THE JOURNAL OF Bmmcrc~~.

CHEMISTRY
Vol. 254, No. 2, Issue of January 25, pp. 391-400, 1979
Printed in U.S.A.

Human Serum Albumin

SPECTROSCOPIC STUDIES OF BINDING AND PROXIMITY RELATIONSHIPS FOR FATTY ACIDS AND
BILIRUBIN*

(Received for publication, January 3, 1978)

Charles B. Berde,$ Bruce S. Hudson,@ Robert D. Simoni, and Larry A. Sklarll

From the Department of Biological Sciences and the Department of Chemistry, Stanford University, Stanford, California
94305

Binding and proximity relationships of hydrophobic the physiological carrier of fatty acids, lysolecithin, and bili-
ligands on human serum albumin have been studied rubin, and it also carries many hydrophobic drugs. Several
using absorption, fluorescence, circular dichroism, and mammalian albumins have recently been sequenced (Behrens
electron paramagnetic resonance spectroscopy. The li- et al., 1975; Brown et al., 1971). Based on sequence data,
gands studied were bilirubin, two conjugated linear analysis of peptide fragments, and hydrodynamic measure-
polyene fatty acids, cis-parinaric acid and cis-eleo- ments, a partial domain model for the tertiary structure of the
stearic acid, and three nitroxide derivatives of stearic protein has recently been proposed (Anderson and Weber,
acid with doxyl groups at positions 5, 10, and 12, re- 1969; Pederson and Foster, 1969; Brown, 1976). According to
spectively. Binding of polyene fatty acids was moni-

Downloaded from http://www.jbc.org/ by guest on September 28, 2017

this model, albumin is comprised of three major domains (see
tored by absorption peak shifts, induced circular di-
Fig. 15) which are in turn comprised of subdomains. The
chroism, enhancement of fluorescence, and energy
subdomains consist of cylindrical segments formed by appo-
transfer between albumins single tryptophanyl resi-
due and the polyene chromophore. Induced circular sition of (Yhelical regions. These helical regions form elongated
dichroism studies indicate excitonic ligand-ligand in- pockets for fatty acids with hydrophobic residues lining the
teraction between bound fatty acids. Fluorescence en- inner surfaces and several positively charged groups at the
hancement of cis-parinaric acid was analyzed using a opening, near the expected location of the fatty acid carboxyl
stepwise multiple equilibrium model, and six binding groups. The strongest binding site for long chain fatty acids is
constants in the range lo8 to lo6 M- were obtained, in thought to be in domain III. The single tryptophanyl residue
agreement with previous measurements for other fatty in human serum albumin lies in domain II.
acids. The temperature dependence of the equilibrium Several investigators (Goodman, 1958; Spector et al., 1969;
constants indicates that the binding enthalpy is nearly Fletcher et al., 1970; Arvidsson et al., 1971; Spector et al.,
zero. Fluorescence energy transfer was similarly used 1971; Ashbrook et al., 1975) have measured the binding of
to quantitate bilirubin binding to albumin. Energy long chain fatty acids to albumins by using partitioning of
transfer, nitroxide quenching of fluorescence, and elec- fatty acids between an organic phase such as heptane and an
tron paramagnetic resonance spectroscopy were used aqueous medium containing the protein. This method of mea-
to elucidate binding geometries which support and ex- surement was used in part because long chain fatty acids
tend proposed structural models for albumin. It is sug- dialyze very poorly. Specific assumptions must be made con-
gested that the first two fatty acids bind side-by-side in cerning the partitioning of fatty acids between organic solvents
an antiparallel fashion in domain III of human serum and water and the activity coefficients of fatty acids in water,
albumin. and these are subjects of much current debate (Mukerjee,
1965; Tanford, 1973; Smith and Tanford, 1973). Organic sol-
vents such as heptane bind to albumin, and this might com-
Albumin is the most plentiful protein in mammalian plasma. plicate the analysis. In any event, corroboration by a single
A monomeric protein of molecular weight 68,000, albumin is phase method would be helpful.
Sklar et al. have recently introduced conjugated linear
* The costs of publication of this article were defrayed in part by polyene fatty acids as probes of membrane structure and of
the payment of page charges. This article must therefore be hereby lipid-protein interactions. The important features of these
marked advertisement in accordance with 18 U.S.C. Section 1734 probes have been described in detail elsewhere (Sklar et al.,
solely to indicate this fact. 1975, 1977, a and b), and their binding to bovine serum
$ Taken from the thesis submitted in partial fulfillment of the albumin has been studied (Sklar et al., 1977c). Briefly, these
Ph.D. requirements, Stanford University (1978). Participant in the probes are isomers of 9,11,13,15-octadecatetraenoic acid (par-
Medical Scientist Training Program.
inaric acid) and 9,11,13-octadecatrienoic acid (eleostearic
Camille and Henry Dreyfus Teacher-Scholar, Alfred P. Sloan
Fellow, and National Institutes of Health Research Career Develop- acid). Parinaric acid fluoresces when bound to albumin, but
ment Awardee GM 00284. Recipient of Grant GM 21149 from the not when it is in water, and the absorption spectra shift to
National Institutes of Health. Present address, Department of Chem- longer wavelengths. The enhancement of fluorescence was
istry and Institute of Molecular Biology, University of Oregon, Eu- used by Sklar et al. to quantitate binding, but binding con-
gene, Ore. 97403. stants were obtained only for the 3rd to 5th mol bound.
1 National Institutes of Health Research Career Development Binding of PnA to bovine serum albumin results in an in-
Awardee GM 00225. Recioient of Grant GM 18539 from the National
Institutes of Health. - The abbreviations used are: cPnA, cis-parinaric acid; HSA, human
11Helen Hay Whitney Postdoctoral Fellow. Present address, Divi- serum albumin: BHT. 2.6-di-tert-butvl-4-methvlphenol; BSA, bovine
sion of Atherosclerosis and Lipoprotein Research, Department of serum albumin; cEsA; cis-eleostearicacid; h NS,-a doxyl derivative of
Medicine, The Methodist Hospital, Baylor College of Medicine, Hous- stearic acid with the doxyl group attached at the hth carbon from the
ton, Texas 77025. COOH terminus.

