by C3 H6 Quadruple mass spectrometry 176
Na on Al2 O3
V2 O5/TiO2 mixtures Solution-based Oxidation on Laser-induced uorescence 177
coprecipitation naphthalene to imaging (LIFI)
naphthoquinone
36 Pt/Zr/V/Al2 O3 Solution-based impregnation Methane oxidation Scanning quadruple mass 178
Co oxidation spectrometry coupled with
monolith multichannel
reactor
52 Pt/Zr/V/Al2 O3 Solution-based impregnation Oxidative dehydro- Scanning quadruple mass
onto Al2 O3 electro- genation of isobutane spectrometry coupled with 35
chemically formed on parallel microreactors
aluminum plate
715 combinations of Pt, Ru, Os, Ir, Rh Solution-based deposition O2 reduction H2 O Fluorescence acidbase indicator 179
oxidation
280 V2 O5 , MoO3 , MnO2 , Fe 2 O3 , Ga2 O3 , Solution-based impregnation Propane dehydrogenation Quadruple mass spectrometry 180
La2 O3 , B2 O3 , MgO on Al2 O3
32 Ti-silsesquioxanes from 8 RSiCl3 /Ti Solution-based sol-gel method Epoxidation of 1-octene Standard GC equipped with 181
(OiPr)4 ternary composition of with tert-butyl- sampling robot
cyclopentyl, cyclohexyl, and phenyl- hydroperoxide (TBHP)
substituted silanes
65 1 mol% main group, transition, and Solution-based sol-gel method Photooxidation of 4- Standard HPLC equipped with 182
rare earth metal dopands on TiO2 , chlorophenol sampling robot
SnO2 , or WO3
a These publications describe only the synthesis of heterogeneous
catalyst libraries.
32.8 Combinatorial Heterogeneous Catalysis
943
944 32 Combinatorial Methods in Catalysis
before they become fully active, and most catalysts deactivate over time. These
properties will also vary according to the reaction conditions. Some parallel screen-
ing schemes oer the capability of monitoring the reactions over time. For example,
a multichannel xed bed reactor coupled with parallel GC and an array micro-
reactor equipped with REMPI allow library screening over a prolonged reaction
time. Many techniques such as MS, IR, and other optical detection schemes can be
applied in both homogeneous and heterogeneous catalysis. Each of these emerg-
ing techniques and tools for high-throughput catalyst screening will be described
in detail in the screening section (see Chapter 32.10). Finally, the heterogeneous
catalysis section will be concluded by case studies (see Section 32.8.2).
32.8.2
Case Studies
sons catalyst formulations on a microgram scale and plotted their primary screen-
ing results versus Thorsteinsons activity diagram [183]. Figure 32.19 shows that
the primary screening data with conversions in the ppm range correlate well with
bulk catalyst activity with percent level conversions.
For greater compositional precision, researchers at Symyx Technologies inves-
tigated additional libraries focusing on the most active area. This ne-tuning of
the catalyst resulted in an ethane conversion of 12.8% and an ethylene selectivity
of 74.4% at an optimal composition of Mo73 V24 Nb3 . Including Sb, Ca, and Li as
dopants improved the catalysts even further. Under identical synthesis and reac-
tion conditions, the optimal composition reported by Thorsteinson, Mo73 V18 Nb9 ,
converts 6.1% at a selectivity of 83.2%.
Considering new mixed metal oxide ternary libraries, for example Cr/Al/Nb/O
systems, researchers at Symyx Technologies were able to identify catalysts that are
signicantly more active than the previously known MoaVaNbaO systems [173,
185, 186].
Both manual and robotic synthesis were used for the library preparations, and
these catalytic materials were tested in parallel. For the best materials, propene
yields of 7% (rst generation) to 9% (fth generation) were achieved.
32.9
Combinatorial Electrocatalysis
32.9.1
Electrocatalysts for Fuel Cells
Fuel cell technology holds the promise of low to zero emission energy for power
plants, backup generators, and in the transportation sector. Fuel cells will also oer
948 32 Combinatorial Methods in Catalysis
fuel eciency in the order of 50% or more. With the increasing interest in fuel
cells as alternative energy sources for stationary power, portable, and automotive
applications, there is a need to develop new anode and cathode electrocatalysts. The
cost of the current precious metal-based electrocatalysts is a considerable hurdle to
the successful commercialization of fuel cells.
The oxidation of methanol in fuel cells has been extensively studied for several
decades. At present, however, the commercialization of the direct methanol fuel
cell (DMFC) is beset by several problems. Given the lower electrochemical activity
of methanol in comparison with hydrogen, much higher loading levels of noble
metal catalysts must be employed. The most ecient anode systems for the com-
plete electro-oxidation of methanol consist of the platinum metals and mixtures
or alloys thereof. Notwithstanding the high cost issues, presently employed anode
systems suer from complications due to catalyst poisoning by CO, which leads to
both an unacceptable loss in cell voltage and degradation in long-term performance.
Numerous studies have indicated that the most promising anode systems consist
of either a supported or unsupported binary alloy of Pt/Ru [189, 190].
The traditional approach to evaluating anode materials has involved the relatively
time-consuming and laborious practice of synthesizing and characterizing mate-
rials one at a time. Mallouk, Smotkin, and coworkers developed a rapid combina-
torial synthesis and uorescent screening method for the exploration of ternary
and quaternary anode electrocatalysts for the DMFC (Fig. 32.20) [157]. The authors
created libraries of alloys by using automated inkjet deposition of metal precursor
aliquots onto a conducting ber paper, followed by reduction of the metal salts
with a borohydride reagent. The authors created 135 unique ternary compositions
Fig. 32.20. Photo images of the dierent overpotential, identifying the most active
stages of discovery of a ternary alloy in array region of composition space. c) Fluorescence
format using a quinine indicator. a) Image in image at high overpotential, where methanol
white light. b) Fluorescence image at low oxidation occurs at every spot in the array.
32.9 Combinatorial Electrocatalysis 949
32.9.2
Combinatorial Electrosynthesis
Ward and coworkers applied two dierent combinatorial approaches to study the
electrochemical reduction of 1,4-benzoquinone to hydroquinone at organosulfur-
modied gold electrodes [193]. The authors prepared an array of physically but not
electrically isolated gold electrodes on a glass substrate. Monolayers of organo-
sulfur reagents were prepared on selected electrodes on the 4 7 array by dispens-
ing aliquots of either hexanethiol or hexadecanethiol dissolved in ethanol at the
respective electrodes. The array was subsequently rinsed and housed in a thin-layer
electrochemical cell. The electrolyte contained 1,4-benzoquinone dissolved in 0.5 M
KOH acidied to pH 3.0 and also a uorescent dye, uorescein, which uoresces
green at pH > 6. A potential of 0.1 V (vs. Ag/AgCl) was applied to the array and
the thin-layer cell illuminated with an ultraviolet lamp. Green uorescence was ob-
served in those electrode regions that had not been modied with an organosulfur
reagent as well as on electrode regions that had been modied with a monolayer
of hexanethiol. Benzoquinone is reduced by two electrons at this potential with the
consumption of two protons. It is the decrease in protons at the electrodes where
benzoquinone is reduced that causes the local pH to increase. No uorescence was
observed at the hexadecanethiol-modied electrodes.
In a second approach, the authors created an electrochemically addressable elec-
trode array from which it was possible to measure directly the electrochemical
activity by monitoring the currentvoltage behavior at each individual electrode.
An 8 8 gold electrode array was created on an insulating thermally oxidized sili-
con substrate ensuring physical and electrical isolation of the individual electrodes
using standard lithographic techniques. Electrical contact was made via standard
64-pin connectors with a single channel potentiostat and multiplexer. Selected elec-
trodes were subsequently modied with hexanethiol, hexadecanethiol, or dodeca-
nethiol monolayers through a series of adsorption and electrochemical desorption
steps. An electrochemical cell was created by attaching the electrode array to a Del-
rin cylinder with a Delrin cap congured for a reference and counter electrode.
Electrolyte containing benzoquinone and KOH acidied to pH 3 was added to the
cell, exposing all 64 electrodes to the electrolyte. In a serial manner, the electro-
chemical reduction of benzoquinone was studied at each electrode by cyclic voltam-
metry in which the potential of each electrode was cycled and the current measured.
The authors proposed that this direct measurement of the electrochemical activity
was more sensitive than the uorescent screen in quantifying the benzoquinone re-
duction at the modied electrodes with the results indicating that the activity in the
library increases in the order of hexadecanethiol < dodecanethiol < hexanethiol <
gold, illustrating suppression of current with increasing alkanethiol chain length.
The authors propose that the uorescent screen allows for measurement of active
zones while electrochemical screens can be used to discriminate smaller dier-
ences in activity.
Yudin and coworkers demonstrated how combinatorial electrochemistry could
be used in the electrosynthesis of small organic molecules [194a]. The authors de-
952 32 Combinatorial Methods in Catalysis
veloped what they term a spatially addressable electrolysis platform (SAEP). Each
electrochemical cell in the 4 4 array was equipped with a stainless-steel cathode
and a graphite rod anode. The cathodes were welded directly onto a stainless-steel
support that provided a common terminal for the current source. The authors ex-
plored the anodic oxidation of carbamates, amides, and sulfonamides leading to
libraries of a-alkoxycarbamates, a-alkoxyamides, and a-alkoxysulfonamides, respec-
tively. The authors have also demonstrated the intramolecular cyclization of hy-
droxyamides yielding heterobicyclic compounds as well as the generation of vicinal
diamines by the reductive hydrocoupling of aldimines.
In another application, Yudin and coworkers generated libraries of catalytic ma-
terials on electrode surfaces by the copolymerization of bithiophene and pyrrole-
containing TEMPO (2,2,6,6-tetramethilpiperidin-1-yloxy) catalysts [194b,c]. Diver-
sity was created by electrochemical copolymerization and by creating surfaces with
dierent ratios of bithiophene/pyrrole. These catalyst lms were utilized in the
electrochemical oxidation of primary alcohols to aldehydes, where cyclic voltamme-
try was used to screen the catalytic activities of the modied electrodes.
Another example of a high-throughput screen for catalyst activity comes from
Hillier and coworkers, who applied a scanning electrochemical microscope to
characterize the hydrogen oxidation reaction on a polycrystalline platinum surface
[195]. This technique utilizes tip-sample feedback and works reliably for determi-
nation of the kinetics of the reaction over a large range of substrate potentials from
the hydrogen adsorption region to the platinum oxidation region.
Hillier explored this technique further and directly demonstrated the measure-
ment of the rate constant for hydrogen oxidation and performed reactivity map-
ping of heterogeneous electrodes consisting of catalytic and noncatalytic domains.
32.10
Novel High-throughput Screening Tools
The process of discovering new catalysts or materials from a large pool of potential
candidates requires a reliable and robust screening process. Traditionally, this pro-
cess presents itself as a bottleneck in the discovery eort, especially if large libraries
of potential catalysts or materials have already been synthesized or prepared. The
rst step in combinatorial catalysis, like all materials design, involves the identi-
cation of a specic chemical transformation of interest. A collection of potential cat-
alysts, prepared from a set of chemically diverse ligand sets and metal precursors,
are combined in a parallel or combinatorial fashion and screened for activity in a
high-throughput primary screen. Although the large volume of potential catalysts
examined in a primary screen typically allows only relative activity to be established,
the goal at this stage of discovery is to identify promising lead catalysts worthy of
further investigation and follow-up. Lead materials are then further examined in
a high-throughput secondary screen designed to screen for specic trends in the
physical and chemical properties in greater detail and typically at a slightly lower
throughput than the primary screen. The information obtained from the second-
32.10 Novel High-throughput Screening Tools 953
ary screen is then used to prepare additional generations of catalysts that can be
optimized into a superior catalyst worthy of commercialization.
Typically, the information necessary to classify new materials cannot be obtained
from a single piece of characterization equipment. Therefore, a series of high-
throughput screening tools is employed at various stages in the combinatorial
process. While a large number of automated commercial systems exist for high-
throughput analysis of microliter quantity samples for medical and pharmaceutical
applications, the vast majority of the screening tools necessary for advanced mate-
rials research are custom-made instruments. Conventional analytical tools such as
mass spectrometry, gas [196] and liquid chromatography, electrophoresis [197, 198],
Raman [199] and nuclear magnetic resonance spectroscopy, X-ray uorescence
microprobe [182], and X-ray diraction [200] have been automated and redesigned
for rapid serial measurements of hundreds of samples per day [201]. A number of
recent reviews of combinatorial catalysis discuss high-throughput screening tech-
niques [20c,d,h, 151, 202204]. Several examples of novel high-throughput screen-
ing systems are discussed below.
32.10.1
Infrared Screening Tools
Infrared detection has become an extremely popular technique for the analysis of
combinatorial libraries because it is generally noncontact, nondestructive, and ame-
nable to very high-throughput screening. Use of the infrared can be divided into
two general categories: thermal imaging (thermography) and infrared spectros-
copy.
Fig. 32.24. Optical layout for an imaging FTIR a bandpass lter (F), a KBr diuser (D), CaF2
spectrometer. The light source is composed of plano-convex lens (L1), reactor (R), and CaF2
an infrared source (S), a KBr beam splitter biconvex lens (L2). The infrared camera with
(BS), a moving mirror (M1), and a stationary Hg/Cd/Te detector array (FPA) acts as the
mirror (M2). The optical set-up is composed of detector.
32.10.2
Optical High-throughput Screening Techniques
The eort to create parallel assays for solution-phase catalysis has led to the devel-
opment of a number of optical techniques. Many techniques include probe mole-
958 32 Combinatorial Methods in Catalysis
cules that change color or uoresce with catalytic behavior. Additional techniques
incorporate ultraviolet or infrared lasers to ionize or heat catalytic products, fol-
lowed by detection with electrodes or probe lasers. Finally, circular dichroism has
been developed to study the eects of chirality on catalysis.
Scheme 32.36. Reactive dyes as a method for rapid screening of homogeneous catalysts.
a large set of parallel chemical reactions in which two molecules are bound by co-
valent interactions. In their study, one substrate (A) is attached to a dye molecule
and the other (B) to a solid support. After a successful coupling reaction, the solid
supports and the dye molecule would be bound together by a covalent interaction
between the two substrates (Scheme 32.37). Analysis under UV illumination fol-
lowing ltering of the reagents allowed the substrate combinations that are capable
of covalent coupling to be identied. In their investigation, an acrylate containing a
tethered coumarin was reacted with an aryl halide supported on a crosslinked poly-
styrene bead. Comparison of the results from the uorescence assay with results
from a standard GC analysis showed that the uorescence assay accurately repre-
sented the trends for the Heck coupling of aryl bromides and chlorides. Two ligands
identied in the assay, tri-(tert-butyl)phosphine and di-(tert-butylphosphino)ferro-
cene, were shown to be the most active systems for the olenation of unactivated
aryl bromides, and di-(tert-butylphosphino)ferrocene the most ecient for olena-
tion of unactivated aryl chlorides.
HPLC that was more than 10% dierent from the FRET result. It was suggested
that the FRET pair developed for this study should have application to many dif-
ferent reactions, such as aryl halide amination, aryl halide etherication, carbonyl
a-arylation, Suzuki coupling, and Hiyama coupling with silanes.
Copeland and Miller developed a similar method for the study of acetic acid
evolution that employs aminomethylanthracenes as pH-sensitive uorophors. Neu-
tral aminomethylanthracenes undergo photoinduced electron transfer (PET) and
uoresce when protonated (Fig. 32.25a) [220]. Attachment of the aminomethylan-
thracenes to partially derivatized resin beads, followed by attachment of peptide
catalysts, created a collection of resin beads related to the structure shown in Fig.
32.25b [221]. Beads functionalized with the most active catalysts appeared bright-
est as a given acylation reaction proceeded. Furthermore, the beads maintained
their relative intensities when examined in separate vessels or pooled together as
catalyst mixtures.
Copeland and Miller have applied the same aminomethylanthracenes as pH
sensors in solution-phase catalyst libraries as well [220]. In this investigation, seven
unique catalysts (Fig. 32.26) were deposited into a standard 96-well plate at three
32.10 Novel High-throughput Screening Tools 961
Fig. 32.27. Fluorescent gel system for the detection of bead-supported catalysts.
loading of approximately 1 mg cm2 [179]. The catalyst array was analyzed using
a three-electrode gas diusion cell, with the Toray carbon substrate linking the
catalyst elements as a working electrode, Pt gauze as the counter electrode, and a
reversible hydrogen electrode as a reference. Electrochemical half-cell reactions
either generate or consume ions, creating a change in the pH in the location of
active catalysts. Utilizing an indicator that is uorescent in the presence of an acid
or conjugate base allows the determination of which elements within the library
were most active for a particular anode or cathode reaction. Mallouk and coworkers
have used quinine and Phloxine B as uorescent indicators for neutral pH, and
Ni 2 complexed with 3-pyridine-2-yl-(4,5,6)triazolo-(1,5-a)pyridine (Ni-PTP) (Fig.
32.28) for low pH [157]. Using the uorescence method, Mallouk and coworkers
screened ternary libraries of metal alloys and identied novel electrocatalysts for
methanol and bifunctional oxygen reduction/water oxidation regenerative fuel cells
[157, 179].
Additional optical screening techniques have been developed for a number of
dierent chemical processes. The methods include colorimetric assays where col-
orless 1-naphthol undergoes an electrophilic aromatic substitution with a diazo-
nium salt to give a bright orange azo product [223], indigo [61], and Prussian blue
staining of reaction products [224].
Fig. 32.30. Schematic of a photothermal deection cell for the detection of ethylene.
and screening) took 2.5 days, including additional checks for reproducibility of the
results.
A similar method utilizing laser-induced uorescence has been developed by Su
and Yueng [177]; however, the technique is only applicable to uorescent species,
which limits its use.
ylene is better than 0.1 ppm and has a very high discrimination factor (@ 10 6 )
against ethane and other species present in the reaction.
