TOWARD A COMPREHENSIVE UNDERSTANDING OF

MICROBIAL METABOLISM

YIFAN XU

A DISSERTATION
PRESENTED TO THE FACULTY
OF PRINCETON UNIVERSITY
IN CANDIDACY FOR THE DEGREE
OF DOCTOR OF PHILOSOPHY

RECOMMENDED FOR ACCEPTANCE

BY THE DEPARTMENT OF
CHEMISTRY
ADVISOR: JOSHUA D. RABINOWITZ

SEPTEMBER 2013
© Copyright by Yifan Xu, 2013. All rights reserved.
 Abstract

A comprehensive understanding of metabolism remains challenging. Even in the best understood model

microbes, some pathways remain ill-defined. For those best defined pathways, their regulation remains

incompletely understood. New tools that allow direct measurement of metabolites and their fluxes by

liquid chromatography-mass spectrometry hold the potential to address these limitations. Using a

combination of metabolomics, genetics, proteomics, biochemistry and modeling, metabolism and its

regulation of model bacterium Escherichia coli and yeast Saccharomyces cerevisiae were investigated.

For example, nucleotide degradation is a universal metabolic capability of any organism. However, the

involved pathway was poorly characterized. Herein, a yeast protein not previously associated with

nucleotide degradation, Phm8, was found to convert nucleotide monophosphates into nucleosides. A

carefully mapping of the downstream steps showed that this pathway eventually salvages carbons into

the pentose phosphate pathway. Deletion of Phm8 or downstream steps of this pathway resulted in

metabolite depletion and impaired survival of starving yeast.

Glycolysis is the best-studied metabolic pathway and its regulation has been extensively characterized.

The fate of the last intermediate of glycolysis, phosphoenolpyruvate (PEP), controls much of cellular

metabolism, e.g. the balance of glycolysis and gluconeogenesis. In both E. coli and yeast, removal of

glucose results in a paradoxical increase in PEP, which goes up the most of any canonical metabolite.

The switch-like inhibition of the PEP consuming enzymes cannot be explained by previously emphasized

regulations. In contrast, the heretofore under-appreciated allosteric regulation predominates in both

organisms, with PEP consumption activated in an ultrasensitive manner by the upstream glycolytic

intermediate, fructose-1,6-bisphosphate. Mutations that eliminate this regulation do not impair growth on

steady glucose, but they render microbes defective in gluconeogenesis and in growth in an oscillating

glucose environment. Thus, microbial central metabolism is intrinsically programmed with ultrasensitive

feed-forward regulation that enables rapid adaptation to changing environmental conditions.

i
 Acknowledgments

First I would like to thank my advisor, Professor Joshua Rabinowitz, for giving me the opportunity to work

in his lab. None of my work could possibly be done without his thoughtful ideas, insights and creativity. He

was always supportive and patient in helping me improve my scientific writing, presentation and providing

me career advice. I was also given freedom to think independently and pursue my own ideas, which

helped me to learn a variety of aspects of metabolism, biochemistry and genomics.

I would like to thank the Rabinowitz lab: Dr. Wenyun Lu was an excellent mentor for me on analytical

chemistry. Without his hard work on solving instrumental problems and helping develop high-throughput

methods, none of the work in this study could have been rapidly finished. Dr. Daniel Amador-Noguez and

Dr. Marshall Reaves taught me experimental techniques and helped me all the way along my thesis. I

would like to thank Jing Fan, for helping me greatly on computation and programming, while providing

insights in my research projects. I would like to thank other researchers in our microbial subgroup: Dr.

Patrick Bradley for system biology, Dr. Xiaoyang Su for the help on the yeast projects and biochemistry,

Sean Hackett for chemostat and statistics, Sarah Johnson for yeast metabolomics, Junyoung Park for

modeling work and helpful discussions, David Glass for kinase signaling, David Thomas for pentose

phosphate pathway regulation, Jonathan Chen for polyol metabolism; these people provided all the

routine help in my research. I also owe a debt of thanks to all other Rabinowitz lab members, Prof. Tomer

Shlomi, Dr. Jurre Kamphorst, Dr. John Purdy, Sisi Zhang, Ling Liu, Michel Nofal, Xin Teng, Zheyun Zhang

for their support on my research, and our lab assistant Marybeth Fedele who helped me greatly on

technical details; and former members: particularly Drs. Jie Yuan, Eugene Melamud, Bryson Bennett,

Chris Doucette, Sophia Lunt, Michelle Clasquin, Livia Vastag, Ludmilla Artistilde, Christopher Crutchfield,

Meytal Higgins, Scott Breunig and Kellen Olszewki for their kind help on my research.

I would like to thank my other mentors greatly supporting this study: Prof. James Broach and Prof. David

Botstein. Both of them patiently instructed me on my research projects, especially on yeast genetics and

microbial physiology. I also benefited a lot by directly participating their lab meeting and working in their

ii
labs. Particularly, I want to thank Broach lab members Drs. Xin Zhao and Farnaz Absalan and Bostein lab

members Drs. Pat Gibney and Sandy Silverman for helping me with yeast assays and strain

constructions. And all the other members in the labs provided me helpful feedbacks.

I would like to thank my colleagues in Princeton, particularly Dr. Alison Hottes and Prof. Saeed Tavazoie

for E. coli strains, Dr. Allison Michaelis, Max Wilson and Prof. Zemer Gitai for helping me on microscopy,

Dr. Megan McClean for microfluidics, Prof. Tom Muir, Prof. Ned Wingreen, Harish and Prof. Jannette

Carey for helpful discussions. I would also like to thank the mass spec facility: Dr. David Perlman

provided kind help and instructed me on proteomics; Kyin Saw and Henry Shwe helped with sample

processing. I also want to thank Dr. Istvan Pelczer for helpful instructions on NMR, Meghan Krause and

Sallie Dunner for their patient help in my visa application, exam and thesis preparations.

My collaborators also provided me with great help and inspired me in my research, in particular: Prof.

Alexander Yakunin provided his valuable insights in biochemistry and crystallography, Prof. Amy Caudy

provided great help both experimentally and intellectually. Prof. Fabien Letisse started various exciting

projects which inspired greatly my later research.

I would also like to thank my committee members, Prof. John Groves, Prof. Michael Hecht for giving me

valuable advice and feedback along my thesis and Prof. Dorothea Fiedler for reading my thesis and

providing me with valuable insights and feedback.

Finally, I would also like to thank my family and friends. I specially thank my wife, Ying Mao for all of her

great support and tolerance, especially in our very busy time at the birth of our baby in 2012; my parents

and parents-in-law especially my mother-in-law Ping Yu for all the assistance and encouragements in my

life.

iii
 Contents

Abstract........................................................................................................................................ i
Acknowledgments ....................................................................................................................... ii
Contents .................................................................................................................................... iv
List of Figures ............................................................................................................................. v
List of Tables ........................................................................................................................... viii
Introduction ................................................................................................................................ 1
Chapter 1 Ultrasensitive regulation of anapleurosis in Escherichia coli ...................................... 7
1.1 Introduction....................................................................................................................... 8
1.2 Results ............................................................................................................................. 9
1.3 Discussion .......................................................................................................................37
1.4 Methods ..........................................................................................................................39
Chapter 2 Regulation of yeast pyruvate kinase by ultrasensitive allostery independent of
phosphorylation.........................................................................................................................53
2.1 Introduction......................................................................................................................54
2.3 Discussion .......................................................................................................................77
2.4 Methods ..........................................................................................................................79
Chapter 3 Nucleotide degradation and ribose salvage in yeast .................................................99
3.1 Introduction....................................................................................................................100
3.2 Results ..........................................................................................................................104
3.3 Discussion .....................................................................................................................122
3.4 Methods ........................................................................................................................125
Chapter 4 Identification of a yeast polyol phosphatase required for maintaining phosphoglucose
isomerase activity ...................................................................................................................128
4.1 Introduction....................................................................................................................129
4.2 Results ..........................................................................................................................131
4.3 Discussion .....................................................................................................................144
4.4 Methods ........................................................................................................................146
References .............................................................................................................................151

iv
 List of Figures
Figure 1.1. Glucose removal results in PEP build-up. ................................................................................ 10
Figure 1.2. Absolute intracellular concentrations of key metabolites during glucose removal and its
replacement with no carbon, acetate, succinate, or glycerol. ..................................................................... 11
Figure 1.3. Heat map of all measured metabolites during glucose removal and its replacement with no
carbon, acetate, succinate, or glycerol. ...................................................................................................... 12
Figure 1.4. Upon glucose removal, decreasing ATP/ADP and NADH/NAD+ ratios drive carbon towards
the bottom of glycolysis (i.e. towards PEP). ................................................................................................ 13
Figure 1.5. Accumulated PEP is not from gluconeogenesis or the PEP-glyoxylate cycle in the first 45
minutes after acetate switch. ...................................................................................................................... 14
Figure 1.6. PEP carboxylase is inhibited by 99% upon glucose removal. .................................................. 14
Figure 1.7. The primary PEP consuming enzymes are the PTS and PEP carboxylase. ........................... 15
Figure 1.8. Metabolic Flux Analysis of wild type and ΔpykA/ΔpykF strains indicates pyruvate kinase flux
accounts for < 5% of PEP consumption in wild type cells. ......................................................................... 16
Figure 1.9. Deletion of pyruvate kinase (both isozymes, pykA and pykF) does not substantially alter E. coli
growth or PEP concentration. ..................................................................................................................... 17
Figure 1.10. Impact of pyruvate kinase deletion (both isozymes, pykA and pykF) on E. coli metabolome
response to glucose removal. ..................................................................................................................... 18
Figure 1.11. Transient labeling experiment as per Figure 1.6, except with shorter labeling duration. ....... 19
Figure 1.12. Fit of experimental data to equations used for quantitation of PEP consumption during
carbon starvation and acetate switch based on transient labeling experiments ......................................... 20
Figure 1.13. Accumulated aspartate is not from anapleurosis. ................................................................... 21
Figure1.14. There is no degradation or stable, inactivating post-translational modification of PEP
carboxylase in the first 10 minutes following glucose removal. .................................................................. 22
Figure 1.15. Activation of PEP carboxylase by FBP is enhanced by physiological levels of acetyl-CoA and
aspartate. .................................................................................................................................................... 23
Figure 1.16. The enhancement of FBP activation by acetyl-CoA and aspartate was also found when
effectors were added at their physiological concentrations in glucose minimal media and carbon
starvation conditions. .................................................................................................................................. 24
Figure 1.17. Inferred PEP carboxylase activity in live cells following glucose removal. ............................. 25
Figure 1.18. Ultrasensitive regulation of PEP carboxylase is required for PEP to increase when glycolytic
influx decreases. ......................................................................................................................................... 26
Figure 1.19. PEP accumulation upon glucose removal enables rapid import of glucose when it reappears.
.................................................................................................................................................................... 27
Figure1.20. PEP carboxylase is not activated by glucose-6-phosphate or fructose-6-phosphate. ............ 28
Figure1.21. Determination of PEP carboxylase wild type and R313Q kinetic parameters. ........................ 29

v
Figure 1.22. Expression of PEP carboxylase R313Q ablates the PEP spike while FBP decreases as
normal. ........................................................................................................................................................ 30
Figure1.23. Heat map of all measured metabolites in the Δppc/pCA24N-ppc (PPC WT) and
Δppc/pCA24N-ppcR313Q (PPC R313Q) during glucose removal and its replacement with no carbon or
acetate......................................................................................................................................................... 31
Figure1.24. The reserve of PEP is required for rapid initiation of anapleurosis upon glucose re-addition. 33
Figure1.25. Quantitation of PEP carboxylase protein concentration in cells growing on glucose or acetate,
with or without inducible expression from plasmid with IPTG. .................................................................... 35
Figure 1.26. The FBP-insensitive variant of PEP carboxylase causes futile cycling on gluconeogenic
media........................................................................................................................................................... 36
Figure 1.27. LC-high resolution MS analysis and LC-triple quadrupole MS/MS analysis yield similar data.
.................................................................................................................................................................... 41
Figure 2.1. Glucose removal results in PEP accumulation. ........................................................................ 58
Figure 2.2, Moving a filter culture to a glucose-free plate is similar to downshift in liquid culture to ~ 0.01%
glucose. ....................................................................................................................................................... 59
Figure 2.3. Accumulated PEP does not derive from TCA cycle intermediates. .......................................... 60
Figure 2.4. Transcriptome and metabolome response to inhibition of glucose-sensing kinases. .............. 62
Figure 2.5, Inhibition of PKA and SCH9 slightly impairs growth but does not change the level of PEP level.
.................................................................................................................................................................... 64
Figure 2.6, Yeast expressing genomically His-tagged pyruvate kinase show the wild-type metabolome
response to glucose removal. ..................................................................................................................... 67
Figure 2.7. PEP accumulation upon glucose removal does not require Cdc19 covalent modification....... 68
Figure 2.8, Ultrasensitive regulation of pyruvate kinase is required for PEP to increase when glycolytic
influx decreases. ......................................................................................................................................... 69
Figure 2.9. Pyruvate kinase exhibits cooperative activation by FBP independent of its phosphorylation
status. .......................................................................................................................................................... 70
Figure 2.10. Allosteric regulation of Cdc19 by FBP is essential for glucose removal-induced PEP
accumulation. .............................................................................................................................................. 71
Figure 2.11, Yeast gluconeogenic growth requires pyruvate kinase regulation by FBP and not by
phosphorylation. .......................................................................................................................................... 73
Figure 2.12. Cdc19 allosteric regulation facilitates energy charge homeostasis upon glucose up-shift and
enhances growth on oscillating glucose. .................................................................................................... 75
Figure 2.13, Cdc19 allosteric regulation facilitates adenylate nucleotide homeostasis during glucose re-
addition and thus growth on oscillating glucose. ........................................................................................ 76
Figure 3.1. Yeast pathway map of nucleotide degradation and ribose salvage. ...................................... 102
Figure 3.2. Starvation triggers accumulation of nucleosides, bases and S7P. ......................................... 105
Figure 3.3. In carbon starvation, nucleosides and PPP compounds are produced by RNA degradation via
autophagy in response to kinase signals. ................................................................................................. 107
Figure 3.4. Macroautophagy triggers RNA degradation ........................................................................... 108

vi
Figure 3.5. Phm8 is the physiological yeast nucleotidase. ....................................................................... 110
Figure 3.6. Confirmation that Pnp1 is the physiological purine nucleoside phosphorylase, Urh1 is the
pyrimidine hydrolase, and Prm15 (Pgm3) is the phosphoribomutase. ..................................................... 113
Figure 3.7. Clarification of downstream steps in ribose salvage pathway ................................................ 114
Figure 3.8. Nucleoside degradation causes S7P accumulation due to depletion of the other transaldolase
substrate glyceraldehyde-3-phosphate, and SBP accumulation due to phosphorylation of S7P into SBP
by Pfk1. ..................................................................................................................................................... 116
Figure 3.9. SBP was produced from S7P. ................................................................................................ 117
Figure 3.10. pfk2 deletion doesn’t show changes of SBP and S7P.......................................................... 118
Figure 3.11. The ribose salvage pathway is essential for yeast’s survival in starvation ........................... 120
Figure 3.12. The accumulation of S7P in carbon starvation facilitates survival of subsequent oxidative
stress ......................................................................................................................................................... 121
Figure 4.1. Metabolomic phenotype of pyp1 deletion cells ....................................................................... 132
Figure 4.2. Formulae deduction of accumulated metabolites in pyp1 deletion strain. .............................. 133
Figure 4.3. Pyp1 is active against compounds with D-glycerol-3-phosphate tail. ..................................... 134
Figure 4.4. Relative level of ribitol-5-phosphate and ribulose-5-phosphate/xylulose-5-phosphate in the
wild type and zwf1 deletion strain. ............................................................................................................ 136
Figure 4.5. pyp1 deletion cells accumulated sorbitol-6-phosphate which inhibited phosphoglucose
isomerase and resulted in growth defect. ................................................................................................. 138
Figure 4.6. Sorbitol-6-phosphate and ribitol-5-phosphate were inhibitors of phosphoglucose isomerase
.................................................................................................................................................................. 140
Figure 4.7. Pyp1’s maintenance on Pgi1’s activity is more important in higher growth rate. ................... 141
Figure 4.8. Specific growth rate of pyp1 and wild type cells growing on trehalose (upper panel) and 3
hours after switching to glucose (lower panel). ......................................................................................... 143
13
Figure 4.9. Deduction of labeling pattern of pentose-5-phosphate in cells growing on 1,2- C2-glucose. 150

vii
 List of Tables
Table 1.1. Intracellular concentrations of PEP, FBP, acetyl-CoA, aspartate, malate, and GTP in E. coli
during different growth conditions as measured by LC-MS. ....................................................................... 25
Table 1.2. Kinetic parameters of purified wild type and R313Q PEP carboxylase. .................................... 29
Table 1.3. Mass, retention time, and signal intensity of measured metabolites by the LC-high resolution
MS analysis. ................................................................................................................................................ 49
Table 1.4. Parent and product ion formulas and masses, collision energy (CE), retention time (RT), and
signal intensity of metabolites measured by the confirmatory LC-triple quadrupole MS/MS analysis ....... 52
Table 2.1. Strains used in this study. .......................................................................................................... 80
Table 2.2. Primers used in this study. ......................................................................................................... 81
Table 2.3 Media used in nitrogen and phosphate starvation ..................................................................... 86
Table 2.4. Mass, retention time, and ion counts of measured metabolites by the LC-high resolution MS
analysis. ...................................................................................................................................................... 92
Table 2.5. Parent and product ion formulas and masses, collision energy (CE), retention time (RT), and
ion counts of metabolites measured by the confirmatory LC-triple quadrupole MS/MS analysis. .............. 96
Table 3.1. Incorrect and missing annotations in the current genome scale metabolic model .................. 103
Table 3.2. Absolute concentration of metabolites in this study. ................................................................ 106

viii
 Introduction
Research Background

Metabolism, a highly conserved biochemical network, consists of several thousand metabolites and

metabolic reactions. Control of metabolite concentrations and fluxes occurs via regulation of enzyme

concentrations, activities and substrate occupancies. Understanding metabolism and its regulation is

important for curing metabolic diseases such as cancer and diabetes and manipulating metabolism in

metabolic engineering. To fulfill these two goals, much effort has been made to try to explore the

regulation of the most important metabolic pathways (e.g. central metabolism and biosynthesis) and to

expand our knowledge towards a more complete metabolic map.

The studies on metabolic regulation have a long history. In the 1950s and 60s, feedback inhibition was

found biochemically and believed to be the most important regulation of linear metabolic pathways, with
1-3
the regulatory point usually occurring at the first committed step . Such design of metabolism is clearly

efficient because once the demand for the end-product decreases, cells don’t need to spend energy and

nutrients to synthesize pathway intermediates, which are also sometimes toxic. Since the 1970s, with the

development of metabolic control analysis, metabolic regulation has been believed to be distributed
4-12
across multiple enzymes . Such findings revealed the complexity of metabolic regulation, leading to

massive biochemical characterizations of enzyme regulators in the 1970s and 80s. Even today, these

sets of biochemistry data are very valuable; however, it will be more intriguing to know which one(s) of

those many regulators play a key role in modulating metabolism in vivo.

The major obstacle in identifying physiologically important regulation was the lacking of sensitive and high

throughput analytical methods for quantitation of intracellular metabolites. One way to circumvent this
13-18
technical difficulty is through Flux Balance Analysis , where flux stands for the rate of metabolic

reactions. This approach predicts metabolic flux in silico based on growth optimization, known constrains

and balanced flux in cells under metabolic steady state. It is very helpful in predicting metabolic flux in

microbes under optimized growth conditions. However, the assumption that microbes maximize their

growth performance is not always plausible. For example, mutant strains with gene deletions are

i
artificially generated without evolutionary pressure. Yet microbes in natural environment are usually under

nutrient limited or stressed conditions where survival, rather than growth is the main goal. Thus,

measuring intracellular metabolites is a must for a more comprehensive understanding of metabolism.

19
Intracellular metabolites exist at low concentrations (ranging from µM to mM ), presenting a significant

analytical challenge. The development of metabolic flux analysis in the 1990s and early 2000s partially

addressed this issue by measuring abundant proteinogenic amino acids in cells growing steadily on
20-29
labeled media using NMR or gas chromatography – mass spectrometry (GC-MS) . The labeling

patterns of proteinogenic amino acids were then used as the input to calculate metabolic fluxes and infer

the source of labeling patterns of metabolic intermediates. Such methods achieved consistent readouts in

steadily grown microbes whose growth rate are controlled by chemostats. However, due to the indirect

nature of proteinogenic amino acids on intermediate metabolism, judging the weights in data fitting is

challenging. For example, we continued to find that the labeling pattern of some intermediate metabolites

can predominantly determine the flux of an entire pathway. However, the lack of direct measurement and

the failure to give this metabolite significant weight renders previous Metabolic Flux analysis unreliable in
30,31
many cases . Therefore, directly measuring important intermediate metabolites is essential for a more

quantitative understanding of metabolism.

The recent development of liquid chromatography – mass spectrometry (LC-MS) facilitates the

measurement of intracellular metabolites. Building from the reliable identification of central metabolites via

Triple Quadruple (QQQ) tandem mass spectrometry (LC-MS/MS) operating in multiple reaction
32
monitoring mode , our lab has developed analytical methods for reliably measuring almost all central
19
carbon metabolites . These results are a significant complement to the currently available biochemistry
33
data since thermodynamic prediction of metabolic reactions become possible .

One limit of these QQQ mass spectrometer is that, although they are capable of determining a

compound’s identity based on its fragmentation patterns, its mass accuracy is not accurate enough to

differentiate isotopically labeled deuterium, carbon or nitrogen. Also, because fragmentation is necessary,

this method requires a considerable amount of time (with the shortest metabolomics method ~ 50

2
minutes), which is unsuitable for high throughput measurement of both large number of metabolites and

large number of samples. A more rapid and accurate quantitation method became available recently,

when more than 150 intracellular metabolites and their deuterium, carbon and nitrogen labeling patterns

were able to be quantitated in a single run on a high-resolution, high-accuracy Orbitrap mass

34
spectrometer (Exactive; Thermo Fisher Scientific) . Its untargeted scanning mode also facilitates the

discovery of previously unidentified metabolites, e.g. novel metabolite sedoheptulose-1,7-bisphosphate in

35
yeast, which participates in a thermodynamically driven riboneogensis pathway . Using this untargeted

scan mode, many other poorly defined pathways could be extensively investigated.

Another challenge to the comprehensive understanding of metabolism was the culturing and extraction

methods used previously. As stated above, metabolic steady state has been investigated thoroughly.

However, all living cells are subject to environmental changes, rendering the understanding of

perturbations and stress important. Our lab had previously developed a culturing and extraction method,
36-39
involving growing microbes on filters placed on top of agarose plates . Such method facilitates fast

perturbation and extraction which are both essential for metabolite quantitation due to the fast changing

nature of metabolism. However, this relies on the diffusion of nutrients in agarose, which might not be

plausible for the addition of drugs or introduction of stress. In addition, the filter switch method for

changing the cell’s growth media could introduce high residual levels of media from the previous agarose

plate, which is not ideal for starvation experiments requiring complete nutrient removal. To address these

issues, a new culturing and extraction method was designed and optimized in this study (described in

detail in methods of Chapter 2). This approach grows cells in batch culture and extracts metabolites using

a rapid filtration method. For nutrient starvation, cells were filtered, washed and resuspended into the new

medium, leaving almost no residual nutrients.

Another type of experiment fulfilled by the high-resolution, high-accuracy Exactive is the LC-MS based

biochemical assays. We found that the previously used colorimetric assays (usually based on the

decrease or increase of NADH) had generally high backgrounds, which is not ideal for quantitation of

enzymatic activity in a wide range or under inhibitory conditions. As a result, previous biochemical assays
40,41
were usually performed in the presence of highly concentrated substrates . From the absolute

3
 19
quantitation data , we learned that concentration of lots of intracellular metabolites was far below

previous estimates. Thus, the kinetic data of many biochemistry assays needs to be reevaluated. Using

Exactive, reaction products of biochemical assays can be directly quantitated without coupling to a NADH

producing or consuming reaction. Such approach could achieve an accurate quantitation of enzyme

activity even when there is a greater than 1000-fold change (described in methods in Chapter 1 and 2).

Using the combination of LC-MS based metabolomics, biochemistry, proteomics, modern genetics and

modeling, metabolism and its regulation in model microbes E. coli and yeast were investigated as

described below.

Overview of Thesis

Central carbon metabolism, consisting of glycolysis, the pentose phosphate pathway and the tricarboxylic

acid (TCA) cycle, is highly conserved across taxa from prokaryotes to mammals. Glycolysis is probably

the most studied metabolic pathway. A common assumption, although subject to some dispute, is that the

committed step, catalyzed by phosphofructokinase-1 (Pfk1), functions as the critical glycolytic control
9,42-44
point . The terminal step in glycolysis, catalyzed by either phosphoenolpyruvate carboxylase (in

gram negative bacteria and some plants) or pyruvate kinase (in most other eukaryotes) is also irreversible

and tightly regulated. Humans encode four isozymes of pyruvate kinase, one of which (PKM2) plays a
45
central role in oncogenesis . Even in the best understood model organism, such as bacteria Escherichia

coli and yeast Saccharomyces cerevisiae, the terminal enzymes of glycolysis have multiple regulators.

Although extensive biochemical studies have been performed to illustrate the relative importance of these
46-50
regulators , no in vivo evidence has substantiated these findings.

With the development of mass spectrometry-based proteomics, covalent modifications have been studied
47,51
to an unprecedented scale. Over 4000 protein phosphorylation sites have been identified in yeast . In

glycolysis, both phosphofructokinase and pyruvate kinase are phosphorylated. Biochemical studies also
47-
showed that, phosphorylation of yeast pyruvate kinase by Protein Kinase A was able to alter its activity
50
, giving rise to the possibility that phosphorylation plays a key role in controlling the glycolytic exit.
4
In Chapter 1 and 2, we characterized and reevaluated the regulation of the terminal step of glycolysis in
52,53
E. coli and yeast . In both organisms removal of glucose results in a paradoxical increase in PEP,

which goes up the most of any canonical metabolite. We show that allostery predominates in both

organisms, with PEP consumption activated in an ultrasensitive (switch-like) manner by the upstream
40,41,46,54-57
glycolytic intermediate, fructose-1,6-bisphosphate, a previously under-appreciated regulator .

Mutations that eliminate this regulation do not impair growth on steady glucose, but they render the

microbes defective in gluconeogenesis and in growth in an oscillating glucose environment. Thus,

microbial central carbon metabolism is intrinsically programmed with ultrasensitive feed-forward

regulation to enable rapid adaptation to changing environmental conditions.

Another intriguing change of the yeast metabolome upon carbon starvation is the accumulation of the

non-oxidative pentose phosphate intermediate, sedoheptulose-7-phosphate. Such accumulation is

unexpected because enzymes involved in the production and consumption of this compound are

reversible and not considered important regulatory points. In Chapter 3, using isotope tracer analysis, we

found that the accumulated sedoheptuloes-7-phosphate was mainly derived from macromolecule break

down, presumably via ribosome degradation. Because degradative pathways, unlike biosynthetic

pathways, are non-essential for growth, they cannot be readily mapped by purely genetic approaches

relying on overt phenotypes such as dependence on a particular added nutrient, or auxotrophy. Moreover,

many degradation enzymes have promiscuous activity against a diversity of substrates, rendering their
58
functions difficult to determine from biochemistry alone . For example, in E. coli, various enzymes have
59-61 62-64
been putatively identified as nucleotidases and nucleosidases , but which one(s) are

physiologically important remains unknown. Mammalian nucleotide degradation and ribose salvage is yet
65
more complex and less well defined .

To analyze the physiological function of nucleotide degradation enzymes in yeast, we employed

66-71
metabolomic analysis of targeted deletion strains , and examined the metabolic response to a

perturbation that strongly induces the degradation pathway. This allowed us to identify the physiologically

relevant enzyme for all but one reaction of the pathway. Identification of these enzymes also provided us

the opportunity to investigate the physiological role of this pathway. We found that this autophagy-

5
mediated pathway is essential for yeast’s survival under poor nutrient conditions. More intriguingly,

deletion of this pathway renders cells unviable when facing oxidative stress, due to its essential role of

degrading nucleotides into central metabolism for driving the production of NADPH, a key anti-oxidant

regenerator. Thus, those poorly characterized pathways, although seemingly less important, provide

intriguing opportunities for drug targeting and metabolic engineering.

In Chapter 4, to further reveal important but uncharacterized pathways, we initiated a metabolomics

screen of yeast deletion strains of genes of unknown function via the identification of accumulated

metabolites as substrate candidates of the corresponding enzyme of the deletion strains. This approach

recently identified a new route of converting glycolytic intermediates into ribose, which yeast cells activate
35
when demand for ribose exceeds that for NADPH . The in vivo substrates of four more enzymes were

discovered and here we show an example of identifying a new class of phosphatase activity, polyol

phosphatase, which is required for maintaining phosphoglucose isomerase activity in fast growing cells.

The enzyme’s substrate, polyol phosphate is a strong inhibitor of phosphoglucose isomerase, the

reversible step in upper glycolysis, upstream of phosphofructokinase. Although not thought of as a key

regulatory point, we show that accumulation of polyol phosphates triggers an intense metabolic response,

which is similar to decreasing the expression level of phosphoglucose isomerase. Such inhibition resulted

in a strong decrease of the flux ratio of glycolysis over the oxidative pentose phosphate pathway,

supporting ongoing work in the field of metabolic control analysis arguing for distributed control of

metabolic flux. Thus, both upstream and downstream regulatory mechanisms work in concert to balance

energy, redox and biomass production.

Studies of this type, by clarifying enzyme functions and pathway regulation hold the potential for a more

comprehensive understanding of metabolism, which will lead to a more predictable and flexible

identification of metabolic drug targets and metabolic engineering pathways.

6
 Chapter 1

Ultrasensitive regulation of anapleurosis in Escherichia coli

Adapted from Yi-Fan Xu, Daniel Amador-Noguez, Marshall Louis Reaves, Xiao-Jiang Feng,

Joshua D. Rabinowitz “Ultrasensitive regulation of anapleurosis in Escherichia coli” Nature

Chemical Biology 2012, 8: 562-8

Abstract

Anapleurosis is the filling of the TCA cycle with four-carbon units. The common substrate for

both anapleurosis and glucose phosphorylation in bacteria is the terminal glycolytic metabolite,

phosphoenolpyruvate (PEP). Here we show that E. coli quickly and almost completely turns off

PEP consumption upon glucose removal. The resulting build-up of PEP is used to quickly import

glucose if it becomes re-available. The switch-like termination of anapleurosis results from

depletion of fructose-1,6-bisphosphate (FBP), an ultrasensitive allosteric activator of PEP

carboxylase. E. coli expressing an FBP-insensitive point mutant of PEP carboxylase grow

normally on steady glucose. However, they fail to build-up PEP upon glucose removal, grow

poorly on oscillating glucose, and suffer from futile cycling at the PEP node on gluconeogenic

substrates. Thus, bacterial central carbon metabolism is intrinsically programmed with

ultrasensitive allosteric regulation to enable rapid adaptation to changing environmental

conditions.

7
1.1 Introduction

The fate of PEP is among the most important governors of overall cellular metabolic activity, dictating the
72
relative flux of glycolysis versus gluconeogenesis . In eukaryotes, pyruvate kinase is the canonical PEP-

catabolizing enzyme. Its importance is indicated by the existence in humans of four isozymes, one of
45
which plays a central role in oncogenesis . In prokaryotes, PEP can be catabolized by two additional

routes: It is used as the phosphate donor for import of glucose and related sugars via the

phosphotransferase system (PTS, net reaction: glucoseextracellular + PEP  glucose-6-phosphateintracellular +

pyruvate). It is also the direct substrate for anapleurosis in some bacteria including E. coli, providing four

carbon units to the TCA cycle via PEP carboxylase (pyruvate is used for anapleurosis in other bacteria,
73
yeast, mammals ). In E. coli growing freely on glucose as the sole carbon source, ~ 82% of PEP is
74
consumed for glucose import and ~ 12% for oxaloacetate synthesis , with pyruvate kinase inessential for
75
cell growth .

In natural environments, microbes face of myriad of potential carbon sources whose availability varies.

This places a premium on the ability to adapt to changing access to carbon. The mechanisms that enable

rapid adaptation, on a time-scale faster than transcription, remain largely unproven. Metabolomics—the
76,77
systems level analysis of metabolites—has opened a new window on such rapid regulation . A striking

finding, conserved across both E. coli and yeast, is that, while upper glycolytic compounds (from glucose
78
to FBP) drop as expected upon glucose removal, PEP rises the most of any canonical metabolite . In E.

coli and other prokaryotes, this rise can in part be accounted for by PEP consumption by the PTS

terminating immediately upon glucose removal. The regulation of other reactions, however, is clearly

required to produce and sustain the dramatic rise in PEP. Here we show that PEP carboxylase shuts off

in a switch-like manner upon glucose removal via ultrasensitive allosteric regulation by FBP, and that this

regulation is essential for E. coli to utilize effectively gluconeogenic substrates or intermittently available

glucose.

8
1.2 Results

Glucose removal results in PEP build-up

We measured the short term (within 90 minutes) metabolome response of E. coli to sudden switches from

glucose to no carbon, acetate, succinate, or glycerol. E. coli cells were grown dispersed on filters on top

of an agarose-media mixture, fed by diffuse of media components from below and oxygen from above.

This culture method allowed rapid, non-disruptive glucose removal, isotope-switches, and metabolome
79
harvesting by transfer to cold organic solvent (40:40:20 acetonitrile:methanol:water at -20 °C ).

Metabolites were extracted in the cold organic solvent mixture and extracts analyzed by liquid

chromatography – mass spectrometry (LC-MS), with compound identities verified by mass and retention

time match to authenticated standards.

Within the first 10 minutes after glucose removal, under each of the four examined conditions, hexose

phosphate concentrations decreased by 5- to 10-fold; FBP and GAP/DHAP decreased by 15- to 30-fold;

and PEP increased by at least 10-fold (Figure 1.1-1.4). The increase in PEP suggests a rapid shut-off of

PEP consumption or rapid onset of gluconeogenesis. To evaluate the possibility that the PEP pool is
13
being fed from TCA components, we switched cells from unlabeled glucose to U- C-acetate and

observed substantial malate labeling (> 80%) but no detectable PEP labeling (< 1%) over the first 45

minutes following glucose removal (Figure 1.5). This rules out substantial PEP formation by
80
gluconeogenesis or the PEP-glyoxylate cycle on this time scale, and indicates that PEP consumption is

blocked.

Routes of PEP consumption

In E. coli, the phosphotransferase system (PTS) and PEP carboxylase (anapleurosis) are two major

consumers of PEP. Assuming optimal aerobic carbon utilization for biomass production, these two

reactions account for > 94% of PEP consumption (Figure 1.6). The other significant consumer of PEP

9
during optimal growth on glucose is aromatic amino acid biosynthesis (~ 2%), with fluxes to cell wall and
81,82
outer membrane biosynthesis having a combined contribution of only ~ 0.5% .

Figure 1.1. Glucose removal results in PEP build-up.

E. coli cells growing freely in glucose minimal media were switched to no carbon, acetate, succinate, or
glycerol as indicated. After the indicated duration of glucose removal, the metabolome was quantitated by
LC-MS. Data are shown in heat map format, with each line reflecting the dynamics of a particular
compound in a particular culture condition. Metabolite levels of biological duplicates were averaged,
normalized to cells growing steadily in glucose (time zero), and the resulting fold changes log transformed.
See Figure 1.2 for absolute concentration changes for key metabolites and Figure 1.3 for a heat map of
all measured metabolites. Note that the decreased ATP/ADP and NADH/NAD+ ratios upon glucose
10
removal thermodynamically drive carbon towards the bottom of glycolysis (i.e. towards PEP) (Figure 1.4).
G6P, glucose-6-phosphate; F6P. fructose-6-phosphate; FBP, fructose-1,6-bisphosphate; GAP,
glyceraldehyde-3-phosphate; DHAP, dihydroxyacetone phosphate; DPG, 1,3-diphosphateglycerate; 3PG,
3-phosphoglycerate; 2PG, 2-phosphoglycerate; OAA, oxaloacetate; αKG, α-ketoglutarate; SucCoA,
succinyl-coenzyme A. G6P and F6P, GAP and DHAP, and 3PG and 2PG were not differentiated by
employed LC-MS method.

Figure 1.2. Absolute intracellular concentrations of key metabolites during glucose removal and
its replacement with no carbon, acetate, succinate, or glycerol.
E. coli cells growing freely in glucose minimal media were switched to no carbon, acetate, succinate, or
glycerol as indicated. After the indicated duration of glucose removal, the metabolome was quantitated by
LC-MS. Cyclic-AMP spikes in all four switches, a canonical response of E. coli to glucose removal that
activates cyclic-AMP receptor protein (CRP) to regulate the transcription of over 180 enzymes.
Gluconeogenesis starts more quickly in the switch to glycerol (~ 30 min) than in the other conditions
(where it takes more than 1.5 h). The x axis represents minutes after the switch, and the y axis represents
absolute intracellular metabolite concentration. Points reflect data (mean ± SD of N = 2 biological
replicates).

11
Figure 1.3. Heat map of all measured metabolites during glucose removal and its replacement with
no carbon, acetate, succinate, or glycerol.
E. coli cells growing freely in glucose minimal media were switched to no carbon, acetate, succinate, or
glycerol as indicated. After the indicated duration of glucose removal, the metabolome was quantitated by
LC-MS. Metabolite levels of biological duplicates were averaged, normalized to cells growing steadily in
glucose (time zero), and the resulting fold changes log transformed.

12
Figure 1.4. Upon glucose removal, decreasing ATP/ADP and NADH/NAD+ ratios drive carbon
towards the bottom of glycolysis (i.e. towards PEP).
+
ATP/ADP and NADH/NAD ratios decrease from 0 – 20 minutes after the switch in all four conditions,
½
which overcomes the dramatic increase of PEP/FBP to drive carbon towards the bottom of glycolysis.
Reaction quotients indicate the observed ratio of reactants to products. Lower quotient values favor
downward reactions (i.e. towards PEP). The x axis represents minutes after the switch, and the y axis
represents ratios and quotients in logarithmic scale. Points reflect data (mean ± SD of N = 2 biological
replicates).

13
Figure 1.5. Accumulated PEP is not from gluconeogenesis or the PEP-glyoxylate cycle in the first
45 minutes after acetate switch.
13
E. coli wild type cells were switched from non-labeled glucose into U- C-acetate. While TCA compounds
13 13 13
became labeled quickly, PEP did not get labeled until 1 h after the switch. C2, C3, and C4 malate
showed up in sequence due to the turning of TCA cycle. The x axis represents minutes after the switch,
and the y axis represents absolute intracellular metabolite concentration. Points reflect data (mean ± SD
of N = 2 biological replicates).

Figure 1.6. PEP carboxylase is inhibited by 99% upon glucose removal.

13
(a, b). Concentrations of unlabeled and U- C-labeled PEP and FBP, showing rapid initial conversion
13
from unlabeled to labeled forms upon the switch to U- C-glucose, and then, upon glucose removal,
accumulation of unlabeled PEP and depletion of labeled FBP. Cells were grown on glucose minimal
13
media as in Figure 1. Glycolytic intermediates were labeled by switching cells to U- C-glucose for 8 min.
Thereafter, glucose was removed and replaced with either no carbon (a, b) or acetate (c, d). Green =
13 13
nonlabeled, red = U- C labeled, blue = sum of nonlabeled and U- C labeled. (c, d) Concentrations of
13
unlabeled and U- C-labeled lower glycolytic intermediates (sum of all measured compounds from FBP to
PEP) starting 10 min after glucose removal. Points reflect data (mean  range of N = 2 biological
replicates); lines represent a fit of the data to equations described in Methods. The fitting enabled
estimation of the decrease in PEP carboxylase flux relative to glucose-fed cells.