391
392 Ligand Binding to Albumin

duced chromophore circular dichroism. When 2 mol of PnA respectively. The absorption maximum of cEsA exhibits a
are bound per mol of protein (V = 2), a negative component of similar shift upon binding, as shown in Fig. 2C. The peaks of
the circular dichroism is observed, which was attributed to the cEsA absorption spectrum are not sufficiently resolved to
excitonic ligand-ligand interaction (Sklar et al., 1977c). The permit measurement of valley to peak ratios. These results
emission spectrum of tryptophan overlaps the absorption are very similar to those previously reported for bovine serum
spectra of cPnA and cEsA, and this feature allowed measure- albumin (Sklar et al., 1977c).
ment of distances from the two tryptophans to the first fatty Circular Dichroism Spectra of cPnA Bound to Human
acid binding site by application of the Fijrster-Dexter theory Serum Albumin-cPnA is not optically active. An intense CD
of energy transfer. is observed in the region of cPnA absorption for several
In this study we extend these absorption, circular dichroism, cPnA.protein complexes (Sklar et al., 1975, 1977c). Fig. 3
and fluorescence measurements to human serum albumin. To shows this induced CD for the cPnA. HSA complex with V =
supplement this information, we also employ electron para- 1 and V = 2. The CD of an equal molar HSA-oleic acid solution
magnetic resonance spectroscopy of nitroxide fatty acids has been subtracted from the CD of the complex. The absorp-
bound to human serum albumin and quenching of the fluo- tion spectrum for V = 2 and the ratio AE/E are also shown for
rescence of cPnA by nitroxide fatty acids. These measure- comparison. The absorption spectrum for V = 1 is very similar
ments permit determination of binding affinities under various to that for V = 2.
conditions, and permit conclusions regarding proximity of Several features of these CD spectra are noteworthy. When
ligands to each other and to the single tryptophanyl residue V I: 1, the CD is entirely positive and has the same general
of human serum albumin. features as the absorption spectrum. The magnitude of this
Albumin carries bilirubin, the breakdown product of heme, positive CD is very similar to that observed for the cPnA. BSA
to the liver for conjugation with glucuronic acid and subse- complex (Sklar et al., 1977c). A comparison of the absorption
quent excretion. Albumin has one very strong binding con- and CD spectra for the cPnA. HSA complex shows that the
stant ( -lo8 M-) and two weaker binding constants for biliru- CD is blue-shifted relative to the absorption and that the