32.10.3
High-throughput Screening Using Mass Spectrometry
the products from the catalytic process are snied from the reaction chamber
and analyzed by a mass spectrometer. In addition to the two chambers, ancillary
equipment is used to control the positioning of the library and to maintain vari-
ables such as the ow rate of reactant gases, temperature, and reaction pressures.
In a typical experiment, a catalyst library is placed onto a platform capable of
translation in three orthogonal directions. Gas ows to and from the library through
a cylindrical delivery tube containing a second concentric internal tube. A reactant
gas ows down the annular region of the tube onto the quartz substrate. Product
gas ows from an element in the library to the analysis chamber through the inner
tube via a capillary connection between the two chambers. Product gas exits the
capillary in the ionization zone of the quadruple mass spectrometer. The tempera-
ture of the individual library elements is controlled using a CO2 laser, an infrared
sensor, and a feedback control loop.
Analysis of the data collected from the entire library (approximately one library
element per minute) allows the relevant parameters associated with the reaction,
such as catalyst activity and selectivity, to be determined for each element in the
library.
Cong and coworkers reported a systematic and integrated approach for the syn-
thesis and screening of libraries of mixed metal alloys containing rhodium, palla-
dium, platinum, and copper [168]. Fully automated gas-phase thin-lm deposition
techniques (RF sputtering) were used to synthesize three unique 120-member li-
braries of Rh/Pt/Cu, Rh/Pd/Cu, and Rh/Pt/Pd alloys. The metal alloy combinato-
968 32 Combinatorial Methods in Catalysis
rial libraries were screened in the scanning mass spectrometer as described pre-
viously for the gas-phase oxidation of carbon monoxide and the reduction of nitric
oxide. In accordance with the literature and despite some discontinuities, rhodium-
rich regions in the Rh/Pt/Pd ternary generally showed enhanced activity for car-
bon monoxide oxidation with oxygen and, interestingly, Rh/Cu binaries such as
Rh/Cu (50:50 atom%) showed promising oxidation activity.
A similar scanning mass spectrometer has been constructed by Maier and co-
workers [169]. This spectrometer incorporated a similar concentric capillary gas
feed and sensing design, but utilized simpler scanning robotics and an open-air
environment.
doenantiomers C and H (Scheme 32.39b). The ratios of mass intensities A/G and
C/H allow the conversion, enantioselectivity, and the selectivity factor to be ob-
tained [229].
32.10.4
Electronic High-throughput Methods
the potential for chemical interference. Advantages of this method include the
elimination of expensive thermal imaging cameras and infrared optics, and a po-
tential increase in the temperature sensitivity.
change in the resistance of the Pt/Rh wire. While there is no gas selectivity, i.e.
any combustible gas will increase the sensor temperature, the detectable gas con-
centration range is between 0.01% and 10% in air.
Semiconductor gas sensors were utilized to analyze catalytic euents for olens
or oxygenates. Flammable gases change the temperature, and therefore the free
carrier density, of an SnO2 semiconductor deposited on an alumina plate held at
600 C. To improve the selectivity, dierent semiconductor layers were deposited
on seven SnO2 sensors. They included 13 wt% SiO2/Al2 O3 , 28 wt% SiO2/Al2 O3 , 28
wt% SiO2/ZnO, 28 wt% SiO2/TiO2 , 28 wt% SiO2/ZrO2 , 3 wt% Mo/SiO2 , and 3 wt%
V/SiO2 . A preliminary investigation revealed that the SnO2 covered with 28 wt%
SiO2/Al2 O3 was most selective for olens, and SnO2 covered with 28 wt% SiO2/
TiO2 or 13 wt% SiO2/Al2 O3 the most selective for oxygenates. A NDIR sensor,
which measures infrared absorption at specic wavelengths, and two SnO2 semi-
conductor sensors sensitized by TiO2 or by 13 wt% SiO2/Al2 O3 were used to eval-
uate catalysts for selective propane oxidation.
The sensors were in a multichannel parallel sensor array for the CO oxidation
and oxidative dehydrogenation studies, and in a serial conguration for the selec-
tive oxidation of propane (Fig. 32.34). Although the accuracy of the gas analysis by
the sensor system was lower than a conventional gas chromatograph, the gas sen-
sors have the advantage of rapid response, continuous measurements, small size,
and lower cost. Furthermore, Yamada and coworkers are investigating new two-
dimensional miniaturized reactors where the solid-state sensor is integrated into
the library.
Fig. 32.34. Schematic showing the gas sensor system for the
evaluation of selective propane oxidation.
32.10 Novel High-throughput Screening Tools 973
32.10.5
Array Reactors
The activity and selectivity of catalysts can change signicantly during extended
exposure to product feedstock, creating the need for a system of high-throughput
secondary screens capable of addressing reaction kinetics. Kinetic studies are indis-
pensable to the catalytic rate as a function of the process variables, such as tem-
perature, pressure, spacetime, and composition of the reaction mixture. Stability
studies that test the catalyst over an extended time period need to be conducted
before a scale-up is initiated. Typically, these extended kinetic studies are performed
in pilot plants with real feed and recycle streams and with the catalyst in its prac-
tical shape. The increased number of potential catalysts generated in a combinato-
rial program has led to the development of a new generation of array reactors that
are capable of testing a larger number of catalysts in parallel for extended periods
of time. Array reactors share common design features, diering primarily in their
construction materials, and in their ow geometries.
ment. Capillaries placed in the euent stream present the major ow resistance in
the system, which means that the largest pressure drop is not over the catalyst bed
but through the capillaries, meaning that dierences in the catalyst packing do
not signicantly inuence the ow through the individual wells. A nondispersive
infrared sensor was used to determine CO and CO2 concentrations, and the au-
thors report a factor of 15 increase in throughput relative to traditional methods.
Senkan and coworkers have applied an array microreactor to a number of cata-
lytic systems including a 66-element Pd/Pt/In library for the catalytic dehydroge-
nation of cyclohexane to benzene. The library consisted of 66 ternary combinations
of Pt, Pd, and In prepared in 0.1 wt% increments, for a total metal loading of
1 wt% that was impregnated on 4-mm-diameter by 1-mm-long alumina supports
[161, 176]. The catalyst library was prepared from solution-phase metal salt pre-
cursors using an automated liquid-handling robot. The compositions were slowly
evaporated, dried, and nally calcined at 500 C for 2 h. The catalyst pellets were
placed in a reactor array consisting of 20 rectangular channels that were micro-
machined on a at nonporous silica slab. The 1-mm-wide, 1-mm-deep, and 20-mm-
long channels had a 4 2 mm cylindrical well to hold the catalyst pellets. A simi-
lar silica slab was micromachined to t onto the top, forming a silica block with
cylindrical channels leading to and from the catalyst sample. Four microarray re-
actor blocks, each containing 20 samples, were stacked and placed into an alumi-
num heating block, making it possible to test up to 80 samples in parallel.
32.10 Novel High-throughput Screening Tools 975
The entire reactor system was mounted on a stand that could be moved in three
dimensions by a computer. The test samples were heated in Ar gas up to 350 C,
then reduced under hydrogen gas, cooled to the desired reaction temperature, and
nally exposed to a feed stream of 10% cyclohexane in Ar. The contact time be-
tween the sample and the feed gas was approximately 4 ms. The level of reactants,
products, and inert carrier gas were determined by withdrawing a small sample
from each microreactor channel using a capillary sampling probe (50 mm diam-
eter) inserted 2 mm into the channels. The gas was analyzed by a quadruple mass
spectrometer. The capillary was inserted into each channel for approximately 5 s via
a computer-controlled positioning system. In this manner, the entire 80-element
array could be screened in approximately 10 min.
Building on an earlier microreactor design that used 15 quartz reactors [234],
Claus and coworkers designed a ceramic monolith reactor with 2.2-mm-square
channels, 150-mm long, that were arranged in a 16 16 array [178]. Each channel
in the monolith represents a single xed bed reactor, which allowed up to 256 cata-
lysts to be tested in parallel up to 600 C. Product gases were analyzed using a cap-
illary mounted on an x/ y/z scanning stage, and attached to a quadruple mass spec-
trometer (Fig. 32.36). A series of control experiments demonstrated that the axial
and radial temperature proles between channels (as measured without catalysts)
was not in excess of 5 K, and the ow rates did not dier more than 10%. The
performance of the monolithic reactor was evaluated using a 36-element library of
Pt/Zr/V/Al2 O3 catalysts for methanol oxidation. The results showed that the reactor
was valuable as a primary screen, distinguishing between poor and good catalysts
at a rate of about 1 min per sample. The accuracy was lowered owing to diculties
in generating equal ows in all channels of the monolith, and, in certain cases,
owing to diusion of exhaust gases from adjacent catalysts into the end of a chan-
nel containing the catalyst currently under investigation.
Fig. 32.36. Monolithic reactor system for the parallel screening of heterogeneous catalysts.
976 32 Combinatorial Methods in Catalysis
trometer. The detected amount of catalytic product in the noncatalytic zone was
less than 2% of that found in the active position. A 52-catalyst methanol oxidation
library used for further verication of the microreactor demonstrated excellent
sensitivity and reproducibility.
32.10.6
Capillary Array Electrophoresis
Recently, capillary electrophoresis (CE) has taken a key role in the parallel screen-
ing of homogeneous catalysts, in addition to the classical methods such as GC
or HPLC. Yeung and coworkers have used multiplexed capillary electrophoresis for
the combinatorial screening of enzyme activity [197a] and homogeneous catalysis
[197b]. In the latter case, this methodology was successfully implemented to opti-
978 32 Combinatorial Methods in Catalysis
32.11
Summary and Outlook
ization, and automation so that large numbers of samples can be synthesized and
screened eciently. Rapid serial and parallelized adaptations of conventional ana-
lytical techniques will become increasingly important in the assay of materials
properties, as will the development and implementation of new and unconven-
tional high-throughput screening tools. Software development and engineering
support in the construction and design of synthesis and screening tools are as
crucial as further advances in chemistry, even when appropriate tools or robots for
synthesis and screening automation are commercially available. Finally, the com-
binatorial methodology generates data much faster than the conventional research
employing empirical and rational approaches to materials discovery, and, inevi-
tably, the proper data handling and storage should accompany the high-throughput
synthesis and screening to maintain the integrity of research and development ef-
forts. Full realization of the combinatorial methodology will require the integration
of chemistry, physics, engineering, and informatics to greatly enhance the proba-
bility of nding a global reaction optimum of a catalyst for a targeted reaction or
materials with desired properties. In the future, the scientists intuition may be
shifted, at least to some degree, toward optimally programming and setting up ap-
propriate experiments and screens, as well as for the analysis of the obtained data
and materials. All these eorts will require an enormous initial investment in all
of these areas [235].
Acknowledgements
The authors are deeply indebted to Ms Silvia Lee (Symyx Technologies) and Ms
Kathryn Boykin (XenoPort) for their invaluable support in reference and patent
searches. The authors thank Mr Ron Krasnow (Symyx Technologies) for checking
the manuscript.
References
News 1998, 76, 4767; c) R. Baum, S. Rees, Tetrahedron 1997, 53, 5643
Borman, A. M. Thayer, J. H. 5678; c) P. H. H. Hermkens, H. C. J.
Krieger, Chem. Eng. News 1996, 74, Ottenheijm, D. Rees, Tetrahedron
2973. 1996, 52, 45274554; d) N. K.
8 a) J. McNeal, Todays Chemist at Work Terrett, M. Gardner, D. W.
1998, 7, 3438; b) J. K. Borchardt, Gordon, R. J. Kobylecki, J. Steele,
Todays Chemist at Work 1998, 7, 35 Tetrahedron 1995, 51, 81358173.
41. 16 For classical reviews about solid-
9 I. Sucholeiki (ed.), High-throughput phase organic synthesis (SPOS), see
synthesis. Marcel Dekker, Inc., New for example: a) J. S. Fruchtel, G.
York, 2001. Jung, Angew. Chem. 1996, 108, 1946;
10 For landmark review articles about the Angew. Chem. Int. Ed. Engl. 1996, 35,
application of combinatorial technol- 1742; b) F. Balkenhohl, C. von
ogies to drug discovery, see: a) M. A. dem Busche-Hunnefeld, A. Lansky,
Gallop, R. W. Barrett, W. J. Dower, C. Zechel, Angew. Chem. 1996, 108,
S. P. A. Fodor, E. M. Gordon, J. Med. 24362488; Angew. Chem. Int. Ed.
Chem. 1994, 37, 12331251; b) E. M. Engl. 1996, 35, 22882337; c) L. A.
Gordon, R. W. Barrett, W. J. Dower, Thompson, J. A. Ellman, Chem. Rev.
S. P. A. Fodor, M. A. Gallop, J. Med. 1996, 96, 555600.
Chem. 1994, 37, 13851401. 17 For accounts and special techniques
11 For chemoinformatics, see: M. Hann, used in combinatorial chemistry, see
R. Green, Curr. Opin. Chem. Biol. for example: a) K. S. Lam, M. Lebl, V.
1999, 3, 379383, and references cited Krchzak, Chem. Rev. 1997, 97, 411
therein. 448; b) A. Nefzi, J. M. Ostreh, R. A.
12 X.-D. Xiang, X.-D. Sun, G. Briceno, Houghten, Chem. Rev. 1997, 97, 449
Y. Lou, K. A. Wang, H. Chang, W. 472; c) M. C. Pirrung, Chem. Rev.
G. Wallace-Freedman, S.-W. Chen, 1997, 97, 473488; d) D. J. Gravert,
P. G. Schultz, Science 1995, 268, K. D. Janda, Chem. Rev. 1997, 97,
17381740. 489509; e) A. W. Czarnik, J. A.
13 a) J. J. Hanak, J. Mater. Sci. 1970, 5, Ellman, Acc. Chem. Res. 1996, 29,
964971; b) J. J. Hanak, Proc. 1er 111170.
Colloq. Int. Pulverisation Cathodiques 18 For a review on recent developments
Ses Appl. 1973, 165, 177197; c) J. J. and applications of polymer-supported
Hanak, J. Vac. Sci. Technol. 1971, 8, reagents and scavangers in synthetic
172175. organic chemistry, see: a) O.
14 F. Zaragoza Dorwald (ed.), Organic Brummer, B. Clapham, K. D. Janda,
Synthesis on Solid Phase, Wiley-VCH, Curr. Opin. Drug Discovery Dev. 2000,
Weinheim 2000. For periodical 3, 462473; b) S. V. Ley, I. R.
selection of signicant papers Baxendale, R. N. Bream, P. S.
published in the recent literature Jackson, A. G. Leach, D. A.
covering the broad area of solid-phase Longbottom, M. Nesi, J. S. Scott,
organic synthesis (SPOS) and R. I. Storer, S. J. Taylor, J. Chem.
advances in linker chemistry, see for Soc., Perkin Trans. 1 2000, 23, 3815
example: C. McCusker, T. Wildman, 4195. For further information about
F. McKerlie, A. Gunn, H. Mandani, polymer-supported reagents, see: c) S.
J. Dalaney, S. Anizon, J. Chem. Soc., W. Kaldor, M. G. Siegel, Curr. Opin.
Perkin Trans. 1 2001, 3, ixxii. Chem. Biol. 1997, 1, 101106; d) S. J.
15 For periodical reviews of recent Shuttleworth, S. M. Allin, P. K.
literature about solid-phase organic Sharma, Synthesis 1997, 12171239;
reactions, see for example: a) S. e) P. Hodge, Chem. Soc. Rev. 1997, 26,
Booth, P. H. H. Hermkens, H. C. J. 417424; f ) D. H. Drewry, D. M.
Ottenheijm, D. Rees, Tetrahedron Coe, S. Poon, Med. Res. Rev. 1999, 19,
1998, 54, 1538515443; b) P. H. H. 97157. For delivery and scavenging
Hermkens, H. C. J. Ottenheijm, D. agents for the synthesis of compounds
References 981
Chem. Int. Ed. 1998, 37, 609 Angew. Chem. 2001, 113, 762765;
611. Angew. Chem. Int. Ed. 2001, 40, 740
167 J. Klein, C. W. Lehmann, H. W. 743.
Schmidt, W. F. Maier, Angew. Chem. 182 C. Lettmann, H. Hinrichs, W. F.
1998, 110, 35573561; Angew. Chem. Maier, Angew. Chem. 2001, 113, 3258
Int. Ed. 1998, 37, 33693372. 3262; Angew. Chem. Int. Ed. 2001, 40,
168 P. Cong, R. D. Doolen, Q. Fan, 31603164.
D. M. Giaquinta, S. Guan, E. W. 183 E. M. Thorsteinson, T. P. Wilson,
McFarland, D. M. Poojary, K. Self, F. G. Young, P. H. Kasai, J. Catal.
H. W. Turner, W. H. Weinberg, 1978, 52, 116.
Angew. Chem. 1999, 111, 507512; 184 F. G. Young, E. M. Thorsteinson,
Angew. Chem. Int. Ed. 1999, 38, 484 US Patent No. 4250346, 1981.
488. 185 Y. Liu, P. Cong, R. Doolen, S. Guan,
169 M. Orschel, J. Klein, H. W. D. M. Poojary, H. W. Turner, W. H.
Schmidt, W. F. Maier, Angew. Chem. Weinberg, Innovation in Selective
1999, 111, 29612965; Angew. Chem. Oxidation by Solid Catalysts, Rimini,
Int. Ed. 1999, 38, 27912803. Italy, 1999, p. 117, Book of Abstracts.
170 O. Buyevskaya, D. Wolf, M. Baerns, 186 Y. Liu, P. Cong, R. Doolen, S. Guan,
Catal. Today 2000, 62, 9199. D. M. Poojary, H. W. Turner, W. H.