14
While pyruvate kinase activity is not required for optimal growth on glucose, it has been recently been
83,84
reported to account for ~ 15% of PEP consumption in batch culture . However, pyruvate kinase activity
85
decreases with oxygenation, and filter cultures have direct exposure to the aerobic environment . To
13
assess pyruvate kinase flux, we performed metabolic flux analysis using 1- C-glucose as the tracer. In

filter cultures, we find that pyruvate kinase flux accounts for less than 5% of PEP consumption (Figure

1.7 and Methods). Consistent with pyruvate kinase making only a minor contribution, side-by-side

comparison of wild type and ∆pykF/∆pykA double knockout cells revealed no differences in in growth rate,
13
PEP concentration, or 1- C-glucose labeling (Figure 1.8-1.9), and only small changes in metabolome

response to glucose removal (Figure 1.10).

Figure 1.7. The primary PEP consuming enzymes are the PTS and PEP carboxylase.
All known PEP consumption routes are shown, along with concentration changes of associated
metabolites upon glucose removal. Arrow sizes indicate steady state net fluxes from PEP for wild type E.
coli cells growing freely on glucose filter cultures. The compound-specific heat maps match exactly the
format of Figure 1. E4P, erythrose-4-phosphate; shikimate-3-P, shikimate-3-phosphate; UDP-GlcNAc,
UDP-N-acetyl-D-glucosamine; ribulose-5-P, ribulose-5-phosphate; arabinose-5-P, arabinose-5-phosphate;
KDO, 2-Keto-3-Deoxy-D-manno-octonate.

15
Figure 1.8. Metabolic Flux Analysis of wild type and ΔpykA/ΔpykF strains indicates pyruvate
kinase flux accounts for < 5% of PEP consumption in wild type cells.

(a). Aspartate, PEP and xylulose-5-phosphate labeling patterns in wild type and ΔpykA/ΔpykF cells
13
growing on 1- C glucose. Colored bars with error bars indicate mean ± SD of N = 6 biological replicates;
dashed horizontal black lines in the aspartate panel indicate the theoretical labeling pattern given the
calculated fluxes. (b). Theoretical distribution of isopotomers (positionally labeled forms, upper panel) and
mass isotopomers (sum of all positionally labeled forms with same number of labeled carbon atoms,
lower panel) of aspartate as a function of TCA turning flux into oxaloacetate (FTCAOAA) divided by sum of
PEP carboxylase flux (FPPC) and FTCAOAA. The dashed vertical line indicates the experimentally observed
value of FTCAOAA / (FPPC + FTCAOAA). (c). Central carbon metabolic fluxes and error estimates in glucose-
fed filter culture wild type and ΔpykA/ΔpykF cells. Reported values are best flux estimates +/- SE. For
associated equations, see Methods.

16
Figure 1.9. Deletion of pyruvate kinase (both isozymes, pykA and pykF) does not substantially
alter E. coli growth or PEP concentration.
(a). Filter culture E. coli wild type and ΔpykA/ΔpykF cells were grown in glucose minimal media. Bars
reflect data (mean ± range of N = 2 biological replicates).(b). E. coli wild type and ΔpykA/ΔpykF cells
were switched from glucose to no carbon. The x axis represents minutes after the switch, and the y axis
represents absolute intracellular PEP concentration. Points reflect data (mean ± range of N = 2 biological
replicates).

Having ruled out a primary contribution of pyruvate kinase, we considered how the other routes of PEP

consumption might be regulated upon glucose removal. As glucose is a substrate of the PTS, glucose

removal immediately stops this largest route of PEP consumption. Absence of glucose also stops protein

synthesis; this led to build-up of tyrosine, tryptophan, and phenylalanine (Figure 1.7), which feedback-
86
inhibit aromatic amino acid biosynthesis . Moreover, absence of glucose decreased the concentration of

substrates involved in the PEP-consuming reactions of cell wall and outer membrane synthesis; such

depletion provides a potential route for turning off these minor fluxes (Figure 1.7). Thus, in order to

explain the increase in PEP concentration upon glucose removal in our culture system, the major

question regards the mechanism of PEP carboxylase inhibition.

46,57
FBP is a known positive regulator of PEP carboxylase , and its depletion could potentially decrease

PEP carboxylase activity and thus cause PEP accumulation. Existing biochemical literature data

suggested, however, that the observed change in FBP concentration would lead to only a five-fold
40,41
decrease in PEP carboxylase activity . This seemed insufficient for PEP to build-up, given that the

PEP carboxylase flux in cells growing steadily in glucose minimal media is 15 mM/min (see Methods),

17
which would consume the observed PEP in 15 s. Other known regulators of PEP carboxylase include
46
acetyl-CoA and GTP (positive regulators) and aspartate and malate (negative regulators) . These

species showed inconsistent and comparatively mild changes in the various carbon switches (Figure 1);

thus, it seemed unlikely that they are the main drivers of the universal spike in PEP.

Figure 1.10. Impact of pyruvate kinase deletion (both isozymes, pykA and pykF) on E. coli
metabolome response to glucose removal.

Metabolite levels of biological duplicates were averaged, normalized to cells growing steadily in glucose
(time zero), and the resulting fold changes log transformed and shown in heat map format.

18
Quantitation of residual PEP consumption flux

To measure directly the extent to which PEP consumption is cut off in vivo, we conducted a metabolomics
13
experiment involving transient feeding of U- C-glucose followed by glucose removal. By labeling
87
glycolytic intermediates but not macromolecule stores (protein, glycogen) , such transient labeling

allowed us to determine the extent to which PEP production via macromolecule breakdown contributes to

the high PEP pool size following glucose removal. For example, degradation of glycogen, glycerolipids, or

nucleic acids could in theory feed unlabeled carbon into the PEP pool. As PEP remained largely labeled
13
after the removal of U- C-glucose (Figure 1.6a and b; Figure 1.11), the PEP build-up is predominantly

from glycolytic intermediates existing at the time of glucose removal. Such build-up requires an almost

complete termination of PEP consumption, whose extent can be quantitated based on the labeling

dynamics.

Figure 1.11. Transient labeling experiment as per Figure 1.6, except with shorter labeling duration.
Methods are the same as Figure 1.6 except with 2 min labeling duration. Data were data analyzed using
Equations (3) and (4) to give decrease in PEP efflux of 99.5%, versus 99.6% in Figure 1.6c.
In the case of the switch to no carbon source, the total concentration of lower glycolytic intermediates
(from FBP to PEP) remained roughly stable from 10 – 60 minutes after glucose removal. The changing
labeling patterns can accordingly be interpreted assuming equal total triose production and consumption

19
fluxes during this period (Figure 1.6c and Figure 1.12a). The resulting fluxes after glucose removal,
based on the rate of decrease of the labeled form, were 0.051 mM/min (99.6% less than PEP
carboxylase flux in the glucose-fed cells; see Methods). In the case of the switch to acetate, the total
lower glycolytic pool dropped while the unlabeled pool remained steady. This is consistent with steady
consumption of lower glycolytic intermediates during the first 10 – 30 minutes after the switch from
glucose to acetate (Figure 1.6d and Figure 1.12b) at a rate of 0.15 mM/min, requiring a 99.0% inhibition
of PEP carboxylase. These decreases far exceed what would be expected based on prior biochemical
40,41,88
literature .

Figure 1.12. Fit of experimental data to equations used for quantitation of PEP consumption
during carbon starvation and acetate switch based on transient labeling experiments with (a) 8
13 13
min C-glucose labeling followed by carbon starvation, (b) 8 min C-glucose labeling followed by
13
acetate switch, and (c) 2 min C-glucose labeling followed by carbon starvation.
T L L
(a). The plot (X ln(X 0/X )) vs. (t – t0) gives a straight line, consistent with a steady unlabeled production
flux and a matching steady consumption flux from PEP (for details see Methods). The slope gives f =
U
0.051 mM/min, which is 99.6% less than PEP carboxylase flux in the glucose-fed cells.(b). The plot (X ln
L L
X + X ) vs. (t – t0) gives a straight line, indicating a decreasing unlabeled production flux and a steady
consumption flux from PEP. The slope gives fex = 0.15 mM/min, which requires a 99.0% inhibition of PEP
U L L
carboxylase. (c). The plot (X ln X + X ) vs. (t – t0) gives a straight line, consistent with a decreasing
unlabeled production flux and a steady consumption flux from PEP. The slope gives fex = 0.071 mM/min,
which is 99.5% less than PEP carboxylase flux in the glucose-fed cells.Points reflect data; lines represent
least square fitting.

The above analysis indicates that PEP consumption in carbon starvation is yet lower than in the acetate

switch condition. This is logical, as carbon starvation triggers depletion of acetyl-CoA, a PEP carboxylase

activator, and accumulation of aspartate, a PEP carboxylase inhibitor. While aspartate is produced in a

single step from oxaloacetate, the isotope labeling pattern of the accumulated aspartate indicates that it

did not come from anapleurosis. A likely source is instead from glutamate, which declined in parallel with

the rise in aspartate (Figure 1.13). Given that acetyl-CoA and aspartate do not change substantially

20
during the acetate switch (Figure 1), it is particularly striking that PEP carboxylase flux nevertheless

terminates 99%.

Figure 1.13. Accumulated aspartate is not from anapleurosis.

13
Glycolytic intermediates were labeled by switching cells to U- C-glucose for 8 min. Thereafter, glucose
13
was removed and replaced with no carbon. The majority of accumulated aspartate is fully ( C4) labeled.
As fully labeled glutamate declines in parallel with the rise in aspartate, the increase in aspartate is likely
fed by turning of the TCA cycle, combined with decreased aspartate consumption to protein and
13
pyrimidine synthesis. Anapleurosis from PEP would result in triply-labeled ( C3) aspartate, containing one
unlabeled carbon atom from bicarbonate. Such aspartate does not accumulate, consistent with the
evidence that anapleurosis terminates almost completely. The x axis represents minutes after carbon
starvation, and the y axis represents absolute intracellular concentration. Points reflect data (mean ±
range of N = 2 biological replicates).

FBP activation of PEP carboxylase is ultrasensitive

Given the almost complete elimination of PEP carboxylase flux, we hypothesized that the enzyme might

be degraded or covalently modified. To examine these possibilities, we quantitated PEP carboxylase

protein concentration and activity before and after glucose removal. Targeted LC-MS revealed no change

in PEP carboxylase protein concentration following glucose removal (Figure 1.14a and Methods).

Similarly, PEP carboxylase activity in cell lysates remained constant, which is inconsistent with PEP
89
carboxylase inactivation by a stable covalent modification event (Figure 1.14b and Methods). To

further rule-out potential regulation at the level of protein concentration, we inducibly over-expressed PEP
21
carboxylase in wild type E. coli. While PEP carboxylase over-expression modestly reduced the basal PEP

concentration in glucose-fed cells, it did not block the PEP spike after glucose removal (Figure 1.14c).

Figure1.14. There is no degradation or stable, inactivating post-translational modification of PEP

carboxylase in the first 10 minutes following glucose removal.
(a). PEP carboxylase protein concentration is the same before and after glucose removal. Wild type E.
coli cells growing freely in glucose minimal media were switched to no carbon or acetate as indicated.
After 10 minutes of glucose removal, PEP carboxylase (PPC) protein concentration was quantitated by
targeted LC-MS (see Methods). The y axis represents relative PEP carboxylase protein concentration
(glucose-fed wild type cells defined as 1). Bars reflect data (mean ± range of N = 2 biological replicates).
(b). PEP carboxylase activity is the same in the cell lysates before and after the switches, indicating no
89
stable, inactivating post-translational modification of PEP carboxylase during glucose removal . PEP
carboxylase activity was measured in the presence of PEP (0.18 mM) and the essential activator acetyl-
CoA (0.61 mM).The y axis represents specific enzyme activity in Units (µmol of product produced per
12
minute) per 10 cells. Bars reflect data (mean ± range of N = 2 biological replicates). (c). While over-
expression of PEP carboxylase modestly reduces the basal PEP concentration in glucose-fed cells, it
does not block the PEP spike after glucose removal. Carbon starvation and acetate switch were
performed on WT/pCA24N-ppc strain with different induction levels of PEP carboxylase. The x axis
reports these levels as fold-increase versus wild type cells as quantitated by targeted LC-MS. IPTG levels
were 20, 50, 100 and 200 µM for the 2-, 5-, 10- and 20-fold over-expression levels, respectively. The y
axis represents absolute intracellular PEP concentration. Bars reflect data (mean ± range of N = 2
biological replicates).

Given the above evidence that PEP carboxylase is present but not actively consuming PEP following

glucose removal, we re-examined PEP carboxylase’s allosteric regulation. A deficiency of prior literature

is that PEP carboxylase activity was measured one effector at a time, not always using the effectors’
40,41,88,90,91
physiological concentrations . Given our evidence from metabolomics that FBP appears to be

the primary regulator, we examined FBP’s effect over its physiological range, alone and in the presence
22
of additional regulators (Figure 1.15 and 1.16). FBP alone showed only modest control of PEP

carboxylase activity (6-fold), and its effect was not ultrasensitive (Hill coefficient 1.0). However, in the

presence of both acetyl-CoA and aspartate, the effects of FBP were magnified (310-fold) and

ultrasensitive (Hill coefficient 2.1). This ultrasensitive regulation is in principle sufficient to produce rapidly

the observed > 99% decrease in PEP carboxylase activity upon glucose removal (Figure 1.21).

Figure 1.15. Activation of PEP carboxylase by FBP is enhanced by physiological levels of acetyl-
CoA and aspartate.

Each small plot shows activity of purified His-tagged PEP carboxylase as a function of FBP concentration.
The six small plots differ in terms of the additional PEP carboxylase effectors present, building from FBP
alone (left-most) to all known effectors (right-most) as indicated. The concentration of PEP (2 mM), as
well as effectors acetyl-CoA (0.63 mM), aspartate (5.2 mM), malate (1.1 mM), and GTP (4.6 mM) were
selected to match the physiological concentration in cells switched from glucose to acetate for 10 minutes
(Table 1.1; see Figure 1.16 for other conditions). Reactions were carried out at 30°C and pH 8.5 in the
presence of 10 mM bicarbonate and magnesium. In the plots, the x axis represents different FBP
concentrations, and the logarithmic y axis represents specific PEP carboxylase activity in enzyme activity
Units (µmol of product produced per minute) per mg of enzyme. Experimental data (mean ± range of N =
2) were fit to Hill-equations (lines). Vmax/Vmin represents observed range of PEP carboxylase activity as a
function of FBP concentration, where Vmax was measured at 15 mM FBP, matching the physiological
concentration in cells growing on glucose, and Vmin was measured at 0.45 mM FBP, matching the
physiological FBP concentration in cells switched from glucose to acetate for 10 minutes. nH and Khalf
represent the Hill-coefficient for FBP and the FBP concentration producing half-maximal activation.

23
Figure 1.16. The enhancement of FBP activation by acetyl-CoA and aspartate was also found
when effectors were added at their physiological concentrations in glucose minimal media and
carbon starvation conditions.

Each small plot shows activity of purified His-tagged PEP carboxylase as a function of FBP concentration.
The concentration of PEP and effectors were selected to match the physiological concentration in cells
growing in steady state on glucose minimal media (top row, cross markers) and switched to no carbon for
10 min (bottom row, triangle markers). Top row concentrations are: PEP, 0.18 mM; acetyl-CoA, 0.61 mM;
aspartate, 4.2 mM; malate, 1.7 mM; GTP, 4.9 mM. Bottom row concentrations are: PEP, 4.7 mM; acetyl-
CoA, 0.21 mM; aspartate, 19 mM; malate, 0.7 mM, GTP, 2.1 mM. Bicarbonate and cofactor magnesium
was added both at 10 mM. The x axis represents different FBP concentrations, and the logarithmic y axis
represents specific PEP carboxylase activity in enzyme activity Units (µmol of product produced per
minute) per mg of enzyme. Experimental data (mean ± range of N = 2) were fitted to Hill-equations (lines).
Vmax/Vmin represents observed range of PEP carboxylase activity as a function of FBP concentration,
where Vmax was measured at 15 mM FBP, matching the physiological concentration in cells growing on
glucose, and Vmin was measured at 0,45 mM FBP, matching the physiological FBP concentration in cells
switched from glucose to acetate for 10 minutes. nH and Khalf represent the Hill-coefficient for FBP and the
FBP concentration producing half-maximal activation.

24
Figure 1.17. Inferred PEP carboxylase activity in live cells following glucose removal.
(a). Fold change (in logarithmic scale) of PEP and PEP carboxylase’s allosteric regulator’s concentrations
over the first 20 minutes after carbon starvation (left) and acetate switch (right), and associated inferred
PEP carboxylase activity. (b). Inferred PEP carboxylase specific activity over the first 20 minutes after
carbon starvation (left) and acetate switch (right) on linear axis.PEP carboxylase activity was calculated
from the biochemistry data shown in Figure 1.15-1.16. Data points are experimental measurements.
Lines connect measured (or, in the case of PEP carboxylase activity, calculated) values at the time points
where measurementswere taken.

PEP FBP acetyl-CoA aspartate malate GTP

Condition
(mM) (mM) (mM) (mM) (mM) (mM)

glucose minimal media 0.18 15 0.61 4.2 1.7 4.9

switch to acetate for 10
2.0 0.45 0.63 5.2 1.1 4.6
min
removal of carbon for
4.7 0.92 0.21 19 0.70 2.1
10 min

Table 1.1. Intracellular concentrations of PEP, FBP, acetyl-CoA, aspartate, malate, and GTP in E.
coli during different growth conditions as measured by LC-MS.

We were also interested in whether allosteric regulation of PEP carboxylase by FBP is in principle

sufficient for PEP to rise when FBP falls. To this end, we developed a simple mathematical model of

glycolytic regulation (Methods). The model includes FBP production, its reversible interconversion with
25
PEP, and PEP consumption by the PTS and PEP carboxylase, with PEP carboxylase flux proportional to
h
[FBP] . Using the model, we examined the steady-state concentrations of FBP and PEP as a function of

glycolytic flux (F). We find that FBP always decreases when F decreases. This matches experimental

observations. For h ≤ 1, the same is true for PEP. For h > 1, however, PEP can increase when F

decreases (Figure 1.18). Thus, ultrasensitive regulation of PEP carboxylase is a prerequisite for PEP to

accumulate when glycolytic flux decreases, as is observed experimentally.

Figure 1.18. Ultrasensitive regulation of PEP carboxylase is required for PEP to increase when
glycolytic influx decreases.
The simplified glycolytic model shown in Scheme 1.1 (methods) was solved quantitatively for [PEP] as a
function of glycolytic flux F. The hill coefficient for activation of PEP carboxylase (h) determines the
qualitative pattern of the PEP response, with h > 1 required for PEP to increase when F decreases. Both
the x and y axis are presented in arbitrary units; the units depend on other parameters of the model, but
the trends are identical irrespective of these parameters.

PEP accumulation results from the depletion of FBP

To determine the physiological significance of PEP carboxylase regulation by FBP, we sought a PEP

carboxylase point mutant desensitized to FBP activation (Figure 1.19). While the FBP binding site on E.

coli’s PEP carboxylase has not been determined, an allosteric regulatory site for glucose-6-phosphate on
92
maize PEP carboxylase has been identified crystallographically . This site involves four arginine residues.

Although E. coli PEP carboxylase is not activated by G6P or F6P (Figure 1.20), all four of these arginine

residues are conserved in the E. coli protein. We thus hypothesized that a phosphate group of FBP might

interact with one or more of these residues. Site-directed mutagenesis led to a mutant form of E. coli PEP
26
carboxylase, R313Q, which was more active than native PEP carboxylase in the absence of FBP and

desensitized to FBP activation (Figure 1.19a). For kinetic parameters of the altered enzyme, see Table

1.2 and Figure 1.21.

Figure 1.19. PEP accumulation upon glucose removal enables rapid import of glucose when it
reappears.
(a). PEP carboxylase mutant R313Q has higher activity than wild type in the absence of FBP and is
desensitized to FBP activation. The concentration of PEP and effectors and the y-axis units match the
right-most plot in Figure 1.15. (b). Expression of PEP carboxylase R313Q ablates the PEP spike upon
carbon starvation (see Figure1.22 for switch from glucose to acetate). E. coli strains in which
endogenous PEP carboxylase has been replaced with inducible expression of wild type enzyme
(Δppc/pCA24N-ppc, shown as PPC WT) or the FBP-insensitive R313Q mutant (Δppc/pCA24N-ppcR313Q,
shown as PPC R313Q) were switched to no carbon. (c). Growth curves of Δppc/pCA24N-ppc and
Δppc/pCA24N-ppcR313Q. Cells were grown in either steady glucose (closed symbols) or alternating
between glucose minimal media and no carbon minimal media every 30 min (open symbols). The x axis
represents time in hours, and the logarithmic y axis represents optical density (A600). (d, e). Cells were
13
switched to U- C-glucose for 8 minutes, followed by no carbon for 30 minutes, and finally re-addition of
unlabeled glucose. Upon glucose re-addition, samples were collected and intracellular metabolites
(labeled and unlabeled) quantitated by LC-MS. The x axis represents seconds after glucose re-addition
(in logarithmic scale), and the y axis represents absolute intracellular concentration (mean ± range of N =
2 biological replicates); (d) Δppc/pCA24N-ppc, (e) Δppc/pCA24N-ppcR313Q. Throughout (b) to (e),
experiments were performed in the presence of 100 µM IPTG.

27
Figure1.20. PEP carboxylase is not activated by glucose-6-phosphate or fructose-6-phosphate.

Each plot shows activity of purified His-tagged PEP carboxylase as a function of the concentration of the
indicated phosphorylated hexose sugar. The concentration of PEP (2 mM), as well as effectors acetyl-
CoA (0.63 mM), aspartate (5.2 mM), malate (1.1 mM), and GTP (4.6 mM) were selected to match the
physiological concentration in cells switched from glucose to acetate for 10 minutes. FBP was either not
added, or added at 0.45 mM (to match the physiological concentration in cells switched from glucose to
acetate for 10 minutes), or 15 mM (to match the physiological concentration in cells growing in steady
state on glucose minimal media). The x axis represents the concentration of the indicated phosphorylated
hexose sugar: F6P (fructose-6-phosphate alone, round markers), G6P (glucose-6-phosphate along,
triangle markers), or H6P (50:50 mix of fructose-6-phosphate : glucose-6-phosphate, square markers).
The logarithmic y axis represents specific PEP carboxylase activity in enzyme activity Units (µmol of
product produced per minute) per mg of enzyme. Points reflect data (mean ± range of N = 2 biological
replicates).

28
Figure1.21. Determination of PEP carboxylase wild type and R313Q kinetic parameters.

Each plot shows the activity of purified His-tagged PEP carboxylase as a function of PEP concentration.
The concentrations of the effectors acetyl-CoA (0.63 mM), FBP (15 mM), aspartate (5.2 mM), malate (1.1
mM), and GTP (4.6 mM) were selected to match the physiological concentration in cells switched from
glucose to acetate for 10 minutes. See Table 1.2 for kinetic parameters. The x axis represents different
PEP concentrations, and the y axis represents specific PEP carboxylase activity in enzyme activity Units
(µmol of product produced per minute) per mg of enzyme. Experimental data (mean ± range of N = 2)
were fitted to Hill-equations (lines).

Turnover number
Km (mM) for PEP -1 Hill-coefficient for PEP
(min )

Wild type R313Q Wild type R313Q Wild type R313Q

condition
acetyl-CoA
1.4 2.3 910 690 1.1 1.3
only
all effectors
9.4 1.8 870 600 2.2 1.3
except FBP
all effectors
0.2 2.3 6200 560 1.2 1.2
including FBP

Table 1.2. Kinetic parameters of purified wild type and R313Q PEP carboxylase. The effector
concentrations of acetyl-CoA (0.63 mM), FBP (15 mM), aspartate (5.2 mM), malate (1.1 mM), and GTP
(4.6 mM) were selected to match the physiological concentration in cells switched from glucose to acetate
for 10 minutes. PEP was added in a range of 0-20 mM. Km represents PEP concentration at which
reaction rate is half-maximum. Three conditions were investigated: acetyl-CoA only, all effectors except
FBP, and all effectors including FBP. Data without any effectors is not reported, as the enzyme activity is
minimal in the absence of acetyl-CoA.

We constructed an E. coli strain that lacked native PEP carboxylase (Δppc) and inducibly expressed the

R313Q mutant. In the presence of inducer, these cells (Δppc/pCA24N-ppcR313Q) grew normally on

glucose, with an intracellular PEP concentration slightly higher than wild type cells (0.21 mM versus 0.16

mM). Upon glucose removal, despite a normal decline in FBP, the mutant cells failed to show a PEP

spike (Figure 1.19b and Figure 1.22 and 1.23). Expression of wild type enzyme in the same manner

(Δppc/pCA24N-ppc) resulted in a normal PEP spike. Thus, the FBP-insensitive R313Q mutant resulted in

inappropriate anapleurotic flux that drained PEP.

29
Figure 1.22. Expression of PEP carboxylase R313Q ablates the PEP spike while FBP decreases as
normal.
(a). E. coli strains in which endogenous PEP carboxylase has been replaced with inducible expression of
wild type enzyme (Δppc/pCA24N-ppc) or the FBP-insensitive R313Q mutant (Δppc/pCA24N-ppcR313Q)
were switched to acetate. (b). The above strains were switched to either no carbon (left) or acetate
(right).The x axis represents minutes after the switch, and the y axis represents absolute intracellular
metabolite concentration. Points reflect data (mean ± range of N = 2 biological replicates).

30
Figure1.23. Heat map of all measured metabolites in the Δppc/pCA24N-ppc (PPC WT) and
Δppc/pCA24N-ppcR313Q (PPC R313Q) during glucose removal and its replacement with no
carbon or acetate.

Metabolite levels of biological duplicates were averaged, normalized to cells growing steadily in glucose
(time zero), and the resulting fold changes log transformed.

31
Accumulated PEP rapidly imports glucose when it reappears

The FBP-insensitive mutant allowed us to test the physiological function of PEP accumulation. We

hypothesized that a reserve of PEP is useful for rapidly importing glucose if it becomes re-available. To
13
test this possibility, we labeled glycolytic intermediates with uniformly C-glucose, removed carbon for 30

minutes, and then added unlabeled glucose. For Δppc/pCA24N-ppc, glucose addition resulted in
13
consumption of the pre-existing C3-PEP within 10 s (Figure 1.19d). Consistent with this PEP being used
13
for glucose import, the labeled carbon appeared in C3-pyruvate within 5 seconds. This pyruvate was
13
then quickly oxidized by pyruvate dehydrogenase to yield C2-acetyl-CoA. The resulting import of

glucose resulted in a rapid rise in FBP, which reached its normal steady state level within 15 seconds.

This increase in FBP activated anapleurosis: malate increased to its normal level within 1 minute. The

labeling pattern of malate confirmed that it was produced via anapleurosis (Figure 1.24a). In contrast, for

Δppc/pCA24N-ppcR313Q, due to the lack of PEP to jump-start glucose assimilation, FBP did not reach

normal levels until 15 minutes after glucose re-addition, 60-fold more slowly than with wild type PEP

carboxylase (Figure 1.19e). The slow increase of FBP also caused a slow turn-on of anapleurosis

(Figure 1.24b).

The above metabolomic data suggested that the PEP accumulation resulting from allosteric regulation of

PEP carboxylase might be required for normal resumption of growth after glucose re-addition. To test this

possibility, we grew Δppc/pCA24N-ppcR313Q and Δppc/pCA24N-ppc on steady versus oscillating

glucose. While the two strains grew identically on steady glucose, the R313Q PEP carboxylase mutant

strain was deficient in growing on oscillating glucose (Figure 1.19c). Thus, the ultrasensitive regulation of

PEP carboxylase enables cellular adaptation to varying carbon availability.

32
Figure1.24. The reserve of PEP is required for rapid initiation of anapleurosis upon glucose re-
addition.
13
Δppc/pCA24N-ppc (PPC WT) and Δppc/pCA24N-ppcR313Q (PPC R313Q) were switched to U- C-
glucose for 8 minutes, followed by no carbon for 30 minutes, and finally re-addition of unlabeled glucose.
Upon glucose re-addition, samples were collected and intracellular metabolites (labeled and unlabeled)
quantitated by LC-MS. Experiments were performed in the presence of 100 µM IPTG.

(a) For Δppc/pCA24N-ppc, malate starts to increase quickly after glucose readdition and reaches normal
13
steady state levels within 1 minute. This increase of malate is not from aspartate: While all the C forms
13
of aspartate decrease after the glucose re-addition, only three- C malate shows a spike, indicating its
glycolytic origin.

(b) For Δppc/pCA24N-ppcR313Q, due to the failure to accumulate PEP during glucose starvation and
consequent slow increase of FBP upon glucose re-addition, malate did not increase until 7 minutes after
glucose re-addition.

The x axis represents seconds after glucose re-addition (in logarithmic scale), and the y axis represents
absolute intracellular concentration. Points reflect data (mean ± range of N = 2 biological replicates).

FBP regulation inhibits futile cycle in gluconeogenesis

In the presence of non-hexose carbon sources, E. coli will eventually transition to gluconeogenesis. For

carbon sources feeding into the TCA cycle (e.g., succinate or acetate), this requires production of PEP

33
 93
via either PEP carboxykinase or malic enzyme-PEP synthase , both of which consume ATP. If PEP

carboxylase is simultaneously active, ATP will be wasted by futile cycling. Because gluconeogenic growth
19
is associated with reduced FBP levels , we hypothesized that another physiological function of PEP

carboxylase regulation is to prevent futile cycling during gluconeogenesis. During steady growth on

acetate, PEP carboxylase protein is present intracellularly at > 25% of the protein level of glucose grown
85 94
cells (Figure1.25). In wild type cells, however, futile cycling is so low as to be undetectable .

To examine the role of PEP carboxylase allosteric regulation in limiting futile cycling, we measured

metabolite levels and growth yields in succinate- or acetate-fed E. coli Δppc/pCA24N-ppcR313Q and

Δppc/pCA24N-ppc at various levels of inducer. Addition of 10 µM IPTG resulted in expression of PEP

carboxylase to a similar level as that found in wild type cells growing on acetate (Figure1.25b). Increasing

expression of wild type PEP carboxylase had only a slight effect on cellular metabolite levels (Figure

1.26a). Growth rate, succinate and acetate uptake rates, and growth yields decreased only modestly with

higher PEP carboxylase expression (Figure 1.26b). In contrast, expression of the R313Q mutant resulted

in substantial depletion of glycolytic intermediates and build-up of TCA cycle four-carbon units, consistent

with marked PEP carboxylase activity and associated futile cycling. Moreover, growth rate fell

monotonically with increasing PEP carboxylase R313Q expression, approaching zero at high expression

levels. This decrease in growth rate reflected a combination of lower acetate and succinate uptake

(presumably due to TCA cycle overloading impairing net carbon import) and worse yield per carbon

consumed (presumably due to futile cycling).

34
Figure1.25. Quantitation of PEP carboxylase protein concentration in cells growing on glucose or
acetate, with or without inducible expression from plasmid with IPTG.
(a). Identical inducible expression wild type and R313Q PEP carboxylase. Wild type E. coli,
Δppc/pCA24N-ppc (PPC WT), and Δppc/pCA24N-ppcR313Q (PPC R313Q) cells were grown in glucose
minimal media, with the latter two strains grown in the presence of 100 µM IPTG.

(b). Left panel: PEP carboxylase protein concentration in wild type E. coli cells growing on acetate is ~25%
compared to protein concentration of wild type cells growing on glucose. Right panel: In the presence of
10 µM IPTG, PEP carboxylase protein concentration in PPC WT and PPC R313Q were approximately
equal to protein concentration of wild type cells growing on acetate. Wild type E. coli, PPC WT, and PPC
R313Q were grown in either glucose or acetate minimal media, with the latter two strains in the presence
of indicated concentration of IPTG.

In both (a) and (b), PEP carboxylase protein concentration was quantitated by SDS-PAGE followed by
targeted LC-MS/MS. The y axis represents the relative enzyme level (PEP carboxylase concentration in
wild type E. coli cells growing on glucose defined as 1). Bars reflect data (mean ± range of N = 2
biological replicates).

35
Figure 1.26. The FBP-insensitive variant of PEP carboxylase causes futile cycling on
gluconeogenic media.

PEP carboxylase does not productively contribute to growth on acetate or succinate, and thus optimal
regulation would eliminate its activity. Persistent activity under these conditions would be expected to
deplete PEP and upstream metabolites, to elevate TCA intermediates, and to reduce growth yields. (a)
Intracellular metabolite concentrations in Δppc/pCA24N-ppc (PPC WT) and Δppc/pCA24N-ppcR313Q
(PPC R313Q) as a function of added IPTG. Data are shown in heat map format, with each line a
particular strain and carbon source. Metabolite levels were averaged, normalized to cells without
recombinant enzyme expression (IPTG = 0), and the resulting fold changes log transformed. Data for
Δppc/pCA24N-ppcR313Q at 200 μM IPTG is missing because the strain would not grow. (b) Growth
parameters of Δppc/pCA24N-ppc and Δppc/pCA24N-ppcR313Q growing on succinate (left) and acetate
(right). The x axis represents different IPTG concentration in μM, and the y axis represents parameters
shown in the plots. Points reflect data (mean ± range of N = 2 biological replicates).

36
1.3 Discussion

Here we show that ultrasensitive allosteric activation of PEP carboxylase by FBP plays a major role in

controlling central metabolic activity in E. coli. At the protein level, such regulation involves sensitization of

PEP carboxylase to FBP by physiological concentrations of acetyl-CoA and aspartate. At the systems

level, it enables accumulation of PEP upon glucose removal, which in turn provides an advantage in

reinitiating glycolysis when glucose reappears.

A key question is whether the ultrasensitive nature of the regulation is truly required to achieve PEP

accumulation. A simple mathematical model of glycolytic regulation demonstrates that the observed

ultrasensitivity is a prerequisite for PEP showing inverse concentration changes to FBP (Figure 1.18),

and thus PEP accumulating when glycolytic flux decreases, as has been observed experimentally in E.
95,96
coli and other organisms . Like PEP carboxylase, the pyruvate kinase isozyme which is expressed in
97
the presence of glucose (PykF) is activated by FBP . We anticipate that, in the presence of physiological

metabolite concentrations, PykF will also respond ultrasensitively to FBP, rendering PEP accumulation a

robust physiological response to glucose removal.

The effects of PEP carboxylase and pyruvate kinase regulation may be augmented by allosteric
98
inactivation of phosphofructokinase by PEP : initial depletion of FBP and build-up of PEP will promote

further inactivation of both enzymes, thereby locking glycolysis in an off state until a suitable glycolytic

substrate reappears.

99
A putative advantage of allosteric regulation is speed . This seems paramount for shut off of PEP

carboxylase and pyruvate kinase upon glucose removal, where termination of glycolytic egress must out

race depletion of upstream glycolytic compounds. In contrast, in cases where enzymes in E. coli are
100
inactivated via covalent modification , there is not a corresponding need for fast regulation. For example,

the central nitrogen metabolic enzyme glutamine synthetase is turned off by covalent modification upon
101
ammonia upshift ; in this case, the delay introduced by the covalent modification step may be

advantageous, favoring uptake of a useful initial burst of nitrogen.

37
An unexpected feature of PEP carboxylase’s allosteric activation is its ability to produce switch-like flux

control that overrides the effects of substrate levels. PEP carboxylase is a tetrameric enzyme, comprised
102 97
of a four identical catalytic units . The same is true of PykF . A likely reason for the evolution of
103-107
multimeric enzymes is to enable ultrasensitivity, to substrate , to feedback inhibitors in
108,109
biosynthesis , or, as shown here, to feed forward activation in central carbon metabolism. The

existence of such intrinsically ultrasensitive enzymes confers metabolism with substantial capabilities for

self-regulation.

38
1.4 Methods

Microbial culture conditions and extraction.

Detailed protocols for preparing filter cultures and extracting metabolites in Escherichia coli have been
36,37 110
published . Briefly, E. coli K-12 strain NCM3722 was grown at 37°C in minimal salts media with 10

mM ammonium chloride as the nitrogen source and 0.4% (w/v) glucose as the carbon source. To obtain

exponential-phase cultures, saturated overnight cultures were diluted 1:30 and then grown in liquid media

in a shaking flask to A600 (absorbance at 600 nm) of ~ 0.1. A portion of the cells (1.6 mL) was then

transferred to nylon membrane filters (Millipore) resting on vacuum filter support. Once the cells were

loaded, the membrane filters were transferred to media-loaded agarose plates. To measure growth, filters

were washed thoroughly with 1.6 mL fresh medium and A600 determined. For metabolomics experiments,

cells were grown to A600 of ~ 0.4 prior to initiation of experimental manipulations (e.g., glucose removal)

and quenching of metabolism. Metabolism was quenched by direct and immediate transfer of the filters

into – 20 °C extraction solvent (40:40:20 acetonitrile/methanol/water with 0.1 M formic acid to quantify
79
nucleotide triphosphates and coenzymes, and without formic acid to quantify other metabolites ), and
37
serial extraction was carried out as described previously . For glucose removal, filters were then

switched from glucose plates to plates containing no carbon source, 0.4% (w/v) acetate, 0.4% (w/v)

succinate, or 0.4% (w/v) glycerol as the carbon source, and extracted at 1, 2, 5, 10, 20, 30, 60, 90, 120

min after the switch. Control experiments in which to cells were switched to a glucose-free plate at t = 0,

and again to a new glucose-free plate at t = 1 min in the case that the first transfer did not eliminate all

glucose, showed identical results to the single transfer (note that for yeast grown in 2% glucose, a single

filter transfer is insufficient to remove all glucose). For experiments involving recombinant enzyme

expression, an appropriate concentration of IPTG was added in both liquid media and agarose plates.

Metabolite and flux measurement.

Cell extracts were analyzed by reversed phase ion-pairing liquid chromatography (LC) coupled by

electrospray ionization (ESI) (negative mode) to a high-resolution, high-accuracy mass spectrometer

39
 5
(Exactive; Thermo Fisher) operated in full scan mode at 1 s scan time, 10 resolution, with compound
34
identities verified by mass and retention time match to authenticated standard . For a list of metabolites,

exact masses, and retention times, see Table 1.3. Confirmatory analysis for a subset of compounds was

conducted by a different reversed phase ion-pairing LC method coupled by negative mode ESI to a

Thermo TSQ Quantum triple quadrupole mass spectrometer operating in multiple reaction monitoring
32
mode . For a list of metabolites, scan events, and retention times, see Table 1.4. All data reported in the

paper is from the high-resolution LC-MS analysis; for the compounds listed in Table 1.4, the LC-MS/MS

analysis confirmed the biological trends (see Figure 1.27 for PEP and FBP data). Absolute intracellular
36
metabolite concentrations in steadily growing E. coli were previously determined . Metabolite

concentrations after perturbations were computed based on fold-change in ion counts relative to steadily-

growing cells (grown and analyzed in parallel) multiplied by the known absolute concentration in the

steadily growing cells.

In Figure 1, isomers such as F6P/G6P, GAP/DHAP, 3PG/2PG was lumped together due to similar

chromatographic retention time. When separate measurements of isomeric species are reported, this is

based on chromatographic resolution of the isomers.

Steady-state fluxes in glucose-fed, exponentially growing cells were inferred based on measured nutrient

uptake and waste excretion, growth rate and associated requirements for biomass production, and
13
metabolic flux analysis with 1- C-glucose tracer. For details, see Methods. Substrate uptake and waste

excretion rates were measured by a combination of LC-MS and NMR.

40
Figure 1.27. LC-high resolution MS analysis and LC-triple quadrupole MS/MS analysis yield similar
data.
Exemplary data are shown for the switch from glucose to acetate. The plots show raw ion counts for the
indicated compound from the targeted LC-triple quadrupole MS/MS analysis (multiple reaction monitoring,
or MRM) (x-axis) and from the LC-high resolution MS analysis (full scan) (y-axis). The results are linearly
2
correlated with R > 0.97. Points reflect means of two independent measurements by each method; lines
represent least square fitting.