Downloaded from http://www.jbc.org/ by guest on September 28, 2017

bin (-lo6 M-l). It has been observed clinically and in experi- relative intensities of the vibronic components are appreciably
mental systems that when the bilirubin to albumin ratio altered. In the CD the higher energy vibronic components
exceeds 1, bilirubin binds to cell membranes. This is a common have relatively more intensity. When V > 1, there is a negative
occurrence in newborn infants, particularly if premature, and component in the CD spectrum at long wavelengths and an
the deposition of bilirubin in the basal ganglia of the brain increase in the positive component at short wavelengths. This
leads to a clinical syndrome, known as kernicterus, involving behavior is qualitatively similar to that observed for the
neurological impairment. There is therefore great clinical
interest in understanding the binding of bilirubin to albumin
and the effects of drugs and other competitors on this binding.
We employ energy transfer between tryptophan and bili-
rubin to extend the measurements of Chen (1973) on binding
of bilirubin and use energy transfer between cPn4 and bili-
rubin to further specify distance relationships between bound
ligands.

MATERIALS AND METHODS

See Miniprint

RESULTS AND DISCUSSION

Absorption Spectroscopy-Fig. 1 shows absorption spectra

of cPnA (2 X 10m6 M) in 0.02 M sodium phosphate, pH 7.4, as
aliquots of human serum albumin are added. As shown in Fig.
lA, cPnA in buffer has its first absorption maximum, Pl, at
about 321.2 nm, and the value for the ratio of the absorbance
at the fist valley, Vl, to the value of the absorbance at the
second peak, P2, is about 0.52. As aliquots of human serum
albumin are added, the observed spectrum reflects the super-
position of spectra due to bound and free fatty acid. When
about 0.12 mol of human serum albumin is added per mol of
cPnA, the valley to peak ratio is greatest, indicating roughly
equal amounts of bound and free fatty acid. This is consistent
with the binding of between 4 and 5 mol of fatty acid/m01 of
human serum albumin. As more human serum albumin is
added, Pl shifts to longer wavelengths and V/P drops again
to a final value of about 0.52, indicating that nearly all of the ..
PnA is bound under these conditions. The positions of peak I I I I I I
270 285 300 315 330 345
Pl and the values of Vl/P2 are plotted in Fig. 2, B and A,
WAVELENGTH (NM.)
Portions of this paper (including Materials and Methods, and FIG. 3. Absorption and circular dichroism
spectra of cPnA bound
Figs. 1,2, 5, 7,9, 11, 13, and 14) are presented in miniprint at the end to human serum albumin. The solid curue is the absorption spectrum
of this paper. Miniprint is easily read with the aid of a standard of cPnA bound to human serum albumin (1 x 10e5 M) when V = 2.
magnifying glass. Full size photocopies are available from the Journal The dashed curue is the CD spectrum for cPnA when V = 1. The
of Biological Chemistry, 9650 Rockville Pike, Bethesda, Md. 20014. contribution due to the protein has been subtracted and the molar
Request Document No. 78M-9, cite author(s), and include a check or elliptic&y is given per mol of PnA. The inset shows AE/E (Kuhn ratio)
money order for $2.25 per set of photocopies. for the V = 2 case ([6]~ = 3300 AE).
 Ligand Binding to Albumin

BSA . cPnA complex, but the magnitude of the negative com-

ponent of the CD is smaller in the case of human serum
albumin by a factor of about 3 (Sklar et al., 1977c).
The induced CD observed for the G = 1 complex is due to
the adoption of a chiral geometry by the fatty acid tetraene
chromophore.3 The strong electric dipole allowed transition
near 310 nm has no magnetic transition moment in the ab-
sence of the chiral perturbation. This perturbation may be
viewed as causing a mixing of low energy magnetic dipole
allowed states with the electric dipole allowed state and vice
versa. The resulting mixed states exhibit circularly dichroic
transitions. One of the magnetic dipole transitions which will
be mixed with the 310 nm transition is the ground state to
excited A, transition whose vibrationless origin is at about
350 nm (Hudson and Kohler, 1974; Sklar et al., 1977a; Andrews
and Hudson, 1978). The vibronic structure of this forbidden
transition extends to about 290 nm. This state will give rise to
a band centered at about 330 nm with a rotatory strength
opposite in sign to that of the 310 nm band. This results in
the shift of the low energy vibronic bands at 308 and 325 nm
and a relative decrease in their intensity compared to the
higher energy bands.
The CD spectrum for the complex with two bound PnAs