171 U. Rodemerck, P. Ignaszewski, M. Weinberg, 4th European Congress on
Lucas, P. Claus, M. Baerns, Chem. Catalysis, Rimini, Italy, 1999, p. 41:
Ing. Tech. 1999, 71, 872. Book of Abstracts.
172 P. Cong, A. Dehestani, R. Doolen, 187 S. Senkan, Nature 1998, 394, 350
D. M. Giaquinta, S. Guan, V. 352.
Markov, D. Poojary, K. Self, H. W. 188 D. Wolf, O. V. Buyevskaya, M.
Turner, W. H. Weinberg, Proc. Natl. Baerns, Appl. Catal. A 2000, 200,
Acad. Sci. USA 1999, 96, 11077. 63.
173 Y. Liu, P. Cong, R. D. Doolen, H. W. 189 K. L. Ley, R. Liu, C. Pu, Q. Fan, N.
Turner, W. H. Weinberg, Catal. Leyarovska, C. Segre, E. S. Smotkin,
Today 2000, 61, 8792. J. Electrochem. Soc. 1997, 144, 1543
174 S. Senkan, K. Krantz, S. O zturk, V. 1548.
Zengin, I. Onal, Angew. Chem. 1999, 190 T. E. Mallouk, E. Reddington, C.
111, 29652971; Angew. Chem. Int. Ed. Pu, K. L. Ley, E. S. Smotkin,
1999, 38, 27942799. Proceedings of the Extended Abstracts
175 U. Rodemerck, D. Wolf, O. V. Fuel Cell Seminar, Orlando, Florida,
Buyevskaya, P. Claus, S. Senkan, M. 1996, 686689.
Baerns, Chem. Eng. J. 2001, 82, 3. 191 a) B. Gurau, R. Viswanathan, R.
176 K. Krantz, S. O zturk, S. Senkan, Liu, T. J. Lafrenz, K. L. Ley, E. S.
Catal. Today 2000, 62, 281289. Smotkin, E. Reddington, A.
177 a) H. Su, E. S. Yeung, J. Am. Chem. Sapienza, B. C. Chan, T. E.
Soc. 2000, 122, 74227423; b) H. Su, Mallouk, S. Sarangapani, J. Phys.
Y. Hou, R. S. Houk, G. L. Schrader, Chem. B 1998, 102, 999710003; b) L.
E. S. Yeung, Anal. Chem. 2001, 73, Liu, R. Viswanathan, R. Liu, E. S.
44344440. Smotkin, Electrochem. Solid State Lett.
178 P. Claus, D. Honicke, T. Zech, 1998, 1, 123125; c) E. Reddington,
Catal. Today 2001, 67, 319339. J.-S. Yu, B. C. Chan, A. Sapienza, G.
179 G. Chen, D. A. Delafuente, S. Chen, T. E. Mallouk, B. Gurau, R.
Sarangapani, T. E. Mallouk, Catal. Viswanathan, R. Liu, E. S. Smotkin,
Today 2001, 67, 341355. S. Sarangapani, Comb. Chem. 2000,
180 O. V. Buyevskaya, A. Bruckner, E. V. 401420; d) Y. Sun, H. Buck, T. E.
Kondratenko, D. Wolf, M. Baerns, Mallouk, Anal. Chem. 2001, 73,
Catal. Today 2001, 67, 369378. 15991604.
181 P. P. Pescarmona, J. C. van der 192 a) P. Strasser, S. Gorer, M.
Waal, I. E. Maxwell, T. Maschmeyer, Devenney, Proceedings of the
References 989
33
Diversity-Based Identication of Ecient
Homochiral Organometallic Catalysts for
Enantioselective Synthesis
Amir H. Hoveyda
33.1
Introduction
throughput strategies [3]. This movement arises from the realization that if the
study of a large number of potential catalysts leads to a discovery, then why not
perform the investigation systematically and more eciently by covering a broad
range of catalyst candidates? Another factor that supports the adoption of diversity-
based approaches is that screening of a large number of compounds may allow
chemists to identify an optimal catalyst for each particular substrate and thus
overcome the problem of generality that often looms large when a small number
of catalyst candidates are available. High-throughput screening can give rise to
new mechanistic scenarios that are by nature more comprehensive, derived from a
more extensive collection of data points. It would be idle to assume that screening
is carried out at random or that diversity approaches to catalyst discovery are devoid
of any mechanistic basis. Mechanistic information gained through examination
of a large set of data can serve as the driving force behind additional investigations;
it can lead to the blossoming of a symbiotic relationship between mechanistic in-
quiry and diversity-based screening that results in the discovery of more powerful
and ecient catalysts of all types, and a deeper and more mature appreciation of the
inner workings of various classes of catalysts.
33.2
Factors Critical to the Success of Diversity-based Reaction Development
are joined by bonds that are readily formed. These ligands are therefore most
suitable for parallel library synthesis. Schi bases and peptide linkages are
among disconnections that fall into the latter category. If CaC, CaP, or CaN
(nonpeptide linkage) bonds are to be modied to alter a ligand (e.g. binol- or
biphen-based systems), then these constructs should be viewed as less readily
modular.
Symmetry. One of the more established dogma in the eld of metal-catalyzed
asymmetric catalysis is that symmetric, particularly C2 -symmetric, chiral li-
gands are preferred so that the number of energetically dierentiable modes
of catalystsubstrate association can be minimized and stereoselectivity can
be more logically designed. Many useful, ecient, and selective transform-
ations have indeed been developed based on C2 -symmetric chiral ligands [1];
it is likely that many more chiral ligands of this type will be developed in the
future. However, recent studies from a number of laboratories indicate that
non-C2 -symmetric chiral ligands can also give rise to outstanding levels of se-
lectivity and eciency in a variety of synthetically important transformations
[4].
As far as diversity-based approaches are concerned, the less symmetric class
of chiral ligands provides a more attractive option. With C2 -symmetric systems,
any structural alteration must be mirrored at the complementary region of the
ligand structure, thus reducing the degree of available diversity.
Multiple binding sites. Another common perception in the eld of asymmetric
catalysis is that the number of binding sites for the metal center within the
chiral ligand should be minimized (often limited to a single point of bind-
ing). As such, chiral ligands have often served a single function in a particular
transformation. The metalligand complex either serves as a Lewis acidic
activator or as a nucleophilic agent. In a limited number of instances, both
functions are delivered by two distinct molecules of the chiral catalyst
(second order in catalyst) [5]. Incorporation of multiple binding sites raises
the intriguing possibility of multimetal systems and multifunctional catalysts
[6]. Care must be taken, however, that dierent sites possess suciently di-
verse coordination properties so that, through association with various metal
salts, chiral ligands give rise to complementary rather than competing eects
[7]. Multiple binding sites are attractive from the point of view of high-
throughput screening studies: they incorporate additional elements of struc-
tural modication and open the possibility of multimetal systems, which, in
itself, represents a critical dimension in diversity.
3 High-throughput catalyst screening. This aspect is often the greatest bottleneck in
assaying each catalyst for asymmetric induction. Although in combinatorial ap-
proaches to biological activity, mixtures of compounds can be analyzed simulta-
neously, such a strategy is by nature problematic in studies that pertain to iden-
tication of eective homochiral catalysts. Because subtle structural variations
can lead to unexpected changes, or even reversal of enantioselectivity, examina-
tion of mixtures of catalysts can lead to conclusions that are misleading (low net
selectivity by two eective catalysts that aord high ee values but in the opposite
sense). Accordingly, recent reports generally involve testing individual systems.
994 33 Diversity-Based Identification of Efficient Homochiral Organometallic Catalysts
It merits mention that the parallel screening strategy has also been applied in
therapeutic discovery eorts because of the diculties involved in the accurate
deconvolution of various mixtures and because of the synergism that may exist
between several active compounds. As such, combinatorial chemistry does not
necessarily involve the generation of mixtures of compounds; it may be better
characterized by the modular nature of the constituent compounds that in dif-
ferent combinations provide large numbers of molecular ensembles. Analysis of
mixtures of candidates for activity (vs. selectivity) can be feasible however and
may be accomplished eectively [3f ].
33.3
Peptidic Schi Bases as Chiral Ligands
33.3.1
Ti-Catalyzed Enantioselective Addition of Cyanide to Meso Epoxides
improve the observed enantioselectivity, each of the three subunits in the modular
ligand was successively optimized (see Fig. 33.2), such that the rst amino acid 1
(AA1 in box) was varied and the other two subunits were kept constant (with
5 as the substrate). Tert-leucine was found to be optimal at AA1 position and this
structural element was retained in successive generations. The second position
(AA2) was then altered, and O-tert-butyl-threonine was identied as the best AA2.
Finally, from a pool of salicylic aldehydes, 3-uorosalicylaldehyde was selected as
the best Schi base (SB). In the end, only a representative sampling of 60 20 3
catalysts was necessary to identify one (see 10 in Table 33.1) that aords nearly a
95:5 ratio of enantiomers (89% ee). The initial catalyst provided the addition prod-
uct with only 26% ee (cyclohexene oxide as substrate). Successive modications of
the ligand structure led us to identify in three steps a chiral ligand that delivers a
synthetically attractive level of enantioselectivity. It is unlikely that any mechanistic
considerations would have pointed to this specic peptidic complex as a more
suitable one.
The above strategy for catalyst screening raises an intriguing question: Is the
optimal catalyst identied by this process truly the very best catalyst? And if it is
not, does the attendant improvement in enantioselectivity (> 99% ee), assuming
there will be no notable dierence in eciency, justify the additional eort that
would be required to achieve it? In the approach described above, we have made
certain assumptions about the additivity and absence of cooperativity between the
three subunits of the ligand structure. At least for this small sampling, these assump-
tions seem to hold true, but without testing every combination we cannot deni-
tively answer this question. Examination of every possibility would be taxing and
detract from the efciency of the general screening method. An important practical
advantage of the above approach is that, in a relatively short amount of time, it al-
lowed us to identify a selective catalyst for an entirely new asymmetric process.
That is, the search strategy is not an open-ended odyssey but a well-structured
Fig. 33.2. Screening by positional optimization of various
peptide Schi base ligands for enantioselectivity in the addition
of TMSCN to cyclohexane oxide (5).
33.3 Peptidic Schiff Bases as Chiral Ligands
997
998 33 Diversity-Based Identification of Efficient Homochiral Organometallic Catalysts
Tab. 33.1. Optimized ligands for catalytic enantioselective addition of TMSCN to meso epoxides.
1 83 72
2 89 65
3 84 68
4 78 69
in Scheme 33.4, a subtle alteration in the structure and not the stereochemical
identity of the peptide ligand leads to inversion of stereochemistry in the epoxide-
opening reaction (compare reaction with ligands 9 and 17). These observations vali-
date our choice of individually synthesizing and testing each catalyst, as mixtures
of catalysts can lead to racemic products.
33.3.2
Ti-Catalyzed Enantioselective Addition of Cyanide to Imines
In the second phase of our program, we applied the above screening technology
to identify specic Tipeptide complexes that catalyze the addition of cyanide to
imines [11, 12]. As the representative cases in Scheme 33.5 illustrate, the reactions
proceed eciently and with outstanding enantioselectivity. In addition to arylimines
(e.g. formation of 19 and 22 in Scheme 33.5), acyclic a,b-unsaturated imines are
eective substrates for these asymmetric CaC bond-forming reactions (cf. forma-
tion of 25 and 28 in Scheme 33.5). Initially, our catalyst-screening approach led
us to identify catalysts that deliver amino nitriles with 9097% ee, but in low
conversion (< 25% conversion in 18 h) [13]. Based on certain mechanistic consid-
erations, which ironically later proved to be incorrect, we argued that addition of
protic additives may lead to an enhancement in reactivity. The latter hypothesis
was based on the fact that when reactions are carried out in relatively large scale,
where adventitious water is more easily avoided, conversions are lower. Thus, the
eect of a variety of alcohols and amines on reaction eciency and selectivity were
systematically screened. These studies led us to establish that the Ti-catalyzed cya-
nide additions are signicantly more ecient if one equivalent of iPrOH is added
to the reaction mixture slowly (see below for mechanistic details). Accordingly, re-
actions of arylimines and unsaturated aliphatic imines were eected with high
enantioselectivity and in good yield, as represented by the examples in Scheme
33.5. It is important to note that in the course of these studies we established that
33.3 Peptidic Schiff Bases as Chiral Ligands 1001
an NnBu amide terminus provides the same levels of reactivity and eciency as
when Gly occupies the AA3 site.
The majority of the optically enriched amino nitrile intermediates can be easily
recrystallized to enantiopurity. The Ti-catalyzed asymmetric process is of notable
utility in organic synthesis, since after a single hydrolysis/deprotection step optically
pure amino acids can be obtained; the example shown in Scheme 33.6 is illustra-
tive. It is worth noting that such amino acids cannot be accessed by the celebrated
catalytic asymmetric hydrogenation protocols [14].
With eective and highly enantioselective catalytic Strecker reactions in hand,
we set out to explore the mechanistic details of these important transformations
[15]. It is our conviction that a better mechanistic appreciation, along with the
ability to prepare and screen large collections of catalyst candidates, will allow us
to extend the scope of this versatile class of chiral ligands to include a number of
other critical catalytic enantioselective processes. Thus, kinetic, structural, and ster-
1002 33 Diversity-Based Identification of Efficient Homochiral Organometallic Catalysts
Various structural features of the chiral peptide ligand were systematically altered
and the corresponding relative rates and enantioselectivities were measured. These
studies, summarized in Scheme 33.7, led to several important ndings: (1) not
only is the presence of the AA2 moiety critical to reactivity and enantioselectivity
(compare reaction of 33 with that of 34), its stereochemical identity is of notable
signicance as well (compare reaction of 33 with those of 35 and 36); (2) the pres-
ence of a more Lewis basic amide carbonyl (vs. a carboxylic ester) has an inuence
on the rate of asymmetric CN addition, e.g. the initial rate of reaction (90 min)
with ligand 7 is 2.3 times faster than that for the derived methyl ester 37.
The importance of AA2 and the inuence of local chirality suggest that the
peptide segment of the ligand actively participates in the asymmetric CaC bond-
forming reaction. That is, the TiSchi base coordinates with the substrate while
an amide moiety within the neighboring peptide segment associates and delivers
HCN to the activated imine. A mechanistic model consistent with the kinetic and
stereochemical data is presented in Fig. 33.3. These ndings thus underline the
signicance of the peptidic moiety of this class of chiral ligands not as passive
providers of a chiral environment, but as active participants in the asymmetric
CaC bond formation. These data provide a solid mechanistic basis regarding the
importance of available diversity at the two peptidic sites and the practical utility of
the non-C2 -symmetric structure of these chiral ligands. The above ndings are
signicant, since they demonstrate that, by virtue of their structural and stereo-
chemical identity, peptidic Schi bases may serve as bifunctional catalysts to de-
liver appreciable reactivity and high enantioselectivity. The above mechanistic par-
33.3 Peptidic Schiff Bases as Chiral Ligands 1003
adigm thus suggests that, in addition to HCN, other reagents such as various alkyl
metals may be delivered eectively and enantioselectively to the bound imine sub-
strate. Such a hypothesis led us to the following aspects of our program.
Fig. 33.3. Delivery of HCN by the AA2 moiety of the peptide ligand.
1004 33 Diversity-Based Identification of Efficient Homochiral Organometallic Catalysts
33.3.3
Zr-Catalyzed Enantioselective Addition of Dialkyl Zincs to Imines
the achiral amine product; when other dialkyl zinc reagents are employed (e.g.
Me 2 Zn), <2% conversion is detected. The facile reduction of substrates with Et2 Zn
may be due to b-H elimination of the metalEt complexes that generate active
metal hydrides that in turn promote amine reduction. The observed ineciency of
alkyl zinc reagents that do not bear a b-H (Me 2 Zn) or carry one that is less active
(n-alkyl2 Zn) support this hypothesis. Therefore, we speculated that the correspond-
ing amine-based peptide ligand (e.g. 44) may be the active catalyst. Thus, only in
cases where it can be eciently generated (i.e. in the presence of Et2 Zn) does the
asymmetric alkylation proceed smoothly. Once again, based on a mechanistic hy-
pothesis and because of the modularity of the chiral Schi base peptide structure,
a variety of ligand candidates were screened. As the data in Scheme 33.9 illustrate,
we established that the amine-based chiral ligands provide appreciable eciency as
well as enantioselectivity in the presence of all dialkyl zincs, regardless of whether
they bear an active b-H or not. It must also be noted that, as was found with the Ti-
catalyzed addition of cyanide to imines (Scheme 33.7), initial studies indicate that
the AA2 moiety is crucial for achievement of both high reactivity and enantio-
selectivity.
33.3.4
Cu-Catalyzed Enantioselective Addition of Dialkyl Zincs to Allylic Phosphates:
Pyridyl Dipeptides as Chiral Ligands
Catalytic enantioselective addition of alkyl metals to olens has been one of the key
problems that our research group had been interested in for the past 10 years [19].
With the success of the peptide-based ligands in promoting the addition of alkyl
zinc reagents to imines, we set out to examine whether a similar process may be
eected with olenic starting materials. One class of principal interest is the allylic
substitution reactions. Within this context, largely due to the scarcity of related pro-
tocols, we are particularly interested in transformations that utilize the less explored
hard alkylating agents which can enantioselectively deliver the problematic qua-
ternary carbon centers [20]. Recently, we have identied a new class of Schi base
peptides that accomplishes part of the above objective [21].
We initiated our search for the optimal conditions by using the more reactive and
readily accessible disubstituted olens (vs. trisubstituted alkenes described below).