Quantitation of PEP carboxylase

E. coli cells were grown on either glucose or acetate minimal media plates as described above. When the

culture reached A600 of ~ 0.4, the cells were washed off the filter with 1.6 mL ice-cold glucose or acetate

minimal liquid media and used as the steady state sample. For the 10 min carbon starvation and 10 min

acetate switch samples, following transfer to the appropriate glucose-free plates for 10 min, ice-cold no

carbon or acetate minimal liquid media were used to wash the cells off the filters. 25 µL cells were

pelleted, re-suspended in 10 µL SDS-running buffer (Bio-Rad), and boiled for 8 minutes. Proteins were

separated by SDS-PAGE on TGX 12% gels (Bio-Rad) and visualized by Coomassie staining (GelCode

Blue, Pierce). The band for PEP carboxylase (indicated by the strong staining of the over-expressed

enzyme) was excised, destained, and washed extensively. The sample was then subjected to in-gel thiol

reduction with DTT, alkylation with iodoacetamide, and overnight trypsin digestion (Trypsin Gold,
111
Promega) . Peptides were eluted from the gel, concentrated, and reconstituted in 3% acetonitrile/0.1%

formic acid. Sample concentration and desalting were performed using a trapping capillary column (150

μm x ca. 40 mm, packed with 5 μm, 100 Å Magic AQ C18 material, Michrom) at a flow rate of 5 μL/min for

3.5 min, while separation was achieved using an analytical capillary column (75 μm x ca. 12 cm, packed

with 1.7 μm, 100 Å C18 BEH material, Waters), under a linear gradient of A and B buffers (buffer A: 3%

acetonitrile/0.2% formic acid/0.1% acetic acid; buffer B: 97% acetonitrile/0.2% formic acid/ 0.1% acetic

acid) over 60 min at a flow rate of approximately 300nL/min. Samples were analyzed by reversed-phase

LC-MS performed on a nano-flow capillary high pressure HPLC system (Eksigent) coupled to a Thermo

LTQ-Orbitrap mass spectrometer, outfitted with a NanoMate ion source robot (Advion). The instrument

was set to operate in pseudo-SRM mode. PEP carboxylase was quantitated based on the most

41
 3+
prominent peptide ion, the [M+3H] ion at m/z 405.22837, derived from the sequence AGIELTLFHGR,
2+
and its most prominent daughter ion, the y8 fragment ion at m/z 487.06. Nanospray ionization was

conducted using the NanoMate in LC-coupling mode at 1.74 kV. The NanoMate was used to feed an

LTQ-Orbitrap mass spectrometer with the LTQ heated capillary at 200 °C. Further information on mass

spectrometry parameters is included in the main text.

Preparation of PEP carboxylase

E. coli PEP carboxylase was purified from bacteria that carried plasmid pCA24N (from ASKA library) that
112
encoded His-tagged PEP carboxylase using Ni-NTA spin columns (QIAGEN). The purification protocol

was modified as below to obtain nearly homogenous expression of tetrameric enzyme. These

modifications were based on our observation by gel filtration that excessive PEP carboxylase expression

causes the formation of octamers, which are less responsive to FBP activation and aspartate inhibition.

The modified protocol is as follows: a saturated overnight culture of E. coli containing the PPC expression

plasmid and grown in glucose minimal media was diluted 1:50 into fresh glucose minimal media. After the

cells reached A600 of ~0.5, IPTG was added to a final concentration of 0.2 mM and PEP carboxylase

expressed for 1.5 hours. Cells were pelleted from 100 mL of culture medium and resuspended in 1 mL

PBS buffer, treated with lysozyme, and lysed in a Misonix 3000 sonicator. Cell lysates (supernatants)

were collected after 30 min of centrifugation at 13,000 g, and applied to Ni-NTA spin columns following

protocols provided by QIAGEN. Eluate containing 200 µg/mL PEP carboxylase was collected and used

for enzyme assays.

For measurement of PEP carboxylase activity in cell extracts, E. coli cells growing on plates were washed

off the filter with ice-cold liquid media, pelleted, re-suspended in PBS buffer, treated with lysozyme, and

lysed in a Misonix 3000 sonicator. The protein preparation process was carried out at 4°C.

42
Measurement of PEP carboxylase activity

The PEP carboxylase reaction was coupled to the malate dehydrogenase (MDH) reaction as described
40
previously . However, instead of using a spectrophotometer to measure the decrease of NADH at 340

nm, we used LC-MS to directly measure the reaction product malate. The rationale is as follows: Even

purified PEP carboxylase enzyme contains contaminants which slowly oxidize NADH. Thus the decrease

of NADH is not solely due to the malate dehydrogenase reaction. This becomes important when

quantitating the low PEP carboxylase activity found in the low FBP conditions which are important

physiologically after glucose removal. Since malate is the product of MDH reaction, it can only come from

oxaloacetate in the purified enzyme system. Also, the background malate signal in purified enzyme

sample before adding PEP carboxylase is almost zero, which facilitates accurate quantitation in low

activity conditions. To differentiate malate produced by MDH (i.e., the product) from malate added as an
2
enzyme inhibitor, we added 2,2,3- D3-malate as the enzyme inhibitor. This is fully resolved from
13
unlabeled malate by high resolution LC-MS. Also, U- C-aspartate was added to avoid potential formation

of malate by the degradation of aspartate.

The basic reaction mixture contained, in a total volume of 0.5 mL, Tris-HCl buffer, pH 8.5, 100 mM; MgCl2,

10 mM; KHCO3, 10mM; NADH, 0.5 mM; malate dehydrogenase, 10 Units/mL; and PEP carboxylase, 0.1-

2 µg/mL (depending on how much PEP was added). Substrate PEP and allosteric regulators acetyl-CoA,

FBP, aspartate, malate, and GTP were added at physiological concentrations based on the metabolomics

experiments. Assays were carried out in 30°C in an Eppendorf Thermomixer with mixing speed of 600

rpm. Reactions were started by adding PEP. At 1, 2, 5, 10, and 20 min thereafter, 20 μL of the reaction

mixture was pipetted into 180 μL of – 20 °C extraction solvent (40:40:20 acetonitrile/methanol/water

without acid) to quench the reaction and extract metabolites. Samples were then further diluted five-fold

followed by LC-MS measurement.

For measurement of PEP carboxylase activity in cell extracts before and after glucose removal (Figure

1.14b), E. coli K-12 strain NCM3722 was grown on glucose minimal media plates as described above.

When the culture reached A600 of ~ 0.4, the cells were washed off the filter with 1.6 mL ice-cold glucose

minimal liquid media and used as the glucose steady state sample. Cells were then pelleted, re-
43
suspended in 0.3 mL PBS buffer, treated with lysozyme and lysed in a Misonix 3000 sonicator. Cell

lysates (supernatants) were collected after 30 min of centrifugation at 13,000 g. The whole process was

carried out at 4 °C. . For the 10 min carbon starvation and 10 min acetate switch samples, following

transfer to the appropriate glucose-free plates for 10 min, ice-cold no carbon or acetate minimal liquid

media were used to wash the cells off the filters. In the enzyme assays, 50 μL cell extract instead of

purified enzyme was used to make 0.5 mL reaction mixture described above. 0.61 mM acetyl-CoA was

added to activate the enzyme and 0.18 mM PEP was added to start the reaction.

Mutant construction and site-directed mutagenesis

The plasmid pCA24N in ASKA library containing PEP carboxylase was used as the template for site-

directed mutagenesis reactions following the protocol of Quickchange Site-Directed Mutagenesis by

Stratagene. The pair of primers used for substitution of amino acids was: 5’-

GAACCGTATCagTATCTGATGAAAAACC-3’ and 5’-CATCAGATActGATACGGTTCTGCGGC-3’ to obtain

the R313Q mutant (R  Q mutants at positions 144, 145, and 192 were constructed analogously but did

not yield as complete desensitization to FBP as the R313Q mutant). The mutation site is described in

lowercase letters. Briefly, the template plasmid was first denatured and annealed with the oligonucleotide

primers containing the desired mutation. Then using the polymerase chain reaction (PCR), the primers

were extended resulting in nicked circular strands. This nicked mutant plasmid was repaired in the

competent cells and then transformed into the NCM3722 ppc deletion strain. The template plasmid was

also transformed to the ppc deletion strain and used as the control. The resulting strains were named as:

Δppc/pCA24N-ppc (control) and Δppc/pCA24N-ppcR313Q.

Metabolic Flux Analysis of wild type and ΔpykA/ΔpykF cells

To determine PEP and TCA cycle fluxes, including pyruvate kinase flux, for cells growing freely on

glucose in filter cultures, metabolite labeling patterns (aspartate, PEP, valine, and xylulose-5-phosphate)
13
were measured after feeding 1- C-glucose. These data were integrated with glucose consumption and

44
acetate excretion rates and cell growth rates, with the combined information sufficient for steady-state flux

quantitation (Figure 1.8). E. coli cells were grown on top of an agarose-media mixture containing 0.4% 1-
13
C glucose. At A600 of ~ 0.4, cells were extracted as described above and the metabolome quantitated

by LC-MS. In addition, glucose and acetate in the underlying agarose-media were measured by NMR;

NMR analysis also ruled out the significant excretion of other metabolites.

Steady-state fluxes were fit to a reduced TCA cycle flux model which omits malic enzyme and PEP

carboxykinase (Figure 1.8c). The former was omitted based on fitting the experimentally observed valine

labeling pattern to a linear combination of pyruvate synthesis from PEP and from malic enzyme, and

obtaining a malic enzyme contribution of zero (upper error bound < 2%; data not shown). The latter was

omitted based on PEP’s labeling pattern being consistent with the combined effects of glycolysis and the

pentose phosphate pathway without significant gluconeogenic flux. Lack of gluconeogenic flux was

confirmed by spiking labeled acetate into cells growing freely on glucose and observing rapid TCA cycle

labeling but no detectable labeling of PEP or other glycolytic compounds (upper error bound < 1%; data

not shown).

Fluxes into biomass, including the PEP carboxylase flux (FPPC), were calculated based on growth rate

from the known metabolite requirements for macromolecular synthesis. The largest route of pyruvate

production from PEP is the PTS, and FPTS equals the experimentally measured glucose uptake rate.

Combined with the experimentally measured acetate excretion rate (FACex), these fluxes are sufficient to

determine the flux from TCA cycle turning into oxaloacetate (FTCAOAA) for the ΔpykA/ΔpykF cells by flux

balancing. The ratio FPPC / FTCAOA is also determined by the aspartate labeling pattern after feeding 1-
13
C-glucose, and agrees within experimental error with the value determined by flux balancing (horizontal

black lines in aspartate panel of Figure1.8a show the calculated labeling pattern based on the FPPC /

FTCAOAA ratio from flux balancing; bars and error bars show the experimental data). Thus, for the

ΔpykA/ΔpykF cells, the system is over-determined with the redundant independent data result in a single

coherent flux set. For the wild type cells, the FPPC / FTCAOAA ratio based on the aspartate labeling pattern

(identical to the ΔpykA/ΔpykF cells based on the indistinguishable labeling) was used to calculate

FTCAOAA. Pyruvate kinase flux (FPYK) can then be calculated by flux balancing, yielding a flux of 0.12

45
mmol/g CDW/h with a standard error of 0.62 mmol/g CDW/h. Flux error estimates were calculated based

on experimental data error by standard rules for propagation of error through the addition, subtraction,

and multiplication operations involved in the flux calculations.

Calculation of aspartate labeling pattern based on FPPC / FTCAOAA ratio

The aspartate labeling pattern was calculated from the PEP labeling pattern based on the above TCA

cycle model by full isotopomer balancing (Figure1.8b). This requires not only knowledge of the number of

labeled carbon atoms in PEP, but also positional labeling, which cannot be directly measured by LC-MS.
13 13
Starting from 1- C1-glucose, glycolysis makes an equal mix of unlabeled and 3- C1-PEP. The oxidative

pentose phosphate pathway selectively burns C1 of glucose as CO 2. This leads to excess unlabeled
13 13
compared to C1 PEP as observed experimentally (Figure1.8a). In addition, we observed some C2-

PEP, which can be made through the non-oxidative pentose phosphate pathway; such activity also yields
13 13
1- C1-PEP, whose abundance was determined by the amount of C2-PEP and the xylulose-5-phosphate

(Xu5P) labeling pattern (the ratio of non-labeled over singly + doubly-labeled Xu5P equals the ratio of 1-
13 13
C1 over 1,3- C2 PEP).

13
Quantitation of PEP consumption flux based on data obtained from transient C-glucose labeling

followed by glucose removal

In the case of the 8 min switch to no carbon, the total PEP pool is roughly stable from 20 – 60 min, and

steadily becomes less labeled during this time. This is consistent with steady production of unlabeled

PEP from macromolecule degradation and matching steady consumption PEP (with labeled and

unlabeled PEP consumed in proportion to their fraction of the total PEP pool). To quantitate the flux from

10 – 60 min, we used the total pool of lower glycolytic intermediates (FBP + GAP/DHAP + DPG/3PG/2PG

+ PEP, which are interconverted via reversible reactions). This pool is roughly steady from 10 – 60

minutes, and has the same labeling dynamics as PEP (Figure 1.6c). As the source of labeled carbon is

depleted after 10 minutes, production of unlabeled PEP from macromolecules (glycogen, glycerolipids, or

nucleic acids) and total consumption of PEP (labeled + unlabeled) were balanced. This steady-state flux

46
 37
is given by the rate of conversion from labeled to unlabeled carbon by kinetic flux profiling . Briefly, the
13 L
kinetics of C lower glycolytic pool X is determined by

dX L XL
  f T (1)
dt X
T
where f is the steady-state flux and X is the total (unlabeled + labeled) lower glycolytic pool size. The

solution of this ordinary differential equation (ODE) starting at t0 = 10 min after carbon removal is

X T  X 0L 
ln  L   t  t 0  (2)
f X 

L 13 T L L
where X 0 is the C pool size at t0. The plot (X ln(X 0/X )) vs. (t – t0) gave a straight line (Figure 1.12a),

consistent with the above analysis. Least square fitting results in f = 0.051 mM/min, which is 99.6% less

than PEP carboxylase flux in the glucose-fed cells.

In the acetate switch case, the total lower glycolytic pool size drops after 10 min. However, total pool size
13
of unlabeled species is steady (Figure 1.6d). Accordingly, the kinetics of the C-labeled lower glycolytic
L
pool X is determined by

dX L XL
  f ex (3)
dt 
X L  XU 
U
where fex is the steady PEP consumption flux (efflux) and X is the constant unlabeled lower glycolytic

pool size. The solution of this ODE starting at t0 = 10 min is

X U
  
ln X L  X L   f ex t  t 0   X U ln X 0L  X 0L (4)

L 13 U L L
where X 0 is the C pool size at t0. The plot (X ln X + X ) vs. (t – t0) gave a straight line (Figure 1.12b)

at 10 – 30 min, consistent with the above analysis. Least square results in fex = 0.15 mM/min, which

corresponds to a 99.0% inhibition of PEP carboxylase.

47
Simplified steady-state model of glycolytic regulation

To examine the regulatory requirements for PEP to increase when glycolytic influx decreases, we used a

simplified model of glycolysis as shown in Scheme 1.1

Scheme 1.1

where F is the PTS flux, k1 – k3 are the rate constants of the indicated reactions, and h is the hill-

coefficient for FBP activation of PEP carboxylase. Given that the bidirectional enzymes in glycolysis have
19
large Km, while unidirectional enzymes are saturated , it is reasonable to approximate the reactions

between FBP and PEP as first-order, and to assume that PEP carboxylase is insensitive to [PEP]. For

this simple model, the steady state kinetics of FBP and PEP are given by:

d [FBP]
0  F  k1[FBP]  k 2 [PEP]2
dt

d [PEP]
0  2k1[FBP]  2k 2 [PEP]2  F  k 3 [FBP]h
dt

These equations have a discrete solution:

1
F h
[FBP]   
 k3 

48
 1
 F h
k1    F
[PEP]   k3 
k2

In the model, when F decreases, [FBP] always decreases. This matches experimental observations. The

qualitative behavior of PEP, however, depends on h. For h = 1, as F decreases, [PEP] decreases (see

Figure 1.18). For h < 1, the decrease is greater. For h > 1, however, [PEP] can increase as F decreases

as is experimentally observed. Thus, ultrasensitive regulation of PEP carboxylase is required to obtain the

experimentally observed inverse behavior of [FBP] and [PEP].

Table 1.3. Mass, retention time, and signal intensity of measured metabolites by the LC-high
resolution MS analysis.

average
intensity in
samples from
retention glucose-fed
Formula m/z time (min) cells
Pyruvate C3H4O3 87.0088 8.3 38200
4-aminobutyrate C4H9NO2 102.0561 4.5 353411
Serine C3H7NO3 104.0353 1 112723
Proline C5H9NO2 114.0561 1.12 15072
Fumarate C4H4O4 115.0037 13.4 34524
2-keto-isovalerate C5H8O3 115.0401 13.06 33289
Valine C5H11NO2 116.0717 1 43363
Succinate C4H6O4 117.0193 11.6 49220
threonine/homoserine C4H9NO3 118.051 1 182229
Thymine C5H6N2O2 125.0357 2.6 5783
citraconic acid C5H6O4 129.0193 12.95 216516
leucine/isoleucine C6H13NO2 130.0874 1.9 79015
asparagines C4H8N2O3 131.0462 1 7816
Ornithine C5H12N2O2 131.0826 1 15785
Aspartate C4H7NO4 132.0302 4.4 30804
Malate C4H6O5 133.0143 12.7 64026
Homocysteine C4H9NO2S 134.0281 1 1716
Hydroxybenzoate C7H6O3 137.0244 10.88 7387
α-ketoglutarate C5H6O5 145.0143 13.1 6383

49
 average
intensity in
samples from
retention glucose-fed
Formula m/z time (min) cells
Glutamine C5H10N2O3 145.0619 1 278323
Glutamate C5H9NO4 146.0459 4.6 836814
2-hydroxy-2-
methylbutanedioic acid C5H8O5 147.0299 12.6 178898
methionine C5H11NO2S 148.0438 1.6 12705
2,3-dihydroxybenzoic acid C7H6O4 153.0193 13.58 12010
orotate C5H4N2O4 155.0098 8.1 65364
phenylalanine C9H11NO2 164.0717 4.1 8084
phosphoenolpyruvate C3H5O6P 166.9751 13.7 1750
DHAP/GAP C3H7O6P 168.9908 9.3 45599
glycerol-3-phosphate C3H9O6P 171.0064 7.3 109925
acetyl-ornithine C7H14N2O3 173.0932 1 29588
citrulline C6H13N3O3 174.0884 0.9 5549
N-carbamoyl-aspartate C5H8N2O5 175.036 12 43638
2-isopropylmalic acid C7H12O5 175.0612 13.94 212239
tyrosine C9H11NO3 180.0666 2 15763
3-phosphoglycerate/2-
phosphoglycerate C3H7O7P 184.9857 13.58 46101
N-acetyl-glutamine C7H12N2O4 187.0724 7 54300
N-acetyl-glutamate C7H11NO5 188.0564 13 228460
citrate C6H8O7 191.0197 13.6 305171
tryptophan C11H12N2O2 203.0826 7.6 1897
pantothenate C9H17NO5 218.1034 11 7309
ribose-5-phosphate C5H11O8P 229.0119 7.2 73902
ribulose-5-
phosphate/xylulose-5-
phsophate C5H11O8P 229.0119 7.8 37402
shikimate-3-phosphate C7H11O8P 253.0119 13.44 8185
glucosamine-1-phosphate C6H14NO8P 258.0384 1.5 6610
hexose-phosphate C6H13O9P 259.0224 6.9 2092317
6-phospho-D-gluconate C6H13O10P 275.0174 13.38 135995
sedoheptulose-1/7-
phosphate C7H15O10P 289.033 6.95 127886
N-acetyl-glucosamine-1/6-
phosphate C8H16NO9P 300.049 7.3 41778
glutathione C10H17N3O6S 306.0765 8 1839115
octoluse bisphosphate C8H17O11P 319.0436 13.64 8372
CMP C9H14N3O8P 322.0446 8.7 6010
UMP C9H13N2O9P 323.0286 10.2 6346
cyclic-AMP C10H12N5O6P 328.0453 12.6 4107
50
 average
intensity in
samples from
retention glucose-fed
Formula m/z time (min) cells
dAMP C10H14N5O6P 330.0609 12 9898
fructose-1,6-bisphosphate C6H14O12P2 338.9888 13.5 381898
trehalose/sucrose C12H22O11 341.109 1.18 27378
AMP C10H14N5O7P 346.0558 11.3 34866
IMP C10H13N4O8P 347.0398 10.7 18055
GMP C10H14N5O8P 362.0507 10.7 6796
sedoheptoluse
bisphosphate C7H15O10P 368.9993 13.64 36022
dCDP C9H15N3O10P2 386.016 13.58 7768
octoluse 8/1P C8H18O14P2 399.0099 6 10120
dTDP C10H16N2O11P2 401.0157 13.58 36699
CDP C9H15N3O11P2 402.0109 13.3 3944
UDP C9H14N2O12P2 402.9949 13.5 20545
trehalose-6-Phosphate C12H23O14P 421.0753 7.2 9221
ADP C10H15N5O10P2 426.0221 13.7 158485
GDP C10H15N5O11P2 442.0171 13.5 28468
dCTP C9H16N3O13P3 465.9823 14.8 25053
dTTP C10H17N2O14P3 480.982 14.94 196251
CTP C9H16N3O14P3 481.9772 14.8 151378
UTP C9H15N2O15P3 482.9612 15 446726
dATP C10H16N5O12P3 489.9936 14.94 41853
ATP C10H16N5O13P3 505.9885 15.1 1621122
GTP C10H16N5O14P3 521.9834 15 217619
UDP-D-glucose C15H24N2O17P2 565.0477 13.25 1483574
UDP-D-glucuronate C15H22N2O18P2 579.027 14.7 766639
UDP-acetyl-glucosamine C17H27N3O17P2 606.0743 13.27 1009912
glutathione disulfide C20H32N6O12S2 611.1447 12.8 288314
NAD+ C21H27N7O14P2 662.1019 8.6 71795
NADH C21H29N7O14P2 664.1175 13.9 6960
NADP+ C21H28N7O17P3 742.0682 13.5 53337
FAD C27H33N9O15P2 784.1498 15 65385
acetyl-CoA C23H38N7O17P3S 808.1185 15.6 500335

51
Table 1.4. Parent and product ion formulas and masses, collision energy (CE), retention time (RT),
and signal intensity of metabolites measured by the confirmatory LC-triple quadrupole MS/MS
analysis

average
intensity in
par pro samples from
parent ion ent CE product ion duct RT glucose-fed
formula m/z (eV) formula m/z (min) cells
- -
fumarate C4H3O4 115 11 C3H3O2 71 30 702
- -
2-keto-isovalerate C5H7O3 115 11 C4H7O 71 28 469
- -
Succinate C4H5O4 117 10 C3H5O2 73 23 933
- -
Malate C4H5O5 133 12 C4H3O4 115 26 19384
- -
Hydroxybenzoate C7H5O3 137 21 C6H5O 93 22 1845
- -
α-ketoglutarate C5H5O5 145 11 C4H5O3 101 28 5739
- -
orotate C5H3N2O4 155 13 C4H3N2O2 111 15 5109
phosphoenolpyruva
- -
te C3H4O6P 167 20 PO3 79 31 574
- -
DHAP/GAP C3H6O6P 169 38 PO3 79 16 1904
glycerol-3-
- -
phosphate C3H8O6P 171 13 PO3 79 13 797
N-carbamoyl-
- -
aspartate C5H7N2O5 175 11 C4H6NO4 132 26 4842
3PG/
- -
2PG C3H6O7P 185 15 H2PO4 97 31 1027
- -
citrate C6H7O7 191 13 C5H3O3 111 32 7440
- -
ribose-5-phosphate C5H10O8P 229 40 PO3 79 13 622
ribulose-5-
phosphate/
xylulose-5-
- -
phsophate C5H10O8P 229 40 PO3 79 14 658
shikimate-3-
- -
phosphate C7H10O8P 253 17 H2PO4 97 12 985
- -
hexose-phosphate C6H12O9P 259 40 PO3 79 13 6568
6-phospho-D- C6H12O10P
- -
gluconate 275 11 H2PO4 97 29 3170
N-acetyl-
glucosamine- C8H15NO9
- -
phosphate P 300 32 PO3 79 13 564
C10H11N5
- -
cyclic-AMP O6P 328 31 C5H4N5 134 29 179
fructose-1,6- C6H13O13P
- -
bisphosphate 2 339 28 H2PO4 97 30 10370
- -
trehalose/sucrose C12H21O11 341 18 C6H11O6 179 3 1103
C15H23N2 C9H12N2O9
- -
UDP-D-glucose O17P2 565 23 P 323 30 103785
C15H21N2 C9H13N2O12
- -
UDP-D-glucuronate O18P2 579 24 P2 403 34 29960
UDP-acetyl- C17H26N3 C9H11N2O11
- -
glucosamine O17P2 606 26 P2 385 30 18346
52
 Chapter 2
Regulation of yeast pyruvate kinase by ultrasensitive allostery

independent of phosphorylation

Adapted from Yi-Fan Xu, Xin Zhao, David S. Glass, Farnaz Absalan, David H. Perlman, James

R. Broach, Joshua D. Rabinowitz “Regulation of yeast pyruvate kinase by ultrasensitive

allostery independent of phosphorylation” Molecular Cell 2012, 48: 52-62; and “Ultrasensitive

allosteric regulation of glycolytic efflux” at International Conference and Exhibition on

Metabolomics and Systems Biology

Abstract

Allostery and covalent modification are major means of fast-acting metabolic regulation. Their

relative roles in responding to environmental changes remain, however, unclear. Here we

examine this issue, using as a case study the rapid decrease in pyruvate kinase flux in yeast

upon glucose removal. The main pyruvate kinase isozyme (Cdc19) is phosphorylated in

response to environmental cues. It also exhibits positively-cooperative (ultrasensitive) allosteric

activation by fructose-1,6-bisphosphate (FBP). Glucose removal causes accumulation of

Cdc19’s substrate, phosphoenolpyruvate. This response is retained in strains with altered

protein-kinase-A or AMP-activated-protein-kinase activity or with CDC19 carrying mutated

phosphorylation sites. In contrast, yeast engineered with a CDC19 point mutation that ablates

FBP-based regulation fail to accumulate phosphoenolpyruvate. They also fail to grow on ethanol

and slowly resume growth upon glucose upshift. Thus, while yeast pyruvate kinase is covalently

modified in response to glucose availability, its activity is controlled almost exclusively by

ultrasensitive allostery.

53
2.1 Introduction

The major mechanisms of rapid metabolic regulation are allostery and covalent modification. Allosteric

regulation is intrinsic to the metabolic network, enabling automatic adjustment of fluxes in response to

metabolite concentration changes. Covalent modification involves distinct signaling enzymes, which

integrate metabolism with growth and other cellular functions. The most thoroughly studied form of

enzyme covalent modification is protein kinase-catalyzed phosphorylation. Both central metabolic

enzymes and major metabolic regulatory kinases are conserved across a wide range of eukaryotes. For

example, yeast and mammals encode the same set of glycolytic and citric acid cycle enzymes as well as

protein kinase A (PKA), protein kinase B (Akt in mammals, Sch9 in yeast), and AMP-activated protein

kinase (AMPK in mammals, SNF1 in yeast).

Glycolysis is the metabolic pathway typically exhibiting the highest flux in both yeast and mammals. A

common assumption, although subject to some dispute, is that the committed step, catalyzed by
9,42-44
phosphofructokinase-1 (Pfk1), functions as the critical glycolytic control point . Pfk1 is stimulated by

fructose-2,6-bisphosphate, produced by phosphofructokinase-2 (Pfk2). The activity of Pfk2 is in turn

regulated by a kinase cascade. In yeast, PKA phosphorylates Pfk2, elevating fructose-2,6-bisphosphate

113
levels and thereby enhancing Pfk1 activity and glycolytic flux . This regulation is physiologically relevant

in yeast during the shift from respiratory to fermentative conditions in order to more rapidly initiate
67,114
glycolysis when glucose suddenly becomes available . Thus, kinase-based signaling and allostery

work in concert to enable rapid regulation of phosphofructokinase.

The terminal step in glycolysis, catalyzed by pyruvate kinase, is also irreversible and tightly regulated.

Humans encode four isozymes of pyruvate kinase, one of which (PKM2) plays a central role in
45
oncogenesis . Yeast encode two isozymes, Cdc19 and Pyk2, with Cdc19 catalyzing most of the flux and
115
therefore essential for fermentative growth . The gene name CDC19 reflects the fact that temperature-

sensitive cdc19 mutants arrest growth in G1 during growth on glucose, for as yet unknown reasons.

116
Both phosphorylation and allostery contribute to pyruvate kinase regulation. El-Maghrabi et al.

reported that PKA deactivates liver pyruvate kinase, while several groups have reported that PKA

54
 47-50
activates Cdc19 in yeast by phosphorylating serine 22 . In addition, most eukaryotic pyruvate kinases,

including Cdc19, are allosterically activated by the product of phosphofructokinase, fructose-1,6-

54-56
bisphosphate (FBP) .

117,118
During gluconeogenic growth in yeast, Cdc19 is present but inactive . This enzyme inactivation limits

futile cycling during steady gluconeogenic growth. To investigate pyruvate kinase regulation further, we

have characterized the response of yeast metabolome to sudden changes of its nutrient environment.

Short-term nitrogen or phosphate starvation has essentially no effect on the concentrations of glycolytic

metabolites. However, glucose removal causes a dramatic accumulation of pyruvate kinase’s substrate,

phosphoenolpyruvate (PEP). This observation suggests that yeast can rapidly suppress Cdc19 enzyme

activity upon glucose deprivation. Here we examine how this inactivation is achieved, with an eye towards

potential general lessons regarding enzyme regulation.

One mechanism that could account for the rapid inactivation of Cdc19 following glucose depletion

involves PKA inactivation. Another involves depletion of FBP (Scheme 2.1). To distinguish between

these possibilities, we use a combination of metabolomic, genetic, and chemical genetic approaches. We

demonstrate that inhibition of PKA selectively alters trehalose and cyclic-AMP (cAMP) levels while not

altering the levels of other metabolites, including PEP. Instead, we find that PEP levels are controlled by

positively cooperative allosteric activation of Cdc19 by FBP. We show that this allosteric regulation is

required for gluconeogenic growth. In addition, it contributes to energy charge homeostasis upon glucose

upshift: accumulation of lower glycolytic intermediates upon glucose removal provides a readily

accessible reservoir of high energy phosphate bonds that can be used to enable rapid phosphorylation of

glucose when it becomes available. Such rapid phosphorylation primes glycolysis on glucose re-addition

and contributes to efficient growth in an oscillating glucose environment. These results support a primary

role in microbial metabolic flux regulation for allosteric mechanisms intrinsic to the metabolic network itself,

and raise the possibility that many cases of enzyme covalent modification may be functionally irrelevant.

55
Scheme 2.1

56
 2.2 Results

Glucose removal results in PEP accumulation due to pyruvate kinase inactivation

We used liquid chromatography – mass spectrometry (LC-MS) to measure short-term changes in the

metabolome profile of S. cerevisiae cells subjected to rapid transition from minimal media to media

lacking glucose, nitrogen or phosphate (Figure 2.1). For glucose removal, yeast cells were grown either

in liquid batch cultures or on filters residing on the surface of an agarose plate. The filter culture technique

enables rapid, non-disruptive nutrient removal and metabolome quenching. However, this method was

unreliable for experiments using ATP analogues to control genetically engineered kinases, due to variable

diffusion of analogues to the cells. Moreover, the filter culture technique retains trace amounts of nutrient

following transfer to nutrient-free media. For example, the metabolome pattern following glucose removal

using filter culture closely mimics that of a downshift in liquid culture to ~ 0.01% glucose (Figure 2.2). For

nitrogen and phosphate starvation, only liquid cultures were examined and these were grown in nitrogen

or phosphate limited media before nutrient deprivation, to deplete the otherwise large intracellular store of

these two nutrients.

Irrespective of the culturing method applied, within 30 minutes of glucose removal, FBP and

dihydroxyacetone phosphate (DHAP) decreased by more than 15-fold, and PEP increased by at least 10-

fold (Figure 2.1A). In contrast, nitrogen or phosphate starvation failed to alter the concentrations of

glycolytic metabolites (Figure 2.1B). Thus, we focused on the metabolic regulatory events induced by

glucose removal. The increase in PEP during glucose removal suggests that cells either rapidly suppress

PEP consumption or rapidly induce gluconeogenesis following glucose deprivation. To evaluate the

possibility that the increased PEP pool results from gluconeogenesis, we switched cells from unlabeled
13
glucose to U- C-ethanol. If PEP were being produced by gluconeogenesis, we would expect
13
incorporation of C into PEP in this experiment. However, we observed no detectable PEP labeling (<

1%) over the first 2 hours following glucose removal, even though a significant amount of labeled malate

(>80%) accumulated during this time, documenting utilization of the labeled ethanol (Figure 2.3). This

result rules out substantial PEP formation by PEP carboxykinase in the gluconeogenesis pathway on this

time scale, leading us to conclude that PEP consumption is curtailed.

57
Figure 2.1. Glucose removal results in PEP accumulation.
(A). S. cerevisiae cells (strain Y3358) growing in minimal media containing 2% glucose were switched to
minimal media containing no carbon, 2% glycerol + 2% ethanol, or 2% ethanol. After the indicated
duration of glucose removal, the metabolome was quantified by LC-MS. Experiments were performed
using either liquid culture (more complete glucose removal) or filter culture (faster sampling). (B).S.
cerevisiae cells (strain Y3358) growing in minimal media were switched to minimal media containing no
nitrogen or no phosphate and metabolome quantitated by LC-MS. Experiments were performed using
liquid culture. In (A) and (B), data are shown in heat map format, with each line reflecting the dynamics of
a particular compound in a particular culture condition. Metabolite levels of biological duplicates were
averaged, normalized to cells growing steadily in glucose (time zero), and the resulting fold changes log
transformed. G6P, glucose-6-phosphate; F6P, fructose-6-phosphate; FBP, fructose-1,6-bisphosphate;
GAP, glyceraldehyde-3-phosphate; DHAP, dihydroxyacetone phosphate; DPG, 1,3-diphosphateglycerate;
3PG, 3-phosphoglycerate; 2PG, 2-phosphoglycerate; G6P and F6P, GAP and DHAP, and 3PG and 2PG
were not differentiated by the LC-MS method employed.

58
Figure 2.2, Moving a filter culture to a glucose-free plate is similar to downshift in liquid culture to
~ 0.01% glucose.
S. cerevsiae cells growing freely in minimal media containing 2% glucose were switched to minimal liquid
media containing either no carbon or a trace amount of glucose as indicated. After the indicated duration
of glucose removal or downshift, the metabolome was quantitated by LC-MS. Data are shown in heat
map format, with each line reflecting the dynamics of a particular compound in a particular culture
condition. Experiments were performed using either liquid culture (more complete glucose removal) or
filter culture (faster sampling). The dynamics of G6P/F6P, FBP and GAP/DHAP in the glucose removal
experiment using filter culture are similar to the switch to 0.01% glucose experiment using liquid culture,
indicating similar residual glucose in these conditions. In all of these glucose downshifts, PEP strongly
increases, indicating similar pyruvate kinase regulation.

59
Figure 2.3. Accumulated PEP does not derive from TCA cycle intermediates.
(A). Experimental design. Yeast filter cultures (strain Y3358) were grown on 2% unlabeled glucose, and
13
suddenly transferred to 2% U- C-ethanol as the sole carbon source. A various time points thereafter,
13
metabolites were quantitated by LC-MS. (B). Exogenous C-ethanol is incorporated into TCA
intermediates but not PEP. The x axis represents minutes after the switch, and the y axis represents
absolute intracellular concentration (mean ± range of N = 2 biological replicates).

Only two known reactions in S. cerevisiae consume PEP: pyruvate kinase and aromatic amino acid
28
biosynthesis, with the former contributing ~99.5% of the total PEP consumption flux in growing cells .

Moreover, because expression of PYK2 is repressed during growth on glucose and Pyk2 does not by
115
itself support growth on glucose , regulation of flux from PEP upon glucose removal must be exerted

through the major pyruvate kinase isozyme, Cdc19. The overall Cdc19 concentration did not change

upon glucose removal (relative abundance 15 min after glucose removal is 0.97 ± 0.09 of initial

abundance, as measured by LC-MS/MS). Thus Cdc19 enzymatic activity must be regulated.

60
Glucose-sensing kinases are not required for pyruvate kinase regulation

119
Protein kinase A (PKA) plays a critical role in the response of yeast cells to glucose availability .

Glucose induces production of cAMP via Ras and Gpa2 signaling pathways, which in turn activate PKA.

Upon glucose removal, cAMP levels decrease and PKA is quickly inactivated, resulting in changes in the
119
phosphorylation state of a variety of metabolic enzymes . To explore the role of PKA in direct metabolic

regulation, we used a yeast strain with genetic modifications to all PKA isozymes
M164G M147G M165G
(tpk1 tpk2 tpk3 ) that result in their specific inhibition by the ATP analogue 1NM-PP1. We
as 120
refer to this strain as the PKA analog sensitive strain (pka ) . Consistent with prior literature, analogue
as
addition to the pka yeast strain induced greater than 10-fold changes in roughly half of transcripts, with

stress response and catabolic gene transcripts increasing and ribosomal ones decreasing (Figure 2.4A).

In addition, growth slowed by ~ 10% (Figure 2.5A).

as
Metabolomic analysis of the pka strain, before and after analog addition, revealed that out of 133

measured metabolites only two showed a substantial concentration changes upon acute PKA inactivation:

cAMP and trehalose (Figure 2.4B). PKA has been proposed to enhance cAMP degradation by activating
119
either or both phosphodiesterases, Pde1 and Pde2 . This regulation may provide feedback inhibition of

the cAMP-PKA signaling circuit. Consistent with this hypothesis, we find that PKA inactivation leads to

dramatically higher cAMP levels. PKA has also been shown to inhibit trehalose-6-phosphate phosphatase,
121-123
whose product is trehalose, and to activate trehalase, which converts trehalose into glucose . Such

regulation would predict that inhibition of PKA should lead to accumulation of trehalose, as we observe.

Importantly, our results demonstrate that PKA inhibition not only promotes higher trehalose, but is alone

sufficient to produce marked trehalose accumulation. The effects of PKA inhibition on cAMP and

trehalose levels document a role for PKA in direct metabolic regulation but suggest that the effect is

limited to a few soluble metabolites.

61
Figure 2.4. Transcriptome and metabolome response to inhibition of glucose-sensing kinases.
(A) Transcriptome. (B) Metabolome. Note that PEP levels are robust to manipulation of glucose-signaling
E392A
kinases but sensitive to a point mutation of Cdc19 that ablates allosteric activation by FBP (cdc19 ). In
the columns marked “glucose”, transcripts and metabolites of yeast cells growing in minimal liquid media
containing 2% glucose were measured during exponential phase starting at A 600 of 0.4. When indicated, 1
as
µM analogue 1NM-PP1 was added at time 0 to deactivate genetically engineered PKA (pka ). In the
columns marked “glucoseno carbon,” cells were switched to no carbon and metabolome quantitated by
LC-MS. When indicated, 1 µM analogue mMe-PP1 was added at time 0 to deactivate genetically
62
 as
engineered SNF1 (snf1 ). Higher analogue concentrations (2.5 uM and 25 uM) were also tested, with
identical results observed. All experiments in this figure were performed using liquid culture, and all
reported values are log2 transformed fold changes relative to time 0; all metabolites data are mean of
as as
duplicate samples at each time point . Strains: Y2864, auxotrophic WT; Y3561, pka ; Y3504, snf1 ;
E392A
Y3358, prototrophic WT; Y3898, cdc19 .