Downloaded from http://www.jbc.org/ by guest on September 28, 2017

has a pattern characteristic of exciton coupling between the
tetraene chromophores, i.e. a negative long wavelength com-
ponent and a positive short wavelength component superim- OP 1 1 I I I /
posed on the positive CD due to the chiral chromophore 0 I 2 3 4 5 6+
geometry. The CD spectrum for the human serum albumin MOLES FATTY ACID
complex with one cPnA molecule (V = 1) is unaffected by the MOLE HSA
addition of oleic acid. This is also the case for the BSA aPnA FIG. 4. Enhancement of PnA fluorescence on binding to human
complex (SkIar et al., 1977c). This demonstrates that the new serum albumin and quenching of human serum albumin fluorescence
features of the V = 2 spectrum are associated with the presence by polyene fatty acids. A solution of human serum albumin (2 x IOm6
of two bound chromophores rather than being a property of M) was prepared in 0.02 M sodium phosphate buffer, pH 7.45. cPnA
was added in ethanol (fmal ethanol concentration 0.2%) and fluores-
the chromophore in the second binding site. The exciton
cence of cPnA was recorded using a microcuvette (0.4-cm path) with
splitting estimated from the b = 2 spectrum is 3500 to 5500 excitation at 335 nm (3 nm slit) and emission scanned from 360 nm to
cm-. If the two transition dipoles are oriented so that the 500 nm (5-nm slit). Intensities were corrected for inner filtering and
exciton splitting is maximized for a given relative distance, are plotted with triungtes. Points shown are averages of four deter-
then that distance is 5 to 7 A; for any other transition dipole minations. Error bars refer to standard deviations. For energy trans-
orientation the chromophores must be closer together. This is fer between human serum albumin and cEsA, a solution of 2 x 1Om6
direct evidence that the fast two fatty acid binding sites are M human serum albumin was used and again additions were made
from an ethanol solution to give a final ethanol concentration of
very close together. 0.1%. Excitation was performed at 294 nm (3 nm slit) and emission
It should be noted that this exciton splitting is due to the scanned from 300 nm to 440 nm (see Fig. 7B). Intensities were
transition dipole-transition dipole interaction of the strong corrected for inner and outer filter effects and plotted with open
transition ( f - 1.3) to the second excited singlet state. The circles. Identical titrations using oleic acid were performed, and the
low energy A, state has a transition dipole which is about 50 plot (squares) indicates that no quenching occurred. A similar titra-
tion of human serum albumin (3 X 1O-7 M) was performed with cPnA
times smaller ( f - 0.025) and will therefore have an exciton
(fiZZed circles). Data points shown for quenching curves are averages
interaction which is about 2500 times smaller. The fluores- of five determinations. Error bars refer to standard deviations.
cence originates in this lAg excited state (Sklar et al., 1977a)
and therefore the fluorescence of the bound PnA will not be of human serum albumin. We note that in this concentration
affected by this exciton interaction. range (1 X lop5 M), the 1st mol of cPnA is essentially all
Fluorescence Enhancement of Bound PnA-The fluores- bound, and the amount of fluorescence is directly proportional
cence enhancement associated with the binding of cPnA to to the amount bound. Fluorescence intensity increases nearly
human serum albumin was used to measure binding constants linearly until 5 or 6 mol of cPnA are added, and then it levels
under a variety of conditions. Fig. 4 shows the enhancement off abruptly. We therefore assume that for any amount of
of cPnA fluorescence as PnA is added to a constant amount cPnA bound, the fluorescence intensity is directly propor-
tional to the amount bound. cPnA was added to human serum
The alternative possibility that the optical activity is due to albumin at mole ratios ranging from 0.1 to 10, and at human
exciton coupling between the polyene chromophore and an albumin serum albumin concentrations ranging from 1.0 X 10m5 M to
aromatic residue can be excluded on two grounds, First, it would
result in a conservative CD pattern with a negative component near 1.6 X lo-* M. Knowing the amount of fluorescence intensity
270 to 280 nm with an intensity equal to the positive polyene com- present for a certain concentration of bound cPnA, the ob-
ponent at 310 nm. Second, the magnitude of the CD is too large to be served intensities at various concentrations and mole ratios
attributed to exciton coupling given the magnitude of the transition were used to determine the amount of bound and free fatty
dipoles and the exciton energy difference for the chromophores. acid in each case. Fig. 5 and Table I summarize the results of
Exciton coupling to higher energy peptide transitions would be even these binding studies and their interpretation in terms of
weaker due to the large excitation energy difference. Furthermore,
this excitonic origin for the polyene CD would not explain the distor- Equation 7. In the experiments shown, binding was measured
tion of the CD relative to the absorption in the long wavelength at 5C!, 23C, and 37C. The medium was a Ca+-free Ringers
region. solution identical with that used by Ashbrook et al. (1975).
394 Ligand Binding to Albumin