Since CuCN has been demonstrated to exhibit a preference for SN 2 0 mode of ad-
dition in related processes with Grignard reagents, it was selected for the prelimi-
nary optimization studies [22]. Examination of potential substrates indicated that
allylic phosphates are the most suitable starting materials (see 51 ! 52 in Scheme
33.10) [23, 24]. The corresponding chlorides react smoothly with Et2 Zn in CH2 Cl2 ,
toluene, or Et2 O (30 C, 18 h, without ligand and Cu salt) and the derived ace-
tates, phenyl ethers, and phenyl carbamates aord < 10% conversion. Since the
previously mentioned mechanistic work suggested that the Schi base portion of
the peptidic ligands is likely a metal-ligation site that is critical to reactivity and
selectivity, we decided to perform catalyst optimization in the following systematic
33.3 Peptidic Schiff Bases as Chiral Ligands 1007
order: (1) identify optimal Schi base type; (2) ascertain the identity of the most
desirable Cu salt; (3) determine the optimal peptidic construct (e.g. di- or tripep-
tide); (4) further enhance enantioselectivity through identication of optimal Schi
base and the amino acid moieties.
As shown in Scheme 33.10 (rst-phase screening), treatment of phosphate 51
(Scheme 33.10) with Et2 Zn in the presence of 10 mol% CuCN and chiral ligands
5356 in tetrahydrofuran (THF) at 30 C leads to >98% conversion within 6 h.
Pyridine dipeptide 56 delivers the highest level of enantioselectivity (34% ee), fol-
lowed by phosphine 55 (26% ee). To ascertain the identity of the most ecient and
selective ligand/Cu salt combination, formation of 52 was examined in two sets of
experiments involving ligands 55 and 56 and a collection of Cu salts (see Scheme
33.10, second-phase screening). This study established that 56 and CuCN, overall,
provide the most ecient regio- and enantioselective process. Next, we secured the
following additional ligand attributes (third-phase screening, Scheme 33.10): (1) an
amide terminus is critical to the enantioselectivity replacement of the NnBu in
56 with an OMe group (57, Scheme 33.10) leads to signicant reduction of ee; (ii)
incorporation of a third amino acid (58, Scheme 33.10) or removal of one (59) is
detrimental to enantioselectivity [25]. The stereochemical outcome from the reac-
tion with 59 (14% ee) once again underlines the importance of the AA2 moiety
and indicates that simple attachment of a chiral group to the pyridyl ligation site is
not alone sucient for high asymmetric induction.
We then prepared chiral ligands 6067 (Scheme 33.10) and examined their abil-
ity to initiate the enantioselective alkylation of 51 under the same conditions men-
tioned above. Thus, the catalytic ability of the derived amine (60), amide (61), and
various a-substituted pyridyl systems (6265) were investigated; in addition, the
related indole-based 66 and C2 -symmetric 67 were probed. All reactions proceed to
>98% conversion and exhibit high degrees of SN 2 0 /SN 2 selectivity, but it is the o-
substituted ligand 64 that generates the highest ee. We thus selected o-OiPr pyridyl
as the Schi base moiety, and continued with the optimization of the AA1 and AA2
segments according to methods reported previously in the context of our work on
the enantioselective addition of cyanide to meso epoxides and imines [3]. These
studies uniformly suggest that l-Phe is the AA2 of choice, and that ligands which
bear l-Val (56), l-t Leu (68, Scheme 33.11) and l-Chg (69, Scheme 33.11) at the
AA1 position oer similarly superior enantioselection. When d-Val is used as AA1,
the sense of enantioselection is reversed, indicating that the stereochemical iden-
tity of AA1 is critical to the sense of induction and that the d,l-ligand may deliver
lower levels of enantioselectivity than the l,l isomer. The above chiral ligands were
subsequently used in catalytic alkylation of a range of aryl olens in the presence
of Et2 Zn and 10 mol% CuCN in THF.
The results shown in Scheme 33.11 are representative and point out which
ligand provides the highest selectivity for a particular substrate (< 2% SN 2 product
in all cases). The selectivity and reactivity levels for the disubstituted alkenes are
competitive with the recently reported catalytic alkylations of allylic chlorides [26],
in which sterically demanding dialkyl zinc reagents (e.g. dineopentyl zinc) are re-
quired for high enantioselectivity (a 50% ee with n-alkyl zinc reagents). The cata-
33.3 Peptidic Schiff Bases as Chiral Ligands 1009
33.3.5
Cu-Catalyzed Enantioselective Conjugate Addition of Dialkyl Zincs to Unsaturated
Ketones: Peptidic Phosphines as Chiral Ligands
33.4
Conclusions and Outlook
The studies summarized in this article represent our research groups attempts
in the past 5 years to establish a reasonably general protocol for the identication
and discovery of chiral catalysts that eect a range of important bond formations
enantioselectively and eciently. By selecting an appropriate class of chiral ligands
that satises the important criteria of modularity, lack of high symmetry, and avail-
ability of multiple but dierentiated binding sites, we have been able to develop cat-
alytic asymmetric CaC bond-forming reactions that involve both early (Ti and Zr)
and late (Cu and Zn) transition metals. Although optimal ligands, metal centers,
protecting groups, and solvents have been typically determined through systematic
screening of parallel libraries, mechanistic knowledge along with basic chemical
intuition are also critical ingredients along the way. These research activities are
the result of the appreciation of the principle that a priori rational design of a
catalyst may be near impossible since highly detailed mechanistic principles are
often not general; such principles can vary with subtle changes in reaction con-
ditions or substrate structure so that, even within a single class of substrates, the
identity of the optimum catalyst may change. These studies are based on the
premise that although the design of a specic catalyst may be impossible, general
mechanistic principles can be utilized to outline a systematic screening protocol
for catalyst identication.
Endnotes and References 1013
The research described above bears testimony to the fact that this line of research
does not advocate that we abandon rational or rigorous investigations of detailed
mechanisms of important processes. Elements of design and a priori decisions are
still required in determining what collections of catalysts need to be prepared; the
framework is simply broader and thus initial bias that may be based on a few ini-
tial observations has less of a chance to point us in the wrong direction.
A diversity-based strategy allows us to base our mechanistic hypotheses on a
much wider pool of data points it discourages us from making naive generalities,
which are more than often revised soon after a few additional experiments. As
demonstrated in the above studies, mechanistic studies based on data that are col-
lected from parallel libraries can provide additional logic and impetus for future
eorts in reaction development.
Acknowledgments
First and foremost, I thank my friend, colleague, and collaborator Professor Marc
Snapper. I am grateful to Dr Joseph Harrity, Dr Ken Shimizu, Dr Bridget Cole, Dr
Clinton Krueger, Mr Kevin Kuntz, Dr Carolyn Dzierba, Mr James Porter, Mr John
Traverse, Ms Courtney Luchaco-Cullis, Dr Hirotake Mizutani, Ms Sylvia Degrado,
Dr Wolfgang Wirschun, Ms Kerry Murphy, Mr Nathan Josephsohn, and Mr John
Gleason for making numerous invaluable intellectual and experimental contribu-
tions to the projects discussed in this article. Research in our laboratories is gen-
erously supported by the National Institutes of Health (GM-47480, GM-57212, and
postdoctoral fellowships F32-GM-17821 and F32-GM-18209). Additional support
has been provided by the National Science Foundation (CHE-9632278), Johnson
and Johnson, Pzer, DuPont, AstraZeneca, Albemarle, ArQule, Dreyfus Founda-
tion, Sloan Foundation, and Deutsche Forschungsgemeinschaft (postdoctoral fel-
lowship to W.W.).
1 Jacobsen, E. N., Pfaltz, A., is implied (as is often the case) that
Yamamoto, H. (eds), Comprehensive numerous catalysts are prepared, and
Asymmetric Catalysis. Springer, Berlin then after a period of trial and error a
1999. catalyst is determined to be optimal,
2 The word design is described in the such an excercise should be referred
Unabridged Webster English Dictionary to as a linear screening. In such a
in the following manner: To conceive case, the plan for the direction of
and plan out in mind. Based on such screening and not the catalyst itself
a denition, it is unlikely that a is conceived based on the existing
chemist can, in general, consider mechanistic data.
mechanistic data and directly design a 3 a) K. D. Shimizu, M. L. Snapper, A.
catalyst that proves to be optimal in H. Hoveyda, Chem., Eur. J. 1998, 4,
selectivity and reactivity. If by design it 18851889; b) M. B. Francis, T. F.
1014 33 Diversity-Based Identification of Efficient Homochiral Organometallic Catalysts
Jamison, E. N. Jacobsen, Curr. Opin. Chem., Int. Ed. Engl. 1997, 36, 1704
Chem. Biol. 1998, 2, 422428; c) H. B. 1707.
Kagan, J. Organomet. Chem. 1998, 11 a) C. A. Krueger, K. W. Kuntz, C. D.
567, 36; d) K. W. Kuntz, M. L. Dzierba, W. G. Wirschun, J. D.
Snapper, A. H. Hoveyda, Curr. Opin. Gleason, M. L. Snapper, A. H.
Chem. Biol. 1999, 3, 313319; e) B. Hoveyda, J. Am. Chem. Soc. 1999, 121,
Jandeleit, D. J. Schaefer, T. S. 42844285; b) J. R. Porter, W. G.
Powers, H. W. Turner, W. H. Wirschun, K. W. Kuntz, M. L.
Weinberg, Angew. Chem. 1999, 111, Snapper, A. H. Hoveyda, J. Am.
26482689; Angew. Chem. Int. Ed. Chem. Soc. 2000, 122, 26572658.
Engl. 1999, 38, 24942532; f ) M. L. 12 For a review on catalytic asymmetric
Snapper, A. H. Hoveyda in: additions to imines, see: S.
Combinatorial Chemistry. Fenniri, H. Kobayashi, H. Ishitani, Chem. Rev.
(ed.), Oxford University Press, Oxford 1999, 99, 10691094.
2000, pp. 433455. 13 For related studies on catalytic
4 For a recent review regarding the use asymmetric cyanide addition to
of some non-C2 -symmetric chiral imines, see: a) M. S. Iyer, K. M.
ligands in metal-catalyzed enantio- Gigstad, N. D. Namdev, M. Lipton,
selective reactions, see: a) A. Pfaltz J. Am. Chem. Soc. 1996, 118, 4910
in: Stimulating Topics in Organic 4911; b) M. S. Sigman, E. N.
Chemistry. Shibasaki, M., Stoddard, Jacobsen, J. Am. Chem. Soc. 1998,
J. F., Vogtle, F. (eds), VCH-Wiley, 120, 49014902; c) M. S. Sigman,
Weinheim 2000, pp. 89103; b) G. E. N. Jacobsen, J. Am. Chem. Soc.
Helmchen, A. Pfaltz, Acc. Chem. 1998, 120, 53155316; d) H. Ishitani,
Res. 2000, 33, 336345. S. Komiyama, S. Kobayashi, Angew.
5 For example, see: K. B. Hansen, Chem., Int. Ed. Engl. 1998, 37, 3186
J. L. Leighton, E. N. Jacobsen, J. 3188; e) E. J. Corey, M. J. Grogan,
Am. Chem. Soc. 1996, 118, 10924 Org. Lett. 1999, 1, 157160; f ) H.
10925. Ishitani, S. Komiyama, Y. Hasegawa,
6 a) H. Steinhagen, G. Helmchen, S. Kobayashi, J. Am. Chem. Soc.
Angew. Chem., Int. Ed. Engl. 1996, 35, 2000, 122, 762766; g) M. Takamura,
23392342; b) M. Shibasaki, H. Y. Hamashima, H. Usuda, M. Kanai,
Sasai, T. Arai, Angew. Chem., Int. Ed. M. Shibasaki, Angew. Chem., Int. Ed.
Engl. 1997, 36, 12361256. Engl. 2000, 39, 16501652.
7 For an example, see: M. Sawamura, 14 a) See reference 12; b) M. C. Hansen,
H. Nagata, H. Sakamoto, Y. Ito, S. L. Buchwald, Org. Lett. 2000, 2,
J. Am. Chem. Soc. 1992, 114, 2586 713715, and references cited
2592. therein.
8 a) H. Nitta, D. Yu, M. Kudo, A. 15 N. S. Josephsohn, K. W. Kuntz, M.
Mori, S. Inoue, J. Am. Chem. Soc. L. Snapper, A. H. Hoveyda, J. Am.
1992, 114, 79697975; b) A. Mori, H. Chem. Soc. 2001, 123, 1159411599.
Abe, S. Inoue, App. Organomet. Chem. 16 J. R. Porter, J. F. Traverse, A. H.
1995, 9, 189197. Hoveyda, M. L. Snapper, J. Am.
9 B. M. Cole, K. D. Shimizu, C. A. Chem. Soc. 2001, 123, 984985.
Krueger, J. P. A. Harrity, M. L. 17 For other catalytic asymmetric
Snapper, A. H. Hoveyda, Angew. alkylations of imines, see: a) S. E.
Chem. 1996, 108, 17761779; Angew. Denmark, C. M. Stiff, J. Org. Chem.
Chem., Int. Ed. Engl. 1996, 35, 1668 2000, 65, 58755878; b) S. E.
1671. Denmark, N. Nakajima, O. J.-C.
10 K. D. Shimizu, B. M. Cole, C. A. Nicaise, J. Am. Chem. Soc. 1994, 116,
Krueger, K. W. Kuntz, M. L. 87978798; c) H. Fujihara, K. Nagai,
Snapper, A. H. Hoveyda, Angew. K. Tomioka, J. Am. Chem. Soc. 2000,
Chem. 1997, 109, 17811785; Angew. 122, 1205512056; for a related review,
Endnotes and References 1015
34
Combinatorial Aspects of Materials Science
Bill Archibald, Oliver Brummer, Martin Devenney,
Daniel M. Giaquinta, Bernd Jandeleit, W. Henry Weinberg, and
Thomas Weskamp
Abstract
34.1
Introduction
Fig. 34.1. Conventional, parallel, and pooled approaches to synthesis and screening.
screen. Methods intermediate between these two extremes are based on parallel or
array syntheses, in a spatially addressable format with usually one compound per
well or region, coupled to automated screens.
A SciFinder9 reference search performed in February 2001 using the key word
combinatorial resulted in more than 9000 hits. An analysis of the literature
search demonstrates the tremendous growth of scientic publications and patent
applications in the combinatorial eld. Although most contributions deal with com-
binatorial applications within pharmaceutical, biological, and medical disciplines,
recent years have witnessed signicant advances in the development of combina-
torial approaches to the discovery and optimization of new materials and catalysts,
as is clearly shown by the increasing number of scientic publications and patent
applications during the last few years within these areas. Additionally, an increas-
ing number of review articles on the application of combinatorial methods to cata-
lyst discovery and optimization, a process dubbed combinatorial catalysis, has
appeared in the recent literature [18]. Furthermore, recent advances of combinato-
rial chemistry and high-throughput screening for chemical process development
have been reviewed [19, 20]. Consequently, several recent reviews address combi-
natorial high-throughput methodologies and experimental strategies in further de-
tail [2132].
This chapter, covering the years 1995 to 2001, summarizes the latest develop-
ments in the application of combinatorial methodologies to the discovery of new
solid-state and organic polymeric materials. Patents, patent applications, and con-
ference proceedings have generally not been included but are well appreciated and
acknowledged. Combinatorial approaches to catalyst discovery and process opti-
mization will be treated in a subsequent chapter.
After this introductory section, the following section, Section 34.2, summarizes
the eorts in combinatorial materials science beginning with a general overview
of combinatorial materials synthesis techniques. Section 34.3 describes examples
of novel high-throughput screening technologies, while Section 34.4 summarizes
some applications of combinatorial methods in the search for novel or improved
electronic, magnetic, and optical properties. Section 34.5 describes case studies
showing the application of combinatorial techniques to the discovery of a new
1020 34 Combinatorial Aspects of Materials Science
34.2
Combinatorial Solid-state Materials Science
The properties of solid-state materials often arise from complex interactions involv-
ing the host structure, dopants, defects, and interfaces. Therefore, they depend sen-
sitively on both composition and processing conditions. Few general principles
have emerged that allow the prediction of structure beyond binary systems and the
resulting properties of such solid-state compounds. Conventional one-at-a-time
synthesis and characterization can be a long and expensive process, and combina-
torial materials science holds great promise in facilitating the materials discovery
and optimization enterprises.
Conventional materials research typically begins with a decision on a general-
phase space which targets a property of interest and which is based on a set of
physical or chemical constraints, many of which may be empirically or intuitively
grounded. This parameter space is then divided into discrete compositions that
must be synthesized and screened for properties of interest. Chemical additions,
substitutions, and modications of synthesis and processing conditions allow the
researcher to optimize the properties of a given system. This process is typically
long and laborious, and may or may not lead to a promising material. The inte-
grated application of rapid synthesis, high-throughput screening, and sophisticated
data analysis allows for a promising alternative to the time-consuming classical
methodology. However, a well thought-through experimental design of the experi-
ments (DOE) is required to reduce the number of samples that will be necessary to
dene sample spaces within the experimental universe or to direct screening to
other spaces (feedback loop).
The combinatorial process relies on the implementation and coupling of high-
speed synthesis and high-throughput screening techniques. These methods facili-
tate more ecient explorations of a given composition space and oer a valuable
tool for the investigation of ternary and higher order systems. However, it is often
impossible to rapidly synthesize materials the physical characteristics of which (e.g.
composition, microstructure, grain size, and density) are exactly the same as mate-
rials made using the nal production process. Similarly, screening of desired prop-
erties is often very slow. Thus, combinatorial studies are based on the predictive
capabilities of synthesis and screening tools, and the challenge of the combinatorial
process is to implement appropriate synthesis and screening techniques. Rather
than comparing the properties of a few specic compositions within a phase space,
entire phase spaces can now be examined in a single experiment. The rst library
of compounds is often a broad compositional search covering an entire phase space,
34.2 Combinatorial Solid-state Materials Science 1021
e.g. an entire ternary composition diagram (Fig. 34.2). A primary screen speci-
cally developed to evaluate the very large number of compositions within that li-
brary identies a particular composition or range of compositions that is of further
interest. Primary screens are typically designed to eliminate a large fraction of the
compositions studied in the rst library, while secondary and follow-up libraries
examine a more narrow range of compositions as well as additional chemical sub-
stitutions and processing conditions. Optical and electronic properties such as
capacitance or luminescence are examples of physical properties that may be e-
ciently examined in high-throughput primary screens.