63
Figure 2.5, Inhibition of PKA and SCH9 slightly impairs growth but does not change the level of
PEP level.
as as as
(A). S. cerevisiae wild type, pka and pka sch9 cells were grown in glucose minimal media. When
indicated, analogue 1NM-PP1 was added at A600 of 0.1. Thereafter, growth rate was determined. Bars
reflect data (mean ± range of N = 2 biological replicates). (B). Pyruvate kinase regulation is robust to
chemical genetic inhibition of both PKA and Sch9. Metabolites of S. cerevisiae cells growing freely in
minimal liquid media containing 2% glucose were measured by LC-MS at different time points during their
exponential growth starting at A600 of 0.4. When indicated, analogue 1NM-PP1 was added at time 0 to
as as
deactivate genetically engineered PKA (pka ) or Sch9 (sch9 ). All experiments in this figure were
performed using liquid culture. Metabolite levels of biological duplicates were averaged, normalized to
metabolite levels at time 0, and the resulting fold changes log transformed.

48,49
Since Cdc19 activity has been reported to be activated by phosphorylation via PKA both in vitro and
124
in vivo , we hypothesized that PKA inactivation would decrease pyruvate kinase activity. However,

analogue addition did not lead to a detectable elevation in PEP levels, arguing against a major role for

PKA-based regulation of pyruvate kinase activity (Figure 2.4). Thus, while inactivation of PKA in vivo

does yield increased cAMP and trehalose levels, it does not significantly affect Cdc19 activity.

Protein Kinase B (Sch9) shares some targets with PKA and can partially compensate for PKA inactivation
119
. It is thus possible that persistent Sch9 activity masked a role for PKA in acute regulation of pyruvate
as as
kinase activity. To rule out such a possibility, we used a yeast strain, pka sch9 , in which both PKA and
T492G M164G M147G M165G
Sch9 were genetically modified (sch9 tpk1 tpk2 tpk3 ) such that they are specifically

inhibited by the ATP analogue 1NM-PP1. Treatment of this strain with 1NM-PP1 resulted in almost
as
identical metabolome changes as those observed for the pka strain (Figure 2.5B). Thus, in cells

growing in glucose, cAMP and trehalose are the only observed metabolites whose levels are strongly and

acutely regulated by PKA/Sch9 inhibition.

as as as
Our results with the pka and pka sch9 strains argue that PKA inactivation is not sufficient to

substantially inhibit pyruvate kinase. To evaluate the possibility that PKA inactivation might be necessary

upon glucose removal for pyruvate kinase inactivation, we used a yeast strain in which the PKA

regulatory subunit Bcy1 is deleted, resulting in constitutive PKA activity. The bcy1 strain grows normally
125
on glucose but fails to grow on gluconeogenic carbon sources . Upon glucose removal, however, like

64
wild type yeast, the bcy1 strain accumulates PEP (Figure 2.4). Thus, PKA inactivation is not necessary

for acute inactivation of pyruvate kinase upon glucose removal.

AMP-activated protein kinase, SNF1, is active in the absence of glucose and is required for growth on

most carbon sources other than glucose. To evaluate a possible role for SNF1 in pyruvate kinase
as
regulation upon glucose removal, we used a yeast strain, snf1 , in which SNF1 was genetically modified
I132G 120
(snf1 ) such that it could be specifically inhibited by the ATP analogue mMe-PP1 . We hypothesized

that activation of SNF1 might be required to inhibit pyruvate kinase upon glucose removal. By adding
as
mMe-PP1 simultaneously with glucose removal in the snf1 strain, we were able to test this possibility.
as
We found that PEP accumulated to the same extent in the snf1 strain upon glucose removal whether or

not mMe-PP1 was added concurrently (Figure 2.4). This result argues against a role for SNF1-based

phosphorylation in regulating pyruvate kinase.

Pyruvate kinase phosphorylation is not required for its regulation

To investigate further a potential role of phosphorylation on pyruvate kinase activity, we inserted a poly-

histidine tag at the N-terminus of the single genomic CDC19 locus to enabled efficient purification of the

endogenous enzyme. Yeast encoding the tagged pyruvate kinase showed a normal metabolomic

response to glucose removal (Figure 2.6). The abundance and covalent modifications of the tagged

enzyme were analyzed in cells growing on glucose and 15 min after glucose removal by purification and

digestion of Cdc19, followed by phosphopeptide enrichment and LC-MS/MS analysis of the resulting

peptide fragments. Quantitative comparison of the phosphopeptides under the two conditions was
15
achieved by mixing cells growing exponentially on glucose and N-ammonium sulfate with an equal

number of glucose-starved cells fed unlabeled ammonium sulfate. Analysis of both tryptic and Lys-C

fragments by LC-MS/MS yielded coverage of the entire protein sequence. Three phosphorylation sites

were identified (S9, T21 and S22), as well as previously unknown enzyme acetylation at K475. The T21

phosphorylation was low abundance, with the LC-MS/MS ion count for the phosphopeptide much less

than for the unmodified peptide, even after phosphoenrichment. The acetylation and phosphorylation of

65
S22 did not change appreciably following glucose removal. We did observe, however, a 30% decrease in

phosphorylation of T21 and an 80% decrease in phosphorylation of S9 in the glucose-starved cells

(Figure 2.7A).

To examine a potential role for the observed phosphorylation in regulating pyruvate kinase, we mutated

the genomic copy of CDC19 to convert S9, T21 or S22 individually to either alanine (A) or glutamate (E),

in an effort to mimic a perpetually dephosphorylated or phosphorylated protein, respectively. We then

analyzed the growth rate and the metabolome of each of the resulting strains before and after acute
T21E
glucose removal. The cdc19 variant (designed to mimic the phosphorylated enzyme, which is low

abundance but somewhat more prevalent in glucose-fed cells) resulted in slow growth and elevated PEP

even prior to glucose removal, due to constitutively impaired pyruvate kinase activity (see below). Each of

the other strains grew normally and, like the wild-type strain, accumulated PEP upon glucose removal,

consistent with these phosphorylation sites playing a minimal role in regulating Cdc19 activity (Figure

2.7B).

Ultrasensitive allostery is sufficient to modulate pyruvate kinase activity

Having shown that pyruvate kinase regulation does not depend on phosphorylation, we re-examined the

allosteric regulation of pyruvate kinase enzyme activity. A simplified mathematical model for glycolysis

indicates that, in the absence of other regulatory mechanisms, FBP must activate Cdc19 in a cooperative

fashion, i.e., with a Hill coefficient of greater than 1, for PEP to accumulate when FBP falls (Figure 2.8

and Methods). However, prior studies indicated that FBP activates Cdc19 but not in a cooperative

manner, i.e., with a Hill coefficient approximately equal to 1. These prior in vitro studies, however,
126-129
measured enzymatic activity with FBP and PEP at non-physiological concentrations . While the

cellular concentrations of PEP range from 0.03 to 0.8 mM, PEP was typically added in saturating

concentrations of 20 mM to facilitate the enzyme activity measurements.

66
Figure 2.6, Yeast expressing genomically His-tagged pyruvate kinase show the wild-type
metabolome response to glucose removal.
Wild type (WT) yeast and yeast expressing genomically His-tagged pyruvate kinase (HIS-Cdc19) were
switched to no carbon and metabolites measured by LC-MS. Experiments in this figure were performed
using liquid culture. Metabolite levels of biological duplicates were averaged, normalized to metabolite
levels at time 0, and the resulting fold changes log transformed.

67
Figure 2.7. PEP accumulation upon glucose removal does not require Cdc19 covalent
modification.
(A). Cdc19 phosphorylation before and after glucose removal. The relative extent of phosphorylation is
defined as: (ion count for the phosphopeptide)/(ion count for the phosphopeptide + ion count for its
unphosphorylated form) with ion counts for phosphopeptides measured in phosphoenriched samples and
those for unphosphorylated peptides in unenriched samples. Due to differences in enrichment and
ionization efficiency, this ratio provides only qualitative information regarding the stoichiometry of the
phosphorylation events. Differences before and after glucose removal reflect isotope ratio-based
measurements and are therefore quantitative. (B). Mutation of Cdc19’s phosphorylation sites to alanine
or glutamate does not prevent PEP accumulation. The x axis represents minutes after glucose removal,
and the logarithmic y axis represents absolute intracellular PEP concentration (mean ± range of N = 2
biological replicates).

68
Figure 2.8, Ultrasensitive regulation of pyruvate kinase is required for PEP to increase when
glycolytic influx decreases.
The simplified glycolytic model shown in Scheme 2.1 was solved quantitatively for [PEP] as a function of
glycolytic flux F. The Hill coefficient for activation of pyruvate kinase (h) determines the qualitative pattern
of the PEP response, with h > 1 required for PEP to increase when F decreases. Both the x and y axis
are presented in arbitrary units; the units depend on other parameters of the model, but the trends are
identical irrespective of these parameters.

We developed a sensitive, mass spectrometry-based assay for pyruvate kinase activity in order to

examine the effect of FBP activation in the presence of PEP at physiological concentrations (Figure

2.9A). We found that FBP is an ultrasensitive activator with Hill coefficients of 2.4 and 1.6 in the presence

of 0.03 and 0.8 mM PEP, respectively, which correspond to the intracellular PEP concentration of cells

growing on glucose and 30 minutes after being switched to no carbon medium. In contrast, a saturating

PEP concentration of 20 mM renders the effect of FBP non-cooperative, with a Hill coefficient of 1.0. Thus,

in the presence of physiological substrate concentrations, FBP cooperatively activates Cdc19, with Cdc19

activity changing by > 1000-fold over the physiological FBP concentration range.

69
Figure 2.9. Pyruvate kinase exhibits cooperative activation by FBP independent of its
phosphorylation status.
(A). Activity of purified His-tagged Cdc19 as a function of FBP concentration. The three small plots differ
in terms of PEP concentration added, increasing from 0.03 mM (left-most) to 20 mM (right-most). (B).
Activity of Cdc19 variants in which the indicated phosphorylation site has been changed to alanine or
glutamate in the presence of 0.8 mM PEP. The x axis represents FBP concentration, and the y axis
represents specific pyruvate kinase activity in enzyme activity Units (µmol of product produced per minute)
per mg of enzyme. Experimental data (mean ± range of N = 2) were fit to Hill-equations (lines). Numbers
in the plots are the Hill-coefficient (nH), and the FBP concentration producing half-maximal activation
(Khalf).

PEP accumulation requires allosteric regulation of pyruvate kinase

As noted above, yeast cells expressing variants of Cdc19 that mimic S9, T21 or S22 in a phosphorylated

or dephosphoryated state still exhibited a significant increase in PEP levels upon glucose removal. This

70
would suggest that each of the mutant enzymes should, like the native enzyme, exhibit cooperative

activation by FBP. We purified each of the mutant enzymes from yeast and determined their enzymatic

properties. As predicted, each variant enzyme exhibited a Hill coefficient for FBP activation of between

1.6 and 2.0 (Figure 2.9B). The only mutant with a significant enzymatic defect was T21E, which exhibited

persistently low enzyme activity and imparted impaired growth on glucose. Since T21E mimics the

phosphorylated enzyme, which is more abundant in the presence of glucose, this low activity cannot

account for the inhibition of pyruvate kinase upon glucose removal.

To determine the physiological significance of pyruvate kinase regulation by FBP, we examined the in

vitro and in vivo properties of a pyruvate kinase mutant desensitized to FBP activation. Fenton and Blair

created a number of mutant variants of yeast pyruvate kinase carrying single amino acid changes of the
130,131
sites likely, based on the enzyme’s structure, to participate in FBP allosteric activation . One such
E392A
mutant enzyme, cdc19 , appeared to diminish the requirement for FBP for enzymatic activation
E392A
without affecting either FBP or PEP binding. Using our assay, we confirmed that cdc19 exhibited full

enzymatic activity at physiological concentrations of PEP even in the absence of FBP (Figure 2.10A).

Accordingly, we constructed a strain in which the genomic CDC19 was mutated to convert E392 to
E392A
alanine. The resulting cells (cdc19 ) grew normally on glucose but failed to show an increase in PEP

levels upon glucose removal (Figure 2.10B). Thus, cells expressing the FBP-insensitive E392A mutant

exhibited inappropriate pyruvate kinase flux, thereby depleting PEP following glucose removal.

Figure 2.10. Allosteric regulation of Cdc19 by FBP is essential for glucose removal-induced PEP
accumulation.

71
(A). The Cdc19 E392A variant is active in the absence of FBP. Pyruvate kinase activity was measured in
the presence of 0.8 mM PEP and varying concentrations of FBP as per Figure 2.9. (B). Genomic
E392A
substitution of wild type pyruvate kinase with E392A (cdc19 ) eliminates PEP accumulation upon
glucose removal. The x axis represents minutes after the indicated switch, and the y axis represents
absolute intracellular PEP concentration (mean ± range of N = 2 biological replicates). Experiments were
performed using filter culture.

Gluconeogenic growth requires pyruvate kinase allosteric regulation

Upon glucose depletion in the presence of a non-fermentable carbon source, S. cerevisiae transitions into

gluconeogenesis. When that carbon source is ethanol, PEP carboxykinase uses ATP to drive the

synthesis of PEP from oxaloacetate, whose production from pyruvate also requires ATP. If pyruvate

kinase were simultaneously active, ATP would be consumed in a futile cycle. However, even though

Cdc19 protein is present during exponential gluconeogenic growth at ~20% of its level in glucose-fed cells,
117,118
futile cycling is undetectable . Because gluconeogenic growth is associated with reduced FBP levels,

we hypothesized that one physiological function of the regulation of pyruvate kinase by FBP is to prevent

futile cycling during gluconeogenesis.

E392A
To explore this hypothesis, we transferred wild type and cdc19 yeast from glucose-containing media

to minimal liquid media containing ethanol as the sole carbon source. Upon sudden glucose removal, wild
E392A
type cells initiated gluconeogenic growth after ~1 day of lag phase. In contrast, cdc19 cells did not

start growing until two weeks later, and those cells that eventually grew out showed evidence of

suppressor mutations, i.e., they quickly initiated gluconeogenesis in repeat experiments (Figure 2.11A).

Yeast cells expressing the Cdc19 variants that mimic S9, T21, S22 in a phosphorylated or

dephosphorylated state exhibited normal initiation of gluconeogenesis and identical growth on ethanol

compared to the wild type cells (Figure 2.11B). These results are consistent with a primary role of

allosteric regulation in controlling futile cycling at the glycolysis-TCA cycle interface during gluconeogenic

growth.

72
Figure 2.11, Yeast gluconeogenic growth requires pyruvate kinase regulation by FBP and not by
phosphorylation.
(A). Cells were transferred from glucose-contained YPED plates to minimal media containing 2% ethanol
E392A
as the carbon source. Wild type cells initiated growth after ~1 day, whereas cdc19 cells cells did not
E392A
grow for two weeks. To test if eventual growth of the cdc19 cells was due to suppressor mutations,
the cells were re-grown on glucose and then ransferred them to ethanol again. In this second round of
growth on ethanol, these cells exhibited only ~1 day of lag phase. (B). Yeast cells expressing variants of
Cdc19 that mimic S9, T21, S22 in a phosphorylated or dephosphorylated state exhibited normal initiation
of gluconeogenesis and identical growth rate on ethanol compared to the wild type cells. The x axis
represents time in hours, and the y axis represents optical density (A600). Results from two independent
experiments are shown.

Pyruvate kinase allosteric regulation improves energy charge homeostasis upon glucose upshift

The FBP-insensitive mutant also allowed us to test the physiological function of the accumulation of PEP

and other lower glycolytic intermediates following glucose removal. We hypothesized that the

accumulated lower glycolytic intermediates might serve as a reservoir of readily accessible high energy

73
phosphate bonds upon glucose addition. Such a reservoir could be physiologically valuable, as glucose

up-shift of yeast results in a marked drop in ATP concentration, since phosphorylation of the newly
132
available glucose rapidly consumes ATP .

To test the possibility that Cdc19 allosteric regulation contributes to energy homeostasis upon glucose
E392A
addition, we added glucose to glucose-starved wild type and cdc19 yeast. In wild type cells, glucose

addition resulted in a transient decrease in ATP and energy charge, accompanied by the consumption of

pre-existing PEP within in 20 s, coincident with the increase in FBP (Figure 2.12A and B, Figure 2.13A
E392A
and B). In the cdc19 mutant, a significant pre-existing PEP pool did not exist and ATP and adenylate

energy charge dropped more severely and recovered more slowly.

To mitigate the decrease in energy change upon glucose up-shift, yeast cells activate the enzyme AMP
132
deaminase, Amd1. Amd1 converts AMP into IMP, which can be later recycled into AMP . Consistent
E392A
with defective energy charge management, cdc19 cells accumulate more IMP than do wild type. To

explore the possibility that pyruvate kinase allosteric regulation and Amd1 work in concert to maintain
E392A
energy charge during glucose upshift, we examined energy management in both wild type and cdc19
E392A
strains from which we had deleted AMD1. The cdc19 amd1 strain manifested a large accumulation of
E392A
AMP and a significantly greater drop in energy charge than that in either cdc19 or amd1 (Figure

2.12C, Figure 2.13C). Thus, we conclude that allosteric regulation of pyruvate kinase plays a significant

role in energy homeostasis in yeast.

To test the functional significance of the impaired energy charge homeostasis during glucose upshift, we
E392A E392A
grew wild type, cdc19 , amd1, and cdc19 amd1 yeast on steady versus oscillating glucose. While
E392A
the point mutation in Cdc19 had no impact on growth rate in steady glucose, cdc19 strains grew

more slowly on oscillating glucose (Figure 2.12D-F, Figure 2.13D and E). Thus, ultrasensitive regulation

of pyruvate kinase facilitates rapid cellular adaptation to varying glucose availability, including rapid

resumption of growth upon glucose upshift.

74
Figure 2.12. Cdc19 allosteric regulation facilitates energy charge homeostasis upon glucose up-
shift and enhances growth on oscillating glucose.
(A). Experimental design for (B) and (C). Filter cultures growing on glucose were switched to no carbon
media for 30 min. Thereafter, cells were switched to glucose and metabolome quantified. (B). [PEP],
energy charge = ([ATP] + 0.5 [ADP]) / ([ATP] + [ADP] + [AMP]), and [IMP] upon glucose re-addition in
E392A
wild type and cdc19 yeast. The logarithmic x axis represents seconds after glucose re-addition, and
the y axis represents absolute intracellular metabolite concentration (mean ± range of N = 2 biological
E392A
replicates). (C). Same as (B) but for amd1 and cdc19 amd1 strains. Amd1 catalyzes conversion of
AMP to IMP. (D). Experimental design for (E) and (F). Filter cultures were alternated between glucose
and no carbon media every 30 min for ~ 10 h and growth monitored. (E). Growth of wild type and
E392A
cdc19 yeast on steady versus oscillating glucose. Closed symbols = steady glucose, open symbols =
E392A
oscillating glucose, blue = wild type, red = cdc19 . The x axis represents time in hours, and the
logarithmic y axis represents optical density (A600) (mean ± range of N = 2 biological replicates). (F).
E392A
Same as (E) but for amd1 (Y3996) and cdc19 amd1 (Y3997).

75
Figure 2.13, Cdc19 allosteric regulation facilitates adenylate nucleotide homeostasis during
glucose re-addition and thus growth on oscillating glucose.
(A). Experimental design for (B) and (C). Filter cultures growing on glucose were switched to no carbon
media for 30 min. Thereafter, cells were switched to glucose and metabolome quantitated. (B).
Consumption of pre-existing PEP (Figure 2.12B) is mirrored by an increase of FBP. ATP decreases more
E392A
and recovers more slowly in the cdc19 mutant strain; AMP increases more, as do compounds in
purine salvage pathway, such as IMP (Figure 2.12B) and inosine. (C). ATP decreases yet more, and
E392A E392A
AMP accumulates yet more, with simultaneous cdc19 mutation and AMD1 deletion (cdc19 amd1),
versus either mutation alone. The logarithmic x axis represents seconds after glucose re-addition, and the
y axis represents absolute intracellular concentration (mean ± range of N = 2 biological replicates). (D and
E). Growth of yeast in a microfluidic flow cell with oscillating glucose. (D) Dividing time of wild type and
E392A E392A
cdc19 cells on oscillating glucose. (E). Distribution of cell division time of wild type and cdc19
cells on oscillating glucose. Yeast cells were switched between glucose and no carbon in a flow cell,
spending 10 minutes on each media. Cell division time represents the time between the appearance of a
bud and the next bud on the same mother cell. The division of 70 cells from 12 different positions under
the microscope was followed. In (D), bars reflect data, mean ± range. In (E), the x axis represents dividing
time in minutes, and the y axis represents percentage of cells falling into the indicated cell division time.

76
2.3 Discussion

Both allosteric activation and covalent modification by phosphorylation have been implicated in regulation

of pyruvate kinase, which catalyzes the main efflux from glycolysis. Here we have examined how each of

these regulatory processes contributes to flux control through the main yeast pyruvate kinase isozyme,

Cdc19. We provide evidence that covalent modification of Cdc19 does not contribute significantly to acute

regulation of flux whereas ultrasensitive allostery provides the dominant flux control mechanism. This

ultrasensitive allostery enables switch-like inhibition of pyruvate kinase during glucose removal (Figures

2.9A and 2.10) and rapid (less than five seconds) re-activation of the enzyme during glucose re-addition

(Figure 2.12B). From a metabolic standpoint, this switch-like behavior results in accumulation of three

carbon units and high energy phosphate bonds in the glycolytic pathway following glucose removal. This

accumulation facilitates the onset of gluconeogenesis (Figure 2.10C) and provides a boost to re-initiating

fermentative growth when glucose becomes available (Figure 2.12).

99
An advantage of allosteric regulation over phosphorylation is speed . In the case of Cdc19, such speed,

in combination with ultrasensitivity, enables pyruvate kinase inactivation to outpace depletion of upstream

glycolytic intermediates, resulting in the seemingly paradoxical accumulation of PEP upon glucose

removal. A similar accumulation of PEP upon glucose removal also occurs in prokaryotes, where PEP is

the substrate for glucose import and phosphorylation. Accordingly, PEP accumulation is required to

enable rapid assimilation of glucose when it becomes available. Here we show that PEP and

phosphoglycerate accumulation in yeast plays a related role, acting as a store of high-energy phosphate

bonds that helps to maintain energy charge upon glucose upshift.

The contrast in dynamics between activation of pyruvate kinase by FBP and activation of

phosphofructokinase by fructose-2,6-bisphosphate is informative. Fructose-2,6-bisphosphate levels are

133
regulated through cyclic-AMP-based signaling, which occurs on the timescale of minutes . In contrast,

the turnover of glycolytic intermediates occurs in seconds. Thus, phosphofructokinase is designed to turn

off more slowly than pyruvate kinase upon glucose removal. Such timing traps residual glucose-derived

carbon in a high energy form as lower glycolytic intermediates.

77
The ultrasensitive activation of pyruvate kinase by FBP is a striking example of intrinsic metabolic

regulation: control of metabolic flux solely by enzymes and core metabolites, without reliance on

transcription, covalent modification, or small molecule signals. The evolution of such intrinsic regulation is

not surprising, given the centrality of pyruvate kinase regulation to the switch between glycolysis and

gluconeogenesis, an ancient biological challenge. More intriguing is the fact that allostery has persisted

as the dominant regulatory mechanism for yeast pyruvate kinase, and perhaps more generally for

microbial metabolism.

The predominance of pyruvate kinase regulation by allostery, despite the enzyme being covalently
134-137
modified by phosphorylation and acetylation, raises a more general question : To what extent is

covalent modification functionally significant? Over 4000 protein phosphorylation sites have been
47,51
identified in yeast . These include PKA sites on ~ 45 metabolic enzymes, e.g., phosphofructokinase

and pyruvate kinase in glycolysis, pyruvate carboxylase in anaplerosis, PEP carboxykinase in

gluconeogenesis, acetyl-CoA oxidase in fatty acid beta-oxidation, glycogen synthase and phosphorylase

in glycogen metabolism, phosphatidylserine synthase and choline kinase in phospholipids biosynthesis,

47
and trehalase in trehalose metabolism . Here we find, however, that the primary soluble metabolic

target of PKA is trehalose. While we cannot rule out a subtle role of PKA in non-transcriptional regulation

of other metabolic pathways, we believe that this finding is a harbinger of a broader paradigm — most

protein covalent modification events, at least in microbes, will prove to have minimal functional impact,

being neither selected for or against evolutionarily. This in no way undermines the importance of covalent

modification as a regulatory mechanism, but merely suggests that biochemical promiscuity may mask

functional specificity. Identifying the functionally important events — which in metabolism can be achieved

using the combination of metabolomics and genetics employed here — holds the potential to simplify and

clarify the signaling-metabolism interface, and perhaps signaling systems in general.

78
2.4 Methods

Yeast strains

Strains used in this study were derived from either a prototrophic or auxotrophic W303-1B strain, Y3358
as as as as
or Y2864, respectively. Analogue sensitive strains pka , pka sch9 and snf1 were described
120
previously . The bcy1 mutant was constructed by transforming Y3358 to geneticin (G418) resistance

with a PCR fragment spanning the bcy1::KanMX allele in the corresponding strain in the Yeast Magic

Marker collection (Open Biosystems, Thermo Fisher Scientific, San Jose, CA). To prevent the selection of

suppressor mutations, a transformant confirmed by PCR analysis was used immediately for the

metabolomics experiment without intervening storage.

Strains with altered CDC19 genomic alleles were generated by insertion of KanMX marker downstream of

genomic CDC19 followed by site-directed mutagenesis. First, the KanMX selectable marker was amplified

by PCR from plasmid pFA6a using primers with 50 bp extensions homologous to the sequences

immediately upstream and downstream to a site lying 200 bp downstream of CDC19. The resulting PCR

fragment was used to transform strain Y3358 (W303-1B prototroph) to geneticin resistance, resulting in

strain Y3886 with a precise insertion of KanMX downstream of CDC19. A two-step fusion PCR process

was used to make CDC19 alleles with the desired point mutation using the genomic DNA of strain Y3886

as the template. Briefly, we used primers containing the desired point mutation and primers containing a

sequence either upstream of N-terminal or downstream of C-terminal of the genomic CDC19::KanMX

locus to amplify by PCR two halves of the altered CDC19 KanMX. The resulting two PCR fragments

carrying mutated base(s) at the C-terminus (for the first half) and N-terminus (for the second half) were

used in a second round of PCR to generate a complete CDC19::KanMX sequence with the desired point

mutation. The resulting sequence was used to transform strain Y3358 to geneticin resistance. Positive

clones were sequenced to confirm the presence of desired mutation.

To generate strains for purification of wild type or mutant Cdc19, the CDC19 locus from strains obtained

as described above was amplified by PCR using primers immediately flanking the 5’ and 3’ coding region

and carrying appropriately positioned BamH1 and Not1 sites to allow cloning into the yeast expression

79
plasmid pYES2/NC-C (Invitrogen, Carlsbad, CA) to create plasmids expressing His-tagged Cdc19

variants. These plasmids were transformed into strain Y3846 for subsequent purification of Cdc19 for

enzymatic characterization. To generate strain Y3979 carrying a genomic His-tagged version of wild type

CDC19, a His-tagged CDC19 KanMX sequence was generated by fusion PCR: a primer sequence in the

middle of CDC19 was used together with a primer either upstream of His-tagged CDC19 on the

expression plasmid or downstream of CDC19 KanMX on the genome to generate two halves of the

desired sequence. The complete sequence was generated by using the two fragments as the template

and then used to transform strain Y3358 to G418 resistance. Positive clones were sequenced to confirm

the presence of the in-frame version of CDC19 His-tagged at its N-terminus. For a list of strains and

primers used in this study, see Table 2.1 and 2.2.

Table 2.1. Strains used in this study.

Strain Designation Genotype Source

Y3358 prototrophic wild type MATa can1
120
Y2092 auxotrophic wild type MATα ade2-1 can1-100 his3-11, 15
(W303-1B) leu2-3, 112 trp1-1 ura3-1
120
Y2864 auxotrophic wild type W303-1B gal1::HIS3
as M164G M147G M165G 120
Y3561 pka W303-1B tpk1 tpk2 tpk3
as as 120
Y3569 pka sch9 Y2864
T492G M164G M147G M165G
sch9 tpk1 tpk2 tpk3
as I132G 120
Y3504 snf1 Y2864 snf1
Y3886 CDC19 KanMX Y3358 CDC19 KAN This study
S9A S9A
Y3888 cdc19 Y3358 cdc19 KAN This study
S9E S9E
Y3889 cdc19 Y3358 cdc19 KAN This study
T21A T21A
Y3890 cdc19 Y3358 cdc19 KAN This study
T21E T21E
Y3891 cdc19 Y3358 cdc19 KAN This study
S22A S22A
Y3892 cdc19 Y3358 cdc19 KAN This study
S22E S22E
Y3893 cdc19 Y3358 cdc19 KAN This study
E392A E392A
Y3898 cdc19 Y3358 cdc19 KAN This study
Y3846 expression strain MATa/α his3Δ1 leu2 trp1-289 ura3-52 INVSc1 from
Invitrogen
Y3862 Cdc19 expression Y3846/pYES2/NC-C-CDC19 This study
strain
S9A S9A
Y3863 Cdc19 expression Y3846/pYES2/NC-C-cdc19 This study
strain

80
 Strain Designation Genotype Source
S9E S9E
Y3864 Cdc19 expression Y3846/pYES2/NC-C-cdc19 This study
strain
T21A T21A
Y3865 Cdc19 expression Y3846/pYES2/NC-C-cdc19 This study
strain
T21E T21E
Y3866 Cdc19 expression Y3846/pYES2/NC-C-cdc19 This study
strain
S22A S22A
Y3867 Cdc19 expression Y3846/pYES2/NC-C-cdc19 This study
strain
S22E S22E
Y3868 Cdc19 expression Y3846/pYES2/NC-C-cdc19 This study
strain
E392A E392A
Y3873 Cdc19 expression Y3846/pYES2/NC-C-cdc19 This study
strain
Y3979 Genomic His-tagged Y3358 6xHis-CDC19 KAN This study
wild type CDC19
Y3996 amd1 Y3886 amd1::NatMX This study
E392A
Y3997 cdc19 amd1 Y3898 amd1::NatMX This study

Table 2.2. Primers used in this study.

Bases that are red and bold are the ones encoding mutated residues. Bases that are red and italic are the
BamH1 and Not1 digestion sites. Blue bases represent the start codon of cdc19.

Name Sequence Notes

CATACATTTTATAAGGTATTCT
kan-F ATAAAAAGAGTATTATGTTATT
GCGGATCCCCGGGTTAATTAA
GCAACACCTCATCGTTATGAC
GACAATTGGAGACAAAAAGGT
kan-R
TAAGAATTCGAGCTCGTTTAAA
C downstream of genomic CDC19
CTTCCACAATTTCGGCTCTATT
F2
G altered CDC19 KanMX sequence in site-
R2 GAAGAATAGGACGGAGTAGC
GATTGACCGCATTAAACGTTGT
S9A-F
TG
CAACAACGTTTAATGCGGTCAA
S9A-R
TC
81
Forward primer containing sequence
downstream of genomic CDC19 and at the
beginning of KanMX on plasmid pFA6a for
inserting KanMX downstream of genomic
CDC19
Reverse primer containing sequence at the
end of KanMX on plasmid pFA6a and
downstream of CDC19 for inserting KanMX
Forward primer containing sequence upstream
st
of genomic CDC19 for generating 1 half of
directed mutagenesis
Reverse primer containing sequence
downstream of genomic CDC19 for generating
nd
2 half of altered CDC19 KanMX sequence in
nd
site-directed mutagenesis, or 2 half of His-
tagged CDC19 KanMX sequence
Forward primer containing mutated sequence
S9A
for cdc19
Reverse primer containing mutated sequence
S9A
for cdc19
Name Sequence Notes
GATTGACCGAGTTAAACGTTG Forward primer containing mutated sequence
S9E-F S9E
TTG for cdc19
CAACAACGTTTAACTCGGTCAA Reverse primer containing mutated sequence
S9E-R S9E
TC for cdc19
TTGAGAAGAGCTTCCATCATTG Forward primer containing mutated sequence
T21A-F T21A
GT for cdc19
ACCAATGATGGAAGCTCTTCT Reverse primer containing mutated sequence
T21A-R T21A
CAA for cdc19
TTGAGAAGAGAGTCCATCATT Forward primer containing mutated sequence
T21E-F T21E
GGT for cdc19
ACCAATGATGGACTCTCTTCTC Reverse primer containing mutated sequence
T21E-R T21E
AA for cdc19
TTGAGAAGAACCGCTATCATTG Forward primer containing mutated sequence
S22A-F S22A
GT for cdc19
ACCAATGATAGCGGTTCTTCTC Reverse primer containing mutated sequence
S22A-R S22A
AA for cdc19
TTGAGAAGAACCGAGATCATT Forward primer containing mutated sequence
S22E-F S22E
GGT for cdc19
ACCAATGATCTCGGTTCTTCTC Reverse primer containing mutated sequence
S22E-R S22E
AA for cdc19
GCTGTTTTCGCTCAAAAGGCC Forward primer containing mutated sequence
E392A-F E392A
AAG for cdc19
CTTGGCCTTTTGAGCGAAAAC Reverse primer containing mutated sequence
E392A-R E392A
AGC for cdc19
Forward primer containing sequence upstream
of genomic CDC19 for ligating two amplified
F1nest GTTTTGCCATCGACAGATTGG
halves for generating altered CDC19 KanMX
sequence
Reverse primer containing sequence
GATTAAACCACCAAACGAAGG downstream of genomic CDC19 KanMX for
R1nest
C ligating two amplified halves for generating
altered CDC19 KanMX sequence
CTTAGACTCTTATACAAAGGAT Forward primer containing BamH1 site and
exp-F CCATGTCTAGATTAGAAAGATT sequence at the beginning of genomic CDC19
G for inserting CDC19 into expression vector
CTACAATTAAGAAACTACGCG Reverse primer containing Not1 site and
exp-R GCCGCAATAATATCTTCATTCA sequence downstream of genomic CDC19 for
ATC inserting CDC19 into expression vector
Forward primer containing sequence on
GAAAAGGCTAAGGAATTCGGT nd
F5 genomic CDC19 for generating 2 half of His-
ATC
tagged CDC19 KanMX sequence
CTTGTTTCTATTTACAAGACAC Forward primer containing sequence upstream
CAATCAAAACAAATAAAACATC of genomic CDC19 and His-tagged CDC19 on
his-F1 st
ATCACAATGGGGGGTTCTCAT the plasmid for generating 1 half of His-
CATCAT tagged CDC19 KanMX sequence
R5 GATACCGAATTCCTTAGCCTTT Reverse primer containing sequence on

82
 Name Sequence Notes
st
TC genomic CDC19 for generating 1 half of His-
tagged CDC19 KanMX sequence
Forward primer for ligating two amplified halves
his-F2 TTCTATTTACAAGACACCAATC for generating His-tagged CDC19 in the
genome
Reverse primer for ligating two amplified
R1 GGACACTAATTGAATCTGCCC halves for generating His-tagged CDC19 in the
genome

Yeast culture conditions and extraction

Detailed protocols for preparing filter cultures and extracting metabolites from Saccharomyces cerevisiae
138
have been published . Briefly, yeast cells were grown at 30 °C in agarose-media mixture with 2 % (w/v)

glucose as the carbon source, and 6.7 g/L Difco Yeast Nitrogen Base (YNB) without amino acids

containing 5 g/L ammonium sulfate as the nitrogen source, resulting in an exponential growth rate of 0.4
-1
h . Saturated overnight cultures were diluted 1:30 and grown in liquid media in a shaking flask to A600

(absorbance at 600 nm) of ~ 0.15. A portion of the cells (1.6 mL) was transferred to 50 mm nylon

membrane filters (Millipore, Billerica, MA) resting on vacuum filter support. Once cells were loaded, the

membrane filters were transferred to media-loaded agarose plates. To measure growth, filters were

washed thoroughly with 1.6 mL fresh medium and A600 determined. For metabolomics experiments, cells

were grown to A600 of ~ 0.6 on filters prior to initiation of experimental manipulations (e.g., glucose

removal) and quenching of metabolism. Metabolism was quenched by direct and immediate transfer of
79
the filters into – 20 °C extraction solvent (40:40:20 acetonitrile/methanol/water , and serial extraction
138
was carried out as described previously . For glucose removal, filters were then switched from glucose

plates to plates containing no carbon source, 3 % (v/v) glycerol + 3 % (v/v) ethanol, or 3% (v/v) ethanol,

and extracted at 0.5, 2, 5, 10, 20, 30, 60, 90 min after the switch.

For growth curves on oscillating glucose (Figure 2.12D, E and F), cells were switched between glucose

and no carbon plates (a new plate each time) spending half an hour on each plate. For steady glucose

growth, cells were switched every 30 min to a new glucose plate.

83
Liquid culture metabolome extraction was used for experiments requiring complete nutrient deprivation or

analogue treatment. For glucose removal, yeast cells were grown at 30°C in liquid minimal media

containing 2% (w/v) glucose + 6.7 g/L YNB. For nitrogen and phosphate starvation, yeast cells were

grown in nitrogen-limited and phosphate-limited liquid media respectively (Table 2.3). For strains with

auxotrophic W303-1B (MATα leu2-3,112 trp1-1 can1-100 ura3-1 ade2-1 his3-11,15) background,

additional 20 mg/L adenine sulfate, 20 mg/L histidine, 100 mg/L leucine, 20 mg/L tryptophan, and 20

mg/L uracil were added to the above media. ATP analogue 1NM-PP1 (Cayman Chemical, Ann Arbor, MI)
as M164G M147G M165G
was added at 1 µM to inhibit genetically altered PKA in the pka (W303-1B tpk1 tpk2 tpk3 )
as as
strain and to inhibit both PKA and Sch9 in pka sch9 strain (W303-1B gal1::HIS3
T492G M164G M147G M165G
sch9 tpk1 tpk2 tpk3 ). ATP analogue 4-methylnapthyl-1-tert-butyl-3-phenylpyrazolo[3,4-d]

pyrimidine (mMe-PP1, a generous gift from Kevan Shokat) was added at 1 µM to inhibit SNF1 in a
as I132G
genetically altered snf1 strain (W303-1B gal1::HIS3 snf1 ). Saturated overnight cultures were diluted

1:30 and grown in liquid media in a shaking flask to A600 of ~ 0.6. A portion of the cells (3 mL) were

filtered onto a 50 mm nylon membrane filter, which was immediately transferred into – 20°C extraction

solvent (40:40:20 acetonitrile/methanol/water). For nutrient removal, 100 mL of cell culture at A600 of

~0.6 were poured onto a 100 mm cellulose acetate membrane filter (Sterlitech, Kent, WA) resting on a

vacuum filter holder with a 1000 mL funnel (Kimble Chase, Vineland, NJ), washed with 100 mL pre-

warmed (30°C) starvation liquid medium lacking the indicated nutrient. Immediately after the wash media

went through, the filter was taken off the holder and cells washed into a new shaking flask containing 100

mL pre-warmed (30 °C) liquid starvation medium. Serial extraction was then carried out at indicated time

points after glucose removal.