TABLE I
Association constants for cPnA and human serum albumin
Binding of cPnA to human serum albumin was determined by the
method of fluorescence enhancement as described under Materials
and Methods. Binding constants were determined by a nonlinear
least squares fit to the coefficients of Equation 7. The buffer is
described under Materials and Methods. The RMS error is the root
mean square deviation between the measured values of V and those
calculated with Equation 7 and the binding constants given. The
enthalpy change on binding to the first site is calculated from the
23C and 37C data to be 2.5 kcal/mol and the entropy change is
roughly 28 Cal. mol- deg-.
5C 23C 37C
liters/m01
Kl 1.3 x los 1.1 x lo8 9.1 x lo7 FIG. 6. Spectral overlap between human serum albumin emission
KZ 3.0 x lo7 2.4 x lo7 2.8 x 10 and the absorption of polyene fatty acids. The broad solid curve
l-6 2.6 x lo7 1.9 x lo7 1.5 x lo7 shows the emission spectrum of human serum albumin upon excita-
K4 3.7 x lo6 6.0 x 10 5.1 x 10 tion at 294 nm. The dashed curve and the three-peaked solid curve
KS 2.8 x 10 2.1 x 10 4.6 X 10 are the absorption spectra of cPnA and cEsA, respectively. The
K6 2.6 x 10 1.3 x lo6 1.0 x lo6 shaded areas indicate that there is sizable spectral overlap between
RMS error 0.193 0.283 0.210 human serum albumin and cPnA and much smaller spectral overlap
between human serum albumin and cEsA.

The binding is apparently only weakly temperature-depend-

ent. This implies that the enthalpy of binding is very small, 4. The addition of comparable amounts of oleic or linoleic acid
and the free energy of binding is due primarily to an entropy fails to alter the fluorescence of human serum albumin until