At the follow-up level, more detailed information can be obtained from high-
throughput secondary screens because the number of compounds that must be
screened has been greatly reduced. This process of synthesis, screening, and opti-
mization continues until a manageable number of compositions has been reached.
1022 34 Combinatorial Aspects of Materials Science
These selected compositions, all of which have passed previous screens, can now be
studied using conventional methods to obtain the more precise chemical and phys-
ical data necessary to characterize a material completely. Combinatorial methods
can thus act as an ecient lter for conventional methods by selecting only the best
candidates for further, more detailed study.
34.2.1
Materials Library Synthesis
Several synthesis techniques have been developed for combinatorial materials li-
brary formation. Some materials can be made using solution deposition methods,
while others are more suited for thin-lm deposition [3338]. By modifying tech-
nologies similar to those used to make integrated circuit (IC) chips, materials
libraries or integrated materials (IM) chips were rst developed and utilized by
Schultz and coworkers [39]. The choice of synthesis technique is based on both the
material being prepared as well as the primary screen employed after synthesis.
Unlike the case of drug discovery, however, the synthesis of solid-state materials
often relies on processing temperatures in excess of 400 C. High-temperature re-
action conditions have been addressed through the creation of two-dimensional,
spatially addressable arrays of samples deposited on thermally stable substrates.
34.2.2
Vapor Deposition Techniques
Thin-lm deposition methods are synthetically quite versatile; they have pro-
gressed to enable atomic- and molecular-layer epitaxy and oer the ability to con-
struct articial lattices, epitaxial overlayers, and patterned lms of a variety of
materials. Dopants are usually sandwiched between layers of the host material to
avoid evaporation and to assure proper interdiusion. Subsequent thermal pro-
cessing results in a library of materials or devices the physical properties of which
can be screened with either contact or noncontact screening probes. The number
of compounds that can be simultaneously synthesized by this technique is limited
by the spatial resolution of the masks and by the degree to which synthesis can be
carried out on a microscale.
One can use either physical shadow masks [39] or photolithographic lift-os [61]
to carry out masking. Photolithography, a standard process in the semiconductor
industry, has a high level of spatial resolution and accuracy and can generate chips
with a high density of diverse compositions (up to 10 6 per square inch). The mul-
tilayer thin-lm deposition may result in the synthesis of multiple binary phases
rather than the desired multielement single-phase material, a result that may be
avoided through the use of an eective two-step annealing process [33]. Early nu-
cleation and thus binary-phase formation is avoided by deposition at relatively low
temperatures. If the thickness of the deposited layer is less than a critical value
(material dependent) of typically 110 nm, diusion is dominant over nucleation
for sequential precursor layers. Johnson and coworkers have advanced this elegant
34.2 Combinatorial Solid-state Materials Science 1025
technique based on the sequential controlled deposition of thin lms [62]. Using
Johnsons method, an extended period of low-temperature (100400 C) annealing is
performed for proper interdiusion of thin-lm precursors. Subsequently, crystalli-
zation of the intermediate, amorphous material is induced by a high-temperature
annealing process allowing growth of entire integrated materials chips [63]. Grow-
ing entire integrated materials chips in this fashion is crucial for many materials,
where the material properties are closely tied to the crystalline quality of the lms.
Reection high-energy electron diraction (RHEED) has been proposed for in
situ monitoring of molecular beam epitaxial (MBE) deposition [35].
Pessaud and coworkers utilized pulsed laser ablation of two compound targets
made of superconducting YBa2 Cu3 O7 (YBCO) and the double-chain insulator
MCuO2 (M Ca, Sr) in an attempt to create new metastable superconducting
compounds [47]. The goal of this exploration was to deposit m consecutive layers
of YBCO with n layers of MCuO2 in an attempt to use the two-dimensional struc-
ture of YBCO as a structural template so that MCuO2 would form additional con-
ducting CuO2 planes upon deposition. The technique has led to the identication
of a number of new phases with enhanced critical temperatures and extends the
multilayer deposition concept from elemental targets to compound targets.
Combinatorial thin-lm methods are also being used to optimize the per-
formance of multiwavelength emitting chips that show potential applications in
wavelength-divisible multiplexing (WDM). Multiwavelength emission has been
achieved using selective area epitaxy and postgrowth selective region intermix-
ing. Layer intermixing techniques have achieved wavelength shifting through
implantation-induced intermixing, which is dicult to control.
Liu and coworkers used arsenic and proton implantation to create a multi-
wavelength-emitting library of doses and species-dependent intermixing,
which they analyzed by microphotoluminescence at room temperature [50].
Al0:35 Ga 0:65 As/GaAs single quantum wells were grown on GaAs (001) substrates
by MBE. Ion implantation was accomplished using four dierent pattern masks
on an 8 8 hollow circle-patterned blank mask in sequence. Four rows of ele-
ments implanted with dierent As doses at an ion energy of 90 keV were obtained.
The four masks were then rotated 90 for proton implantation at an ion energy of
40 keV. The fth column and the fth row were implanted with protons only and
As only. The implanted chip was annealed in a rapid thermal annealer (RTA) at
950 C, and the microphotoluminescence measured at room temperature. Using
this technique Liu and coworkers were able to generate more than 20 dierent
wavelengths from a single chip.
Amorphous materials have been investigated in a combinatorial optimization of
hydrogenated amorphous silicon (a-Si:H)-based thin-lm transistor (TFT) devices
by Koinuma and coworkers using plasma-enhanced chemical vapor deposition
[57b]. A contact mask placed over the indiumtin oxide substrate and sequential
depositions were carried out through a moving slit mask. A 90 substrate rotation
followed by deposition of hydrogenated amorphous silicon nitride (a-SiN:H) allows
for systematic investigation of the eects of thickness and compositional variation
across the library. Source and drain contacts were deposited on the top of the bi-
1026 34 Combinatorial Aspects of Materials Science
34.2.3
Alternative Library Synthesis Techniques
for the parallel synthesis of in the order of 1000 combinations in one experiment.
The system was tested by reproducing the phase diagram of the intensively studied
ternary phase system Na2 O/Al2 O3/SiO2/H2 O in a single experiment and under
identical reaction conditions. With a total volume of each sample gel of not more
than 0.5 mL, crystalline phases of zeolite A, faujasite, and gmelinite were obtained,
with the region of zeolite A coinciding almost perfectly with earlier results for bulk
materials. Good agreement was also obtained in the case of faujasite. The forma-
tion of sodalite was only observed in a more recent investigation. This partial agree-
ment was attributed to better control of the water content in the parallel autoclave.
The attractiveness of combinatorial methods for hydrothermal synthesis was
also realized by Klein et al., who claim that the advantage of their reactor over the
one described above is their much smaller reaction volume, in the order of 2 mL, as
well as the direct preparation of a library of materials the components of which
can be identied automatically on the library substrate by X-ray microdiraction
(Fig. 34.5) [69]. The central feature of Kleins reactor is a silicon wafer that, after
hydrothermal synthesis and calcination, contains the sintered reaction products and
represents the library. The identication of the individual products may be directly
performed with a GADDS (General Area Detector Diraction System) micro-
diractometer. Choi and coworkers designed a centrifuge apparatus that allows
quantitative product recovery onto lter paper for X-ray microdiraction without
manipulation of the individual samples [70]. Furthermore, Lai and coworkers
demonstrated parallel synthesis of zeolite lms in a 21-well reactor using a vertical
substrate conguration to provide uniform wetting that favored heterogeneous
lm growth in an organic-free clear solution [71]. The results of the synthesis of
Nanocrystal assembly occurs with high selectivity on just one type of the spatially
and chemically distinct sites. The production of binary, ternary, and quaternary
patterns of nanocrystals has been reported, and the procedure appears applicable
for the preparation of multicolored nanocrystal-based color pixels. Moreover, the
technique should be compatible with other lithographic patterning steps, such as
those involved in the formation of electrical contacts, capacitors, and other device
components of the photonics and microelectronics industries. Furthermore, the
procedure has been demonstrated on a variety of substrates.
A novel approach by Wright and coworkers has shown that combinatorially
derived peptide and peptidomimetic ligands based on phytochelatins can stabilize
CdS nanoclusters [7476]. Using the 7-mer sequence XaCysaYaCysaZaCysaGly
of n3 -phytochelatin as a lead structure, Wright and coworkers employed standard
Fmoc protocols (Fmoc 9-uorenylmethoxycarbonyl) for combinatorial peptide
library synthesis. In their study, the cysteine residues are held constant while the
spacer residues X, Y, and Z are varied over all combinations of the selected spacer
amino acids a-Glu, g-Glu, g-amino butyric acid (GABA), SerGly, and e-Ahx (Fig.
34.6). This approach limits the number of peptides to 125, as opposed to the over
one billion potential ligands obtained if all seven positions are varied over all 20
naturally occurring amino acids. The ve selected spacer residues chosen have be-
tween three and seven atoms in the backbone and can span an idealized extended
C a aC a distance between the cysteines. Using commercial software, quantitative
structureactivity relationship (QSAR) methods were employed to calculate the
molecular numerical properties (descriptors) of the combinatorial peptide library.
The positional analysis of individual amino acids allows the correlation between
the properties of the stabilizing (iso)peptidic ligand and the resultant peptide-
encapsulated CdS nanocluster to be determined.
Many new advanced materials ranging from magnetic materials to bulk metallic
glasses involve alloying multiple elements to control both intrinsic and extrinsic
properties. Permanent magnets, for example, may have a dozen or more elements
included to control grain boundaries and domain wall movement. Cohen-Adad
1030 34 Combinatorial Aspects of Materials Science
and coworkers utilized a melting zone technique to investigate AlCo and AlCoFe
alloys [77]. In their method, source materials are cut into wedges and then welded
together into a bar. The bar is then placed in an alumina crucible and drawn
through a horizontal zone furnace to facilitate mixing. The resulting bar is diced
into pieces and the composition at each point along its length measured with
microprobe analysis. The morphology and crystal structure of the materials were
investigated with optical microscopy, dierential thermal analysis, and X-ray dif-
fraction. Inuoe has reviewed similar techniques used to search for new glasses,
which typically have ve or more elements present to inhibit crystallization [25].
34.3
High-throughput Screening
The large numbers of compounds and the speed with which they may be prepared
using the combinatorial synthesis methods described above require automated
high-throughput screening techniques that are capable of equally rapid character-
ization. Typically, the information necessary to classify new materials cannot be
obtained from a single piece of characterization equipment. Therefore, a series of
high-throughput screening tools is employed at various stages in the combinatorial
process to create a screening hierarchy. While a large number of automated com-
mercial systems exist for the high-throughput analysis of microliter samples for
medical and pharmaceutical applications, the vast majority of the screening tools
necessary for advanced materials research are custom-made instruments. Conven-
tional analytical tools such as mass spectrometry, gas and liquid chromatography,
optical and nuclear magnetic resonance spectroscopy, X-ray uorescence micro-
probe [78], and X-ray diraction have been automated and redesigned for rapid
serial measurements of hundreds of samples per day [79, 80]. Several examples of
custom-made high-throughput screening systems are discussed below.
34.3.1
Optical Screening
determine the band gap of various oxide thin lms. Jin and coworkers [56] and
Makino and coworkers [83] calculated the optical band-gap from the measured
optical absorbance of 81 elements in a library of ZnO thin lms doped with nine
independent 3d-metal ions. Similar transmission and reection methods have been
used in the infrared to estimate band-gaps for thermoelectric and photovoltaic
materials.
Furthermore, Xiang describes an apparatus for the measurement of electro-optic
coecients of thin-lm samples that functions by measuring the polarization mod-
ulation of light passing through a sample [21]. The variation in the index of refrac-
tion upon application of an electric eld is typically small, approximately 0.1%,
making thin-lm measurements dicult. The electro-optic measurements are con-
ducted using the library, a compensator, and a quarter-wave plate which are located
between two crossed polarizers and are illuminated by a HeNe laser. An electric
eld is applied to the sample at 45 to the incident polarization. Given the index of
refraction and some geometric factors, the electro-optic coecients can be deter-
mined.
34.3.2
X-ray Characterization
34.4
Applications
34.4.1
Superconductivity
34.4.2
Ferromagnetic Semiconductors
ductors by adding dilute amounts of cobalt to the anatase form of TiO2 [88]. Ana-
tase TiO2 was chosen as a candidate semiconductor because of its high electron
carrier mobility at room temperature.
TiO2 and Ti 0:5 Co0:5 O2 ceramic targets were ablated using a KrF exicmer laser
in a combinatorial laser molecular beam epitaxy system [55]. Anatase lms of
Tix Co1x O2 were grown on LaAlO3 (001) and SrTiO3 (001) substrates and the com-
position mapped by microprobe analysis. X-ray diraction studies show that the Co-
doped anatase compositions obey Vegards law up to a solubility limit of x 0:08.
Magnetic images obtained by a scanning Squid microscope at 3 K showed do-
mains of around 20 mm in size for the Co-doped compositions. The spontaneous
magnetization systematically increases with increasing Co content to 0.32 mB (Bohr
magnetron) observed for the x 0:07 sample. The ferromagnetic ordering tem-
perature was estimated to be in excess of 400 K, a substantial increase over the
Mn-doped GaAs compounds.
34.4.3
Magnetoresistant Materials
34.4.4
Dielectric and Ferroelectric Materials
a
Fig. 34.9. a) Dopant map for the materials. A quaternary masking scheme with
(Ba 0:8 Sr0:2 )TiO3 quadrant of a ferroelectric four masks was used to deposit 256 dierently
thin-lm library; the same map applies to the doped thin lms of Ba x Sr1x TiO3 0:5 < x < 1
other three quadrants containing dierent host on a 2.54 cm 2.54 cm LaAlO3 substrate.
(BST) were determined [63, 95]. Ferroelectric materials are being widely used in
the development of new microwave devices such as frequency agile lters, phase
shifters, and tunable resonators, and (Ba x Sr1x )TiO3 compounds have been inten-
sely studied for such applications owing to their low loss and high dielectric co-
ecients. Four libraries of 256 dierently doped thin lms of (Ba x Sr1x )TiO3 ,
where x 1, 0.8, 0.7, and 0.5, were generated on a 2.5 cm 2.5 cm LaAlO3 sub-
strate using multistep thin-lm RF sputtering together with a quaternary masking
strategy with four physical shadow masks. These four hosts were then doped with
dierent combinations of up to three out of nine dierent metallic elements with
each dopant added in excess of 1 mol% with respect to the BST host; the resultant
dopant map for the (Ba 0:8 Sr0:2 )TiO3 quadrant is shown in Fig. 34.9a. The same map
applies to the other three quadrants of the library with the three other host mate-
rials, BaTiO3 , (Ba 0:7 Sr0:3 )TiO3 , and (Ba 0:5 Sr0:5 )TiO3 .
Appropriate postannealing aorded high-quality epitaxial thin lms, similar to
those made by conventional in situ methods from stoichiometric targets on hot
substrates. The rapid, quantitative and nondestructive characterization of the mi-
crowave dielectric properties, the dielectric coecient, and loss tangent of samples
in the library was achieved with a scanning-tip microwave near-eld microscope,
as shown in Fig. 34.8, at 1 GHz, and the results were found to be consistent with
measurements made with interdigital electrodes [21, 33, 34, 59, 96]. In Fig. 34.9b
the averaged dielectric coecient and tangent loss data are displayed for the BaTiO3
(BTO) quadrant; each square corresponds to a dierent thin-lm sample site, and
darker shades represent lower dielectric coecient. Specic dopants were found to
34.4 Applications 1037
b
Fig. 34.9. b) Dielectric coecient and loss submicron spatial resolution. Each square
tangent images of doped BaTiO3 lms, corresponds to a dierent thin-lm sample site
measured using a scanning-tip microwave as given in the composition map above.
near-eld microscope (STMNM) with
signicantly aect the dielectric coecient and the tangent loss, with La (sample 2)
and Ce (sample 3) increasing the dielectric coecient, whereas most dopants result
in decreased values when compared with the undoped material. In the tangent
loss data, a lighter shade implies a lower loss tangent. In the case of microwave
applications, where the values of the dielectric coecient are considered to be high,
a minor reduction may be tolerable as long as there is a noticeable improvement
in the tangent loss, which is especially important for the microwave application of
tunable dielectric materials. Possible applications include radar, which steers the
beam and quickly locates objects electrically rather than mechanically, and cellular
phones, which instantaneously tune into the incoming signals.
Subsequently, Xiang and coworkers evaluated the eects of dopants on device
performance by incorporating a layer of the doped dielectric (Ba x Sr1x )TiO3 in a
parallel-plate capacitor library between Pt and La 0:5 Sr0:5 CoO3 electrodes [97]. Like
van Dovers work on thin-lm dielectrics, this investigation aims to aid in the iden-
1038 34 Combinatorial Aspects of Materials Science
34.4.5
Luminescent Materials
tion source, host lattice, sensitizer, and luminescent center. Two groups reported
application of combinatorial methods to the discovery and optimization of several
luminescent metal oxide materials. Oxides are attractive host materials for the de-
velopment of advanced phosphors owing to their ease of synthesis and stability. For
example, the ecient refractive oxide phosphors Y2 O3 :Eu 3 (red), Y3 Al5 O12 :Tb 3
(green), and BaMgAl10 O17 :Eu 2 (blue), have found applications in tricolor lamps,
projection TVs, and plasma displays.