For a confirmatory growth rate study on oscillating glucose, cells were switched between glucose and no

carbon in a microfluidic flow cell (Bioptechs, Butler, PA), spending 10 minutes in each medium. Cells were

visualized continuously using a Nikon Ti microscope. Before growing cells, a calibration experiment was

performed using fluorescein to test the functionality of the pumps and make sure media switched as
E392A
expected. To obtain exponentially growing cells, wild type and cdc19 cells were first grown at 30°C in

liquid media containing 2% (w/v) glucose + 6.7 g/L YNB until reaching an A600 of 0.2. The resulting

culture were diluted to A600 of 0.05, mildly sonicated, and injected into the flow cell primed with
84
concanavalin A solution (2 mg/mL concanavalin A, MP Biomedicals, Solon, OH; 5 mM MnCl 2 and 5 mM

CaCl2, Sigma-Aldrich, St. Louis, MO) which makes yeast cells stick to the coverslip. The whole

experiment was performed at 30°C under the control of FCS2 temperature controller. Before switching

the media, cells were equilibrated in the flow cell with the glucose-containing media for 2 h, and 12

positions were selected and followed under the microscope. Thereafter, pumps were programmed to

change media every 10 minutes and pictures were taken every 2 minutes. 70 healthy cells were selected

after collecting all the pictures to obtain the average and distribution of the dividing time, defined as the

time between the appearance of a bud and the next bud on the same mother cell. (Figure 2.13D and E).

This whole experiment was performed in McClean lab, Princeton University under the generous help of

Patrick Bradley and Megan McClean.

Metabolite measurement

Cell extracts were analyzed by reversed phase ion-pairing liquid chromatography (LC) coupled by

electrospray ionization (ESI) (negative mode) to a high-resolution, high-accuracy mass spectrometer

5
(Exactive; Thermo Fisher Scientific) operated in full scan mode at 1 s scan time, 10 resolution, with
34
compound identities verified by mass and retention time match to authenticated standard . For a list of

metabolites, exact masses, and retention times, see Table 2.4. Confirmatory analysis for a subset of

compounds was conducted by a different reversed phase ion-pairing LC method coupled by negative

mode ESI to a Thermo TSQ Quantum triple quadrupole mass spectrometer operating in multiple reaction
32
monitoring mode (see Table 2.5). All data reported in the paper is from the high-resolution LC-MS

analysis. Isomers are reported separately only where they fully chromatographically resolved. Absolute

intracellular metabolite concentrations in steadily growing S. cerevisiae were determined as described

36
previously . Metabolite concentrations after perturbations were computed based on fold-change in ion

counts relative to steadily-growing cells (grown and analyzed in parallel) multiplied by the known absolute

concentration in the steadily growing cells.

85
 139
Table 2.3 Media used in nitrogen and phosphate starvation
Comparison of salts composition in the media used
Minimal Carbon Nitrogen Nitrogen Phosphat Phosphate
Media starvation limited starvation e limited starvation
Glucose 20 g - 20 g 20 g 20 g 20 g
Ammonium
5g 5g 625 mg - 5g 5g
Sulfate
Sodium
- - 5g 5g - -
Sulfate
Potassium
phosphate 1g 1g 1g 1g - -
monobasic
Potassium
- - - - 1g 1g
Chloride
Magnesium
500 mg 500 mg 500 mg 500 mg 500 mg 500 mg
Sulfate
Sodium
100 mg 100 mg 100 mg 100 mg 100 mg 100 mg
Chloride
Calcium
100 mg 100 mg 100 mg 100 mg 100 mg 100 mg
Chloride
From
Vitamins In yeast In yeast In yeast In yeast
1000X From 1000X
and other nitrogen nitrogen nitrogen nitrogen
(see (see below)
metals base base base base
below)

1000X vitamins and metals

1000X vitamins (per 1L) 1000X Metals (per 1L)
Biotin 1 mg Boric acid 500 mg
Calcium pantothenate 400 mg Copper Sulfate.5H2O 40 mg
Folic acid 2 mg Potassium Iodide 100 mg
Myo-inositol 2000 mg Ferric Chloride.6H2O 200 mg
Nicotinic acid 400 mg Manganese Sulfate.H2O 400 mg
p-aminobenzoic acid 200 mg Sodium Molybdate.2 H2O 200 mg
Pyridoxine HCl 400 mg Zinc Sulfate.7H2O 400 mg
Riboflavin 200 mg
Thiamine HCl 400 mg

Transcript measurement
120
The transcript analysis of liquid culture yeast cells were performed as described previously . Briefly,

liquid culture samples for transcript analysis were collected by filtration and frozen in liquid nitrogen. Total

RNA was extracted using the Qiagen RNeasy Mini Kit (Valencia, CA). 100 ng of total RNA was used as a

template for first and second strand cDNA synthesis with reverse transcriptase using a primer containing

86
poly dT and T7 polymerase promoter. Labeled cRNA was synthesized from cDNA using T7 RNA

polymerase and cyanine3- (Cy3-) or Cy5-labeled CTP

(PerkinElmer Life and Analytical Sciences, Boston, MA). The amount of cRNA synthesized and

incorporation of Cy3- and Cy5-CTP into cRNA were measured using a NanoDrop (NanoDrop

Technologies, Wilmington, DE). Equal amounts of Cy3- and Cy5-labeled cRNA were combined, mixed

with the control target and fragmented for 30 min. Each sample was then hybridized to an Agilent yeast

oligo microarray for 17 h at 60°C. The arrays were washed and scanned using Agilent Microarray

Scanner (Agilent Technologies) at 100% PMT for red and green channels and at 10 mm resolution. The

feature information was extracted from the microarrays using Agilent Feature Extraction Software with the

default settings. The data were stored and analyzed on the Princeton University Microarray database.

Dye normalization for each array was determined by the rank consistency method

and then spot intensities were calculated by the LOWESS method. Spots were retained for further

analysis only if both the Cy3 and Cy5 channels were greater than 2.6 s of mean background intensity and

were uniform in intensity. Only those genes for which 80% of the arrays yielded good data were retained

for analysis. All data described in this study can be viewed and downloaded from the PUMAdb website

(http://puma.princeton.edu/cgi-bin/publication/viewPublication.pl?pub_no=544).

Preparation of purified pyruvate kinase

S. cerevisiae pyruvate kinase used in biochemistry assays was purified from yeast that carried plasmid

encoding His-tagged wild type or point mutant Cdc19. Genomic wild type or altered CDC19 was used as

template DNA to create His-tagged CDC19 sequence on a plasmid. BamH1 and Not1 binding sites were

included in forward and reverse primers, respectively. The resulting PCR product and vector pYES2/NC-

C (Invitrogen) were digested with BamH1 and Not1, ligated at room temperature for 10 minutes and

transformed into TOP10 (Invitrogen) competent cells. Quick and Dirty screening was used for positive

clones due to the requirement of screening large number of colonies. Briefly, a colony was first lysed in

lysis buffer (20% w/v sucrose, 200mM NaOH, 120mM KCl, 10mM EDTA, 0.5% SDS, 0.08% w/v

Bromophenol blue). The resulting extract was loaded on DNA gel and the plasmid with inserts was

87
detected by their slower movements on the gel. Thereafter, plasmids from positive colonies were

sequenced (Genewiz, South Plainfield, NJ) to confirm the presence of correct mutation and in frame

fusion of N-terminal His tag. Confirmed plasmid was then transformed into the yeast expression strain

INVSc1 (Invitrogen).

To obtain purified pyruvate kinase, the above yeast expression strain was grown in 50 mL synthetic

complete (SC) media with added uracil and 2% w/v raffinose as the carbon source. After reaching an

A600 of 0.5, cells were spun down, re-suspended in SC with uracil and 2% galactose to induce the GAL1

promoter, and incubated at 30°C for 7 hours. The resulting cells were harvested by centrifugation,

washed once with Native Purification Buffer (Invitrogen ProBond Purification Systems) with protease

inhibitor (Roche, Indianapolis, IN), and re-suspended in 3 mL of the same buffer. Glass beads were

added to 300 µL of aliquoted cell suspension to lyse the cells. The resulting mixture was spun at 5,000 g

for 10 minutes at 4°C and the supernatant was collected. It was diluted in Native Purification Buffer and

applied to Invitrogen ProBond Purification Systems following protocols provided by Invitrogen. Eluate

containing 200 µg/mL pyruvate kinase was collected and used for enzyme assays.

Measurement of pyruvate kinase activity

The 0.5 mL reaction mixture contained 100 mM Tris-HCl buffer, pH 8.5, 10 mM MgCl2, 10mM KCl, and

0.1-2 µg/mL pyruvate kinase, depending on the concentration of PEP. PEP, ADP and FBP were added at

physiological concentrations based on the metabolomics experiments. Specifically, ADP was added at

0.43 mM, which corresponds to its in vivo concentration in glucose grown and glucose starved cells. PEP

was added to 0.8 mM or 0.03 mM as indicated in Figure 2.9 and 2.10, which corresponds to its

concentration in glucose grown cells and glucose deprived cells, respectively. To compare to prior

literature, PEP was also added to 20 mM (Figure 2.9A). FBP was either absent or added at

concentrations from 0.3 to 12 mM to quantitate its regulatory effects. Assays were carried out at 30°C in

an Eppendorf Thermomixer with mixing speed of 600 rpm. Reactions were initiated by adding PEP. At 1,

2, 5, 10, and 20 min thereafter, 20 μL of the reaction mixture were transferred into 180 μL of – 20 °C

88
extraction solvent (50:50 methanol/water) to quench the reaction and extract metabolites. Samples were

then further diluted five-fold followed by LC-MS measurement.

LC-MS/MS analysis of pyruvate kinase concentration and covalent modification

Liquid cultures of S. cerevisiae strains with His-tagged CDC19 on the chromosome were grown in
15 15
glucose minimal media containing 5 g/L unlabeled or N-labeled (NH4)2SO4. Cultures grown in N-

(NH4)2SO4 were harvested at a density of A600 = 0.6 by pouring cell cultures into pre-cooled -20 °C ice.

Cells growing in unlabeled (NH4)2SO4 was subjected to carbon starvation as described above, and

harvested 15 minutes after the removal of glucose.

15
Unlabeled and N-labeled cells were spun down, frozen in liquid nitrogen, mixed, and ground with a Cryo

Mill (Retsch, Newtown, PA) at -196 °C. The resulting mixture was suspended in Native Purification Buffer

(Invitrogen), and cell lysate was collected by centrifugation. The lysate was then diluted in Native

Purification Buffer and applied to ProBond Purification Systems following protocols provided by Invitrogen.

Eluate containing purified pyruvate kinase was collected and used for phosphoproteomics. The whole

process except grinding was carried out at 4 °C.

Purified pyruvate kinase samples were subjected to buffer exchange, thiol reduction and alkylation, and
111
trypsin or endoproteinase Lys-C digestion in gel following SDS-PAGE . Peptides were desalted using
140
StageTip micro-scale reversed-phase chromatography , and then subjected to phosphopeptide
141,142
enrichment by titanium dioxide metal oxide affinity chromatography .

Peptide samples were analyzed by a reverse-phase nano-LC-MS/MS performed on a nano-flow capillary

high pressure HPLC system (Eksigent, Dublin, CA) coupled to an LTQ-Orbitrap hybrid mass spectrometer

(Thermo Fisher Scientific), outfitted with a NanoMate ion source robot (Advion, Ithaca, NY). Sample

concentration and desalting were performed online using a trapping capillary column (200μm x ca. 20

mm, packed with 5μm, 100Å Magic AQ C18 material, Michrom, Auburn, CA) at a flow rate of 7μL/min for

3.5 minutes, while separation was achieved using an analytical capillary column (75μm x ca. 15cm,

packed with 1.7μm, 130 Å BEH C18 material, Waters, Billerica, MA) under a linear gradient of A and B
89
solvents (solvent A: 3% acetonitrile/ 0.1% formic acid/ 0.1% acetic acid; solvent B: 97% acetonitrile/ 0.1%

formic acid/ 0.1% acetic acid) over 180 min at a flow rate of approximately 0.4 μL/ min. Electrospray

ionization was carried out using the NanoMate ion source at 1.74 kV, with the LTQ heated capillary set to

200 °C.

15
Identification and relative quantitation of N-labeled and unlabeled peak pairs was carried out using
143 15 14
PVIEW software , which accomplishes N- N MS peak pair determination together with database
15
searching of MS/MS to distinguish the sequence-dependent mass shifts incurred through N metabolic
144
labeling . Phosphopeptide peak pair quantitation was confirmed by manual inspection of summed mass

spectra over a time range spanning the peptide chromatographic elution.

Additionally, peptide assignments to tandem MS were visualized by peaklist file generation through

Proteome Discoverer™ software (Thermo Fisher Scientific), database searching using the Mascot™

search engine (Matrix Science, Cambridge, MA), and results integration using Scaffold™ software

(Proteome Software, Portland, OR). The SGD S. cerevisiae protein database was used for the database

searches, stipulating trypsin cleavage with up to 5 missed cleavages, a precursor mass window of ±8

ppm, fixed cysteine carbamidomethylation, and variable oxidation of methionine, N-terminal protein

acetylation, and phosphorylation at serine, threonine, and tyrosine residues. Mascot™ Error Tolerant

searching was also conducted, limiting the variable modifications to methionine oxidation and N-terminal

protein acetylation, in order to identify potential novel modifications on Cdc19. Peptide and protein

assignments were thresholded at a false discovery rate of 1%, based on reverse database searching.
145
Positional assignments of modification sites were confirmed using MaxQuant software , in label-free

quantitation mode using similar parameters.

Simplified steady-state model of glycolytic regulation

To examine the regulatory requirements for PEP to increase when glycolytic influx decreases, we used a

simplified model of glycolysis as shown in Scheme 2.1

90
 Scheme 2.1

where F is the influx of FBP, k1 – k3 are the rate constants of the indicated reactions, and h is the hill-

coefficient for FBP activation of pyruvate kinase. Given that the bidirectional enzymes in glycolysis have
146-148 129
large Km , while unidirectional enzymes are saturated , it is reasonable to approximate the

reactions between FBP and PEP as first-order, and to assume that pyruvate kinase is insensitive to [PEP].

For this simple model, the steady state kinetics of FBP and PEP are given by:

d [FBP]
0  F  k1 [FBP]  k 2 [PEP]2
dt

d [PEP]
0  2k1[FBP]  2k 2 [PEP]2  k 3 [FBP]h
dt

These equations have a discrete solution:

1
 2F  h
[FBP]   
k
 3 

1
 2F  h
k1    F
[PEP]   k3 
k2
91
In the model, when F decreases, [FBP] always decreases. This matches experimental observations. The

qualitative behavior of PEP, however, depends on h. For h = 1, as F decreases, [PEP] decreases (see

Figure 2.8). For h < 1, the decrease is greater. For h > 1, however, [PEP] can increase as F decreases

as is experimentally observed. Thus, ultrasensitive regulation of pyruvate kinase is required to obtain the

experimentally observed inverse behavior of [FBP] and [PEP].

Table 2.4. Mass, retention time, and ion counts of measured metabolites by the LC-high resolution
MS analysis.

retention average ion counts in

time samples from glucose-
compound formula m/z (min) fed cells
pyruvate C3H4O3 87.0088 8.3 361179
alanine C3H7NO2 88.0404 1 46869
4-aminobutyrate C4H9NO2 102.056 4.5 697619
serine C3H7NO3 104.035 1 351700
glycerate C3H6O4 105.019 6.3 6932
uracil C4H4N2O2 111.02 1.5 513190
proline C5H9NO2 114.056 1.12 43348
fumarate C4H4O4 115.004 13.4 32434
2-keto-isovalerate C5H8O3 115.04 13.06 89397
indole C8H7N 116.051 7.5 8594
valine C5H11NO2 116.072 1 491860
succinate C4H6O4 117.019 11.6 71779
threonine C4H9NO3 118.051 1 959882
cysteine C3H7NO2S 120.013 1.1 1257
nicotinate C6H5NO2 122.025 11.2 12924
citraconic acid C5H6O4 129.019 12.95 55301
N-acetyl-alanine C5H9NO3 130.051 7.97 4129
hydroxyproline C5H9NO3 130.051 1.1 7989
leucine/isoleucine C6H13NO2 130.087 1.9 2369055
asparagine C4H8N2O3 131.046 1 591559
hydroxyisocaproic acid C6H12O3 131.071 14.35 229851
ornithine C5H12N2O2 131.083 1 1355800
aspartate C4H7NO4 132.03 4.4 584509
malate C4H6O5 133.014 12.7 61362
homocysteine C4H9NO2S 134.028 1 2552
92
 retention average ion counts in
time samples from glucose-
compound formula m/z (min) fed cells
hypoxanthine C5H4N4O 135.031 3.6 137774
anthranilate C7H7NO2 136.04 13.3 41010
hydroxybenzoate C7H6O3 137.024 10.88 1184
α-ketoglutarate C5H6O5 145.014 13.1 44933
glutamine C5H10N2O3 145.062 1 9122109
lysine C6H14N2O2 145.098 0.8 93175
O-acetyl-serine C5H9NO4 146.046 1.1 2994
glutamate C5H9NO4 146.046 4.6 2050381
2-hydroxy-2-
methylbutanedioic acid C5H8O5 147.03 12.6 62643
methionine C5H11NO2S 148.044 1.6 62839
guanine C5H5N5O 150.042 3.4 3333
xanthine C5H4N4O2 151.026 1.18 2908
histidine C6H9N3O2 154.062 1.6 522520
orotate C5H4N2O4 155.01 8.1 38938
dihydroorotate C5H6N2O4 157.026 6.8 6219
indole-3-carboxylic acid C9H7NO2 160.04 14.4 84890
aminoadipic acid C6H11NO4 160.062 4.33 49741
phenylpyruvate C9H8O3 163.04 15.1 28891
phenylalanine C9H11NO2 164.072 4.1 159486
phenyllactic acid C9H10O3 165.056 14.82 25202
quinolinate C7H5NO4 166.015 13.58 189
phosphoenolpyruvate C3H5O6P 166.975 13.7 852
pyridoxamine C8H12N2O2 167.083 0.82 821
dihydroxy-acetone-
phosphate C3H7O6P 168.991 9.3 33915
glycerol-3-phosphate C3H9O6P 171.006 7.3 121490
aconitate C6H6O6 173.009 13.8 20603
N-acetyl-ornithine C7H14N2O3 173.093 1 123922
arginine C10H14N5O7P 173.104 0.8 319794
citrulline C6H13N3O3 174.088 0.9 1208020
N-carbamoyl-aspartate C5H8N2O5 175.036 2 5657
2-isopropylmalic acid C7H12O5 175.061 13.94 969718
hydroxyphenylpyruvate C9H8O4 179.035 13.58 1603
tyrosine C9H11NO3 180.067 2 98995
3-phosphoglycerate C3H7O7P 184.986 13.58 4446
N-acetyl-glutamine C7H12N2O4 187.072 7 115332
acetyllysine C8H16N2O3 187.109 1.2 43954
93
 retention average ion counts in
time samples from glucose-
compound formula m/z (min) fed cells
kynurenic acid C10H7NO3 188.035 14.22 8218
N-acetyl-glutamate C7H11NO5 188.056 13 124910
citrate C6H8O7 191.02 13.6 458915
gluconate C6H12O7 195.051 5.8 122475
tryptophan C11H12N2O2 203.083 7.6 397371
xanthurenic acid C10H7NO4 204.03 13.9 41045
kynurenine C10H12N2O3 207.078 3.87 1588
cystathionine C7H14N2O4S 221.06 1 18338
deoxyuridine C9H12N2O5 227.067 3.54 1129
ribose-5-phosphate C5H11O8P 229.012 7.2 172703
ribulose-5-
phosphate/xylulose-5-
phosphate C5H11O8P 229.012 7.8 76696
ribose-1-phosphate C5H11O8P 229.012 8.4 4720
thymidine C10H14N2O5 241.083 5.4 1333
cytidine C9H13N3O5 242.078 1.2 9400
uridine C9H12N2O6 243.062 2.1 55241
shikimate-3-phosphate C7H11O8P 253.012 13.44 367
glucono-δ-lactone-6-
phosphate C6H11O9P 257.007 6.17 1593
glucosamine-phosphate C6H14NO8P 258.038 1.5 5312
fructose-6-
phosphate/glucose-6-
phosphate C6H13O9P 259.022 6.9 744524
fructose-1-phosphate C6H13O9P 259.022 8.1 19987
adenosine C10H13N5O4 266.09 7 2373
inosine C10H12N4O5 267.074 3.6 208308
6-phospho-gluconate C6H13O10P 275.017 13.38 25374
guanosine C10H13N5O5 282.084 4.2 7001
xanthosine C10H12N4O6 283.068 7.5 1124
sedoheptulose-7-
phosphate C7H15O10P 289.033 7.3 150747
sedoheptulose-1-
phosphate C7H15O10P 289.033 8.2 581
N-acetyl-glucosamine-
1/6-phosphate C8H16NO9P 300.049 7.3 32309
glutathione C10H17N3O6S 306.077 8 1302124
dUMP C9H13N2O8P 307.034 10.28 186
dTMP C10H15N2O8P 321.049 11.27 365

94
 retention average ion counts in
time samples from glucose-
compound formula m/z (min) fed cells
CMP C9H14N3O8P 322.045 8.7 6114
UMP C9H13N2O9P 323.029 10.2 13223
cyclic-AMP C10H12N5O6P 328.045 12.6 654
dAMP C10H14N5O6P 330.061 12 373
fructose-1,6-
bisphosphate C6H14O12P2 338.989 13.5 671479
trehalose C12H22O11 341.109 1.18 71117
AMP C10H14N5O7P 346.056 11.3 157225
dGMP C10H14N5O7P 346.056 11 165196
IMP C10H13N4O8P 347.04 10.7 10571
GMP C10H14N5O8P 362.051 10.7 12935
xanthosine-5-phosphate C10H13N4O9P 363.035 12.7 1091
sedoheptoluse
bisphosphate C7H15O10P 368.999 13.64 6538
farnesyl diphosphate C15H28O7P2 381.124 16.74 1878
S-adenosyl-
homocysteine C10H12N5O6P 383.114 6.1 2204
5-phosphoribosyl-1-
pyrophosphate C5H13O14P3 388.945 14.94 3245
octulose-8-phosphate C8H18O14P2 399.01 6 20738
dTDP C10H16N2O11P2 401.016 13.58 2873
CDP C9H15N3O11P2 402.011 13.3 4506
UDP C9H14N2O12P2 402.995 13.5 64871
trehalose-6-Phosphate C12H23O14P 421.075 7.2 30015
ADP C10H15N5O10P2 426.022 13.7 335824
dGDP C10H15N5O10P2 426.022 14 335824
GDP C10H15N5O11P2 442.017 13.5 16384
CDP-ethanolamine C11H20N4O11P2 445.053 6.38 5360
FMN C17H21N4O9P 455.097 14.2 3744
dUTP C9H15N2O14P3 466.966 14.81 856
dTTP C10H17N2O14P3 480.982 14.94 12998
CTP C9H16N3O14P3 481.977 14.8 26922
UTP C9H15N2O15P3 482.961 15 261985
dATP C10H16N5O12P3 489.994 14.94 3089
ATP C10H16N5O13P3 505.989 15.1 890368
dGTP C10H16N5O13P3 505.989 15.46 897194
GTP C10H16N5O14P3 521.983 15 40387
UDP-glucose C15H24N2O17P2 565.048 13.25 288172

95
 retention average ion counts in
time samples from glucose-
compound formula m/z (min) fed cells
UDP-N-acetyl-
glucosamine C17H27N3O17P2 606.074 13.27 801167
glutathione disulfide C20H32N6O12S2 611.145 12.8 95099
+
NAD C21H27N7O14P2 662.102 8.6 291478
NADH C21H29N7O14P2 664.118 13.9 47386
dephospho-CoA C21H35N7O13P2S 686.142 14.8 675
+
NADP C21H28N7O17P3 742.068 13.5 26546
NADPH C21H30N7O17P3 744.084 14.87 4846
coenzyme A C21H36N7O16P3S 766.108 15.5 6431
FAD C27H33N9O15P2 784.15 15 25398
acetyl-CoA C23H38N7O17P3S 808.119 15.6 172443
3-hydroxy-3-
methylglutaryl-CoA C27H44N7O20P3S 910.15 15.72 5740

Table 2.5. Parent and product ion formulas and masses, collision energy (CE), retention time (RT),
and ion counts of metabolites measured by the confirmatory LC-triple quadrupole MS/MS analysis.

average ion
counts in
samples
pare product from
parent ion nt CE ion produ RT glucose-fed
compound formula m/z (eV) formula ct m/z (min) cells
2-keto-
- -
isovalerate C5H7O3 115 11 C4H7O 71 25 1950
- -
fumarate C4H3O4 115 11 C3H3O2 71 30 980
- -
succinate C4H5O4 117 10 C3H5O2 73 23 326
- -
nicotinate C6H4NO2 122 14 C5H4N 78 23 28422
- -
malate C4H5O5 133 12 C4H3O4 115 27 102585
- -
hypoxanthine C5H3N4O 135 16 C4H2N3 92 3 21888
- -
anthranilate C7H6NO2 136 16 C6H6N 92 31 116696
α-
- -
ketoglutarate C5H5O5 145 11 C4H5O3 101 30 290088
- -
xanthine C5H3N4O2 151 21 C4H2N3O 108 3 440
- -
orotate C5H3N2O4 155 13 C4H3N2O2 111 15 140867
- -
dihydroorotate C5H5N2O4 157 12 C4H5N2O2 113 13 10147
- -
PEP C9H7O3 163 11 C7H7 91 35 42978
phosphoenolp
- -
yruvate C3H4O6P 167 20 PO3 79 32 3080
96
 average ion
counts in
samples
pare product from
parent ion nt CE ion produ RT glucose-fed
compound formula m/z (eV) formula ct m/z (min) cells
dihydroxy-
acetone-
- -
phosphate C3H6O6P 169 38 PO3 79 16 95691
glycerol-3-
- -
phosphate C3H8O6P 171 13 PO3 79 14 89497
- -
aconitate C6H5O6 173 10 C5H5O4 129 32 17818
N-carbamoyl-
- -
aspartate C5H7N2O5 175 11 C4H6NO4 132 26 47866
hydroxyphenyl
- -
pyruvate C9H7O4 179 11 C7H7O 107 30 1446
- -
3PG C3H6O7P 185 15 H2PO4 97 31 3686
- -
citrate C6H7O7 191 13 C5H3O3 111 32 331510
- -
gluconate C6H11O7 195 15 C5H5O4 129 12 13735
ribose-5-
- -
phosphate C5H10O8P 229 40 PO3 79 13 90116
ribulose-5-
phosphate/xyl
ulose-5-
- -
phosphate C5H10O8P 229 40 PO3 79 14 70248
ribose-1-
- -
phosphate C5H10O8P 229 40 PO3 79 15 11789
- -
uridine C9H11N2O6 243 19 C8H10NO5 200 3 199
shikimate-3-
- -
phosphate C7H10O8P 253 17 H2PO4 97 12 6564
fructose-6-
phosphate/glu
cose-6-
- -
phosphate C6H12O9P 259 40 PO3 79 13 111495
fructose-1-
- -
phosphate C6H12O9P 259 40 PO3 79 14 14164
- -
inosine C10H11N4O5 267 25 C5H3N4O 135 10 35223
6-phospho-
- -
gluconate C6H12O10P 275 11 H2PO4 97 31 7531
- -
xanthosine C10H11N4O6 283 22 C5H3N4O6 151 14 1420
sedoheptulose
- -
-7-phosphate C7H14O10P 289 0 H2PO4 97 13 57185
sedoheptulose
- -
-1-phosphate C7H14O10P 289 0 H2PO4 97 14 4540
N-acetyl-
glucosamine-
- -
1/6-phosphate C8H15NO9P 300 32 PO3 79 14 5571
97
 average ion
counts in
samples
pare product from
parent ion nt CE ion produ RT glucose-fed
compound formula m/z (eV) formula ct m/z (min) cells
C9H12N2O8P
- -
dUMP 307 16 C5H8O6P 195 21 136
C10H11N5O6
- -
cyclic-AMP P 328 31 C5H4N5 134 28 246
fructose-1,6-
- -
bisphosphate C6H13O13P2 339 28 H2PO4 97 31 60268
- -
trehalose C12H21O11 341 18 C6H11O6 179 3 6700
sedoheptoluse
- -
bisphosphate C7H14O10P 369 0 H2PO4 97 31 918
- -
PRPP C5H12O14P3 389 18 C5H9O10P2 291 35 387
C10H14N5O10
- -
dGDP P2 426 25 HO6P2 159 33 20092
C17H20N4O9
- -
FMN P 455 19 C5H10O7P 213 33 270
C15H23N2O17 C9H12N2O9
- -
UDP-glucose P2 565 23 P 323 30 140271
UDP-N-acetyl- C17H26N3O17 C9H11N2O11
- -
glucosamine P2 606 26 P2 385 30 70466

Accession numbers

Microarray data in this study can be viewed and downloaded from the PUMAdb website

(http://puma.princeton.edu/cgi-bin/publication/viewPublication.pl?pub_no=544).

98
 Chapter 3

Nucleotide degradation and ribose salvage in yeast

Adapted from Yi-Fan Xu, Fabien Létisse, Farnaz Absalan, Wenyun Lu, Ekaterina Kuznetsova,

Greg Brown, Amy A. Caudy, Alexander F. Yakunin, James R. Broach, Joshua D. Rabinowitz

“nucleotide degradation and ribose salvage in yeast” Molecular Systems Biology 2013, 9: 665

Abstract

Nucleotide degradation is a universal metabolic capability. Here we combine metabolomics,

genetics and biochemistry to characterize the yeast pathway. Nutrient starvation, via PKA,

AMPK/SNF1, and TOR, triggers autophagic breakdown of ribosomes into nucleotides. A protein

not previously associated with nucleotide degradation, Phm8, converts nucleotide

monophosphates into nucleosides. Downstream steps, which involve the purine nucleoside

phosphorylase Pnp1 and pyrimidine nucleoside hydrolase Urh1, funnel ribose into the non-

oxidative pentose phosphate pathway. During carbon starvation, the ribose-derived carbon

accumulates as sedoheptulose-7-phosphate, whose consumption by transaldolase is impaired

due to depletion of transaldolase’s other substrate, glyceraldehyde-3-phosphate. Oxidative

stress increases glyceraldehyde-3-phosphate, resulting in rapid consumption of sedoheptulose-

7-phosphate to make NADPH for antioxidant defense. Ablation of Phm8 or double deletion of

Pnp1 and Urh1 prevent effective nucleotide salvage, resulting in metabolite depletion and

impaired survival of starving yeast. Thus, ribose salvage provides means of surviving nutrient

starvation and oxidative stress.

99
3.1 Introduction

In natural environments, microbes face myriad environmental stressors including starvation for essential

nutrients, such as carbon, nitrogen, phosphorous. Yeast has evolved metabolic responses to these

stressors which enable their survival and effective resumption of growth when conditions improve. Recent

evidence suggests that these metabolic responses are substantially hardwired in metabolic enzymes
52,53,149
themselves . In parallel, with this intrinsic adaptation of metabolic network, proteins such as

kinases sense metabolic conditions and activate signaling cascades to coordinate metabolism and overall

cellular activity. Important examples of such signaling enzymes include protein kinase A (PKA, activated

by glucose), AMP-activated protein kinase (SNF1, repressed by glucose), and target of rapamycin (TOR,
120,150-152
activated by abundant nitrogen) .

A particularly important function of these kinases is to control levels of ribosomes, which constitute about

10% of yeast dry weight. This is achieved through regulation both of ribosome production and of
153-159
ribosome degradation via autophagy (ribophagy) . Ribosomes are composed of roughly equal
160
amounts of protein and RNA . Thus, their degradation yields both amino acids and nucleotides.

Catabolism of nucleotides involves their dephosphorylation into nucleosides, and subsequent hydrolysis
65
into ribose and bases . While many hydrolytic enzymes with nucleotidase and nucleosidase activities
161-166
exist in various organisms , the ones that are actually responsible for nucleotide degradation in cells

have remained undefined.

The carbonaceous product of nucleotide degradation is ribose-5-phosphate, which must be converted

into glycolytic intermediates in order to extract energy and generate anabolic precursors. Such conversion

is carried out by the non-oxidative pentose phosphate pathway, a series of five reversible reactions

whose net stoichiometry is 3 ribose-5-phosphate  2 fructose-6-phosphate + 1 glyceraldehyde-3-

phosphate (Figure 3.1). Both fructose-6-phosphate and glyceraldehyde-3-phosphate can be catabolized

by glycolysis to yield NADH and ATP. In addition, fructose-6-phosphate can isomerize into glucose-6-

phosphate, which in turn can be catabolized by the oxidative pentose phosphate pathway (PPP) to

100
generate NADPH, a key high energy electron donor required for reductive biosynthesis and redox
167,168
defense .

A central intermediate in the non-oxidative PPP is sedoheptulose-7-phosphate. When net flux is from

ribose-5-phosphate towards glycolytic intermediates, S7P is produced by the enzyme transketolase and

consumed by transaldolase. It was previously shown that, in the absence of transaldolase, xylose grown

E. coli could also utilize the glycolytic enzyme phosphofructokinase (Pfk1) to phosphorylate S7P into SBP

which is further converted to erythrose-4-phosphate and glycolytic intermediate dihydroxyacetone

169
phosphate . No such reaction has been reported for wild type cells or other organisms, although it has

recently been found that SBP can be produced in wild type yeast from erythrose-4-phosphate and

dihydroxyacetone phosphate and subsequently dephosphorylated by the phosphatase Shb17 to yield

35
S7P and thereby drive ribose production .

Here we use a combination of metabolomics, genetics, and biochemistry to map the nucleotide

degradation pathway in yeast (Figure 3.1), and to explore its regulation and physiological function. We

show that during nutrient starvation, yeast accumulate both nucleosides and bases, indicating active

nucleotide degradation. Surprisingly, during carbon starvation they also accumulate S7P and SBP. By

tracking the impact of gene deletions on the levels of these metabolites, we identify Phm8, which is
170
currently annotated as a lysophosphatidic acid phosphatase , as the main yeast nucleotidase. In

addition, we determine the cause and physiological function of the S7P accumulation. We show that S7P

accumulation and more generally nucleotide degradation enable metabolic homeostasis and survival of

starving yeast.

101
Figure 3.1. Yeast pathway map of nucleotide degradation and ribose salvage.
The pathway begins with the degradation of RNA. In starvation, this is triggered by macroautophagy
(ribophagy). The resulting nucleotides are degraded into nucleosides except for AMP, which is first
132
converted to IMP . Purine nucleosides are then converted into ribose-1-phosphate and base, whereas
pyrimidines are hydrolyzed into ribose and base. Both ribose and ribose-1-phosphate are converted into
ribose-5-phosphate, which enters the non-oxidative branch of PPP. Gene annotations that are taken from
definitive prior literature are shown in black; those confirmed by experiments in this study are shown in
blue, and those newly assigned in this study are shown in red. Incorrect and missing annotations in the
171
current genome scale metabolic model (http://www.comp-sys-bio.org/yeastnet/) are listed in Table 3.1.
The enzyme converting ribose to R5P remains unclear. G6P, glucose-6-phosphate; FBP, fructose-1,6-
bisphosphate; GAP, glyceraldehyde-3-phosphate; DHAP, dihydroxyacetone phosphate; R5P, ribose-5-
phosphate; X5P, xylulose-5-phosphate; Ru5P, ribulose-5-phosphate; E4P, erythrose-4-phosphate.

102
Table 3.1. Incorrect and missing annotations in the current genome scale metabolic model

Description in Substrates in
Gene Systematic the current the current Annotations and
name name genome scale genome scale corrections in this study
model model

Major 5’-nucleotidase
PHM8 YER037W None acting on GMP, UMP and
CMP

Minor 5’-nucleotidase
SDT1 YGL224C 5’-nucleotidase CMP, UMP (insufficient to substitute
for PHM8)

adenosine,
Nucleoside guanosine, No adenosine
PNP1 YLR209C
phosphorylase inosine, phosphorylase activity
xanthosine

Nucleoside adenosine, No purine hydrolase

URH1 YDR400W
hydrolase uridine, cytidine activity

glucose-1- Phosphoribomutase
PGM3
Phosphoglucomut phosphate, converting ribose-1-
(PRM15 YMR278W
ase glucose-6- phosphate to ribose-5-
)
phosphate phosphate

Minor ribokinase; other

RBK1 YCR036W Ribokinase ribose unknown major ribokinase
exists

103
3.2 Results

Starvation triggers degradation of RNA into nucleosides, bases, and sedoheptulose-7-phosphate

To characterize the nucleotide degradation pathway, we first identified physiological conditions that

robustly induced its activity. To evaluate whether nutrient starvation would do so, we used liquid

chromatography – mass spectrometry (LC-MS) to measure the changes in the metabolome profile of S.

cerevisiae cells subjected to abrupt transition from complete minimal medium to media lacking carbon,

nitrogen or phosphate. Nucleotide monophosphates (NMPs), the direct product of RNA degradation,

increased only slightly. Nucleosides and bases, however, rapidly accumulated to greater than 10-fold

basal levels (Figure 3.2). Moreover, while sugar phosphates in upper glycolysis decreased by more than

10-fold upon carbon starvation, PPP intermediates sedoheptulose-7-phosphate (S7P) and

sedoheptulose-1,7-bisphosphate (SBP) accumulated, with S7P becoming the most abundant sugar

phosphate in the cell (concentration ~ 5 mM).

To explore the origin of the accumulating metabolites, we conducted a metabolomics experiment

13
involving feeding of U- C-glucose for 70 min followed by carbon starvation (Figure 3.3A). This regimen

yields complete labeling of metabolic intermediates, including free nucleotides, but only partial labeling of

macromolecule stores (protein, glycogen, RNA and DNA). Such transient labeling allowed us to
87
determine the extent to which metabolites arise from macromolecule breakdown . We found that the

proportion of unlabeled nucleosides, bases and sedoheptulose species increased over time after carbon

starvation (Figure 3.3B, closed symbols), indicating that these metabolites were predominantly derived

from degradation of macromolecules. Consistent with those macromolecules being RNA, the total cellular

RNA concentration decreased over the same time period (Figure 3.4A).

104
Figure 3.2. Starvation triggers accumulation of nucleosides, bases and S7P.
Wild type S. cerevisiae cells growing in minimal media were switched to minimal media containing no
carbon, no nitrogen, or no phosphate. After the indicated duration of starvation, the metabolome was
quantified by LC-MS. For media composition, see Table 2.3. For the list of absolute concentration of
metabolties, see Table 3.2. Data are shown in heat map format, with each line reflecting the dynamics of
a particular compound in a particular culture condition. Metabolite levels of biological duplicates were
averaged, normalized to cells growing steadily in glucose (time zero), and the resulting fold changes log 2
transformed.

105
Table 3.2. Absolute concentration of metabolites in this study. Absolute concentrations were
19
measured by an isotope ratio based method in a single condition where metabolites abundance is high.
These direct measurements were shown in bold. Other values were inferred by fold change relative to the
directly measured condition.