Symyx Technologies red phosphor, Y0:845Al0:07 La 0:06 VO4 :Eu 0:025 , with a quantum
eciency approaching those of existing commercial red phosphors, was discovered
in a library of 25,000 dierent compositions produced using electron-beam evapo-
ration to deposit thin lms onto a 7.5-cm silicon wafer [81]. In this discovery library,
four constant thickness columns were rst deposited consisting of SnO2 (480 nm),
V2 O5 (160 nm), Al2 O3 V2 O5 (150 nm 80 nm), or Al2 O3 (300 nm). On top of
these layers were deposited four rows of linearly varying thickness of La2 O3 , Y2 O3 ,
MgO, and SrCO3 to create 16 host lattice subregions. Finally, within each of the 16
subregions linearly varying thicknesses of the rare earths Eu2 O3 , Tb4 O7 , Tm2 O3 ,
and CeO2 were deposited, resulting in approximately 600 dierent chemical com-
positions per square centimeter, see Fig. 34.10. A stainless-steel primary mask con-
sisting of 230-mm 2 elements spaced 420 mm apart was attached to the substrate to
separate individual library elements. The spatial variation of materials deposited
on the library was created using stationary and movable physical masks to control
the thickness of specic evaporants in selected regions of the substrate. Because
the constituents of each library element were deposited in layers, oxidative thermal
processing at various temperatures was needed to mix the layers and create the
desired chemical compounds. High-throughput screening for UV-excited photo-
luminescence was performed by photographing the visible emission of the library
with a charge-coupled device (CCD) camera while exciting with a 254-nm broad-
band source. A quantitative measure of the chromaticity relative to calibration stan-
dards was calculated from three images obtained using, respectively, red, green,
and blue tristimulus emission lters. To optimize the best host compositions of
the Eu-doped red phosphors, Y1m Alm VO4 , identied in the initial high-density
exploration library, a second library was deposited on a triangular grid, including
La as a third host component. First, Eu2 O3 and V2 O5 were deposited uniformly
over the entire substrate, followed by linearly varying the amount of Y2 O3 , Al2 O3 ,
and La2 O3 along three axes oset by a rotation of 120 , thus exploring all possible
Y0:95mn Aln Lam VO4 :Eu 0:05 compounds. The maximum intensity with red chro-
maticity suitable for a commercial material corresponds to the phosphor composi-
tion Y0:82Al0:07 La 0:06 VO4 :Eu 0:05 . The activator concentration in this lead host was
subsequently optimized with a third library in which the Eu 3 concentration was
varied between 0% and 20% using a single movable mask, probing the composi-
tions Y0:87m Al0:07 La 0:06 VO4 :Eu m . Screening of this optimization library identied
2.5% Eu as the most ecient dopant, resulting in the nal optimized composition
Y0:845Al0:07 La 0:06 VO4 :Eu 0:025 with improved red chromaticity (x 0:67, y 0:32)
compared with the more orange standard commercial red phosphor Y1:95 O3 :Eu 0:05
(x 0:64, y 0:35). Synthesis of bulk samples by conventional methods produced
1040
34 Combinatorial Aspects of Materials Science
materials indistinguishable from the thin-lm library samples and conrmed the
performance of the newly identied phosphor composition.
Xiang and coworkers prepared several 128-member libraries of aluminates of
the formula Gd(La,Sr)AlOx doped with rare earth activators, and have evaluated the
phosphorescence of all the samples simultaneously by photographing the entire
library under broad wavelength UV light (254 nm) [99]. The resulting color photo-
graph allowed a convenient qualitative evaluation of the library. The combinatorial
libraries were used to survey a wide range of oxide compositions under a variety of
processing conditions, facilitated by the simultaneous deposition of seven identical
libraries on LaAlO3 , MgO, and Si single-crystalline substrates followed by dierent
annealing and sintering conditions.
Subsequently, Xiang and coworkers demonstrated that a scanning multi-inkjet
delivery system (see above) can be used to achieve results identical to the vapor
deposited libraries [64].
Multicomposition materials libraries of thin-lm phosphors
(Gd,La,Y)m (Ta,Zr,W,Mo,Zn)n (Al,Mg,Sr)k Ox :(Tm,Eu,Tb,Ce)y (with 0 a m; n; k a 1,
0:005 a y a 0:1) were generated by Sun and Xiang using sequential RF to create
three identical libraries simultaneously on LaAlO3 [100]. Dierent annealing con-
ditions and atmospheres completed the synthesis of the phosphor libraries, and
photoluminescence images were acquired under 254-nm centered broad wave-
length UV radiation of a Hg lamp using color photography. A new red phosphor
with the composition (Gd1:54 Zn 0:46 )O3d :Eu 30:06 is claimed, and reported to have
characteristics similar to the state-of-the-art orangered phosphor Y2 O3 :Eu 3 and
also to possess excellent X-ray and cathodoluminesent properties.
Schultz and coworkers used RF/DC sputtering to prepare three 1024-member
libraries of silicate and gallate host materials on Si substrates, each library entity
containing a number of dierent dopants [61]. After annealing, a parallel imaging
system and a scanning spectrophotometer were used to identify and characterize
compositions in the library. Few blue phosphor were identied by their emission
and excitation spectra, but neither bulk nor thin lm samples on LaAlO3 sub-
strates exhibited blue luminescence. Subsequently, gradient libraries of
Gd x Ga1x Oz :(SiO2 )y on LaAlO3 substrates were generated using pulsed laser de-
position (PLD) and broadband blue emission was observed from these libraries.
34.5
Case Studies
34.5.1
Materials Discovery
The group from Symyx Technologies followed the red vanadate phosphor with a
report on a novel luminescent oxide, Sr2 CeO4 , with a bluish white emission [82].
Whereas previous studies of luminescent materials had primarily focused on the
synthesis of compositional variants within classes of materials with known struc-
1042 34 Combinatorial Aspects of Materials Science
34.5.2
Device Optimization
Fig. 34.11. The crystal structure of the blue phosphor Sr2 CeO4
consists of linear chains of trans edge-sharing distorted CeO6
octahedra with four equatorial O atoms and two terminal CeaO
bonds per octahedron, separated by interchain Sr 2 cations.
34.6
Organic Materials and Polymers
34.6.1
Schi Bases for Nonlinear Optical (NLO) Materials
compounds can be synthesized and screened. Five crystal structures were deter-
mined, all with centrosymmetric space groups that preclude the compounds from
being NLO candidates.
34.6.2
Articial Receptors for Small Organic Molecules
Since the rational design of nanoporous materials and their inclusion cavities based
on molecular structures is dicult, combinatorial methods were implemented to
nd superior lattice inclusion compounds [105]. When screened against a guest
molecule (iPrOH), seventeen new ammonium carboxylate host compounds were
discovered from a combinatorial library of 100 salts formed by mixing ten commer-
cially available amines and ten carboxylic acids. One host compound was identied
that forms inclusion compounds with 25 dierent organic solvents. Application to
chiral recognition by the proper choice of amines and carboxylic acids to form a
variety of host compounds is under investigation.
Leblanc and coworkers applied combinatorial principles to surface chemistry by
synthesizing and screening libraries of amphiphilic lipid molecules [106]. At the
airwater interface, the hydrophilic polar moiety of these molecules is embedded
into the water phase and the hydrophobic alkyl chain is oriented toward the air
phase. When compressed at the interface, the amphiphilic molecule will move and
assemble into organized supramolecular structures, known as Langmuir mono-
layers. The polar moiety was functionalized with a peptide library such that the self-
assembly of these functional lipids formed protein-like supramolecular structures
(Fig. 34.14). One library with 250 members and three sublibraries of 54 members
each were assayed for molecular recognition of maltose as well as other sugar mol-
ecules. One sublibrary, containing both polar charged (Glu) and aromatic (Tyr)
amino acid residues, was identied that selectively bound maltose but not d-glucose
and sucrose. This selectivity can be explained by the three-dimensional combina-
tion of those two amino acids in the binding site with Tyr playing a more signi-
cant role for the binding activity.
34.6.3
New Materials for the Separation of Enantiomers
Fig. 34.14. The structure of a combinatorial peptide lipid library (AA i amino acid).
34.6 Organic Materials and Polymers 1047
The same group also applied the principle of reciprocity to the identication of
new CSPs [109]. This approach is based on the following assumption: if a ligand of
the stationary phase is selective toward an analyte, then the same analyte immobi-
lized onto the proper support should result in a stationary phase that is very likely
to be selective toward the original ligand [110]. Following this principle, the chiral
substrate (S)-(3,5-dinitrobenzoyl)leucine acted as an inverse selector and was im-
mobilized on synthetic polymer beads to give CSP 2 (Fig. 34.15). Three-component
1048 34 Combinatorial Aspects of Materials Science
the library and allowed to equilibrate. The enantiomeric ratio of the analyte in the
supernatant is analyzed after the equilibration period using circular dichroism
measurements or other methods. A selective absorption of one of the two en-
antiomers to the resin is indicative of a chiral selector.
Welch and coworkers introduced an improved method for rapid liquid chroma-
tography/mass spectrometry (LC/MS) screening of chiral stationary phases based
on the use of isotopically labeled enantiomers [114]. Pseudoracemates made up of
a pair of isotopically dierentiated pseudoenantiomers were prepared and screened
against dierent CSPs. The enantioenrichment of the supernatants can be rapidly
estimated by comparing the MS abundance of the ions corresponding to each of
the pseudoenantiomers.
A visual screen for identifying chiral selectors was introduced by Still and co-
workers [115]. The idea was to treat an equimolar mixture of a blue l-amino acid L-7
and a red d-amino acid D-7 with a library of chiral selectors on synthesis beads
in which each bead carries a dierent selector (Fig. 34.16). Highly enantioselec-
tive binding results in red or blue beads, whereas unspecic binding yields brown
beads. The reddest and the bluest beads found in the assay were picked and de-
coded to determine the structure of the bound chiral selector.
34.6.4
Molecular Imprinting
Molecularly imprinted polymers (MIPs) are another class of articial receptors for
small organic molecules. Their binding performances give these materials great
potential in combinatorial approaches as recognition matrices for the screening
and rapid selection of ligands from a combinatorial library [116]. Molecular im-
printing is a technology where recognition sites can be generated within a macro-
molecular matrix by use of a molecular template in a casting procedure [117]. The
selected ligand (print molecule) is rst allowed to interact freely in solution via
bond formation with one or more functional monomers. The resulting adducts are
subsequently copolymerized with a large excess of crosslinker to give a rigid, in-
soluble polymer. Following extraction of the print molecule, specic recognition
sites are left in the polymer. The technique has been applied to the screening of
a combinatorial steroid library [116a], and a combinatorial library of molecularly
imprinted polymers has been tested in chiral separations [116c]. Combinatorial
methods proved to be advantageous since the preparation and evaluation of MIPs
is usually very tedious and time-consuming.
In 1999, Takeuchi and coworkers reported a rapid combinatorial approach to
synthesize and evaluate articial receptors [118]. The preparation of the MIPs was
automatically performed using programmable liquid-handling equipment with a
new in situ molecular imprinting protocol. The freshly synthesized MIP library
(polymer lms) was then subjected to a primary screening to estimate the anity
of the template to the resultant polymer. A secondary screening on puried MIPs
then allows for a more precise evaluation of the anities and selectivities of the
new materials. MIPs that are articial receptors for triazine herbicides ametryn 8
and atrazine 9 were prepared by the combinatorial molecular imprinting approach
using variable amounts of two functional monomers, methacrylic acid (MAA) 10
and 2-(triuoromethyl)acrylic acid (TFMAA) 11, as well as ethylene glycol dime-
thacrylate as a crosslinker and 2,2 0 -Azo-bis-isobutyryl-nitrile (AIBN) as the poly-
merization initiator (Fig. 34.17). On examination of the 49-member MIP library, it
appears that, depending on the functional monomer used, the imprinting e-
ciency is dierent for each triazine herbicide. This result suggests that high-
throughput synthesis and screening of large sets of polymers will provide the most
ecient molecularly imprinted receptors for given target molecules.
34.6.5
Polymers with Novel Topologies and Functionalization
Combinatorial approaches have also been applied to polymer chemistry. For ex-
ample, a small library of polyarylates (0.2 g per sample) (14) was prepared from 14
distinct tyrosine-derived diphenols (12) and eight aliphatic diacids (13), resulting in
34.6 Organic Materials and Polymers 1051
112 structurally related strictly alternating copolymers (Scheme 34.3) [119]. These
copolymers are biodegradable and potentially useful as medical implant materials.
The structural variations provided incremental dierences in polymer free volume,
bulkiness, exibility, and hydrophobicity. Up to 32 polymers were synthesized in
parallel under identical reaction conditions and conventional methods were used
to measure the molecular weight by GPC (gel permeation chromatography), Tg by
DSC (dierential scanning calorimetry), and the airwater contact angle (sessile
drop) on a sample-by-sample basis, providing structureproperty correlations. Di-
verse and focused combinatorial libraries of synthetic biodegradable polymers
could also be designed by using molecular topology and genetic algorithm opti-
mized quantitative structureproperty relationships (QSPR) [120].
Scheme 34.3. Library of polyarylates 14 derived from diphenols 12 and diacids 13.
tiators of the type 15 were prepared that provide for two independent rounds of po-
lymerization. Block copolymers 22 could be obtained in a temperature-controlled
manner through sequential normal and living polymerizations. To add another
dimension to the combinatorial diversity, initiator 16 participates as a monomer in
the rst polymerization with monomer A, resulting in statistic copolymers of
type 23. The TEMPO-functionalized residues then mediate the living poly-
merization at 130 C, giving access to combinatorial polymers of type 24. The
members of the block copolymer library were then screened for their ability to
serve as a soluble polymer support for organic reactions. One block copolymer
could be identied that has a solubility prole that is complementary to the cur-
rent soluble polymer of choice in liquid-phase organic synthesis (LPOS) poly-
ethylene glycol (PEG). R1 and the nitrile groups in block copolymers of type 22
could be used to functionalize these molecules.
Scheme 34.4. Parallel polymer synthesis via sequential normal/living free radical process.
1054 34 Combinatorial Aspects of Materials Science
34.7
Summary and Outlook
Acknowledgments
The authors wish to thank Ms Silvia Lee (Symyx Technologies) and Ms Kathryn
Boykin (XenoPort) for their kind and prompt support in reference and patent
searches. The authors thank Ron Krasnow (Symyx Technologies) for checking the
manuscript.
References
Angew. Chem. Int. Ed. Engl. 1996, 35, Reetz, Angew. Chem. 2001, 113, 292
1742; b) F. Balkenhohl, C. von 320; Angew. Chem. Int. Ed. 2001, 40,
dem Busche-Hunnefeld, A. Lansky, 285310, and references cited therein;
C. Zechel, Angew. Chem. 1996, 108, c) F. Gennari, P. Seneci, S. Miertus,
24362488; Angew. Chem. Int. Ed. Catal. Rev.-Sci. 2000, 42, 385402, and
Engl. 1996, 35, 22882337; c) L. A. references cited therein; d) B.
Thompson, J. A. Ellman, Chem. Rev. Jandeleit, D. J. Schaefer, T. S.
1996, 96, 555600. Powers, H. W. Turner, W. H.
15 For accounts and special techniques Weinberg, Angew. Chem. 1999, 111,
used in combinatorial chemistry, see 26482689; Angew. Chem. Int. Ed.
for example: a) K. S. Lam, M. Lebl, V. 1999, 38, 24952532, and references
Krchzak, Chem. Rev. 1997, 97, 411 cited therein.
448; b) A. Nefzi, J. M. Ostreh, R. A. 19 M. J. Cannarsa, T. Uno, C. Larsen,
Houghten, Chem. Rev. 1997, 97, 449 Curr. Opin. Drug Discov. Dev. 2000, 3,
472. c) M. C. Pirrung, Chem. Rev. 743749.
1997, 97, 473488. d) D. J. Gravert, 20 H. Weinmann, Nachr. Tech. Chem.
K. D. Janda, Chem. Rev. 1997, 97, 2001, 49, 150154.
489509; e) A. W. Czarnik, J. A. 21 a) X.-D. Xiang, Biotechnol Bioeng.
Ellman (eds), Special Issue on 1999, 61, 227241; b) X.-D. Xiang,
Combinatorial Chemistry, Acc. Chem. Annu. Rev. Mater. Sci. 1999, 29, 149
Res. 1996, 29, 111170. 171; c) D. R. Liu, P. G. Schultz,
16 For a review on recent developments Angew. Chem. 1999, 111, 3656;
and applications of polymer-supported Angew. Chem. Int. Ed. 1999, 38, 36
reagents and scavangers in synthetic 54.
organic chemistry, see: a) O. 22 a) X.-D. Xiang, Chem. Ind. 1998, 19,
Brummer, B. Clapham, K. D. Janda, 800802; b) X.-D. Xiang, Materials
Curr. Opin. Drug Discovery Dev. 2000, Today 1998, 1, 2326; c) I. Takeuchi,
3, 462473; b) S. V. Ley, I. R. X.-D. Xiang, Gendai Kagaku 1998,
Baxendale, R. N. Bream, P. S. 332, 2430; d) P. G. Schultz, X.-D.
Jackson, A. G. Leach, D. A. Xiang, Curr. Opin. Sol. State Mat. Sci.
Longbottom, M. Nesi, J. S. Scott, R. 1998, 3, 153158.
I. Storer, S. J. Taylor, Special Issue: J. 23 C. Schmitz, P. Poesch, M.
Chem. Soc., Perkin Trans. 1 2000, 23, Thelakkat, H.-W. Schmidt, Polym.
38154195; for further information Prep. 1999, 40, 11821183.
about polymer-supported reagents, 24 I. Bozovic, V. Matijasevic, Mat. Sci.
see: c) S. W. Kaldor, M. G. Siegel, Forum 2000, 352 (Trends in Advanced
Curr. Opin. Chem. Biol. 1997, 1, 101 Materials and Processes), 18.
106; d) S. J. Shuttleworth, S. M. 25 S. Inoue, Kagaku Kogyo 2000, 51,
Allin, P. K. Sharma, Synthesis 1997, 128133.