Absolute concentration (mM) Absolute concentration (mM)

in replete condition in carbon starvation

(t = 0 for C-starvation) (t = 1 h for C-starvation)

Standard Standard
Compound Concentration Concentration
Deviation Deviation
AMP 8.40E-02 1.70E-02 1.20E-01 3.60E-02
CMP 5.20E-03 2.00E-03 1.80E-02 7.70E-05
DHAP 8.10E-01 9.40E-02 3.90E-02 9.20E-03
FBP 4.00E+00 5.20E-01 3.60E-02 3.10E-03
G6P 5.90E+00 8.70E-01 3.30E-01 1.80E-02
GMP 1.00E-02 3.50E-03 3.40E-02 2.00E-03
IMP 3.80E-02 1.30E-02 1.20E-02 1.40E-03
R5P 1.50E-01 3.30E-02 3.80E-02 8.70E-03
SBP 8.20E-02 2.70E-02 9.50E-01 1.90E-01
S7P 3.60E-01 2.80E-02 3.50E+00 8.10E-01
UMP 1.50E-02 5.20E-03 9.90E-02 1.50E-03
R1P 1.50E-02 1.00E-03 8.30E-03 3.70E-04

adenosine 3.30E-04 3.70E-05 5.20E-03 3.40E-04

cytidine 3.40E-04 9.60E-05 9.00E-03 1.30E-03

guanine 2.60E-03 0.00E+00 1.50E-01 9.70E-03

guanosine 3.30E-04 6.90E-05 6.80E-03 3.70E-04

hypoxanthine 4.40E-03 3.30E-04 1.20E+00 9.00E-02

inosine 3.20E-03 7.40E-05 8.90E-03 7.10E-04

uracil 3.70E-02 1.10E-02 4.10E+00 7.20E-02
uridine 7.20E-03 5.90E-04 2.80E-01 1.30E-02
xanthine 3.00E-02 9.10E-03 1.80E+00 1.40E-01

106
Figure 3.3. In carbon starvation, nucleosides and PPP compounds are produced by RNA
degradation via autophagy in response to kinase signals.
(A). Experimental design for demonstrating metabolite production via macromolecule degradation. Yeast
13
cells were grown on 2% unlabeled glucose, and then switched to U- C-glucose for 70 minutes, which
completely labels free metabolite but only partially labels macromolecules. Thereafter, glucose was
removed and the metabolome analyzed by LC-MS. (B). Fraction of unlabeled nucleosides, nucleic bases
and PPP intermediates as a function of starvation time, in wild-type and autophagy deficient (atg7
deletion) yeast. The x axis represents minutes after carbon starvation, and the y axis represents fraction
of unlabeled metabolites (mean ± range of N = 2 biological replicates). (C). Ratio of metabolite levels in
atg7 strain vs. wild type strain in carbon starvation. (D). Ratio of metabolite levels in bcy1 strain and snf1
strain vs. wild type strain in carbon starvation. (E). Ratio of metabolite levels in rapamycin treatment vs.
nitrogen starvation for wild type yeast. In (C) to (E), all reported values are log2 transformed ratios; data
are mean of duplicate samples at each time point.

107
Figure 3.4. Macroautophagy triggers RNA degradation
(A). Total RNA level before and 1 hour after carbon starvation. The y axis represents RNA level in the
units of ribose concentration (mean ± range of N = 2 biological replicates). (B). Ratio of metabolite level in
atg7 strain vs. wild type strain in nitrogen and phosphate starvation. All reported values are log 2
transformed ratios; data are mean of duplicate samples at each time point.

A likely mechanism of RNA degradation is via macroautophagy. We accordingly measured metabolite

concentrations in the atg7 deletion strain, which fails to elicit macroautophagy upon starvation. Consistent

with a requirement for macroautophagy in RNA catabolism, during carbon starvation the atg7 deletion

strain did not accumulate nucleosides, bases and sedoheptulose species, as normally observed in a wild

type strain (Figure 3.3C). Moreover, the deletion strain showed less increase in unlabeled species when
13
transient feeding of C-labeled glucose was followed by carbon starvation (Figure 3.3B, open symbols).

Similarly, the increase in nucleosides and bases during nitrogen starvation was absent in the atg7

deletion strain (Figure 3.4B). In contrast, their accumulation during phosphate starvation was not

impaired by atg7 deletion, indicating that phosphate starvation activates RNA degradation via an

alternative mechanism. Nonetheless, our results demonstrate that RNA degradation upon carbon and

nitrogen starvation requires macroautophagy.

We examined whether the mechanism by which nutrient starvation triggers autophagy and subsequent

nucleoside and base accumulation is through canonical nutrient sensing kinases, such as cyclic-AMP

108
dependent protein kinase (protein kinase A, PKA), AMP-activated protein kinase (SNF1), and target of

rapamycin complex 1 (TORC1). PKA is an inhibitor of macroautophagy but is inactivated upon glucose
158,159
removal . SNF1 is active in the absence of glucose and is required for growth on most carbon
119,172
sources other than glucose ). To test if ribophagy requires inactivation of PKA and/or activation of

SNF1, we performed carbon starvation on a bcy1 deletion strain, which exhibits constitutively activated

PKA activity, and on a snf1 deletion strain, which lacks SNF1 activity. Both strains had substantially

decreased accumulation of nucleosides, bases, and sedoheptulose species (Figure 3.3D). Thus, both

activation of SNF1 and inactivation of PKA are required for robust induction of autophagy in response to

carbon starvation. The target of rapamycin complex I (TORC1) is an inhibitor of autophagy which is
173
inactivated upon nitrogen removal . To test if TORC1 inactivation is sufficient to initiate RNA

degradation, we treated the wild type cells growing in the presence of nitrogen with rapamycin and

analyzed metabolites over time following treatment. Rapamycin treatment resulted in accumulation of

nucleosides and bases to an extent even higher than those observed following nitrogen starvation

(Figure 3.3E). Thus, TORC1 inactivation is per se sufficient to activate autophagy.

Phm8 is the physiological yeast nucleotidase

The significant accumulation of nucleosides upon nutrient starvation or rapamycin addition indicates not

only that starvation induces RNA degradation but also that the resulting nucleotides are

dephosphorylated to nucleosides. The enzyme catalyzing this step under physiological conditions has not

been identified in any organism. Sdt1 has been proposed as the responsible enzyme in yeast, since Sdt1
164
dephoshorylates UMP and CMP into uridine and cytidine rapidly in vitro . Contrary to this expectation,

our metabolomics analysis of an sdt1 strain subjected to carbon starvation did not reveal altered

nucleotide salvage pathway metabolite concentrations (Figure 3.5A). The enzyme Phm8 is homologous

to Sdt1 (42% amino acid sequence identity). Phm8 has been proposed to be a cytosolic lysophosphatidic
170
acid (LPA) phosphatase involved in LPA hydrolysis in response to phosphate starvation . Over-

expression of Phm8 decreased LPA in yeast but deletion of PHM8 did not markedly increase LPA levels,

prompting the hypothesis that Phm8 is a minor LPA phosphatase. Interestingly, PHM8, but not SDT1, is
109
 139,174,175
expressed under various stress conditions including carbon, nitrogen and phosphate starvation

(Figure 3.5B). These expression data together with PHM8’s sequence similarity to SDT1 suggested to us

that it might serve as a nucleotidase in the ribose salvage pathway.

Figure 3.5. Phm8 is the physiological yeast nucleotidase.

(A). Ratio of metabolite levels in sdt1 and phm8 strains vs. wild type strain in carbon starvation. All
reported values are log2 transformed ratios; data are mean of duplicate samples at each time point. (B).
139,174,175
Summary of transcripts data of PHM8, SDT1 and other enzymes in the pathway . All values are
log2 transformed fold changes. (C). Screening of Phm8’s phosphatase activity against 90 phosphorylated
2+
compounds. Phosphatase activity was measured in the presence of 0.5 mM substrate and 5 mM Mg
(pH = 7.0, 30°C). Compounds with specific activity higher than 0.1 µmol/mg/min are shown. The x axis
represents specific phosphatase activity (µmol of phosphate produced per minute per mg of enzyme,
mean ± range of N = 2 replicates). (D). Top table: absolute intracellular concentration of nucleotide
monophosphates in carbon starvation. Bottom plots: Phosphatase activity of Phm8 and Sdt1 as a function
of CMP concentration. The x axis represents CMP concentration and the y axis represents specific
phosphatase activity (µmol of phosphate produced per minute per mg of enzyme, mean ± range of N = 2
replicates).

110
We measured the enzymatic activity of purified Phm8 against 90 different phosphorylated compounds. Of
-1 -1
these, the preferred substrates were nucleotide monophosphates (specific activity > 500 nmol mg min
-
for AMP, GMP, CMP, and UMP) (Figure 3.5C). In contrast, the activity toward LPA was only ~2 nmol mg
1 -1 170
min . Thus, Phm8’s primary enzymatic activity is as a nucleotidase. Kinetic characterization of Phm8

and Sdt1 toward their most preferred substrate CMP shows that Phm8 has higher kcat and lower Km

(Figure 3.5D); thus, Phm8 is more active than Sdt1. To evaluate Phm8’s physiological role, we measured

metabolite levels in the phm8 strain upon carbon starvation. Although nucleosides and bases still

increased, phm8 deletion blocked the increase in S7P and SBP levels (Figure 3.5A). Thus, while other

nucleotidases can apparently partially substitute for Phm8, Phm8 is required to maintain sufficient

nucleotide catabolic flux to elevate S7P and SBP.

Nucleosides are converted into ribose-5-phosphate by Pnp1, Urh1, and Pgm3

Nucleosides are cleaved into nucleic bases and ribose moieties via either a phosphorylase, which yields

ribose-1-phosphate, or a hydrolase, which yields unphosphorylated ribose. In yeast, purine nucleoside

161,162
phosphorylase (Pnp1) and uridine hydrolase (Urh1) have been shown to catalyze these reactions .

To investigate if Pnp1 and Urh1 are required for ribose salvage, we examined strains deleted for PNP1 or

URH1 or both. Following carbon starvation the pnp1 strain accumulated more of the nucleosides inosine

and guanosine and less of the bases hypoxanthine and guanine than did wild type cells; similarly urh1

accumulated more uridine and cytidine and less uracil (Figure 3.6). Adenosine did not accumulate in the

pnp1 strain, consistent with the previous finding that Pnp1 exhibits no activity with adenosine. In addition,

guanine, cytosine, and cytidine concentrations were markedly lower than xanthine, uracil, and uridine,

presumably due to the activities of guanine, cytosine, and cytidine deaminases Gud1, Fcy1, and Cdd1

respectively (Figure 3.7A). Strains with individual deletions of PNP1 or URH1 exhibited decreased S7P

and SBP levels upon carbon starvation compared to wild type, indicating that both enzymes contribute to

ribose salvage. The double deletion strain accumulated all nucleosides and depleted all bases, and

111
exhibited decreased S7P and SBP levels following carbon starvation to an extent similar as the phm8

deletion. Thus, Pnp1 and Urh1 work in concert to convert purine and pyrimidine nucleosides into bases

and ribose or ribose-1-phosphate.

Further metabolism of ribose-1-phosphate requires its conversion to ribose-5-phosphate.

Phosphoglucomutase 3 (Pgm3) has been proposed to be a phosphoribomutase capable of catalyzing this

176
conversion . Consistent with this proposal, we found that deletion of PGM3, but not PGM1 or PGM2,

led to accumulation of ribose-1-phosphate and upstream purines in the ribose salvage pathway (Figure

3.6). Thus, Pgm3 is the physiological phosphoribomutase required for ribose salvage from purines. We

accordingly refer to the gene by the new name PRM15.

The enzyme responsible for phosphorylation of ribose generated from pyrimidine degradation remains,

however, undefined. An open reading frame (Rbk1) has been annotated as a ribokinase due to its

similarity to ribokinases in other organisms. While S. cerevisiae cannot utilize ribose as the sole carbon

source, in the absence of glucose, it did assimilate extracellular ribose, albeiet at a slow rate (Figure 3.7C,

D). Rbk1 deletion only slightly slows extracellular ribose assimilation and does not alter S7P or SBP

accumulation during carbon starvation (Figure 3.7B). Thus, another enzyme presumably phosphorylates

the ribose produced by pyrimidine nucleoside degradation. As no other yeast enzyme exhibits strong

homology to known ribokinases, it is possible that an annotated hexokinase might fulfill this role.

112
Figure 3.6. Confirmation that Pnp1 is the physiological purine nucleoside phosphorylase, Urh1 is
the pyrimidine hydrolase, and Prm15 (Pgm3) is the phosphoribomutase.
Ratio of metabolite levels in pnp1/urh1, pnp1, urh1 and prm15 (pgm3) strains vs. wild type strain during
carbon starvation. Data are shown in heat map format, with each line reflecting the dynamics of the ratio
of the metabolite level in a particular strain vs. wild type strain. All reported values are log2 transformed
ratios; data are mean of duplicate samples at each time point.

113
Figure 3.7. Clarification of downstream steps in ribose salvage pathway
(A). In carbon starvation, the total concentration of guanine, cytosine and cytidine is much lower than the
concentration of their deaminated form. The y axis represents the sum of intracellular and extracellular
metabolite amount divided by cell dry weight (mean ± range of N = 2 biological replicates) at 1 hour after
carbon starvation. (B). rbk1 deletion did not change the concentration of pentose phosphate
intermediates following carbon starvation. (C). Experimental design of data shown in (D). Yeast cells
13 12
growing on 2% U- C-glucose were switched to 2% U- C-ribose, and metabolome was analyzed by LC-
12
MS. (D). The increase of unlabeled R5P and S7P made from U- C-ribose. In (B) and (D), the x axis
represents hours after glucose removal, and the logarithmic (B) and linear (D) y axis represents absolute
intracellular concentration (mean ± range of N = 2 biological replicates).

Sedoheptulose species accumulate due to substrate limitation of transaldolase

To confirm that the sedoheptulose species accumulating during carbon starvation derive from ribose

phosphate produced from RNA degradation, we deleted both transketolase isozymes (TKL1/TKL2),

preventing conversion of pentose phosphates into sedoheptulose7-phosphate. Unlike wild-type, the

tkl1/tkl2 strain did not accumulate S7P and SBP upon carbon starvation (Figure 3.8A and B). Thus,

ribose salvage requires a functional non-oxidative PPP.

114
The non-oxidative PPP should in principle be capable of converting ribose phosphate entirely into

glycolytic intermediates. Thus, the cause of S7P and SBP accumulation under carbon starvation

conditions is unclear. In the canonical non-oxidative PPP, S7P is synthesized from two pentose

phosphates by transketolase, and then converted by transaldolase by condensation with glyceraldehyde-

3-phosphate (GAP) into fructose-6-phosphate (F6P) and erythrose-4-phosphate (E4P). Indeed, genetic

inactivation of both transaldolase isozymes (tal1/nqm1) led to chronic S7P accumulation (Figure 3.8C).

Accordingly, depletion of GAP during carbon starvation might prevent S7P consumption by transaldolase,

and thus lead to S7P accumulation. Although we could not directly measure GAP due to its low

abundance, it is in rapid exchange with DHAP, which decreased more than 20-fold in carbon starvation.

To explore whether GAP depletion underlies the S7P accumulation, in lieu of complete carbon starvation,

we transferred cells from minimal media containing glucose as the sole carbon source to minimal media

containing dihydroxyacetone (DHA) as sole carbon source. DHA can be transported into yeast cells and

phosphorylated into DHAP, which then forms GAP. These reactions can support yeast growth, but only
177
after several days of adaptation . In the shorter time scale of our experiments, we found that the switch

to DHA produces the same metabolite concentration changes as overt starvation, with only three

exceptions: DHAP was two-fold higher and S7P and SBP decreased rather than increased (Figure 3.8D).

Thus, during carbon starvation, GAP depletion precludes S7P catabolism by transaldolase and

consequently results in S7P accumulation.

The results from the above experiments present a conundrum regarding the source of SBP. Previous

studies have shown that SBP is produced in yeast via a reversible reaction catalyzed by FBP aldolase,
35
whose substrates are DHAP + erythrose-4-phosphate (E4P) . However, DHAP decreases substantially

under carbon starvation. Moreover, as evident from Figures 3.3, 3.5, and 3.6, S7P and SBP

concentrations are highly correlated. We also observed that, after feeding yeast with different labeled

forms of glucose, S7P and SBP always exhibited identical labeling patterns during subsequent carbon

starvation (Figure 3.9). Accordingly, we explored the possibility that SBP can also be formed from S7P.

115
Figure 3.8. Nucleoside degradation causes S7P accumulation due to depletion of the other
transaldolase substrate glyceraldehyde-3-phosphate, and SBP accumulation due to
phosphorylation of S7P into SBP by Pfk1.
(A). Diagram of proposed reactions. (B-E). DHAP, SBP and S7P levels in wild type, tkl1/tkl2 (B),
tal1/nqm1 (C) and pfk1 (E) strains upon carbon starvation, and wild type strain upon abruptly switching
from glucose to no carbon (WT) versus to dihydroxyacetone (WT+DHA) as the sole carbon source (D).
The x axis represents hours after carbon starvation, and the logarithmic y axis represents absolute
intracellular concentration (mean ± range of N = 2 biological replicates).

116
Figure 3.9. SBP was produced from S7P.

When feeding different labeled form of glucose before starvation, S7P and SBP exhibited identical
labeling patterns in the following carbon starvation. The top panel of each block represents the
experimental design of the data shown in the bottom panel of the same block. Left block: Yeast cells
13
growing on unlabeled glucose were switched to U- C-glucose for 70 minutes. Thereafter, glucose was
removed and metabolome analyzed by LC-MS. Middle block: Yeast cells growing on unlabeled glucose
13
were switched to 1,2- C2-glucose for 70 minutes. Thereafter, glucose was removed and metabolome
13
analyzed by LC-MS. Right block: Yeast cells growing on 1,2- C2-glucose for more than 10 generations
were switched to no carbon and metabolome analyzed. The y-axis in the data represents percentage of
each labeled form.

169
E. coli phosphofructokinase can phosphorylate S7P to SBP . To test whether this can occur in yeast we

created strains lacking either phosphofructokinases isozyme, Pfk1 or Pfk2, and measured metabolites

following carbon starvation. The pfk1 but not pfk2 deletion strain had lower SBP and higher S7P (Figure

3.8E and Figure 3.10), indicating that Pfk1 phosphorylates S7P to SBP, perhaps selectively during

117
carbon starvation when the S7P/F6P ratio (reflected by S7P/G6P ratio, as G6P and F6P are linked via a

single reversible reaction) greatly increases (Figure 3.3). As the pfk1 strain exhibited lower DHAP

compared to the wild type strain upon carbon starvation, SBP is presumably degraded by aldolase into

DHAP, providing an alternative route into glycolysis.

Figure 3.10. pfk2 deletion doesn’t show changes of SBP and S7P
DHAP, SBP and S7P levels in wild type and pfk2 strains upon carbon starvation. The x axis represents
hours after carbon starvation, and the logarithmic y axis represents absolute intracellular concentration
(mean ± range of N = 2 biological replicates).

Ribose salvage is essential for yeast’s survival in starvation and oxidative stress

The identification of the enzymes of ribose salvage provided us the opportunity to investigate the

pathway’s physiological role. Ribose salvage could serve to maintain energy charge and levels of key

metabolites during nutrient starvation. We confirmed this hypothesis by metabolomic analysis of the

ribose salvage mutants phm8 and pnp1/urh1 (Figure 3.11A). Specifically, there was a marked

impairment in energy charge in carbon starvation, glutamine abundance (a hallmark of nitrogen

119,178 179
availability; in nitrogen starvation, and ATP abundance (a hallmark of phosphate availability; in

phosphate starvation. These metabolite concentration changes were also associated with altered cell

growth and viability: While the phm8 and pnp1/urh1 deletion strains grew identically to wild type on

complete minimal media, they were less effective than wild type at sustaining growth when nitrogen or

phosphorous was removed (Figure 3.11B). All strains ceased growth immediately upon carbon starvation.
118
However, phm8 and pnp1/urh1 strains exhibited substantially slower initiation of gluconeogenesis if

switched from glucose to glycerol+ethanol (Figure 3.11D), presumably due to the inferior carbon supply

in the interval between fermentative and respiratory growth. The phm8 and pnp1/urh1 strains also

exhibited decreased viability in long-term starvation (Figure 3.11C). These findings indicate that

nucleotide degradation and ribose salvage is essential for yeast’s growth and survival under nutrient

deprivation.

The above results provide a clear physiological role for the ribose salvage pathway. They do not, however,

clarify the function of the S7P accumulation during carbon starvation. Motivated by our observation that

S7P can be consumed if GAP is available, we looked for a stress condition under which DHAP and GAP
180
might accumulate. Since oxidative stress inhibits glyceraldehyde-3-phosphate dehydrogenase in yeast ,

carbon stored as S7P could be released by a transient increase in GAP triggered by oxidative stress.

Moreover, conversion of S7P to F6P could in turn lead to production of glucose-6-phosphate by

phosphoglucose isomerase and NADPH production by the oxidative PPP. Subsequent consumption of

NADPH to regenerate reduced glutathione could promote survival of oxidative stress. To test this

possibility, we added H2O2 to carbon-starved wild type, phm8 and pnp1/urh1 yeast cells (Figure 3.12A

and B). Indeed, DHAP, which is in rapid exchange with GAP, accumulated immediately upon oxidative

stress, mirrored by a rapid decrease in S7P and rise in G6P in the wild-type cells (Figure 3.12C). The

phm8 and pnp1/urh1 deletion strains, in contrast, exhibited lower G6P, NADPH/NADP+ ratio, reduced

glutathione, and viability (Figure 3.12D). Thus, S7P accumulated during short-term carbon starvation

enables rapid NADPH production in response to oxidative stress.

119
Figure 3.11. The ribose salvage pathway is essential for yeast’s survival in starvation
(A). Level of hallmark metabolites and energy charge in wild type, phm8 and pnp1/urh1 strains in
starvation. The x axis represents hours after starvation, and the logarithmic y axis represents energy
charge ([ATP]+0.5[ADP]/([ATP]+[ADP]+[AMP]) or absolute intracellular concentration (mean ± range of N
= 2 biological replicates). (B, C). Growth and viability of wild type, phm8 and pnp1/urh1 strains in
starvation. (D). Growth of wild type, phm8 and pnp1/urh1 strains in the switch from glucose to glycerol +
ethanol. The x axis represents hours after starvation and the logarithmic y axis represents optical density
(A600) in (B, D) and percentage of live cells calculated by dividing the number of colonies formed after
starvation by the number of colonies formed before starvation in (C) (mean ± range of N = 2 biological
replicates).

120
Figure 3.12. The accumulation of S7P in carbon starvation facilitates survival of subsequent
oxidative stress
(A). Experimental design. Yeast cells growing on glucose were switched to no carbon media for 30 min.
Thereafter, 20 mM H2O2 was added and metabolome analyzed by LC-MS. (B). Diagram of proposed
reactions and their regulation. S7P accumulates during carbon starvation due to low level of
glyceraldehyde-3-phosphate (GAP). Oxidative stress blocks GAP consumption via glycolysis by inhibiting
GAPDH. This provides sufficient GAP to react with S7P, producing F6P via transaldolase. F6P is further
converted to G6P which enters the oxidative PPP. The resulting NADPH is utilized for regeneration of
reduced glutathione. 1,3-DPG, 1,3-diphosphoglycerate; GSSG, glutathione disulfide; GSH, reduced
+
glutathione. (C, D). DHAP, S7P, G6P, and glutathione levels and NADPH/NADP ratio in wild type, phm8
and pnp1/urh1 strains in the experiment shown in (A). The x axis represents minutes after oxidative
stress, and the y axis represents absolute intracellular concentration or ratio of intracellular concentration
(mean ± range of N = 2 biological replicates). (E). Viability of wild type, phm8 and pnp1/urh1 strains in
oxidative stress. The y axis represents percentage of survived cells calculated by dividing the number of
colonies formed after oxidative stress by the number of colonies formed under the same starvation
condition but without H2O2 treatment (mean ± range of N = 2 biological replicates).

121
3.3 Discussion

Our characterization of nucleotide degradation in yeast represents the most comprehensive metabolomic

investigation to date, in any organism, of the pathway from nucleotide monophosphates to central

metabolism. Such investigation has substantially clarified the enzymatic steps of the pathway. Because

degradative pathways, unlike biosynthetic pathways, are inessential for growth, they cannot be readily

mapped by purely genetic approaches relying on overt phenotypes such as dependence on a particular

added nutrient, or auxotrophy. Moreover, many degradation enzymes have promiscuous activity against a
58
diversity of substrates, rendering their functions difficult to determine from biochemistry alone . For
59-61
example, in E. coli, various enzymes have been putatively identified as nucleotidases and
62-64
nucleosidases , but which one(s) are physiologically important remains unknown. Mammalian
65
nucleotide degradation and ribose salvage is yet more complex and less well defined .

To analyze the physiological function of nucleotide degradation enzymes in yeast, we employed

66-71
metabolomic analysis of targeted deletion strains , and examined the metabolic response to a

perturbation that strongly induces the degradation pathway. This allowed us to identify the physiologically

relevant enzyme for all but one reaction of the pathway. The significance of this effort is highlighted by its

rectifying multiple different misannotations in the current genome-scale yeast metabolic model, including

clarifying the lack of Pnp1 activity against adenosine (Figure 3.6), the specificity of Urh1 to pyrimidines

(Figure 3.6), the phosphoribomutase activity of Pgm3 (now Prm15) (Figure 3.6), and most importantly

identifying the physiological yeast nucleotidase to be Phm8 (Figure 3.5), a gene of the haloacid

dehalogenase-like hydrolase family. Intriguingly, the strongest metabolic response to PHM8 deletion

occurred in S7P and SBP levels, which are more sensitive to flux through the ribose salvage pathway

than are Phm8’s direct substrates and products. Measurement of these more distal metabolites was a key

benefit of metabolomics over a more targeted analytical approach.

We initiated our study of the ribose salvage pathway to explore an initial observation that, during

starvation, mutants defective in nutrient signaling (bcy1 and snf1) exhibited a distinct pattern of PPP
53
metabolites (Figure 3.4D) . This suggested that nutrient signaling kinases might regulate metabolic

122
enzymes of the PPP through direct phosphorylation. Similar to our recent observations regarding control

of glycolytic enzymes upon glucose removal and re-feeding (Xu et al., 2012b), however, we do not find

evidence that direct enzyme phosphorylation is controlling PPP metabolite concentrations or fluxes.

Rather, the nutrient signaling kinases serve to coordinate metabolism and cell ribosome contents, through

regulation of autophagy. The effect of these kinases on the ribose salvage pathway intermediates thus

arises from mass action as a consequence of the increased production of nucleotide monophosphates via

ribophagy.

Beyond being a marker of nucleotide degradation flux, S7P accumulation during carbon starvation

provides a potentially useful nutrient reserve of carbon and phosphate. We find that this reserve helps

carbon-starved yeast to respond to an acute oxidative stressor. The initial cause of the S7P accumulation

in carbon starvation is depletion of glyceraldehyde-3-phosphate, which, when present, reacts with S7P to

produce fructose-6-phosphate and erythrose-4-phosphate. Oxidative stress blocks the glycolytic

180
consumption of glyceraldehyde-3-phosphate , thereby providing the substrate required to react with

S7P. The resulting fructose-6-phosphate production in turn drives NADPH production via the oxidative

PPP and subsequent glutathione reduction.

The accumulation of S7P during carbon starvation is reminiscent of similar accumulation of PEP upon

glucose removal. We have recently shown that the increase in PEP is caused by rapid depletion of

fructose-1,6-bisphosphate (FBP), an essential activator of the PEP-consuming enzyme pyruvate kinase.

The elevated PEP provides a readily accessible reserve of high-energy phosphate bonds which facilitate
52,53
fast uptake of glucose if it becomes re-available . Analogously, S7P accumulation is caused by

depletion of FBP’s product glyceraldehyde-3-phosphate, a substrate of the S7P-consuming enzyme

transaldolase. Like the elevated PEP, the increased S7P prepares the cell for a possible subsequent
181
event, oxidative stress in this case, a major threat for microbes growing on plants . In both cases

intrinsic properties of the metabolic network result, during carbon starvation, in carbonaceous metabolite

accumulation to provide energy reserves for quick response to further environmental changes.

Beyond the specific role of the S7P accumulation in stress protection, we find that nucleotide degradation

is critical to metabolic homeostasis and thus survival of starving yeast. In previous work, during the
123
transient period following glucose upshift when ATP consumption by hexokinase and

phosphofructokinase precedes ATP generation by lower glycolysis, AMP degradation was found to help
132
maintain energy charge . In starvation, the goal is not elimination of AMP, but access to the carbon,

nitrogen, and phosphorous within it and other nucleotide monophosphates. That such constituents would

be fed into central metabolism via nucleotide degradation is not unexpected. The significance of their

contribution to starvation survival is nevertheless surprising, given that proteins are more copious and

carbohydrate stores more accessible. The strikingly lower energy charge, altered metabolite levels, and

poor survival in the phm8 deletion strain, however, show that nucleotide salvage plays an indispensable

role. This can be rationalized based on the abundance of ribosomes, which are needed in large amounts
182-184
only during rapid growth . Thus they may be preferentially used as fuel during the transition from fast

growth to slow growth or stationary phase.

Building from careful mapping of the nucleotide degradation pathway, we have been able to demonstrate

an important role of the pathway in starvation survival of yeast. Orthologs of Phm8 are found in most fungi

and plants, and some bacteria and animals but not humans (http://yeast-phylogroups.princeton.edu/). In

organisms where Phm8 orthologs are present, our results may directly facilitate mapping of their

nucleotide degradation pathways. In other organisms, the approach laid out here may translate even the

gene product does not. For example, although mammalian cells are not normally subject to acute
185,186
starvation, autophagy is essential for survival of the neonatal period and contributes to the growth of
187-189
Ras-driven tumors . The latter observation raises the intriguing possibility that a more complete

understanding of nucleotide degradation in humans would reveal new drug targets for treatment of

aggressive cancers.

124
3.4 Methods

Yeast strains and media

Yeast strains were derived from prototrophic S288C or W303. Prototrophic deletions were created by

homologous recombination using the allele amplified by PCR from the synthetic genetic analysis (SGA)
190
deletion set . Double deletion strains were made by sporulation and tetrad dissection. Cells were grown

in minimal media comprising 6.7g/L Difco Yeast Nitrogen Base without amino acids plus 2% (w/v) glucose.

For nitrogen and phosphate starvation, cells were grown in nitrogen or phosphate limited media (Table

2.3) before nutrient deprivation to deplete the otherwise large intracellular store of these two nutrients.

Yeast culture conditions and extraction

53
The metabolome of Saccharomyces cerevisiae was characterized as described previously . Briefly,

saturated overnight cultures were diluted 1:30 and grown in liquid media in a shaking flask to A600 of ~

0.6. A portion of the cells (3 mL) were filtered onto a 50 mm nylon membrane filter (Millipore, Billerica,

MA), which was immediately transferred into -20 °C extraction solvent (40:40:20

acetonitrile/methanol/water). For nutrient removal, 100 mL of cell culture at A600 of ~0.6 were poured

onto a 100 mm cellulose acetate membrane filter (Sterlitech, Kent, WA) resting on a vacuum filter holder

with a 1000 mL funnel (Kimble Chase, Vineland, NJ) and were washed with 100 mL pre-warmed (30°C)

starvation liquid medium lacking the indicated nutrient. Immediately after the wash media went through,

the filter was taken off the holder and the cells were washed into a new flask containing 100 mL pre-

warmed (30 °C) starvation medium. Samples were then taken at the indicated time points after nutrient

removal and filtered and quenched as described above.

For RNA extraction and quantitation, yeast cells were grown in the same condition as for metabolomics

experiments. Total RNA was extracted using the hot acid phenol method followed by ethanol precipitation
191
. The resulting sample was dissolved in water and total RNA concentration was measured by a

fluorometer (Qubit; Invitrogen, Carlsbad, CA).

125
For cell viability assay, 200 µL of culture were diluted in a 10 fold series, spread onto YPD agar plates

and incubated for 24 hours. The percentage of cells that survived under a stress was calculated by

dividing the number of colonies formed after a stress exposure by the number of colonies formed under

the same growth condition but without the stress exposure.

Metabolite measurement

Cell extracts were analyzed by reversed phase ion-pairing liquid chromatography (LC) coupled by

electrospray ionization (ESI) (negative mode) to a high-resolution, high-accuracy mass spectrometer

(Exactive; Thermo Fisher Scientific, Waltham, Massachusetts) operated in full scan mode at 1 s scan time,
5
10 resolution, with compound identities verified by mass and retention time match to authenticated
34
standard . Isomers are reported separately only where they fully chromatographically resolved.

Absolute intracellular metabolite concentrations in steadily growing S. cerevisiae were determined as

36
described previously . Metabolite concentrations after perturbations were computed based on fold-

change in ion counts relative to steadily-growing cells (grown and analyzed in parallel) multiplied by the

known absolute concentration in the steadily growing cells, as determined using an isotope ratio-based
19 192
approach . Energy charge was calculated as ([ATP]+0.5[ADP])/([ATP]+[ADP]+[AMP]) .

Protein purification and enzymatic assays

The genes encoding Phm8 (Yer037w) and Sdt1 (Ygl224c) were amplified by PCR using S. cerevisiae

genomic DNA. The amplified fragments were cloned into a modified pET15b vector (Novagen, Darmstadt,

Germany) and overexpressed in the E. coli BL21(DE3) Gold strain (Stratagene, La Jolla, California) as
193
previously described . The recombinant proteins were purified using metal ion affinity chromatography

on nickel affinity resin (Qiagen, Hilden, Germany) to high homogeneity and stored at -80 ºC. Purified

Phm8 and Sdt1 were screened for the presence of phosphatase activity against the general phosphatase

substrate p-nitrophenyl phosphate (pNPP) and 90 phosphorylated metabolites as described previously .

For the determination of kinetic parameters (Km and kcat), phosphatase activity was determined in the

126
presence of 5 mM MgCl2 over the range of substrate concentrations (0.001 to 9.5 mM), and the kinetic

parameters were calculated by non-linear regression analysis of raw data to fit to the Michaelis-Menten

function using the GraphPad Prism Software (GraphPad Software, San Diego, CA).

127
 Chapter 4

Identification of a yeast polyol phosphatase required for maintaining

phosphoglucose isomerase activity

Adapted from Yi-Fan Xu, Jonathan Chen, Wenyun Lu, Sarah Johnson, Patrick A. Gibney, David

Botstein, Joshua D. Rabinowitz “Identification of a yeast polyol phosphatase required for

maintaining phosphoglucose isomerase activity” in preparation

Abstract

Sugar alcohols (Polyols) exist widely in nature. Saccharomyces cerevisiae accumulates glycerol,

a three-carbon polyol, as an intracellular solute during osmotic stress. Here we show that

YNL010W, a gene conserved across all fungi species and some plants, encodes a polyol

phosphatase (Pyp1). Unlike sn-(L)-glycerol-3-phosphatase, this enzyme acts on D-polyol

phosphates without specificity with respect to the number of carbons. Polyol phosphates are

structural analogs of the intermediate of the phosphoglucose isomerase reaction. Therefore,

Pyp1 activity is important for yeast to maintain phosphoglucose isomerase (Pgi1) activity in the

presence of polyols or pentose sugars: cells with PYP1 deleted fail to grow on sorbitol and grew

defectively in the presence of ribitol or ribose. Such inhibition results in a greater growth defect

in faster growing cells, which is consistent with PYP1’s expression levels being highly correlated

with growth rate, due to the higher demand for glycolytic flux compared to pentose phosphate

pathway flux in faster growing cells. Thus, microbial sugar phosphate metabolism is dynamically

protected from polyol phosphate analog inhibitors.

128
4.1 Introduction

194,195
Polyols, the reduced forms of sugar, are a common family of carbohydrates in nature . Glycerol, the

simplest form of a polyol, is the backbone of phospholipids and the excretion product of many microbes
196,197
as a stress response . Common longer chain polyols include erythritol, ribitol, xylitol, arabitol, sorbitol,

and mannitol, all of which exist only in the D- form and are usually found in plants. Due to the inability of

most organisms to assimilate long chain polyols, they are often regarded as inert solutes with unclear

physiological functions. However, polyols have gained more interest in the last decade with the discovery
198
in the food science industry for their sugar substitution ability .

199
Many fungi species are able to produce polyols via reduction of the corresponding sugar . The involved

pathway has been overexpressed for engineered production of polyols via the dephosphorylation of
200-203
sugar phosphates and the following reduction of the resulting sugar . This method is clearly different

from the glycerol production pathway, where sn-(L)-glycerol-3-phosphate is first made from the reduction
204,205
of dihydroxyacetone phosphate, followed by the dephosphorylation of the phosphorylated polyol .

Thus, one should expect an alternative pathway for polyol production, which involves the reduction of

sugar phosphates and subsequent dephosphorylation of the resulting polyol phosphate. However, no

fungal enzymes have been identified so far carrying these two catalytic activities.

Phosphoglucose isomerase locates at the branch point between glycolysis and the oxidative pentose

phosphate pathway (PPP). Glycolysis provides energy and the precursors for amino acid biosynthesis

and the tricarboxylic acid cycle, with extra carbons flowing into fermentation pathways to maintain redox

balance. A common assumption, although subject to some dispute, is that the committed step, catalyzed
9,42-44
by phosphofructokinase-1 (Pfk1), functions as the critical glycolytic control point . The oxidative PPP

provides NADPH and precursor for ribosome biogenesis, both of which are required for biomass

production. To optimize biomass production and growth, this branch point must be tightly regulated. The

regulation of the oxidative PPP has been intensively studied due to the importance of NADPH in fast
206
proliferating cells such as cancer cells . Again, most studies have focused on the regulation of the first

129
 192
committed step, glucose-6-phosphate dehydrogenase . Thus, despite of its important location on the

metabolic map, phosphoglucose isomerase is currently not thought to be a key regulatory object.

Phosphoglucose isomerase catalyzes the inter-conversion between glucose-6-phosphate and fructose-6-

phosphate through a mechanism which involves the opening of the glucose pyranose ring, isomerization
207
of glucose into fructose through an enediol intermediate, and the closing of the fructose furanose ring .

We noticed that polyol phosphates mimic the structure of the enediol intermediate (Scheme 4.1). Thus, it

is likely that polyol phosphates could function as a competitive inhibitor of phosphoglucose isomerase.

Here we show that YNL010W, a gene conserved across all fungi species and some plants, encodes a

polyol phosphatase (Pyp1). This enzyme is responsible for regulating the level of polyol phosphates in

yeast. Cells lacking this enzyme cannot grow on sorbitol and grow poorly in the presence of ribitol and

other polyols because of the accumulation of intracellular polyol phosphates. We show biochemically that

these polyol phosphates are indeed strong inhibitors of phosphoglucose isomerase. Expression of PYP1

is highly correlated with growth rate, consistent with the increased demand for phosphoglucose

isomerase activity, which drives more carbon into glycolysis in faster growing cells.

Scheme 4.1

130
4.2 Results

A yeast Pyp1 mutant accumulates octulose-8-phosphate and polyol phosphates

We initiated our study via a metabolomics screen of yeast deletion strains with genes of unknown function.

Yeast metabolome was then measured by liquid chromatography – mass spectrometry (LC-MS). We

found that the deletion of PYP1, formerly known as the uncharacterized gene YNL010W, led to the

accumulation of three metabolites in yeast grown on glucose (Figure 4.1).

13 15 208
The metabolites’ formulae were obtained by labeling cells with C and N and observing no shift from

nitrogen labeling and a shift of +8, +5, or +4 daltons from carbon labeling (Figure 4.2A). Exact masses of

these compounds matched putative formulae of C8H17O11P1, C5H13O8P1 and C4H11O7P1. Upon searching

these formulae in the KEGG database, we found matching metabolites as octulose-8-phosphate (O8P)

for C8H17O11P1, ribitol-5-phosphate (ribitol-5P)/arabitol-5-phosphate (aol5P)/xylitol-5-phosphate (xol5P) for

C5H13O8P1, and erythritol-4-phosphate (erythritol-4P) for C4H11O7P1.

35
The existence of octulose-8-phosphate in yeast has been reported previously . Indeed, the

chromatographic retention time of the accumulated C8H17O11P1 exactly matched the synthesized O8P

standard. To find the identity of the five- and four-carbon compounds, we performed an isotope labeled
13
experiment involving switching yeast cells from U- C-glucose to unlabeled ribitol, arabitol, xylitol or

erythritol (Figure 4.2B). Although baker’s yeast cannot utilize these polyols as effective carbon sources,

they can transport and phosphorylate them slowly in the absence of glucose. We found that feeding these

five and four-carbon polyols to yeast resulted in the accumulation of intracellular metabolites with masses

exactly matching formulae of the above polyol phosphates. Moreover, those derived from ribitol and

erythritol exactly matched the retention time of the endogenously accumulated C5H13O8P1 and C4H11O7P1

respectively (Figure 4.2C). Feeding of arabitol and xylitol resulted in the accumulation of polyol

phosphates with different retention times (Figure 4.2C). Thus, the five- and four-carbon compounds

accumulated in the ynl010w strain were identified as ribitol-5P and erythritol-4P.