12171239; e) P. Hodge, Chem. Soc. 26 J. N. Cawse, Acc. Chem. Res. 2001, 34,
Rev. 1997, 26, 417424. 213221.
17 T. R. Boussie, K. Hall, A. M. 27 J. D. Hewes, SPIE-Int. Soc. Opt. Eng.
LaPointe, V. Murphy, T. S. Powers, 2000, 3941 (Combinatorial Composi-
J. A. M. van Beek (Symyx Techno- tion Spread Techniques in Materials
logies), Delivery and scavenging and Device Development), 4661.
agents for the synthesis of compounds 28 a) W. F. Maier, Nachr. Chem. Tech.
and catalysts, PCT Int. Appl. WO Lab. 1999, 47, 12071210; b) W. F.
98/56796. Meier, Angew. Chem. 1999, 111, 1294
18 For reviews on combinatorial methods 1296; Angew. Chem. Int. Ed. 1999, 38,
of combinatorial catalyst development 12611218.
and high-throughput screening in 29 a) E. W. McFarland, W. H.
catalysis, see: a) S. Senkan, Angew. Weinberg, Trends Biotechnol. 1999, 17,
Chem. 2001, 113, 322341; Angew. 107115; b) E. W. McFarland, W. H.
Chem. Int. Ed. 2001, 40, 312329, and Weinberg, Mater. Technol. 1998, 3,
references cited therein; b) M. T. 153158.
References 1059
35
Reprogramming Combinatorial Biology for
Combinatorial Chemistry*
Sean V. Taylor
35.1
Introduction
As the other chapters in this book have nicely illustrated, combinatorial chemistry
has made dramatic technical advances since its inception, with signicant im-
provements in the design, synthesis, purication, and evaluation of combinatorial
libraries. It is now commonplace for companies focusing on drug discovery to
have 0.51.5 million compounds arrayed throughout their combinatorial libraries,
which can be screened in a matter of weeks or months for activity against any
number of biological targets. The potential success of this strategy is underscored
by reports from numerous pharmaceutical companies of compounds that were
identied or optimized using combinatorial chemistry and that have now entered
clinical drug trials. It is quite likely that, over the next 100 years, a signicant per-
centage of pharmaceuticals will be derived in some manner from combinatorial
libraries.
If we expand our denition of combinatorial libraries, however, the current im-
portance of the combinatorial approach to drug discovery, as well as to most life
processes, is more immediately clear. For instance, eight of the 20 top-selling
pharmaceuticals are natural products or derived from natural products [13]. It is
not a large leap in logic to consider these compounds as already being derived
from combinatorial libraries biological, or perhaps evolutionary combinatorial
libraries. In contrast to combinatorial libraries produced by synthetic means, the
combinatorial libraries that give rise to natural products are not produced at the
compound level, but at the genetic level during the process of evolution, through
mutation, gene duplication, and interspecies genetic transfer. The evaluation of
these biological combinatorial libraries is simply the process of natural selection.
Those compounds that provide some value toward the survival of the producing
organism will be propagated and further improved upon through additional evo-
lutionary rounds.
*Portions of this article were previously published (Angew. Chem. 113 (2001) 34083436; Angew.
Chem. Int. Ed. 40 (2001) 33103335). [Ref. 100]
Nature has utilized such evolutionary libraries over billions of years, but mo-
lecular biologists, biochemists, and chemists have only recently learned how to
harness this combinatorial power, and for a number of dierent uses. In this
chapter, the various eorts to engineer and apply biological combinatorial libraries
for use in enzymology, drug discovery, bioprocessing, and also for addressing fun-
damental issues in chemistry and biology will be explored. Instead of an exhaus-
tive review, specic examples will be used to highlight various combinatorial as-
pects of biology. In addition to asking the question Where are we now?, I also
hope to ask Where are we (or where could we be) going?
35.2
Combinatorial Biosynthesis: Creation of Novel Small-molecule Natural Products
Fig. 35.2. Biosynthetic pathways. (A) The A. Seven amino acid incorporation modules
modular erythromycin polyketide synthase (17), coded by three genes srfA-A, srfA-B, and
produces 6-deoxyerythronolide B, which is then srfA-C, are shown with the growing chain. The
converted to erythromycin A by downstream b-hydroxy fatty acid side-chain is incorporated
nonmodular processing enzymes. The modular at the beginning of the biosynthesis by an
polyketide synthase has a loading module (L) unknown mechanism. P, incorporation of b-
and six chain elongation modules (16) hydroxy fatty acid (not part of module); A,
expressed from three genes DEBS13. AT, acyl amino acid adenylation domain; T, thiolation
transferase domain; CP, acyl carrier protein; domain; C, condensation; E, epimerization
KS, ketosynthase domain; KR, ketoreductase domain; TE, thioesterase (cyclization) domain.
domain; DH, dehydratase; ER, enoyl reductase; (C) Biosynthesis pathway for d-desosamine,
TE, thioesterase (cyclization) domain. (B) The the deoxysugar of the secondary metabolites
modular peptide synthase-producing surfactin methamycin/neomethamycin.
35.2.1
Polyketide Combinatorial Biosynthesis
Polyketide biosynthesis has been the focus of intense study for several years. Sev-
eral excellent reviews summarizing the current state of knowledge in this area
have been published elsewhere [2023]. The enzyme systems responsible for pol-
yketide biosynthesis can be roughly assigned to four categories: the type I modular,
type I iterative, type II, and higher plant PKSs, each with dierent characteristics.
Type I modular PKSs are bacterial in origin and produce mostly macrolide-type
natural products, such as erythromycin, and have a separate catalytic module for
each chain elongation step (including modications along the way, such as reduc-
tions). Type I iterative PKSs, generally found in fungi, have only one module
which iteratively processes the polyketide chain through numerous steps [24]. In
35.2 Combinatorial Biosynthesis: Creation of Novel Small-molecule Natural Products 1067
these PKSs, some activities can be turned on and o at dierent points in the bio-
synthesis [25], and this process probably occurs through the intercession of
helper proteins outside the module [26]. The complexity of the type I iterative
systems has to this point precluded attempts to reengineer them for combinatorial
biosynthesis. Type II PKSs are multisubunit enzymes that usually make aromatic
polyketides, such as actinorhodin (see Fig. 35.3). Each subunit is responsible for
catalyzing a specic type of reaction. Finally, the higher plant PKSs utilize para-
coumaroyl-CoA as a starter unit and produce complex hydroxylated aromatics. In-
terestingly, they utilize a single enzyme active site for the chain elongation, de-
carboxylation, cyclization, and aromatization steps along the biosynthetic pathway,
1068 35 Reprogramming Combinatorial Biology for Combinatorial Chemistry
aromatic polyketides, some of which are shown in Fig. 35.4. Subsequent reports of
combinatorial biosynthesis using type II PKSs, while generally successful, showed
that the hybrid enzyme systems did not always perform as expected [35, 36], which
has reduced the ability to design combinatorial biosynthesis libraries from type II
PKSs with predictable structures. Additionally, the assignments of the roles of dif-
ferent parts of the minimal PKS in specifying chain length and point of initial
cyclization have been shown to be incorrect, at least for some type II PKSs [3739].
Despite this limitation, or perhaps because of it, several previously unknown aro-
matic polyketides have been created, sometimes with quite intricate and unexpected
structures (Fig. 35.5). Work on combinatorial biosynthesis of aromatic polyketides
1070 35 Reprogramming Combinatorial Biology for Combinatorial Chemistry
from type II PKSs continues, but has perhaps been overshadowed by develop-
ments in engineering type I modular PKSs for production of combinatorial bio-
synthesis libraries [14, 18, 20, 40, 41].
Fig. 35.7. Replacing the wild-type ery loading module with the
avermectin loading module, which displays reduced specicity,
results in iso-propyl- and sec-butyl-containing erythromycin
variants.
(Fig. 35.8) [5558]. For two or more plasmids (circular segments of DNA) to coex-
ist in a single cell, they must be compatible (containing compatible origins of rep-
lication). In the multiplasmid PKS approach, compatible plasmids are introduced
into an appropriate expression host, i.e. one that produces no polyketides, or has
had the polyketide-producing pathways it normally contains completely eliminated.
These plasmids contain dierent sets of genes coding for dierent PKS modules,
and each plasmid may harbor PKS genes that have mutations providing further
permutations. By introducing dierent combinations of plasmids through cotrans-
formation, each host cell has the capability to produce a dierent polyketide. Such
an approach has led to the simplied production of a library of novel polyketides
based around the DEBS architecture (Fig. 35.9) [55]. As the collection of suitable
PKS module-containing plasmids multiplies, the ability to produce more novel
35.2 Combinatorial Biosynthesis: Creation of Novel Small-molecule Natural Products 1073
Fig. 35.8. The multiplasmid approach to possible combinations are transformed into a
combinatorial biosynthesis. For each module a suitable host organism. Growth on all three
separate plasmid is used. Each plasmid also antibiotics ensures that each host cell contains
contains a dierent antibiotic resistance gene. all three types of modules.
Modications are made to each module, then
35.2.2
Combinatorial Biosynthesis of Nonribosomal Peptide Products
Nonribosomal peptides are produced over a wide range of species and probably
function in their native organisms mainly as signaling, siderophoric, and defensive
molecules [4]. In contrast to the PKSs, all known nonribosomal peptide synthases
(NRPSs) appear to utilize a modular structure. Several excellent reviews summarize
the state of knowledge concerning NRP biosynthesis from genetic, enzymologic,
and structural viewpoints [22, 6467]. NRP-containing products have a number
of useful medical and biological functions, including biocidal, antibiotic, anti-
cholesterolic, and immunosuppressive [8]. As with polyketide natural products,
the ability to produce new NRPs would thus be a boon to drug discovery. While the
35.2 Combinatorial Biosynthesis: Creation of Novel Small-molecule Natural Products 1075
35.2.3
Combinatorial Biosynthesis from Hybrid PKS/NRPS Systems
Fig. 35.10. (A) Surfactin A, which is produced domain immediately downstream of the
by the seven modules. Two new analogs were srfA-A gene (which encodes the rst three
produced by replacing the d-Leu module with amino acid incorporation modules) results in
phenylalanine and ornithine incorporation the production of the expected truncated
modules. (B) Placing the thioesterase peptide.
35.2 Combinatorial Biosynthesis: Creation of Novel Small-molecule Natural Products 1077
35.2.4
Combinatorial Biosynthesis of Carbohydrates
Deoxysugars are essential appendages for many secondary metabolites. Their bio-
synthesis has been reviewed extensively elsewhere [911, 82, 83]. In natural prod-
ucts they often act as guidance systems, determining the specicity and biological
activity toward a target [84, 85]. The ability to create diverse deoxysugars and incor-
porate them into larger structures provides an opportunity to devise natural prod-
uct analogs with altered targets, specicity, activity, and physical properties.
As indicated in Fig. 35.2C, three types of enzymes are necessary for the incor-
poration of deoxysugar moieties into larger natural products: (1) nucleotidyl-
transferases are responsible for adding nucleotide diphosphate appendages to car-
bohydrates; (2) modifying enzymes, which, for example, alter the oxidation state of
the molecule or replace hydroxyls with amines; and (3) glycosyltransferases, which
couple the fully derivatized sugar with its intended aglycon partners. To produce
novel deoxysugars, one can alter or relax the specicity of enzymes through the
introduction of mutations, replace a natural enzyme in one pathway with those
from other pathways, or take advantage of the relaxed specicity some of these
enzymes already demonstrated to incorporate non-natural substrates.
In one example, non-natural analogs of desosamine, the sugar moiety found
in methymycin and neomethymycin, were produced (Fig. 35.12). The desosamine
biosynthetic cluster was altered by deletion of the desV (reductive amination) and
desVI (dimethylation) genes [86, 87]. Several new macrolides were produced from
the deletion mutants, including some that were quite unexpected. In addition to
1078 35 Reprogramming Combinatorial Biology for Combinatorial Chemistry
producing potentially useful new macrolides, this work also demonstrates that the
glycosyltransferase involved has a relaxed substrate specicity, in that it could in-
corporate a variety of deoxysugars.
The ability to swap genes between pathways is another important demonstration
of the potential for deoxysugar combinatorial biosynthesis. In the methymycin
pathway, the desVI gene can be replaced without diculty by an analogous gene,
tylM1, that is involved in the biosynthesis of mycaminose in the tylosin pathway
(Fig. 35.13) [88]. Similarly, the tylB gene can replace desV [87]. The ease of such
replacements might suggest that very few proteinprotein interactions are essential
in the deoxysugar biosynthetic pathway, thus reducing the chances that communi-
cation problems may be stumbling blocks when mixing and matching deoxysugar
biosynthetic proteins in combinatorial biosynthesis.
The appending of alternative sugar moieties to aglycons for combinatorial bio-
synthesis also rests upon our ability to reengineer or utilize the existing reduced
specicity of the nucleotidyltransferases and glycosyltransferases involved. For
instance, the wild-type a-d-glucopyranosyl phosphate thymidylyltransferase (Ep )
shows very broad acceptance of sugar substrates for coupling to thymidylate (Fig.
35.14) [89, 90]. Structural analysis suggested that introducing specic point muta-
tions should further relax the specicity and better allow incorporation of deoxy-
sugars modied at C2, C3 or C6 [91]. The pooled Ep mutants were capable of
incorporating non-natural deoxysugars, including some that are unable to be in-
35.2 Combinatorial Biosynthesis: Creation of Novel Small-molecule Natural Products 1079
Fig. 35.14. (A) Reaction catalyzed by Ep , the pooled mutants (described in the text). WT
glycosylation of thymine by glucose phosphate. percentage, incorporation by the WT (wild
(B) Ep shows broad specicity for a variety of type) Ep ; Pool percentage, the eciency with
glucose phosphate analogs. Analogs with which the pool was able to incorporate the
numbers below them were evaluated by the sugar.
35.3 Other Combinatorial Aspects of Biology 1081
Fig. 35.15. (A) Mithramycin and urdamycin A, decient mutants of Streptomyces argillaceus
naturally produced by Streptomyces argillaceus and Streptomyces fradiae results in the
and Streptomyces fradiae respectively. (B) production of a variety of novel carbohydrate-
Expressing the elloramycin PKS genes in PKS- containing elloramycins.
35.3
Other Combinatorial Aspects of Biology
versity, and quality can be readily analyzed (most commonly via DNA sequencing)
[98, 99]. This approach is in contrast to the wholly synthetic production of combi-
natorial biosynthesis libraries, in which complex product mixtures must be con-
tended with, or sophisticated spatial addressing methods must be employed, if one
is to productively screen compounds for activity. Second, several methods have
been developed for the in vivo or in vitro selection or screening from DNA-encoded
libraries, and so complex biopolymer mixtures can be screened or selected en masse
and readily separated into active and inactive components [100]. Those members
of biopolymer libraries with interesting properties can be easily identied by se-
quencing of the DNA template, and further optimization of a hit in a biopolymer
library can be readily undertaken via simple mutagenesis and directed evolution
methods.
While combinatorial biosynthesis certainly contains potential for the drug dis-
covery process through the generation of novel natural products, it has also been
very useful as a tool for studying the enzymology and organization of the biosyn-
thetic systems involved. Combinatorial libraries of DNA, RNA, small peptides, and
proteins may also one day be very useful for pharmaceutical or medicinal purposes,
but they are currently used mostly for the study of catalysis, folding, and molecular
recognition. What follows are brief discussions of how these combinatorial libraries
are currently being created and applied.
35.3.1
Combinatorial Libraries of Random DNA and RNA
Oligonucleotide molecules with a wide variety of functions have been selected from
RNA or DNA combinatorial libraries. Such libraries are routinely generated with as
many as 10 15 unique members. Because the structure, function, and genetic infor-
mation is self-contained for each molecule, the in vitro recovery and amplication
of rare genes with a desired phenotype is straightforward and very ecient. Con-
sequently, combinatorial libraries of oligonucleotides can be easily subjected to
multiple cycles of selection and enrichment.
The selection of RNA aptamers with a broad range of receptor-like activities
from large combinatorial RNA libraries has been particularly successful [101104].
Using such an approach, and a number of clever in vitro selection methods, RNA
aptamers have been identied with binding ability for small molecules [105109],
sometimes with very high specicity [110, 111]. Aptamers have also been selected
from large oligonucleotide combinatorial libraries with anity to a number of
medically and biologically important protein targets, including the CD4 antigen
[112], surface proteins specic to African trypanosomes [113], the cellular prion
protein PrP c (but not the infectious scrapie prion form PrP Sc ) [114], the hepatitis
C virus protein 3 protease [115], the human immunodeciency virus (HIV) TAT-I
protein [116], and human angiogenin [117].
RNA and DNA catalysts, termed ribozymes or DNAzymes, have also been se-
lected from combinatorial ensembles for a wide array of reactions [118122], in-
cluding oligonucleotide modications [123, 124], alkylation and acylation reactions
35.3 Other Combinatorial Aspects of Biology 1083
[125, 126], peptide bond formation [127], and DielsAlder cycloadditions [128].
While these ribozymes are often poorly active, in some cases very high levels of
catalysis have been achieved through the further optimization of an initial lead
[129].