131
Figure 4.1. Metabolomic phenotype of pyp1 deletion cells

(A). Metabolite structures associated with metabolic phenotype of pyp1Δ. O8P, ribitol-5P and erythritol-4P
accumulated in pyp1 cells grown on glucose. Sorbitol-6P accumulated in pyp1Δ cells in the presence of
sorbitol. Glycerol-1-phosphate, but notglycerol-3-phosphate, is in vitro substrate of the Pyp1 enzyme,
indicating that Pyp1 is a D-polyol phosphatase. (B). Relative quantitation of accumulated metabolites in
cells grown on glucose. The y axis represents the ratio of metabolite level in pyp1Δ cells vs. wild type
cells (mean ± range of N = 4 biological replicates). (C). The negative ionization mode-extracted ion
chromatogram for O8P in pyp1Δ and wild type cells. Inset: Mass spectrum displaying the accurate mass
13
for the parent ion (M) and natural C abundance ion (M+1) observed for O8P in negative ionization mode
via LC/MS. (D). Table of [M-H] ions with altered abundance between pyp1Δ and wild type cells.

While ribitol-5P and erythritol-4P are polyol phosphates, O8P is a sugar phosphate. Thus, how these

different phosphorylated metabolites accumulated simultaneously was intriguing. The most common
209
conformer of the seven carbon sugar sedoheptulose and its derivatives in solution is β-furanose . It is

likely that O8P has a similar β-furanose ring structure (Figure 4.1A), which has three carbons (6’-8’) of

octulose out of the ring, forming a D-glycerol-3-phosphate-like tail. Interestingly, this moiety was

conserved in all three accumulated metabolites (Figure 4.1A), indicating that Ynl010w is a non-specific

132
D-polyol phosphatase. Although D-polyols are excretion products of fungi and plants, polyol phosphates

were not previously thought to be an intermediate in this pathway (sugar phosphate  sugar  polyol).

We hereafter name this newly identified enzyme polyol phosphatase (Pyp1).

Figure 4.2. Formulae deduction of accumulated metabolites in pyp1 deletion strain.

13 14
(A). Table of [M-H] ions in cells growing on C or N labeled media. (B). Experimental design for
13
identification of five and four carbon polyol phophates. pyp1Δ cells grown on U- C-glucose were
switched to minimal media without glucose and in the presence of unlabeled polyol. (C). The negative
ionization mode-extracted ion chromatogram for five (left panel) and four (right panel) carbon polyol
13
phosphates. The upper chromatogram of each panel shows C labeled polyol phosphate and the lower
chromatogram of each panel shows unlabeled polyol phosphate. Internal comparison of the retention time
identifies the five carbon polyol phosphate as ribitol-5-phosphate and the four carbon polyol phosphate as
erythritol-4-phosphate.

133
Pyp1 is a D-polyol phosphatase

To determine if the accumulated compounds were indeed the substrates of Pyp1, we synthesized O8P

and ribitol-5P and tested the biochemical activity of recombinant Pyp1 on these compounds. Incubation

with Pyp1 led to the depletion of ribitol-5P and the accumulation of ribitol (Figure 4.3). We however, didn’t

observe detectable phosphatase activity on O8P. O8P is an intermediate of the non-oxidative PPP with
35
very low producing and consuming flux . Thus, it is likely Pyp1’s activity on O8P is also low, which is

below the detection level in our assay. It is also likely that Pyp1 needs to bind an activator or highly active

substrate first in order to act on O8P.

Figure 4.3. Pyp1 is active against compounds with D-glycerol-3-phosphate tail.

2+
Purified Pyp1’s phosphatase activity was assayed in the presence of 0.5 mM substrate and 5 mM Mg
(pH = 7.0, 30°C) by monitoring the appearance of phosphate or the polyol. See Methods for details. The
y axis represents specific phosphatase activity (µ mol product made per mg of protein per minute) (mean
± range of N = 2 replicates).

To test if Pyp1 is indeed a D-polyol phosphatase with no specificity for the length of the overall carbon

chain, we also measured its activity against sorbitol-6-phosphate, L-glycerol-3-phosphate (glycerol-3P,

134
the common sn-glycerol-3-phosphate) and D-glycerol-3-phosphate (glycerol-1P, sn-glycerol-1-phosphate).

Incubation with Pyp1 led to the depletion of sorbitol-6P and glycerol-1P, but not glycerol-3P (Figure 4.3).

Unlike the known glycerol-3-phosphatases (Hor2 and Rhr2) in the glycerol biosynthetic pathway, which
205
act on both glycerol-3P and glycerol-1P , Pyp1 is specific to D-polyol phosphates.

Pyp1 deletion mutant accumulates sorbitol-6-phosphate and fails to grow on sorbitol

Although D-polyol phosphates were not thought to be present in baker’s yeast, they have been
210
discovered in other fungi species . There are two likely routes for the in vivo biosynthesis of these

compounds. The first involves the reduction of the corresponding sugar phosphate and the second

involves phosphorylation of polyols. Both of these activities were discovered in fungi but the responsible
210
genes have yet to be discovered . As there are no detectable polyols except glycerol in baker’s yeast,

polyol phosphates in cells grown on glucose are more likely made from the reduction of sugar phosphates.

In more distant organisms, ribitol-5-phosphate dehydrogenase has been found in gram positive
211
pathogens including Haemophilus influenza and Staphylococcus aureus . Mannitol-1-phosphate (same

as mannitol-6-phosphate due to the symmetric structure) dehydrogenase and sorbitol-6-phosphate

212
dehydrogenase have been found in various bacteria including Escherichia coli . None of these

enzymes have homologs in yeast, so the responsible enzyme for polyol phosphate dehydrogenase

activity remains unknown . However, we do find that deletion of Zwf1, the first step in the oxidative PPP,

resulted in the disappearance of ribitol-5-phosphate (Figure 4.4), presumably due to the strain’s inability

to synthesize ribulose-5-phosphate. This confirms that ribitol-5-phosphate was made from the reduction

of ribulose-5-phosphate in yeast grown on glucose.

135
Figure 4.4. Relative level of ribitol-5-phosphate and ribulose-5-phosphate/xylulose-5-phosphate in
the wild type and zwf1 deletion strain.

The metabolomes of wild type and zwf1Δ strains growing on 2% glucose were measured by LC-MS. The
y axis represents relative metabolite level. (mean ± range of N = 2 replicates).

When polyols are present in the growth environment, polyol phosphates can be made from the

phosphorylation of polyols. Of all the natural polyols, sorbitol is the only polyol that yeast can utilize as the
213 214
sole carbon source . Unlike bacteria, which use a phosphotransferase system , yeast utilizes sorbitol

by first oxidizing it into fructose. Because yeast could phosphorylate various polyols in the absence of

glucose (Figure 4.2B and C), we hypothesized that sorbitol-6-phosphate could be made in yeast grown

on sorbitol. To test this, we grew both wild type and pyp1Δ strains on minimal media containing sorbitol as

the sole carbon source. We found that while wild type cells managed to grow up after a long lag phase,

pyp1Δ cells couldn’t grow (Figure 4.5A). Metabolome profiling revealed that sorbitol-6-phosphate levels

in the pyp1Δ strain were much higher than the wild type strain (Figure 4.5B). These results confirmed

that Pyp1 dephosphorylates sorbitol-6-phosphate in vivo and suggested that sorbitol-6P is toxic to yeast

cells.

136
Sorbitol-6-phosphate slows down yeast growth due to its inhibition on phosphoglucose

isomerase

To investigate the physiological effect of sorbitol-6P, we expressed apple aldose-6-phosphate reductase

215
(A6PR) which converts fructose-6-phosphate (F6P) to sorbitol-6P under a gal promoter on a high copy

plasmid. The resulting cells (yA6PR) and cells transformed with the same plasmid but without the A6PR

gene (control cells) were grown on galactose. We found that yA6PR cells showed consistently slower

growth rate than the control cells (Figure 4.5C) and accumulated ~1 mM intracellular sorbitol-6P (Figure

4.5D). Metabolome profiling revealed that yA6PR cells also accumulated glucose-6-phosphate (G6P)

while manifesting lower levels of fructose-1,6-bisphosphate (FBP) and dihydroxyacetone phosphate

(DHAP) (Figure 4.5D), suggesting that a step between G6P and FBP was inhibited. We also observed

that pentose phosphate intermediates ribose-5-phosphate (R5P) and sedoheptulose-7-phosphate (S7P)

accumulated (Figure 4.5D), indicating a relatively higher pentose phosphate pathway flux. These

metabolome data suggested that either phosphoglucose isomerase (Pgi1 in yeast) or phosphofrutokinase

(Pfk1,2 in yeast) was inhibited.

137
Figure 4.5. pyp1 deletion cells accumulated sorbitol-6-phosphate which inhibited phosphoglucose
isomerase and resulted in growth defect.

(A). Growth of pyp1Δ and wild type cells on sorbitol. Yeast cells grown on YPD medium were switched to
minimal media containing 2% sorbitol as the carbon source. The x axis represents days after the switch
and the y axis represents optical density (A600). (B). Absolute concentration of sorbitol-6-phosphate in
pyp1Δ and wild type cells in the experiment shown in (A). The x axis represents days after the switch and
the y axis represents absolute intracellular concentration (mean ± range of N = 2 replicates). (C). Specific
growth rate (mean ± range of N = 2 replicates) for yeast cells expressing A6PR enzyme (yA6PR) and
empty vector (control). Yeast cells grown on YPD medium were diluted into minimal media containing
galactose as the carbon source. (D). Metabolome profiling of yA6PR (pink) and control cells (dark blue )
grown on galactose indicates an upregulated PPP and downregulated glycolysis. (E). Left panel: Labeling
fractions of pentose-5-phosphate (sum of ribose-5-phosphate, ribulose-5-phosphate and xylulose-5-
13 13
phosphate) in cells with high (5 nM estradiol) and low (1 nM estradiol) Pgi1 induction, where C2 and C4
labeling fraction indicate relative oxidative PPP level. The y axis represents labeling fractions (mean ±
range of N = 2 replicates). Right panel: Relative oxidative PPP level in cells with high and low Pgi1
induction. The y axis represents relative oxidative PPP level (mean ± range of N = 2 replicates). See
Methods for calculation deductions. (F). Metabolome profiling of cells with high (black) and low (grey)
Pgi1 induction. In (D) and (F), the y axis represents absolute intracellular concentration, except for the
+
panel displaying the NADPH/NADP ratio (mean ± range of N = 2 replicates).

138
The mechanism of phosphofructokinase involves the phosphorylation of the β-furanose form of F6P into
216
FBP without opening of the furanose ring . The mechanism of phosphoglucose isomerase involves the

opening of the glucose pyranose ring, isomerization of glucose into fructose through an enediol
207
intermediate, and the closing of the fructose ring . We noticed that the structure of sorbitol-6P exactly

mimics the structure of the enediol intermediate of phosphoglucose isomerase (Figure 4.6A), indicating

that sorbitol-6P is likely to be an inhibitor of phosphoglucose isomerase. Indeed, sorbitol-6P and other

compounds with structures mimicking the enediol intermediate were previously reported as strong
217
inhibitors of phosphoglucose isomerase in vitro . To confirm this, we performed biochemical assays of

purified Pgi1 in the absence or presence of different concentrations of sorbitol-6P. We found that sorbitol-

6P is a very strong inhibitor of Pgi1 activity with an IC50 of ~50 µM (Figure 4.6B). Upon adding 1 mM

sorbitol-6P, Pgi1 activity was inhibited ~85%. Such inhibition resulted in the hyperactive oxidative PPP

flux and hypoactive glycolytic flux observed in yA6PR cells and the growth defect in pyp1Δ cells grown on

sorbitol. To further confirm that Pgi1 indeed regulates the branch point between glycolysis and the

oxidative PPP, we expressed Pgi1 under an inducible promoter. Indeed, lower induction of Pgi1 resulted

in a similar metabolic phenotype as the yA6PR strain, as well as a relatively higher oxidative PPP flux

(Figure 4.5E and F).

139
Figure 4.6. Sorbitol-6-phosphate and ribitol-5-phosphate were inhibitors of phosphoglucose
isomerase
(A). Structures of sorbitol-6P and the enediol intermediate of the phosphoglucose isomerase reaction. (B).
2+
Purified Pgi1’s activity was assayed in the presence of 0.6 mM fructose-6-phosphate, 5 mM Mg (pH =
7.0, 30°C) and a range of 0.05-5 mM of sorbitol-6P or ribitol-5P by monitoring the appearance of glucose-
6-phosphate. See Methods for details. The y axis represents relative Pgi1 activity (mean ± range of N = 2
replicates).

Pyp1 maintains Pgi1 activity according to cell’s growth rate

Although yeast cannot utilize polyols other than sorbitol as a sole carbon source, polyols widely exist in

nature. We sought a condition where polyols could be transported and phosphorylated at a high rate in

growing cells. Trehalose is a natural dimer of glucose that can be slowly utilized by yeast by its cleavage
218
into two glucose monomers by trehalase . The cleavage and growth rate is slow, so the growth is
132
usually considered a glucose- limited culturing condition . We found that in the presence of ribitol,

ribitol-5P accumulates but does not affect the growth rate of wild type cells on trehalose (Figure 4.7A,

upper panel; and B, time zero). In contrast, the presence of ribitol resulted in a ~2.5-fold higher

accumulation of ribitol-5P and a ~20% decrease of the growth rate of pyp1 cells (Figure 4.7A, upper

panel; and B, time zero). We also performed biochemical assays of Pgi1 on ribitol-5P (Figure 4.5B).

Although ribitol-5P did not inhibit Pgi1 as strongly as sorbitol-6P, its much higher intracellular level still

resulted in a strong inhibitory effect.

140
Figure 4.7. Pyp1’s maintenance on Pgi1’s activity is more important in higher growth rate.

(A). Relative growth rate of pyp1Δ and wild type cells growing on trehalose (upper panel) and 3 hours
after switching to glucose (lower panel). Yeast cells growing on 1% trehalose in the absence of presence
of ribitol were switched to media with trehalose substituted by 2% glucose. The y axis represents relative
growth rate (mean ± range of N = 2 replicates) of cells growing in the presence of ribitol vs. cells growing
in the absence of ribitol. (B). Absolute concentration of ribitol-5-phosphate in pyp1Δ and wild type cells in
the experiment shown in (A). The x axis represents hours after the switch and the y axis represents
absolute intracellular concentration (mean ± range of N = 2 replicates). (C). Relative expression level of
219
PYP1 at different growth rates in C-, N-, P- and S-limited chemostats . The x axis represents different
growth rates and the y axis represents relative expression level of PYP1. (D). Metabolome profiling of
pyp1Δ (dark red) and wild type (blue) cells at t = 1h after glucose upshift indicates an upregulated PPP
and downregulated glycolysis. The y axis represents absolute intracellular concentration, except for the
+
panel displaying the NADPH/NADP ratio (mean ± range of N = 2 replicates).

141
Since all of the above results indicate that Pyp1 is essential for yeast in the presence of polyols in slow
219
growth conditions, we checked its expression levels at different growth rates controlled by chemostats .

Surprisingly, Pyp1’s expression is highly correlated to growth rate, regardless of the limiting nutrient used

(Figure 4.7C). To address this paradox, we investigated the mechanism of polyol phosphate inhibition on

cell growth. Pgi1 locates at the branch point of glycolysis and the oxidative PPP. The oxidative PPP

produces NADPH, which is required for biomass production. Thus, the flux of oxidative PPP is highly
29
regulated and strongly correlated with growth rate, with a relatively lower flux at faster growth rates .

This is due to the fact that in faster growing microbes, carbon is mainly converted to fermentation
220
products. For yeast growing exponentially on glucose, >90% of glucose is converted to ethanol . In

contrast, under slow growth conditions, carbons are mostly directed to the TCA cycle and biomass.

Such difference in efficiency results in a lower biomass yield in faster growing cells which require

relatively less NADPH made from the oxidative PPP. A dysregulated branch point between glycolysis and

oxidative PPP would result in imbalance of biomass production, redox and growth rate. Thus, Pyp1 could

serve to protect Pgi1 activity at this branch point, driving more flux into glycolysis at higher growth rates.

To test this hypothesis, we grew pyp1Δ cells and wild type cells on trehalose in the presence of ribitol and

then switched them to glucose. This carbon upshift boosted the cell’s growth rate (Figure 4.8). Although

the transportation of ribitol is repressed by glucose, the accumulated ribitol-5P could not be degraded

instantly in pyp1Δ cells, resulting in an eight-fold higher concentration of ribitol-5P compared to wild type

cells after one hour of switching to glucose (Figure 4.7B, time = 1h). The accumulated ribitol-5P resulted

in ~50% lower growth rate in pyp1Δ cells, which is much higher than the growth defect seen in slower

growth conditions. In comparison, the growth of wild type cells in the same conditions was not affected

(Figure 4.7A, lower panel). Metabolome profiling indicates a strong inhibition of Pgi1 upon carbon upshift

(Figure 4.7D). These results proved that Pyp1 was indeed more important at higher growth rates due to

the relatively smaller demand for oxidative PPP and a higher demand for driving glycolysis.

142
Figure 4.8. Specific growth rate of pyp1 and wild type cells growing on trehalose (upper panel)
and 3 hours after switching to glucose (lower panel).

Yeast cells growing on 1% trehalose in the absence of presence of ribitol were switched to media with
trehalose substituted by 2% glucose. The y axis represents specific growth rates. (mean ± range of N = 2
replicates).

143
4.3 Discussion

221 222
Sorbitol-6-phosphatase activity has been found in apple leaves and silk worms . However, the gene

encoding this enzyme has not been discovered. In engineered fungi, sorbitol production has been

achieved by expressing bacterial sorbitol-6-phosphate dehydrogenase. The phosphatase activity that is

223
clearly required for the subsequent step was however, unclear . Here we report a previously

unidentified enzymatic activity encoded by PYP1 that dephosphorylates D-polyol phosphates. Pyp1 is

conserved across all fungi species and some plants and bacteria, including Ricinus communis,

Dehalococcoides ethenogenes, and Bacillus anthracis (http://yeast-phylogroups.princeton.edu/). The

highly conserved nature of Pyp1 in fungi indicates its importance in cells facing a polyol-containing natural

environment. Indeed, we show that the activity of Pyp1 is essential for yeast cells to maintain

phosphoglucose isomerase activity when growing on sorbitol or in the presence of other polyols. The

expression level of Pyp1 is regulated in response to the change of growth rate due to the higher demand

for glycolysis flux vs. oxidative PPP flux in higher growth rate.

The activity of Pyp1 against different compounds indicates that Pyp1 identifies its substrate by binding the

D-glycerol-3-phosphate tail. Thus, it is very likely that Pyp1 could also dephosphorylate other polyol

phosphates, such as xylitol-5-phosphate and mannitol-1-phosphate. Indeed, while wild type yeast could

grow on mannitol, pyp1Δ cells cannot (data not shown). Interestingly, the sugar phosphate erythrose-4-

phosphate is also the substrate of Pyp1, although evidence that Pyp1 has physiological impact on this

compound is lacking.

Apart from their existence in plants, polyols can also accumulate in humans when glucose catabolism is
224
dysfunctional (hyperglycemia, diabetes). . Since polyol phosphates have not been identified as a

family of naturally existing metabolites in most organisms (http://metacyc.org), their existence and

physiological impact needs to be systematically re-evaluated. The common feature of the polyol

phosphates identified in this study is their strong inhibitory effect on the glycolytic enzyme

phosphoglucose isomerase. Compared to phosphofructokinase, which is heavily regulated via allostery

9,42-44,113
thus believed to be the key control point of glycolysis , phosphoglucose isomerase is not an

144
 217,225
allosteric enzyme, with only few competitive regulators known . However, the critical branch point it

locates gives rise to the possibility that this enzyme can be a control point of sugar catabolism. This idea

was augmented by the finding that the metabolic product of 2-deoxyglucose, a glucose analog, is 2-
226
deoxyglucose-6-phosphate, a competitive inhibitor of phosphoglucose isomerase . However, recent

studies showed that 2-deoxyglucose has physiological effects other than being a catabolic blocker,
227
rendering the need for more investigation of this potential anti-cancer drug . Our unpublished data also

showed that treatment of yeast cells with 2-deoxyglucose resulted in a different metabolic phenotype

compared to cells with low Pgi1 induction or with high polyol phosphates. Given that pyp1Δ cells and cells

with accumulated polyol phosphates share similar metabolic phenotypes as cells with low Pgi induction,

polyol phosphates represent a class of more specific phosphoglucose isomerase inhibitors, raising an

intriguing possibility to design polyol phosphate analogs as new anti-catabolism drugs.

145
4.4 Methods

Yeast strains and media

Yeast strains were derived from prototrophic S288C. Prototrophic deletions were created by homologous
190
recombination using the allele amplified by PCR from the synthetic genetic analysis (SGA) deletion set .
228
Pgi1 inducible strain was generated as described . Briefly, human hormone estradiol was used to

induce PGI1 specifically via insertion of a synthetic promoter involving chimeric transcription factor-

estrogen receptor in front of PGI1. 1 nM and 5 nM estradiol were added as low and high expression
228
levels of yeast Pgi1, resulting in a 2.5-fold difference in Pgi1 protein level .

Cells were grown in minimal media comprising 6.7g/L Difco Yeast Nitrogen Base without amino acids plus

2% (w/v) glucose or sorbitol. Trehalose minimal media were prepared by mixing 6.7g/L Difco Yeast

Nitrogen Base without amino acids and 1% (w/v) trehalose, and adjusting pH to 4.8 by adding 15g/L

succinic acid.

Yeast culture conditions and extraction

The metabolome of batch culture Saccharomyces cerevisiae was characterized as described previously
53
. Briefly, saturated overnight cultures were diluted 1:30 and grown in liquid media in a shaking flask to

A600 of ~ 0.6. A portion of the cells (3 mL) were filtered onto a 50 mm nylon membrane filter (Millipore,

Billerica, MA), which was immediately transferred into -20 °C extraction solvent (40:40:20

acetonitrile/methanol/water). For carbon upshift, 100 mL of cell culture grown on trehalose at A600 of

~0.6 were poured onto a 100 mm cellulose acetate membrane filter (Sterlitech, Kent, WA) resting on a

vacuum filter holder with a 1000 mL funnel (Kimble Chase, Vineland, NJ) and were washed with 100 mL

pre-warmed (30°C) glucose minimal medium. Immediately after the wash media went through, the filter

was taken off the holder and the cells were washed into a new flask containing 100 mL pre-warmed

(30 °C) glucose minimal medium. Samples were then taken at the indicated time points after the switch

and filtered and quenched as described above.

146
LC/MS Metabolite measurement

Cell extracts were analyzed by reversed phase ion-pairing liquid chromatography (LC) coupled by

electrospray ionization (ESI) (negative mode) to a high-resolution, high-accuracy mass spectrometer

(Exactive; Thermo Fisher Scientific, Waltham, Massachusetts) operated in full scan mode at 1 s scan time,
5
10 resolution. Peaks differing between wild-type and pyp1Δ strain were determined using the in-house
34,229
developed, open-source software MAVEN . Compounds’ identities were verified by mass and

retention time matched to authenticated standards. Isomers are reported separately only where they are

fully chromatographically resolved.

Absolute intracellular metabolite concentrations in steadily growing S. cerevisiae were determined as

36
described previously . Metabolite concentrations after perturbations were computed based on fold-

change in ion counts relative to steadily-growing cells (grown and analyzed in parallel) multiplied by the

known absolute concentration in the steadily growing cells, as determined using an isotope ratio-based
19
approach .

The number of C and N atoms in each accumulated compound was determined by the method described
208
. Yeast batch cultures were grown with uniformly labeled glucose or ammonium sulfate (Cambridge

Isotopes, Andover, MA) for > 20 generations to ensure complete labeling of the metabolome.

Synthesis of ribitol-5-phosphate and octulose-8-phosphate

230
Synthesis of ribitol-5-phosphate was performed as described previously . Synthesis of D-Glycero-D-
231
Altro-Octulose-8-phosphate was performed enzymatically as described .

Protein purification and enzymatic assays

In the initial screening of Pyp1’s phosphatase activity, the gene encoding Pyp1 was amplified by PCR

using S. cerevisiae genomic DNA. The amplified fragments were cloned into a modified pET15b vector

(Novagen, Darmstadt, Germany) and overexpressed in the E. coli BL21(DE3) Gold strain (Stratagene, La
193
Jolla, California) as previously described . The recombinant proteins were purified using metal ion
147
affinity chromatography on nickel affinity resin (Qiagen, Hilden, Germany) to high homogeneity and stored

at -80 ºC. Purified Pyp1 was screened for the presence of phosphatase activity against the general

phosphatase substrate p-nitrophenyl phosphate (pNPP) and 90 phosphorylated metabolites as described

61
previously .

For the determination of kinetic parameters (Km and kcat) of Pyp1, a yeast Open Reading Frame (ORF)

strain with an expression vector containing C-terminal His-tagged Pyp1 (Open Biosystems, Thermo

Fisher Scientific, San Jose, CA) was grown on galactose to induce Pyp1 expression. The resulting cells

were lysed using glass beads and His-tagged Pyp1 was purified using Qiagen Ni-NTA spin columns

follow the protocol provided by Qiagen. Phosphatase activity against ribitol-5-phosphate, sorbitol-6-

phosphate and octulose-8-phosphate was determined by monitoring the increase of ribitol, sorbitol or

octulose using LC-MS. The reaction mixture contained 100 mM Tris-HCl at the PH of 8.0, 10 mM MgCl2

and a range of substrate concentrations (0.01 to 5 mM). Kinetic parameters were calculated by non-linear

regression analysis of raw data to fit to the Hill equation using the GraphPad Prism Software (GraphPad

Software, San Diego, CA).

For the determination of IC50 of sorbitol-6-phosphate and ribitol-5-phosphate on phosphoglucose

isomerase, yeast Pgi1 was purchased from Sigma Aldrich (St. Louis, MO). Phosphoglucose isomerase

activity was determined by adding fructose-6-phosphate and monitoring the appearance of glucose-6-

phosphate using LC/MS. We found such LC/MS based assay is consistently more sensitive and accurate

than the colometric-based assay, which involves coupling the Pgi1 activity with glucose-6-phosphate

dehydrogenase activity and monitoring the appearance of NADH. The reaction mixture contained 100 mM

Tris-HCl at a pH of 8.0, 10 mM MgCl2, 0.6 mM fructose-6-phosphate (concentration in cells grown

exponentially on glucose), and a range of sorbitol-6-phosphate and ribitol-5-phosphate concentrations

(0.05 to 5 mM). The resulting data were again fitted to the Hill equation using the GraphPad Prism

Software.

148
Oxidative pentose phosphate pathway flux quantitation

13
Yeast cells were grown in 1,2- C2-glucose in the presence of 1 nM or 5 nM estradiol. The labeling pattern

of pentose-5-phosphates (ribose-5-phosphate, ribulose-5-phosphate and xylulose-5-phosphate) was in

13
fast equilibrium and regarded as the same pentose-5-phosphate pool. 1,2- C2-glucose labeling resulted
13
in almost 100% C2-labeled hexose-6-phosphate (glucose-6-phosphate and fructose-6-phosphate) and
13
50:50 unlabeled and C2-labeled triose-phosphate (dihydroxyacetone phosphate, glyceraldehyde-3-

phosphate and 3-phosphoglycerate). Under the assumption that transketolase reaction (fructose-6-

phosphate + glyceraldehyde-3-phosphate <-> xylulose-5-phosphate + erythrose-4-phosphate) rate is

near steady, the relative level of the oxidative PPP flux could be ascertained by the labeling pattern of

pentose-5-phosphate.

13
Specifically, C1-pentose-5-phosphate can only be derived from the oxidative PPP, where the first carbon
13
on glucose-6-phosphate is lost (see Figure 4.9A). The transketolase reaction converts two 1- C1-
13
pentose-5-phosphates to generate 1,3- C2-sedoheptulose-7-phosphate and unlabeled glyceraldehyde-3-
13
phosphate. The labeling pattern of glyceraldehyde-3-phosphate (50:50 unlabeled and 2,3- C2-labeled) is

predominantly determined by the high flux through glycolysis. Thus, the reversible reaction of
13 13 13
transketolase also gives 1,4,5- C3-pentose-5-phosphate. Therefore, both C1- and C3-pentose-5-

phosphate indicate the relative flux of oxidative PPP.

The other transketolase reaction converts fructose-6-phosphate and glyceraldehyde-3-phosphate to

13 13
generate either 1,2- C2- or 1,2,4,5- C4-pentose-5-phosphates (Figure 4.9B). This reaction also

generates unlabeled erythrose-4-phosphate, which can be further converted to sedoheptulose-7-

phosphate (by transaldolase) and then pentose-5-phosphates. As fructose-6-phosphate is mainly 1,2-

13 13
C2-labeled, the erythrose-4-phosphate-derived sedoheptulose-7-phosphate is mainly 1,2- C2-labeled.

Such forms of sedoheptulose-7-phosphate and glyceraldehyde-3-phosphate (50:50 unlabeled and 2,3-

13 13 13
C2-labeled) give 1,2- C2- or 1,2,4,5- C4-xylulose-5-phosphates and an unlabeled ribose-5-phosphate
13
(Figure 4.9C). Thus, the unlabeled fraction of pentose-5-phosphates should be subtracted from the C2
13
and C4 fractions to give the true representation of the transketolase reaction (fructose-6-phosphate +

149
glyceraldehyde-3-phosphate <-> xylulose-5-phosphate + erythrose-4-phosphate). Accordingly, relative
13 13 13 13 12
oxidative PPP flux is given by the ratio of labeling fractions: ( C1 + C3)/( C2 + C4 - C).

13
Figure 4.9. Deduction of labeling pattern of pentose-5-phosphate in cells growing on 1,2- C2-
glucose.
13 13
(A). Oxidative PPP gives 1- C1- and 1,4,5- C3-pentose-5-phosphates. (B). The erythrose-4-phosphate
13 13
forming transketolase reaction gives 1,2- C2- and 1,2,4,5- C4-pentose-5-phosphates. (C). The
transaldolase reaction followed by the transketolase reaction also gives unlabeled pentose-5-phosphate
13 13
whenever a 1,2- C2- or 1,2,4,5- C4-pentose-5-phosphate is made. Labeling patterns are represented by
blue (unlabeled) and red (labeled).

150
 References
1. Gerhart, J.C. & Pardee, A.B. The enzymology of control by feedback inhibition. J Biol Chem 237,
891-6 (1962).

2. Pardee, A.B. & Yates, R.A. Control of pyrimidine biosynthesis in Escherichia coli by a feed-back
mechanism. J Biol Chem 221, 757-70 (1956).

3. Umbarger, H.E. Evidence for a negative-feedback mechanism in the biosynthesis of isoleucine.

Science 123, 848 (1956).

4. Fell, D. Understanding the Control of Metabolism, (Portland Press, 1997).

5. Kacser, H., Burns, J.A. & Fell, D.A. The Control of Flux. Biochemical Society Transactions 23,
341-366 (1995).

6. Heinrich, R. & Rapoport, T.A. Linear Steady-State Treatment of Enzymatic Chains - General
Properties, Control and Effector Strength. European Journal of Biochemistry 42, 89-95 (1974).

7. Kacser, H. & Porteous, J.W. Control of Metabolism - What Do We Have to Measure. Trends in
Biochemical Sciences 12, 5-& (1987).

8. Crabtree, B. & Newsholme, E.A. The Derivation and Interpretation of Control Coefficients.
Biochemical Journal 247, 113-120 (1987).

9. Small, J.R. & Kacser, H. Responses of Metabolic Systems to Large Changes in Enzyme-
Activities and Effectors .1. The Linear Treatment of Unbranched Chains. European Journal of
Biochemistry 213, 613-624 (1993).

10. Kell, D.B. & Westerhoff, H.V. Metabolic Control-Theory - Its Role in Microbiology and
Biotechnology. Fems Microbiology Reviews 39, 305-320 (1986).

11. Hofmeyr, J.H.S. & Cornishbowden, A. Quantitative Assessment of Regulation in Metabolic

Systems. European Journal of Biochemistry 200, 223-236 (1991).

12. Savageau, M.A. Optimal Design of Feedback-Control by Inhibition - Dynamic Considerations.

Journal of Molecular Evolution 5, 199-222 (1975).

13. Schilling, C.H., Edwards, J.S., Letscher, D. & Palsson, B.O. Combining pathway analysis with flux
balance analysis for the comprehensive study of metabolic systems. Biotechnol Bioeng 71, 286-
306 (2000).

14. Calik, P. & Akbay, A. Mass flux balance-based model and metabolic flux analysis for collagen
synthesis in the fibrogenesis process of human liver. Med Hypotheses 55, 5-14 (2000).

15. Edwards, J.S. & Palsson, B.O. Metabolic flux balance analysis and the in silico analysis of
Escherichia coli K-12 gene deletions. BMC Bioinformatics 1, 1 (2000).

16. Ramakrishna, R., Edwards, J.S., McCulloch, A. & Palsson, B.O. Flux-balance analysis of
mitochondrial energy metabolism: consequences of systemic stoichiometric constraints. Am J
Physiol Regul Integr Comp Physiol 280, R695-704 (2001).

17. Mahadevan, R., Edwards, J.S. & Doyle, F.J., 3rd. Dynamic flux balance analysis of diauxic
growth in Escherichia coli. Biophys J 83, 1331-40 (2002).
151
18. Kauffman, K.J., Prakash, P. & Edwards, J.S. Advances in flux balance analysis. Curr Opin
Biotechnol 14, 491-6 (2003).

19. Bennett, B.D. et al. Absolute metabolite concentrations and implied enzyme active site occupancy
in Escherichia coli. Nature Chemical Biology 5, 593-599 (2009).

20. Yu, X. et al. Kinetic analysis of dynamic 13C NMR spectra: metabolic flux, regulation, and
compartmentation in hearts. Biophys J 69, 2090-102 (1995).

21. Chatham, J.C., Forder, J.R., Glickson, J.D. & Chance, E.M. Calculation of absolute metabolic flux
and the elucidation of the pathways of glutamate labeling in perfused rat heart by 13C NMR
spectroscopy and nonlinear least squares analysis. J Biol Chem 270, 7999-8008 (1995).

22. Schmidt, K. et al. 13C tracer experiments and metabolite balancing for metabolic flux analysis:
comparing two approaches. Biotechnol Bioeng 58, 254-7 (1998).

23. de Graaf, A.A. et al. Determination of full 13C isotopomer distributions for metabolic flux analysis
using heteronuclear spin echo difference NMR spectroscopy. J Biotechnol 77, 25-35 (2000).

24. Wiechert, W., Mollney, M., Petersen, S. & de Graaf, A.A. A universal framework for 13C
metabolic flux analysis. Metab Eng 3, 265-83 (2001).

25. Wiechert, W. 13C metabolic flux analysis. Metab Eng 3, 195-206 (2001).

26. Wiechert, W. An introduction to 13C metabolic flux analysis. Genet Eng (N Y) 24, 215-38 (2002).

27. Zhao, J. & Shimizu, K. Metabolic flux analysis of Escherichia coli K12 grown on 13C-labeled
acetate and glucose using GC-MS and powerful flux calculation method. J Biotechnol 101, 101-
17 (2003).

28. Blank, L.M., Kuepfer, L. & Sauer, U. Large-scale 13C-flux analysis reveals mechanistic principles
of metabolic network robustness to null mutations in yeast. Genome Biol 6, R49 (2005).

29. Frick, O. & Wittmann, C. Characterization of the metabolic shift between oxidative and
fermentative growth in Saccharomyces cerevisiae by comparative 13C flux analysis. Microb Cell
Fact 4, 30 (2005).

30. Toya, Y. et al. 13C-metabolic flux analysis for batch culture of Escherichia coli and its Pyk and
Pgi gene knockout mutants based on mass isotopomer distribution of intracellular metabolites.
Biotechnol Prog 26, 975-92.

31. Fiaux, J. et al. Metabolic-flux profiling of the yeasts Saccharomyces cerevisiae and Pichia stipitis.
Eukaryot Cell 2, 170-80 (2003).

32. Lu, W., Bennett, B.D. & Rabinowitz, J.D. Analytical strategies for LC-MS-based targeted
metabolomics. J Chromatogr B Analyt Technol Biomed Life Sci 871, 236-42 (2008).

33. Flamholz, A., Noor, E., Bar-Even, A. & Milo, R. eQuilibrator--the biochemical thermodynamics
calculator. Nucleic Acids Res 40, D770-5 (2012).

34. Lu, W.Y. et al. Metabolomic Analysis via Reversed-Phase Ion-Pairing Liquid Chromatography
Coupled to a Stand Alone Orbitrap Mass Spectrometer. Analytical Chemistry 82, 3212-3221
(2010).

35. Clasquin, M.F. et al. Riboneogenesis in yeast. Cell 145, 969-80 (2011).
152
36. Bennett, B.D., Yuan, J., Kimball, E.H. & Rabinowitz, J.D. Absolute quantitation of intracellular
metabolite concentrations by an isotope ratio-based approach. Nature Protocols 3, 1299-1311
(2008).

37. Yuan, J., Bennett, B.D. & Rabinowitz, J.D. Kinetic flux profiling for quantitation of cellular
metabolic fluxes. Nat Protoc 3, 1328-40 (2008).

38. Zhou, L., Yuan, J., Yin, J. & Wang, E. Kinetic study of prolidase activity in erythrocytes against
different substrates using capillary electrophoresis with electrochemiluminescence detection. J
Chromatogr A 1200, 255-9 (2008).

39. Yuan, J. et al. Metabolomics-driven quantitative analysis of ammonia assimilation in E. coli. Mol
Syst Biol 5, 302 (2009).

40. Morikawa, M., Izui, K., Taguchi, M. & Katsuki, H. Regulation of Escherichia-Coli
Phosphoenolpyruvate Carboxylase by Multiple Effectors Invivo .1. Estimation of the Activities in
the Cells Grown on Various Compounds. Journal of Biochemistry 87, 441-449 (1980).

41. Izui, K., Taguchi, M., Morikawa, M. & Katsuki, H. Regulation of Escherichia-Coli
Phosphoenolpyruvate Carboxylase by Multiple Effectors Invivo .2. Kinetic-Studies with a Reaction
System Containing Physiological Concentrations of Ligands. Journal of Biochemistry 90, 1321-
1331 (1981).

42. Moreno-Sanchez, R., Saavedra, E., Rodriguez-Enriquez, S. & Olin-Sandoval, V. Metabolic

control analysis: A tool for designing strategies to manipulate metabolic pathways. Journal of
Biomedicine and Biotechnology (2008).

43. Galazzo, J.L. & Bailey, J.E. Fermentation Pathway Kinetics and Metabolic Flux Control in
Suspended and Immobilized Saccharomyces-Cerevisiae. Enzyme and Microbial Technology 12,
162-172 (1990).

44. Torres, N.V., Mateo, F., Melendezhevia, E. & Kacser, H. Kinetics of Metabolic Pathways - a
System Invitro to Study the Control of Flux. Biochemical Journal 234, 169-174 (1986).

45. Christofk, H.R. et al. The M2 splice isoform of pyruvate kinase is important for cancer metabolism
and tumour growth. Nature 452, 230-U74 (2008).

46. Izui, K., Matsumura, H., Furumoto, T. & Kai, Y. Phosphoenolpyruvate carboxylase: A new era of
structural biology. Annual Review of Plant Biology 55, 69-84 (2004).

47. Ptacek, J. et al. Global analysis of protein phosphorylation in yeast. Nature 438, 679-684 (2005).

48. Portela, P., Howell, S., Moreno, S. & Rossi, S. In vivo and in vitro phosphorylation of two Isoforms
of yeast pyruvate kinase by protein kinase A. Journal of Biological Chemistry 277, 30477-30487
(2002).

49. Portela, P., Moreno, S. & Rossi, S. Characterization of yeast pyruvate kinase 1 as a protein
kinase A substrate, and specificity of the phosphorylation site sequence in the whole protein.
Biochemical Journal 396, 117-126 (2006).