The direct selection of highly ecient self-ligating ribozymes (i.e. phosphoryl
transferases) from large RNA combinatorial libraries is one of the most impressive
accomplishments thus far in the eld of nucleotide-based catalysis (Fig. 35.16)
[130]. A library of @10 15 dierent RNAs with 220 random nucleotides was con-
structed, and from this ensemble roughly 1 in 10 13 members had the desired ligase
activity. Further cycles of diversication and selection were then used to evolve the
initially selected group of ligase ribozymes, and one of these was subsequently
1084 35 Reprogramming Combinatorial Biology for Combinatorial Chemistry
35.3.2
Peptide Combinatorial Libraries
Short peptides are good inhibitors for a number of enzymes, and are thus suitable
lead structures for developing potent nonpeptide inhibitors, which may have in-
creased bioavailability and half-lives in vivo. Peptide ligands and inhibitors are also
valuable for collecting a wealth of information about the structure, specicity, and
size of receptor binding pockets and enzyme active sites. Since it is often dicult
to predict which peptide sequence will have optimal inhibitory activity, selection
from combinatorial peptide libraries is attractive. Additionally, the selection results
may reveal new, previously unidentied components of the targeted pathway, which
may, in turn, represent formidable targets for inhibition by small organic mole-
cules.
Because short peptides are often quickly degraded in the cell, peptide libraries
for in vivo selection are usually displayed inside a larger protein scaold. For ex-
ample, combinatorial hexadecameric peptide libraries were displayed (by insertion
at the genetic level) into a surface loop of a biologically inert carrier protein (an
inactive mutant of staphylococcal nuclease) [132]. From a @10 6 -member yeast-
transformed library, three peptides that inhibited the spindle checkpoint and 29
peptides that inhibited a mating pheromone signaling pathway (29 peptides) were
identied by in vivo genetic selection in cleverly engineered yeast strains. The pu-
tative in vivo targets were subsequently identied using yeast two-hybrid analysis
[133] and genetic dissection of the target pathways. A similar in vivo genetic selec-
tion approach was recently used to identify members from a combinatorial non-
apeptide library that block intracellular dimerization of HIV-1 protease, albeit in E.
coli [134].
Peptide combinatorial libraries can also be evaluated in vitro for binding or
inhibitory activity. As opposed to in vivo peptide selections, in vitro selection has
three particular advantages. First, no carrier protein is required to protect the short
peptides from in vivo degradation. Of course, this freedom also increases the con-
formations available to the peptide, which may reduce the anity of the peptide to
a potential target [135, 136]. Secondly, much larger (@ 10 13 ) peptide libraries can
be assayed in vitro than in vivo (maximally 10 9 , which is the practical upper limit
because of plasmid transformation eciencies [137, 138]). Finally, unnatural amino
35.3 Other Combinatorial Aspects of Biology 1085
acids can be utilized in the peptide combinatorial libraries through the use of sup-
pressor codons or other strategies [139142].
Perhaps the most powerful method for in vitro selection of peptide combinato-
rial libraries is mRNA display, which works by utilizing direct covalent attachment
of each member of the peptide combinatorial library to its encoding mRNA (Fig.
35.17) [143, 144]. Such covalent constructs are formed by in vitro translation of 3 0 -
puromycin mRNA templates. After mRNA translation, puromycin enters the pep-
tidyl transferase site of the ribosome and is covalently attached to the C-terminus
of the nascent peptide. The peptide moiety of the resulting mRNApeptide conju-
gate can then be selected as described for the other formats.
In a recent example, 20 dierent peptide aptamers to streptavidin were identi-
ed from a combinatorial library of 10 13 dierent 88 amino acid sequences [145].
These aptamers had binding anities (KD 510 nM) that were three orders of
magnitude better than a commercially available streptavidin-binding decapeptide
(KD 1372 mM). Truncation analysis showed that one of these peptides could
be reduced to only 38 amino acids, and also indicated that the sequence motif of
histidineprolineglutamine (HPQ) was essential for binding activity (as in other
peptides with anity for streptavidin [146]). Interestingly, all 20 aptamers that
were identied were not from the intended reading frame but from the third read-
ing frame that results from a two-base insertion. This outcome is probably because
the HPQ motif occurs more frequently in the third reading frame.
35.3.3
Protein Combinatorial Libraries
functional protein from rst principles is not yet an achievable goal. As an alter-
native, researchers have begun producing combinatorial libraries of proteins with
varying levels of randomization, and then screening or selecting from these libraries
for valued function.
Given the size of typical protein domains, fully randomized protein combinatorial
libraries are, for practical purposes, innitely large. For a 100-amino-acid protein,
for example, there are 20 100 (10 130 ) possible sequences, and it is thus impossible
to examine a substantial portion of this diversity. To evaluate as large a fraction as
possible, researchers have turned to in vitro selection methods (discussed in Sec-
tion 35.3.2) such as mRNAprotein fusions. Using the same mRNA display system
outlined in Section 35.3.2, but selecting for protein behavior instead of peptide
behavior, Keefe and Szostak recently identied four ATP binders starting from a
library of 6 10 12 proteins in which 80 contiguous amino acids were randomized
(Fig. 35.18) [147]. One of these variants showed very high anity (KD 5 nm) and
specicity for ATP, and was capable of being further truncated to a protein of only
45 residues. Like most in vitro selection strategies, the ATP-binding proteins were
identied based on their ability to elicit binding responses, but it is possible to
select proteins on the basis of their catalytic activity [148150], which may also
allow the directed evolution of novel enzymatic activities.
Fully randomized combinatorial protein libraries are of limited practical utility,
since the frequency of foldable, much less functional, sequences in them is ex-
pected to be very low. As an alternative, partially randomized protein combinatorial
libraries can provide the same benets of randomization but in a sequence that is
predicted to fold into a stable structure. For example, the utilization of basic struc-
tural information, such as the sequence preferences of helices and sheets, termed
35.3 Other Combinatorial Aspects of Biology 1087
placement of two loop sequences, and then combinatorial protein libraries were
created from the remodeled protein by randomizing positions in the loops. In vivo
genetic selection was used to identify members from the combinatorial library
capable of behaving as PRAIs.
To further probe the mutability of the (a/b)8 scaold, Harbury and coworkers
created combinatorial protein libraries by randomizing portions of an (a/b)8 barrel
triose phosphate isomerase (TIM) (Fig. 35.21) [163]. They then selected for mem-
bers of the libraries that were capable of complementing a TIM-decient strain.
Their results suggest that the (a/b)8 barrel is quite mutable in some of the loop
regions, which would be the probable site for introducing combinatorial random-
ization for the creation of new catalytic activities. This study has set the platform
for the rational design and creation of further combinatorial libraries from the
(a/b)8 barrel from which new function can be selected.
A third type of protein combinatorial library can be created through minor
changes designed to optimize or alter activity or to improve physical properties,
and then using selection or screening techniques to identify members of the library
with the desired characteristic. Such a procedure is usually called directed evolution,
and has been reviewed extensively elsewhere [164170]. These directed evolution
protein libraries could be considered analogous to lead optimization combinatorial
libraries that are created to nd an initial hit and optimize the scaold for use as a
drug.
In directed evolution experiments, mutations are introduced by error-prone PCR
[171] or DNA shuing [172, 173] and related techniques [174, 175]. After a suit-
able selection step, selected library members can be recombined to blend additively
positive mutations with the selection procedure carried out on the further evolved
library (Fig. 35.22). Using such a strategy, a number of proteins have been further
evolved to improve the activity [176], alter the selectivity [177190], or increase the
thermal stability [191, 192]. In some cases, even multistep processes utilizing sev-
1090 35 Reprogramming Combinatorial Biology for Combinatorial Chemistry
35.4
Perspectives
use. For instance, combinatorial biosynthesis holds great potential for the produc-
tion of potential lead compounds for drug discovery. These leads could then be
further optimized by more traditional synthetic and combinatorial chemistry.
Additionally, combinatorial biosynthesis could be used to produce starting material
intermediates (perhaps for further combinatorial chemistry) and nal products
that are not easily made using current synthetic methods. One example of this
potential utility is the use of a thioesterase domain from the tyrocidine NRPS for
catalyzing the cyclization of a variety of short, synthetic peptides [194, 195].
Combinatorial chemistry could play similar roles for combinatorial biology. For
instance, an attractive scaold that is identied through combinatorial chemistry
could also act as the lead for which a more complexly functionalized (and syn-
thetically challenging) combinatorial biosynthesis library could be prepared. Deoxy-
sugar libraries prepared by combinatorial chemistry could be used as feeding pre-
cursors for more complex polyketide- or NRP-based combinatorial biosynthesis
libraries utilizing low-specicity glycosyltransferases.
Peptide combinatorial libraries, as mentioned above, are very attractive for pro-
ducing initial leads for drug development, as well as for providing insight into the
structure, activity, and specicity of biological receptors and enzymes. But their
poor behavior in vivo often precludes their use as eective drugs. Combinatorial
chemistry is thus essential for producing nonpeptide drugs based around the ini-
tial peptide leads. Clearly, combinatorial biology will never supplant combinatorial
chemistry; rather, both elds will continually augment each other.
Large-molecule combinatorial libraries of proteins and DNA or RNA may also
nd eventual use as drugs, but currently they are of interest to researchers mainly
for their ability to probe issues of structure and catalysis. In addition to this, how-
ever, the production of such libraries has dramatically increased our knowledge
and ability to engineer and develop suitable selection and screening systems, some
of which have found their way to the combinatorial chemistry bench. Additionally,
our ability to engineer and create proteins or oligonucleotides with desired func-
tion may one day provide combinatorial chemists with another catalytic repertoire,
one that is based around environmentally friendly enzymatic reactions. Current
work on structural redesign of nucleotidyltransferases [91], and on the utilization
of glycosyltransferases with relaxed specicity [93], is a step in this direction.
What does the future hold for our utilization of combinatorial biology for human
benet? Given the promise of the multiplasmid approach in polyketide combina-
torial biosynthesis, and the predictive nature of modular biosynthetic systems, it is
conceivable that we will someday have the ability to produce virtually any polyketide-
or NRP-derived product, appended with appropriate deoxysugars, in high yield,
merely by transforming the right combination of plasmids into a suitable produc-
tion strain. It is also likely, given our ever-increasing understanding of protein
structure and function, that we will someday be able to design and engineer novel
enzymes with catalytic activities that are useful for applications in chemistry (in-
cluding combinatorial chemistry) or medicine. Additionally, our ability to rapidly
select useful or interesting variants from combinatorial libraries steadily improves.
Coupling binding or catalytic activity with genotypic information is currently pos-
1092 35 Reprogramming Combinatorial Biology for Combinatorial Chemistry
sible with DNA-encoded libraries, and the same linkage may someday also be pos-
sible with natural product combinatorial biosynthesis libraries.
With the advent of the genomic, or postgenomic era, there will be a ood of
potential new medical targets. The ability to utilize biological tools can only help
our combinatorial eorts in designing drugs to control them.
Acknowledgments
The helpful suggestions and support of Professor Don Hilvert during the prepara-
tion of this chapter are gratefully acknowledged, as well as Kinya Hotta and Gavin
Macbeath for the preparation of graphics.
References
Index
chiral auxiliary 234, 232, 120, 446, 928 critical temperatures 1025
chiral catalyst 991 Crixivan2 912
chiral Lewis acids 450 cross metathesis 94, 587
chiral pentamines 912 cross-coupling reactions 86, 531
chiral separation methods 7 crosslinked (insoluble) polymers 24
chiral stationary phases (CSPs) 921, 1046 crosslinked acrylamides 41
chiral synthon chiron 910 crosslinked PEGs 44
chirality multiplication 906 crosslinked polymer beads 27
2-chlorotrityl resin 67 crosslinked-ethoxylate acrylate resins
chorismate mutase 1087 (CLEAR) 43
Christ vacuum centrifuge 222 crosslinkers 25, 32
chromaticity 1030 cross-metathesis 664, 929
chromatography 284 crystallization 212
chromenones 75 C2 -symmetric chiral ligands 993
chromium arene complexes 108 C3 -symmetric triarylphosphines 920
cinnolines 86 cuprates 482
Ciprooxacin2 262 Curtis rearrangement 336
circular dichroism (CD) 1049 Curtius degradation 356
detection 966 2-cyanoethylphosphate protecting group 291
Claisen 515 4-cyanotrityl protection group 68
condensations 515, 792 cyanuric chloride 255
reaction 791 cyclative cleavage 103, 526, 629, 631, 633,
rearrangements 465 654, 789
Claisen-Schmidt reaction 661 to hydantoins 791
clearance 744, 752 cyclic alkenes 588
clinical candidates 10 cyclic amide 143
clinical research 12 cyclic peptides 592, 594
C log P 726 cyclization reaction 314
clustering 732 5-exo cyclizations 228, 232
cobalt enolates 521 2+2 cycloaddition 463, 586
b-C,O bond scission 292 3+2 cycloadditions 453
cocatalysts 992 4+2 cycloadditions 441
colorimetric spot test 921 6+3 cycloaddition 464
colossal magnetoresistance 1034 cyclo cleavage 789
CombiFlash TM from Isco 212 cycloaddition 649, 651, 698, 701
combinatorial antibody library 892 cycloaddition imine 459
combinatorial biosynthesis 1064, 1075 cycloading 636
combinatorial catalysis 886f, 927, 1019 cyclobutanones 464
combinatorial libraries 764 cyclofragmentation 295
combinatorial semisynthesis 614 cyclopentenones 1010
combinatorial total synthesis 626 cyclorelease 631, 633
composition conditions 1020 cycloreversions 299, 676
composition spread technique (CCS) 1035 cyclotriphosphazene 49
computer-controlled multiplexer 969 cysteine 289
condensation reactions 685 cytochrome P450 753
contamination 94 cytochrome P450 oxygenase 904
continuous composition spread (CCS) 1022 cytomegalovirus 628
controlled pore glass (CPG) support 46
copper alkynes 329 Danishefskys diene 298, 325, 441, 480, 909,
Coreys chiral oxazaborolidine catalyst 928 922
coumarin 519 DART (development automated reaction
para-coumaroyl-CoA 1067 toolkit) 833, 870
coupling reagents 347f data-handling system 854
C-radicals 231 daughter libraries 925
Index 1103
ring-closing metathesis 587, 600, 629, 659, self metathesis 587, 589
662 self-ligating ribozymes 1083
ring-closing reactions 254 self-similarity 763
ring-opening cross-metathesis 603 SEM linker 75, 83, 73, 144
ring-opening metathesis polymerization semiconductor gas sensors 972
(ROMP) 587, 595 sensing chemistry 894
Rink resin 136, 123, 67 serial processing 14
RNA aptamers 1082 serine proteases 769
RNA combinatorial libraries 1083 Sharpless asymmetric dihydroxylation 311
robot-arm-based systems 190 shelf life characterization 806
ROMP spheres 36 shikimic acid 455, 635
ROMP-based polymers 51 SiaH insertion reaction 926
route scouting 834, 840, 869, 869 sialyl trimetric Lewis X 713
RP-HPLC 214 sigmatropic rearrangements 465
Rutherford backscattering spectroscopy 1035 silyl amide linkers (SAL) 289
silyl enol ethers 441, 498, 505f, 520f
SAC linker 66, 75 silyl ketene acetals 505, 525
Saccharopolyspora erythraea 1079 silyl linkers 73
safety-catch linkers 73, 77, 80, 90, 99, 624 silylated resins 73
safety-catch procedure 444 similarity 770
safety-catch resin 282 similarity radius 734
Sakurai reaction 653, 327 similarity searching 763
SAL linker 66, 75 single compounds 3
salicylaldiminato ligands 934 size exclusion chromatography 47
samarium iodide 229, 284 sol-gel processing 1026, 940
sample logistics 219 solid-phase Bischler-Napieralski reaction 277
Sandmeyer reaction 129 solid-phase extraction (SPE) 203, 208, 211,
sarcodictyns 375, 620 908
SASRIN resin 63, 650, 65 solid-phase Friedel-Crafts acylation 272
Savant vacuum centrifuge 222 solid-phase iodination 274
Sc(OTf )3 502, 505, 521 solid-phase Mitsunobu reaction 253
SCAL (safety-catch) 67 solid-phase phenolate alkylation 250
scanning lectron microscope (SEM) 33 solid-phase Williamson ether synthesis 249
scanning electrochemical microscope 952 solid-state materials 1019
scanning evanescent microwave microscope solubility 727, 744, 748f, 755
(SEMM) 1031 soluble polymeric supports 46
scanning-tip microwave near-eld microscope soluble polystyrene polymer 626
1036 solution deposition methods 1022
scavenger resins 20, 63, 212, 452 Sonogashira reaction 329, 544, 688, 297
Schi base ligands 908 spacer group 61
screening intensity 839 spatial encoding 170
screening multisubstrate/one catalyst 927 spatially addressable electrolysis platform
screening strategies 851 (SAEP) 952
screening technologies 1017 spin eld eect transistors 1033
secondary amine binary coding 175 split-and-combine concept 4, 5, 17, 176, 886
selection algorithms 732 sporochnol 1009
selenenyl bromide 623 sputtering 1022
selenides 237, 286 srfA-A 1075
selenium 77 stannane-based linkers 81, 77
selenium bromide 636 Staphylococcus auerus 623
selenium reductions 397 statistical design 852
selenium-based linkers 81, 398 Staudinger reaction 407, 463, 650
selenol 293 stereoselective alkylations 510
selenoxides 286 enolates 510
Index 1113
Yb resin 298
UDP-galactose 714
Yb(OTf )3 503, 657, 670
Ugi reactions 409, 599, 601, 697, 699, 299,
yield determination 200
461, 487
yohimbinic acid 615
Ullmann aryl ether-forming reactions 925
ytterbium triate 451f, 456
ultraltration 48
ultrasound 85
zearalenone 474
g,d-unsaturated acids 231
(S)-zearalenone 630
a,b-unsaturated ketones 441, 494, 503
Zenyx Magellan synthesizer 205
urazines 451
zeolite synthesis 1026
urdamycin A 1079
zinc borohydride polymers 416
UV lasers 964
zinc organyles 328
UV-excited photoluminescence 1030
Zinsser Sophas 206
zirconium borohydride polymers 416
vacuum centrifuges 221 Zn enolates 494, 526
vancomycin 69, 110, 590, 623f zoom screens 949
vapor deposition techniques 1022 Zr-catalyzed 1004, 1006
Varian-HPLC 215 zymark 199