50. Galello, F., Portela, P., Moreno, S. & Rossi, S. Characterization of Substrates That Have a
Differential Effect on Saccharomyces cerevisiae Protein Kinase A Holoenzyme Activation. Journal
of Biological Chemistry 285, 29770-29779 (2010).

153
51. Bodenmiller, B. et al. Phosphoproteomic Analysis Reveals Interconnected System-Wide
Responses to Perturbations of Kinases and Phosphatases in Yeast. Science Signaling 3(2010).

52. Xu, Y.F., Amador-Noguez, D., Reaves, M.L., Feng, X.J. & Rabinowitz, J.D. Ultrasensitive
regulation of anapleurosis via allosteric activation of PEP carboxylase. Nat Chem Biol 8, 562-8
(2012).

53. Xu, Y.F. et al. Regulation of yeast pyruvate kinase by ultrasensitive allostery independent of
phosphorylation. Mol Cell 48, 52-62 (2012).

54. Dombrauckas, J.D., Santarsiero, B.D. & Mesecar, A.D. Structural basis for tumor pyruvate kinase
M2 allosteric regulation and catalysis. Biochemistry 44, 9417-29 (2005).

55. Blair, J.B. & Walker, R.G. Rat liver pyruvate kinase: influence of ligands on activity and fructose
1,6-bisphosphate binding. Arch Biochem Biophys 232, 202-13 (1984).

56. Jurica, M.S. et al. The allosteric regulation of pyruvate kinase by fructose-1,6-bisphosphate.
Structure 6, 195-210 (1998).

57. Sanwal, B.D. & Maeba, P. Regulation of Activity of Phosphoenolypyruvate Carboxylase by

Fructose Diphosphate. Biochemical and Biophysical Research Communications 22, 194-& (1966).

58. Saghatelian, A. & Cravatt, B.F. Global strategies to integrate the proteome and metabolome. Curr
Opin Chem Biol 9, 62-8 (2005).

59. Innes, D., Beacham, I.R. & Burns, D.M. The role of the intracellular inhibitor of periplasmic UDP-
sugar hydrolase (5'-nucleotidase) in Escherichia coli: cytoplasmic localisation of 5'-nucleotidase is
conditionally lethal. J Basic Microbiol 41, 329-37 (2001).

60. Proudfoot, M. et al. General enzymatic screens identify three new nucleotidases in Escherichia
coli. Biochemical characterization of SurE, YfbR, and YjjG. J Biol Chem 279, 54687-94 (2004).

61. Kuznetsova, E. et al. Genome-wide analysis of substrate specificities of the Escherichia coli
haloacid dehalogenase-like phosphatase family. J Biol Chem 281, 36149-61 (2006).

62. Jensen, K.F. & Nygaard, P. Purine nucleoside phosphorylase from Escherichia coli and
Salmonella typhimurium. Purification and some properties. Eur J Biochem 51, 253-65 (1975).

63. Petersen, C. & Moller, L.B. The RihA, RihB, and RihC ribonucleoside hydrolases of Escherichia
coli. Substrate specificity, gene expression, and regulation. J Biol Chem 276, 884-94 (2001).

64. Dandanell, G., Szczepanowski, R.H., Kierdaszuk, B., Shugar, D. & Bochtler, M. Escherichia coli
purine nucleoside phosphorylase II, the product of the xapA gene. J Mol Biol 348, 113-25 (2005).

65. Tozzi, M.G., Camici, M., Mascia, L., Sgarrella, F. & Ipata, P.L. Pentose phosphates in nucleoside
interconversion and catabolism. FEBS J 273, 1089-101 (2006).

66. Fraenkel, D.G. Genetics and Intermediary Metabolism. Annual Review of Genetics 26, 159-177
(1992).

67. Raamsdonk, L.M. et al. A functional genomics strategy that uses metabolome data to reveal the
phenotype of silent mutations. Nature Biotechnology 19, 45-50 (2001).

68. Fiehn, O. Metabolomics--the link between genotypes and phenotypes. Plant Mol Biol 48, 155-71
(2002).
154
69. Forster, J., Gombert, A.K. & Nielsen, J. A functional genomics approach using metabolomics and
in silico pathway analysis. Biotechnol Bioeng 79, 703-12 (2002).

70. Allen, J. et al. High-throughput classification of yeast mutants for functional genomics using
metabolic footprinting. Nature Biotechnology 21, 692-6 (2003).

71. Saghatelian, A. & Cravatt, B.F. Assignment of protein function in the postgenomic era. Nature
Chemical Biology 1, 130-142 (2005).

72. Kotte, O., Zaugg, J.B. & Heinemann, M. Bacterial adaptation through distributed sensing of
metabolic fluxes. Molecular Systems Biology 6, 355 (2010).

73. Sauer, U. & Eikmanns, B.J. The PEP-pyruvate-oxaloacetate node as the switch point for carbon
flux distribution in bacteria. Fems Microbiology Reviews 29, 765-794 (2005).

74. Palsson, B.O. et al. A genome-scale metabolic reconstruction for Escherichia coli K-12 MG1655
that accounts for 1260 ORFs and thermodynamic information. Molecular Systems Biology 3, 121
(2007).

75. Siddiquee, K.A., Arauzo-Bravo, M.J. & Shimizu, K. Effect of a pyruvate kinase (pykF-gene)
knockout mutation on the control of gene expression and metabolic fluxes in Escherichia coli.
Fems Microbiology Letters 235, 25-33 (2004).

76. Ishii, N. et al. Multiple high-throughput analyses monitor the response of E-coli to perturbations.
Science 316, 593-597 (2007).

77. Fiehn, O. Metabolomics - the link between genotypes and phenotypes. Plant Molecular Biology
48, 155-171 (2002).

78. Brauer, M.J. et al. Conservation of the metabolomic response to starvation across two divergent
microbes. Proceedings of the National Academy of Sciences of the United States of America 103,
19302-19307 (2006).

79. Rabinowitz, J.D. & Kimball, E. Acidic acetonitrile for cellular metabolome extraction from
Escherichia coli. Analytical Chemistry 79, 6167-6173 (2007).

80. Fischer, E. & Sauer, U. A novel metabolic cycle catalyzes glucose oxidation and anaplerosis in
hungry Escherichia coli. Journal of Biological Chemistry 278, 46446-46451 (2003).

81. Salleh, H.M., Patel, M.A. & Woodard, R.W. Essential cysteines in 3-deoxy-D-manno-octulosonic
acid 8-phosphate synthase from Escherichia coli: Analysis by chemical modification and site-
directed mutagenesis. Biochemistry 35, 8942-8947 (1996).

82. Kim, D.H. et al. Characterization of a Cys115 to Asp substitution in the Escherichia coli cell wall
biosynthetic enzyme UDP-GlcNAc enolpyruvyl transferase (MurA) that confers resistance to
inactivation by the antibiotic fosfomycin. Biochemistry 35, 4923-4928 (1996).

83. Siddiquee, K.A., Arauzo-Bravo, M.J. & Shimizu, K. Metabolic flux analysis of pykF gene knockout
Escherichia coli based on C-13-labeling experiments together with measurements of enzyme
activities and intracellular metabolite concentrations. Applied Microbiology and Biotechnology 63,
407-417 (2004).

84. van Rijsewijk, B.R.B.H., Nanchen, A., Nallet, S., Kleijn, R.J. & Sauer, U. Large-scale (13)C-flux
analysis reveals distinct transcriptional control of respiratory and fermentative metabolism in
Escherichia coli. Molecular Systems Biology 7, 477 (2011).
155
85. Peng, L. & Shimizu, K. Global metabolic regulation analysis for Escherichia coli K12 based on
protein expression by 2-dimensional electrophoresis and enzyme activity measurement. Applied
Microbiology and Biotechnology 61, 163-178 (2003).

86. Herrmann, K.M. The Shikimate Pathway as an Entry to Aromatic Secondary Metabolism. Plant
Physiology 107, 7-12 (1995).

87. Yuan, J. & Rabinowitz, J.D. Differentiating metabolites formed from de novo synthesis versus
macromolecule decomposition. Journal of the American Chemical Society 129, 9294-+ (2007).

88. Kodaki, T., Fujita, N., Kameshita, I., Izui, K. & Katsuki, H. Phosphoenolpyruvate Carboxylase of
Escherichia-Coli - Specificity of Some Compounds as Activators at the Site for Fructose-1,6-
Bisphosphate, One of the Allosteric Effectors. Journal of Biochemistry 95, 637-642 (1984).

89. Garnak, M. & Reeves, H.C. Phosphorylation of Isocitrate Dehydrogenase of Escherichia-Coli.

Science 203, 1111-1112 (1979).

90. Wohl, R.C. & Markus, G. Phosphoenolpyruvate Carboxylase of Escherichia-Coli - Purification and
Some Properties. Journal of Biological Chemistry 247, 5785-& (1972).

91. Silverst.R & Willis, M.S. Concerted Regulation in-Vitro of Phosphoenolpyruvate Carboxylase from
Escherichia-Coli. Journal of Biological Chemistry 248, 8402-8407 (1973).

92. Takahashi-Terada, A. et al. Maize phosphoenolpyruvate carboxylase - Mutations at the putative

binding site for glucose 6-phosphate caused desensitization and abolished responsiveness to
regulatory phosphorylation. Journal of Biological Chemistry 280, 11798-11806 (2005).

93. Oh, M.K., Rohlin, L., Kao, K.C. & Liao, J.C. Global expression profiling of acetate-grown
Escherichia coli. Journal of Biological Chemistry 277, 13175-13183 (2002).

94. Zhao, J., Baba, T., Mori, H. & Shimizu, K. Effect of zwf gene knockout on the metabolism of
Escherichia coli grown on glucose or acetate. Metabolic Engineering 6, 164-174 (2004).

95. Schaub, J. & Reuss, M. In Vivo Dynamics of Glycolysis in Escherichia coli Shows Need for
Growth-Rate Dependent Metabolome Analysis. Biotechnology Progress 24, 1402-1407 (2008).

96. Theobald, U., Mailinger, W., Baltes, M., Rizzi, M. & Reuss, M. In vivo analysis of metabolic
dynamics in Saccharomyces cerevisiae .1. Experimental observations. Biotechnology and
Bioengineering 55, 305-316 (1997).

97. Zhu, T., Bailey, M.F., Angley, L.M., Cooper, T.F. & Dobson, R.C.J. The quaternary structure of
pyruvate kinase type 1 from Escherichia coli at low nanomolar concentrations. Biochimie 92, 116-
120 (2010).

98. Ogawa, T., Mori, H., Tomita, M. & Yoshino, M. Inhibitory effect of phosphoenolpyruvate on
glycolytic enzymes in Escherichia coli. Research in Microbiology 158, 159-163 (2007).

99. Kern, D. & Zuiderweg, E.R. The role of dynamics in allosteric regulation. Curr Opin Struct Biol 13,
748-57 (2003).

100. Walsh, K. & Koshland, D.E. Branch Point Control by the Phosphorylation State of Isocitrate
Dehydrogenase - a Quantitative Examination of Fluxes during a Regulatory Transition. Journal of
Biological Chemistry 260, 8430-8437 (1985).

156
101. Stadtman, E.R. The story of glutamine synthetase regulation. Journal of Biological Chemistry 276,
44357-44364 (2001).

102. Kai, Y., Matsumura, H. & Izui, K. Phosphoenolpyruvate carboxylase: three-dimensional structure
and molecular mechanisms. Archives of Biochemistry and Biophysics 414, 170-179 (2003).

103. Gonzalez, C.F. et al. Molecular Basis of Formaldehyde Detoxification CHARACTERIZATION OF

TWO S-FORMYLGLUTATHIONE HYDROLASES FROM ESCHERICHIA COLI, FrmB AND YeiG.
Journal of Biological Chemistry 281, 14514-14522 (2006).

104. Brown, G. et al. Structural and Biochemical Characterization of the Type II Fructose-1,6-
bisphosphatase GlpX from Escherichia coli. Journal of Biological Chemistry 284, 3784-3792
(2009).

105. Lee, M., Chan, C.W., Guss, J.M., Christopherson, R.I. & Maher, M.J. Dihydroorotase from
Escherichia coli: Loop movement and cooperativity between subunits. Journal of Molecular
Biology 348, 523-533 (2005).

106. Song, W.J. & Jackowski, S. Kinetics and Regulation of Pantothenate Kinase from Escherichia-
Coli. Journal of Biological Chemistry 269, 27051-27058 (1994).

107. Bhasin, M., Billinsky, J.L. & Palmer, D.R.J. Steady-state kinetics and molecular evolution of
Escherichia coli MenD [(1R,6R)-2-succinyl-6-hydroxy-2,4-cyclohexadiene-1-carboxylate
synthase], an anomalous thiamin diphosphate-dependent decarboxylase-carboligase.
Biochemistry 42, 13496-13504 (2003).

108. Broglie, K.E. & Takahashi, M. Fluorescence studies of threonine-promoted conformational

transitions in aspartokinase I using the substrate analogue 2'(3')-O-(2,4,6-
trinitrophenyl)adenosine 5'-triphosphate. J Biol Chem 258, 12940-6 (1983).

109. Eisenstein, E., Yu, H.D. & Schwarz, F.P. Cooperative binding of the feedback modifiers
isoleucine and valine to biosynthetic threonine deaminase from Escherichia coli. J Biol Chem 269,
29423-9 (1994).

110. Gutnick, D., Calvo, J.M., Klopotow.T & Ames, B.N. Compounds Which Serve as Sole Source of
Carbon or Nitrogen for Salmonella Typhimurium Lt-2. Journal of Bacteriology 100, 215-& (1969).

111. Shevchenko, A., Tomas, H., Havlis, J., Olsen, J.V. & Mann, M. In-gel digestion for mass
spectrometric characterization of proteins and proteomes. Nat Protoc 1, 2856-60 (2006).

112. Kitagawa, M. et al. Complete set of ORF clones of Escherichia coli ASKA library (A complete Set
of E. coli K-12 ORF archive): Unique resources for biological research. DNA Research 12, 291-
299 (2005).

113. Kretschmer, M. & Fraenkel, D.G. Yeast 6-Phosphofructo-2-Kinase - Sequence and Mutant.
Biochemistry 30, 10663-10672 (1991).

114. Boles, E., Gohlmann, H.W.H. & Zimmerman, F.K. Cloning of a second gene encoding 6-
phosphofructo-2-kinase in yeast, and characterization of mutant strains without fructose-2,6-
bisphosphate. Molecular Microbiology 20, 65-76 (1996).

115. Boles, E. et al. Characterization of a glucose-repressed pyruvate kinase (Pyk2p) in

Saccharomyces cerevisiae that is catalytically insensitive to fructose-1,6-bisphosphate. Journal of
Bacteriology 179, 2987-2993 (1997).

157
116. El-Maghrabi, M.R., Claus, T.H., McGrane, M.M. & Pilkis, S.J. Influence of phosphorylation on the
interaction of effectors with rat liver pyruvate kinase. J Biol Chem 257, 233-40 (1982).

117. Wilson, A.J. & Bhattacharjee, J.K. Regulation of Phosphoenolpyruvate Carboxykinase and
Pyruvate-Kinase in Saccharomyces-Cerevisiae Grown in the Presence of Glycolytic and
Gluconeogenic Carbon-Sources and the Role of Mitochondrial-Function on Gluconeogenesis.
Canadian Journal of Microbiology 32, 969-972 (1986).

118. Cortassa, S., Aon, J.C. & Aon, M.A. Fluxes of Carbon, Phosphorylation, and Redox Intermediates
during Growth of Saccharomyces-Cerevisiae on Different Carbon-Sources. Biotechnology and
Bioengineering 47, 193-208 (1995).

119. Zaman, S., Lippman, S.I., Zhao, X. & Broach, J.R. How Saccharomyces Responds to Nutrients.
Annual Review of Genetics 42, 27-81 (2008).

120. Zaman, S., Lippman, S.I., Schneper, L., Slonim, N. & Broach, J.R. Glucose regulates
transcription in yeast through a network of signaling pathways. Mol Syst Biol 5, 245 (2009).

121. Francois, J., Neves, M.J. & Hers, H.G. The Control of Trehalose Biosynthesis in Saccharomyces-
Cerevisiae - Evidence for a Catabolite Inactivation and Repression of Trehalose-6-Phosphate
Synthase and Trehalose-6-Phosphate Phosphatase. Yeast 7, 575-587 (1991).

122. De Mesquita, J.F., Panek, A.D. & de Araujo, P.S. In silico and in vivo analysis reveal a novel
gene in Saccharomyces cerevisiae trehalose metabolism. BMC Genomics 4, 45 (2003).

123. Arguelles, J.C., Carrillo, D., Vicentesoler, J., Garciacarmona, F. & Gacto, M. Lack of Correlation
between Trehalase Activation and Trehalose-6 Phosphate Synthase Deactivation in Camp-
Altered Mutants of Saccharomyces-Cerevisiae. Current Genetics 23, 382-387 (1993).

124. Rayner, T.F., Gray, J.V. & Thorner, J.W. Direct and novel regulation of cAMP-dependent protein
kinase by Mck1p, a yeast glycogen synthase kinase-3. Journal of Biological Chemistry 277,
16814-16822 (2002).

125. Toda, T. et al. Cloning and Characterization of Bcy1, a Locus Encoding a Regulatory Subunit of
the Cyclic Amp-Dependent Protein-Kinase in Saccharomyces-Cerevisiae. Molecular and Cellular
Biology 7, 1371-1377 (1987).

126. Mesecar, A.D. & Nowak, T. Metal-ion-mediated allosteric triggering of yeast pyruvate kinase .1. A
multidimensional kinetic linked function analysis. Biochemistry 36, 6792-6802 (1997).

127. Mesecar, A.D. & Nowak, T. Metal-ion-mediated allosteric triggering of yeast pyruvate kinase .2. A
multidimensional thermodynamic linked function analysis. Biochemistry 36, 6803-6813 (1997).

128. Nowak, T. & Bollenbach, T.J. Thermodynamic linked-function analysis of Mg2+-activated yeast
pyruvate kinase. Biochemistry 40, 13088-13096 (2001).

129. Susan-Resiga, D. & Nowak, T. Proton donor in yeast pyruvate kinase: Chemical and kinetic
properties of the active site Thr 298 to cys mutant. Biochemistry 43, 15230-15245 (2004).

130. Fenton, A.W. & Blair, J.B. Kinetic and allosteric consequences of mutations in the subunit and
domain interfaces and the allosteric site of yeast pyruvate kinase. Archives of Biochemistry and
Biophysics 397, 28-39 (2002).

131. Friesen, R.H.E., Castellani, R.J., Lee, J.C. & Braun, W. Allostery in rabbit pyruvate kinase:
Development of a strategy to elucidate the mechanism. Biochemistry 37, 15266-15276 (1998).
158
132. Walther, T. et al. Control of ATP homeostasis during the respiro-fermentative transition in yeast.
Mol Syst Biol 6, 344 (2010).

133. Dospassos, J.B. et al. Glucose-Induced Activation of Plasma-Membrane H+-Atpase in Mutants of

the Yeast Saccharomyces-Cerevisiae Affected in Camp Metabolism, Camp-Dependent Protein-
Phosphorylation and the Initiation of Glycolysis. Biochimica Et Biophysica Acta 1136, 57-67
(1992).

134. Zhao, S.M. et al. Regulation of Cellular Metabolism by Protein Lysine Acetylation. Science 327,
1000-1004 (2010).

135. Wang, Q.J. et al. Acetylation of Metabolic Enzymes Coordinates Carbon Source Utilization and
Metabolic Flux. Science 327, 1004-1007 (2010).

136. Rikova, K. et al. Global survey of phosphotyrosine signaling identifies oncogenic kinases in lung
cancer. Cell 131, 1190-1203 (2007).

137. Huttlin, E.L. et al. A Tissue-Specific Atlas of Mouse Protein Phosphorylation and Expression. Cell
143, 1174-1189 (2010).

138. Crutchfield, C.A., Lu, W.Y., Melamud, E. & Rabinowitz, J.D. Mass Spectrometry-Based
Metabolomics of Yeast. Methods in Enzymology, Vol 470: Guide to Yeast Genetics: 470, 393-426
(2010).

139. Klosinska, M.M., Crutchfield, C.A., Bradley, P.H., Rabinowitz, J.D. & Broach, J.R. Yeast cells can
access distinct quiescent states. Genes & Development 25, 336-349 (2011).

140. Rappsilber, J., Ishihama, Y. & Mann, M. Stop and go extraction tips for matrix-assisted laser
desorption/ionization, nanoelectrospray, and LC/MS sample pretreatment in proteomics.
Analytical Chemistry 75, 663-670 (2003).

141. Wu, J., Shakey, Q., Liu, W., Schuller, A. & Follettie, M.T. Global profiling of phosphopeptides by
titania affinity enrichment. Journal of Proteome Research 6, 4684-4689 (2007).

142. Sano, A. & Nakamura, H. Titania as a chemo-affinity support for the column-switching HPLC
analysis of phosphopeptides: Application to the characterization of phosphorylation sites in
proteins by combination with protease digestion and electrospray ionization mass Spectrometry.
Analytical Sciences 20, 861-864 (2004).

143. Khan, Z., Bloom, J.S., Garcia, B.A., Singh, M. & Kruglyak, L. Protein quantification across
hundreds of experimental conditions. Proceedings of the National Academy of Sciences of the
United States of America 106, 15544-15548 (2009).

144. Khan, Z. et al. Accurate proteome-wide protein quantification from high-resolution 15N mass
spectra. Genome Biol 12, R122 (2011).

145. Cox, J. & Mann, M. MaxQuant enables high peptide identification rates, individualized p.p.b.-
range mass accuracies and proteome-wide protein quantification. Nature Biotechnology 26,
1367-72 (2008).

146. Fonvielle, M., Weber, P., Dabkowska, K. & Therisod, M. New highly selective inhibitors of class II
fructose-1,6-bisphosphate aldolases. Bioorganic & Medicinal Chemistry Letters 14, 2923-2926
(2004).

159
147. White, M.F. & Fothergillgilmore, L.A. Development of a Mutagenesis, Expression and Purification
System for Yeast Phosphoglycerate Mutase - Investigation of the Role of Active-Site His181.
European Journal of Biochemistry 207, 709-714 (1992).

148. Poyner, R.R., Laughlin, L.T., Sowa, G.A. & Reed, G.H. Toward identification of acid/base
catalysts in the active site of enolase: Comparison of the properties of K345A, E168Q, and E2114
variants. Biochemistry 35, 1692-1699 (1996).

149. Link, H., Kochanowski, K. & Sauer, U. Systematic identification of allosteric protein-metabolite
interactions that control enzyme activity in vivo. Nature Biotechnology 31, 5 (2013).

150. Ratnakumar, S. & Young, E.T. Snf1 dependence of peroxisomal gene expression is mediated by
Adr1. J Biol Chem 285, 10703-14 (2010).

151. Komeili, A., Wedaman, K.P., O'Shea, E.K. & Powers, T. Mechanism of metabolic control. Target
of rapamycin signaling links nitrogen quality to the activity of the Rtg1 and Rtg3 transcription
factors. J Cell Biol 151, 863-78 (2000).

152. Wilson, W.A. & Roach, P.J. Nutrient-regulated protein kinases in budding yeast. Cell 111, 155-8
(2002).

153. Ertugay, N. & Hamamci, H. Continuous cultivation of bakers' yeast: Change in cell composition at
different dilution rates and effect of heat stress on trehalose level. Folia Microbiologica 42, 463-
467 (1997).

154. von der Haar, T. A quantitative estimation of the global translational activity in logarithmically
growing yeast cells. Bmc Systems Biology 2(2008).

155. Lee, Y.S., Mulugu, S., York, J.D. & O'Shea, E.K. Regulation of a cyclin-CDK-CDK inhibitor
complex by inositol pyrophosphates. Science 316, 109-12 (2007).

156. Oliveira, A.P. & Sauer, U. The importance of post-translational modifications in regulating
Saccharomyces cerevisiae metabolism. FEMS Yeast Res 12, 104-17 (2012).

157. Martin, D.E., Soulard, A. & Hall, M.N. TOR regulates ribosomal protein gene expression via PKA
and the Forkhead transcription factor FHL1. Cell 119, 969-79 (2004).

158. Yorimitsu, T., Zaman, S., Broach, J.R. & Klionsky, D.J. Protein kinase A and Sch9 cooperatively
regulate induction of autophagy in Saccharomyces cerevisiae. Mol Biol Cell 18, 4180-9 (2007).

159. Stephan, J.S., Yeh, Y.Y., Ramachandran, V., Deminoff, S.J. & Herman, P.K. The Tor and PKA
signaling pathways independently target the Atg1/Atg13 protein kinase complex to control
autophagy. Proc Natl Acad Sci U S A 106, 17049-54 (2009).

160. Ben-Shem, A. et al. The Structure of the Eukaryotic Ribosome at 3.0 angstrom Resolution.
Science 334, 1524-1529 (2011).

161. Lecoq, K., Belloc, I., Desgranges, C., Konrad, M. & Daignan-Fornier, B. YLR209c encodes
Saccharomyces cerevisiae purine nucleoside phosphorylase. J Bacteriol 183, 4910-3 (2001).

162. Kurtz, J.E., Exinger, F., Erbs, P. & Jund, R. The URH1 uridine ribohydrolase of Saccharomyces
cerevisiae. Curr Genet 41, 132-41 (2002).

160
163. Mitterbauer, R., Karl, T. & Adam, G. Saccharomyces cerevisiae URH1 (encoding uridine-cytidine
N-ribohydrolase): functional complementation by a nucleoside hydrolase from a protozoan
parasite and by a mammalian uridine phosphorylase. Appl Environ Microbiol 68, 1336-43 (2002).

164. Nakanishi, T. & Sekimizu, K. SDT1/SSM1, a multicopy suppressor of S-II null mutant, encodes a
novel pyrimidine 5'-nucleotidase. J Biol Chem 277, 22103-6 (2002).

165. Canduri, F. et al. Structures of human purine nucleoside phosphorylase complexed with inosine
and ddI. Biochem Biophys Res Commun 313, 907-14 (2004).

166. Hunsucker, S.A., Mitchell, B.S. & Spychala, J. The 5 '-nucleotidases as regulators of nucleotide
and drug metabolism. Pharmacology & Therapeutics 107, 1-30 (2005).

167. Grant, C.M. Role of the glutathione/glutaredoxin and thioredoxin systems in yeast growth and
response to stress conditions. Mol Microbiol 39, 533-41 (2001).

168. Ralser, M. et al. Dynamic rerouting of the carbohydrate flux is key to counteracting oxidative
stress. J Biol 6, 10 (2007).

169. Nakahigashi, K. et al. Systematic phenome analysis of Escherichia coli multiple-knockout mutants
reveals hidden reactions in central carbon metabolism. Mol Syst Biol 5, 306 (2009).

170. Reddy, V.S., Singh, A.K. & Rajasekharan, R. The Saccharomyces cerevisiae PHM8 gene
encodes a soluble magnesium-dependent lysophosphatidic acid phosphatase. J Biol Chem 283,
8846-54 (2008).

171. Herrgard, M.J. et al. A consensus yeast metabolic network reconstruction obtained from a
community approach to systems biology. Nature Biotechnology 26, 1155-60 (2008).

172. Broach, J.R. Nutritional control of growth and development in yeast. Genetics 192, 73-105 (2012).

173. Noda, T. & Ohsumi, Y. Tor, a phosphatidylinositol kinase homologue, controls autophagy in yeast.
J Biol Chem 273, 3963-6 (1998).

174. Gasch, A.P. et al. Genomic expression programs in the response of yeast cells to environmental
changes. Mol Biol Cell 11, 4241-4257 (2000).

175. Bradley, P.H., Brauer, M.J., Rabinowitz, J.D. & Troyanskaya, O.G. Coordinated Concentration
Changes of Transcripts and Metabolites in Saccharomyces cerevisiae. PLoS Comput Biol
5(2009).

176. Walther, T. et al. The PGM3 gene encodes the major phosphoribomutase in the yeast
Saccharomyces cerevisiae. FEBS Lett 586, 4114-8 (2012).

177. Molin, M., Norbeck, J. & Blomberg, A. Dihydroxyacetone kinases in Saccharomyces cerevisiae
are involved in detoxification of dihydroxyacetone. J Biol Chem 278, 1415-23 (2003).

178. Crespo, J.L., Powers, T., Fowler, B. & Hall, M.N. The TOR-controlled transcription activators
GLN3, RTG1, and RTG3 are regulated in response to intracellular levels of glutamine. Proc Natl
Acad Sci U S A 99, 6784-9 (2002).

179. Boer, V.M., Crutchfield, C.A., Bradley, P.H., Botstein, D. & Rabinowitz, J.D. Growth-limiting
intracellular metabolites in yeast growing under diverse nutrient limitations. Mol Biol Cell 21, 198-
211 (2010).

161
180. Shenton, D. & Grant, C.M. Protein S-thiolation targets glycolysis and protein synthesis in
response to oxidative stress in the yeast Saccharomyces cerevisiae. Biochem J 374, 513-9
(2003).

181. Baker, C.J. & Orlandi, E.W. Active oxygen in plant pathogenesis. Annu Rev Phytopathol 33, 299-
321 (1995).

182. Mager, W.H. & Planta, R.J. Coordinate expression of ribosomal protein genes in yeast as a
function of cellular growth rate. Molecular and Cellular Biochemistry 104, 181-7 (1991).

183. Bremer, H. & Dennis, P.P. Modulation of chemical composition and other parameters of the cell
by growth rate. in Escherichia coli and Salmonella (ed. Neidhardt, F.C.) 1553-1569 (ASM Press,
Washington, D.C., 1996).

184. Scott, M., Gunderson, C.W., Mateescu, E.M., Zhang, Z. & Hwa, T. Interdependence of cell growth
and gene expression: origins and consequences. Science 330, 1099-102 (2010).

185. Kuma, A. et al. The role of autophagy during the early neonatal starvation period. Nature 432,
1032-1036 (2004).

186. Efeyan, A. et al. Regulation of mTORC1 by the Rag GTPases is necessary for neonatal
autophagy and survival. Nature (2012).

187. Guo, J.Y. et al. Activated Ras requires autophagy to maintain oxidative metabolism and
tumorigenesis. Genes Dev 25, 460-70 (2011).

188. Mathew, R. & White, E. Autophagy, stress, and cancer metabolism: what doesn't kill you makes
you stronger. Cold Spring Harb Symp Quant Biol 76, 389-96 (2011).

189. Yang, S. et al. Pancreatic cancers require autophagy for tumor growth. Genes Dev 25, 717-29
(2011).

190. Tong, A.H. et al. Systematic genetic analysis with ordered arrays of yeast deletion mutants.
Science 294, 2364-8 (2001).

191. Spellman, P.T. et al. Comprehensive identification of cell cycle-regulated genes of the yeast
Saccharomyces cerevisiae by microarray hybridization. Mol Biol Cell 9, 3273-97 (1998).

192. Berg, J.M., Tymoczko, J.L. & Stryer, L. Metabolism: Basic Concepts and Design. in Biochemistry
(W.H. Freeman and Co., New York, 2007).

193. Kuznetsova, E. et al. Structure and activity of the metal-independent fructose-1,6-bisphosphatase

YK23 from Saccharomyces cerevisiae. J Biol Chem 285, 21049-59 (2010).

194. Teo, G. et al. Silencing leaf sorbitol synthesis alters long-distance partitioning and apple fruit
quality. Proc Natl Acad Sci U S A 103, 18842-7 (2006).

195. Shindou, T. et al. Determination of Erythritol in Fermented Foods by High-Performance Liquid-

Chromatography. Journal of the Food Hygienic Society of Japan 29, 419-422 (1988).

196. Pahlman, A.K., Granath, K., Ansell, R., Hohmann, S. & Adler, L. The yeast glycerol 3-
phosphatases Gpp1p and Gpp2p are required for glycerol biosynthesis and differentially involved
in the cellular responses to osmotic, anaerobic, and oxidative stress. J Biol Chem 276, 3555-63
(2001).

162
197. Nevoigt, E. & Stahl, U. Osmoregulation and glycerol metabolism in the yeast Saccharomyces
cerevisiae. FEMS Microbiol Rev 21, 231-41 (1997).

198. Bradshaw, D.J. & Marsh, P.D. Effect of Sugar Alcohols on the Composition and Metabolism of a
Mixed Culture of Oral Bacteria Grown in a Chemostat. Caries Research 28, 251-256 (1994).

199. Chang, Q., Griest, T.A., Harter, T.M. & Petrash, J.M. Functional studies of aldo-keto reductases
in Saccharomyces cerevisiae. Biochimica Et Biophysica Acta-Molecular Cell Research 1773,
321-329 (2007).

200. Moon, H.J., Jeya, M., Kim, I.W. & Lee, J.K. Biotechnological production of erythritol and its
applications. Applied Microbiology and Biotechnology 86, 1017-1025 (2010).

201. Povelainen, M. & Miasnikov, A.N. Production of D-arabitol by a metabolic engineered strain of
Bacillus subtilis. Biotechnol J 1, 214-9 (2006).

202. Povelainen, M. & Miasnikov, A.N. Production of xylitol by metabolically engineered strains of
Bacillus subtilis. J Biotechnol 128, 24-31 (2007).

203. Toivari, M.H., Maaheimo, H., Penttila, M. & Ruohonen, L. Enhancing the flux of D-glucose to the
pentose phosphate pathway in Saccharomyces cerevisiae for the production of D-ribose and
ribitol. Appl Microbiol Biotechnol 85, 731-9 (2010).

204. Albertyn, J., Hohmann, S., Thevelein, J.M. & Prior, B.A. GPD1, which encodes glycerol-3-
phosphate dehydrogenase, is essential for growth under osmotic stress in Saccharomyces
cerevisiae, and its expression is regulated by the high-osmolarity glycerol response pathway. Mol
Cell Biol 14, 4135-44 (1994).

205. Norbeck, J., Pahlman, A.K., Akhtar, N., Blomberg, A. & Adler, L. Purification and characterization
of two isoenzymes of DL-glycerol-3-phosphatase from Saccharomyces cerevisiae - Identification
of the corresponding GPP1 and GPP2 genes and evidence for osmotic regulation of Gpp2p
expression by the osmosensing mitogen-activated protein kinase signal transduction pathway.
Journal of Biological Chemistry 271, 13875-13881 (1996).

206. Vander Heiden, M.G. et al. Metabolic pathway alterations that support cell proliferation. Cold
Spring Harb Symp Quant Biol 76, 325-34 (2011).

207. Read, J. et al. The Crystal Structure of Human Phosphoglucose Isomerase at 1.6 angstrom
resolution: Implications for catalytic mechanism, cytokine activity and haemolytic anaemia.
Journal of Molecular Biology 309, 447-463 (2001).

208. Hegeman, A.D. et al. Stable isotope assisted assignment of elemental compositions for
metabolomics. Anal Chem 79, 6912-21 (2007).

209. Kuchel, P.W. et al. C-13 and P-31 Nmr-Studies of the Pentose-Phosphate Pathway in Human
Erythrocytes. Biomedica Biochimica Acta 49, S105-S110 (1990).

210. Jennings, D.H. Polyol Metabolism in Fungi. Advances in Microbial Physiology 25, 149-193 (1984).

211. Pereira, M.P. & Brown, E.D. Bifunctional catalysis by CDP-ribitol synthase: Convergent
recruitment of reductase and cytidylyltransferase activities in Haemophilus influenzae and
Staphylococcus aureus. Biochemistry 43, 11802-11812 (2004).

163
212. Novotny, M.J., Reizer, J., Esch, F. & Saier, M.H. Purification and Properties of D-Mannitol-1-
Phosphate Dehydrogenase and D-Glucitol-6-Phosphate Dehydrogenase from Escherichia-Coli.
Journal of Bacteriology 159, 986-990 (1984).

213. Sarthy, A.V., Schopp, C. & Idler, K.B. Cloning and Sequence Determination of the Gene
Encoding Sorbitol Dehydrogenase from Saccharomyces-Cerevisiae. Gene 140, 121-126 (1994).

214. Postma, P.W., Lengeler, J.W. & Jacobson, G.R. Phosphoenolpyruvate - Carbohydrate
Phosphotransferase Systems of Bacteria. Microbiological Reviews 57, 543-594 (1993).

215. Figueroa, C.M. & Iglesias, A.A. Aldose-6-phosphate reductase from apple leaves: Importance of
the quaternary structure for enzyme activity. Biochimie 92, 81-88 (2010).

216. Banaszak, K. et al. The Crystal Structures of Eukaryotic Phosphofructokinases from Baker's
Yeast and Rabbit Skeletal Muscle. Journal of Molecular Biology 407, 284-297 (2011).

217. Milewski, S., Janiak, A. & Wojciechowski, M. Structural analogues of reactive intermediates as
inhibitors of glucosamine-6-phosphate synthase and phosphoglucose isomerase. Archives of
Biochemistry and Biophysics 450, 39-49 (2006).

218. Jules, M., Guillou, V., Francois, J. & Parrou, J.L. Two distinct pathways for trehalose assimilation
in the yeast Saccharomyces cerevisiae. Appl Environ Microbiol 70, 2771-8 (2004).

219. Brauer, M.J. et al. Coordination of growth rate, cell cycle, stress response, and metabolic activity
in yeast. Mol Biol Cell 19, 352-367 (2008).

220. Fraenkel, D.G. Yeast Intermediary Metabolism, (Cold Spring Harbor Laboratory, 2011).

221. Zhou, R., Cheng, L.L. & Wayne, R. Purification and characterization of sorbitol-6-phosphate
phosphatase from apple leaves. Plant Science 165, 227-232 (2003).

222. Oda, Y., Iwami, M. & Sakurai, S. Membrane-bound sorbitol 6-phosphatase in fat body cells
controls the dynamics of sorbitol 6-phosphate, a major hemolymph sugar in the silkworm. Insect
Biochemistry and Molecular Biology 35, 1284-1292 (2005).

223. Ladero, V. et al. High-level production of the low-calorie sugar sorbitol by Lactobacillus plantarum
through metabolic engineering. APPLIED AND ENVIRONMENTAL MICROBIOLOGY 73, 1864-
1872 (2007).

224. Lee, A.Y.W., Chung, S.K. & Chung, S.S.M. Demonstration That Polyol Accumulation Is
Responsible for Diabetic Cataract by the Use of Transgenic Mice Expressing the Aldose
Reductase Gene in the Lens. Proceedings of the National Academy of Sciences of the United
States of America 92, 2780-2784 (1995).

225. Noltmann, E.A. Aldose-ketose isomerases, (Academic Press, 1972).

226. Pelicano, H., Martin, D.S., Xu, R.H. & Huang, P. Glycolysis inhibition for anticancer treatment.
Oncogene 25, 4633-46 (2006).

227. Ralser, M. et al. A catabolic block does not sufficiently explain how 2-deoxy-D-glucose inhibits
cell growth. Proc Natl Acad Sci U S A 105, 17807-11 (2008).

228. McIsaac, R.S. et al. Synthetic gene expression perturbation systems with rapid, tunable, single-
gene specificity in yeast. Nucleic Acids Res 41, e57 (2013).

164
229. Melamud, E., Vastag, L. & Rabinowitz, J.D. Metabolomic Analysis and Visualization Engine for
LC-MS Data. Analytical Chemistry 82, 9818-9826 (2010).

230. Egan, W., Schneerson, R., Werner, K.E. & Zon, G. Structural Studies and Chemistry of Bacterial
Capsular Polysaccharides - Investigations of Phosphodiester-Linked Capsular Polysaccharides
Isolated from Hemophilus-Influenzae Types a, B, C, and F - Nmr Spectroscopic Identification and
Chemical Modification of End Groups and the Nature of Base-Catalyzed Hydrolytic
Depolymerizationlu. Journal of the American Chemical Society 104, 2898-2910 (1982).

231. Kapuscinski, M., Franke, F.P., Flanigan, I., Macleod, J.K. & Williams, J.F. Improved Methods for
the Enzymic Preparation and Chromatography of Octulose Phosphates. Carbohydr Res 140, 69-
79 (1985).

165