Downloaded from http://www.jbc.org/ by guest on September 28, 2017

increase upon binding, which is a general characteristic
called hydrophobic interactions in other systems (Klotz and
Urquhart, 1949; Tanford, 1973). Similar experiments
shown very little change in the binding affinity as the salt
concentration is varied over a range from 0 to 300 IIIM.
Removal of Mg2+ results in a slight increase in binding affinity.
The binding constants obtained are quite similar to those
obtained by Ashbrook et al. (1975), despite differences
experimental design and use of a different fatty acid. This
lends corroboration to their findings.
The binding constants given in Table I refer to the binding
affinity for the ith cPnA molecule to human serum albumin
with i - 1 cPnA molecules bound. There is no direct connec-
tion between these binding constants and the affinities for
individual sites unless sequential binding is assumed. An al-
ternative description in terms of site affinities can be obtained
from Equation 8 which is valid in the absence of site-site
interactions. The fit obtained with Equations 7 and 8 (which
involve the same number of parameters) is equally good. The
values obtained for K1 through kg at 23C are 4.4 x 107, 2.4 x
107, 1.3 X 107, 8.1 X 106, 2.8 X 106, and 2.3 X lo6 liters mol-.
The values for 37C are 5.2 X 107, 1.3 X 107, 9.1 x 106, 8.0 x
106, 6.8 X 106, and 3.1 X 106. These individual site binding
constants are sufficiently close together in many cases, espe-
cially when the errors in the constants are considered, that
the assumption of strictly sequential binding is only approxi-
mately valid. (This does not affect the validity of the constants
obtained using Equation 7, only their interpretation.)
means that the properties associated with a given human
serum albumin complex may actually represent an average
over several species with the ligands in different sites.
Energy Transfer between Human Serum Albumin
Polyene Fatty Acids-The emission spectrum of tryptophan
and the absorption spectra of cPnA and cEsA are shown in
Fig. 6. It can be seen that both fatty acids have spectral
overlap with tryptophan, and this affords the possibility of
dipole-dipole excitation transfer via the mechanism described
by Forster (1948). The emission spectra of human serum
albumin upon addition of aliquots of cPnA and cEsA are
shown in Fig. 7; the emission is decreased by the presence of
the polyene fatty acid. cEsA reduces the emission intensity of
human serum albumin less efficiently than cPnA. The relative
intensities of human serum albumin emission in the presence
of these two probes for various values of 6 are plotted in Fig.
of so- V is greater than 3. These fatty acids cannot quench human
serum albumin fluorescence by Forster transfer. When V > 3,
have the human serum albumin emission spectrum is shifted to
slightly shorter wavelengths presumably due to a minor con-
formational change. In calculating spectral overlap integrals
for V > 3, we use the emission spectrum of human serum
albumin in the presence of comparable amounts of linoleic
in acid since the conformational effect of this fatty acid on
human serum albumin emission is presumably very similar to
that of cPnA and cEsA. In any case, this conformational effect
is so small for human serum albumin that this assumption is
not important. We calculate the spectral overlap integral J
for HSA-cPnA transfer to be 1.49 X lo-l4 cm3 M-l. Using
values for the refractive index n = 1.45 (Sklar et al., 1977c)
and the quantum yield Q = 0.3 (Burstein et al., 1973; Teale,
1960), we calculate that for the case in which the orientation
factor IC equals 2/3,then Rc,, the distance at which 50% transfer
occurs, is about 30 A for cPnA. Similarly, when V < 3, the
spectral overlap integral for human serum albumin and cEsA
is 5.2 x lo-l6 and R. is 17 A. As V increases to 5, the calculated
value of R. for cEsA increases by about 1 A.
Using these calculations, Equation 4, and assuming that the
orientation factor a2 takes the average value of 2/3in each case,
we determine the series of tryptophan-PnA distances shown
in Table II. The use of Equation 4 requires that the binding
of subsequent ligands does not affect the geometric factor Gi
= Ki2/Ri6 for the ith ligand. This was demonstrated to be the
This case for i = 1 since addition of 1 mol of oleic acid to the
complex of human serum albumin with one cPnA did not
change the tryptophan quenching. The correspondence be-
tween the distances given for cPnA and cEsA for the first two
and binding sites is quite close. When V 2 3, cPnA quenches
virtually all of the human serum albumin fluorescence; thus,
for the third and fourth fatty acids, only the cEsA quenching
is used to calculate distances. The calculated distances shown
in Table II indicate that the first two fatty acids bind to
human serum albumin equally distant from the tryptophan.
This is consistent with the exciton interaction evident in the
CD. Also, the first two fatty acids bind further from the
tryptophanyl residue than the third and fourth fatty acids.
Distance measurements by energy transfer are subject to
some uncertainty due to the fact that the orientation factor
K, which ranges from 0 to 4, is not known. In practice, this is
only a serious problem if K is extremely small, which only
 Ligand Binding to Albumin 395

occurs for a very small range of the relative orientation bovine serum albumin, while there is sizable quenching of
variables and which requires that there is no relative orienta- human serum albumin. This difference may be ascribed in
tional motion of the chromophores during the time between part to the fact that the emission spectrum of human serum
excitation and emission. This is because the distance meas- albumin is shifted to shorter wavelengths than the bovine
urements only depend on the sixth root of K. For instance, if serum albumin emission spectrum, and spectral overlap with
K is 4 instead of 2/, the distance will be underestimated by cEsA is greater. Another reason for the relatively smaller
35%. Similarly, if K is 0.1, the distance will be overestimated degree of quenching of bovine serum albumin is that roughly
by 35%. To eliminate the possibility of large overestimates of half the emission intensity for bovine serum albumin is
distances, which occur when K is close to 0, we employed thought to be due to the tryptophanyl residue, not present in
quenching of human serum albumin fluorescence by the ni- human serum albumin, which resides in domain I. It is to be
troxide fatty acids 5 doxyl stearate (5 NS), 10 doxyl stearate expected that this tryptophanyl residue is very far from the
(10 NS), and 12 doxyl stearate (12 NS). (In our notation, the first fatty acid bound, and would not be quenched effectively.
number indicates the carbon bearing the doxyl group.) Our In both human serum albumin and bovine serum albumin,
data is in general agreement with the data of Morrisett et al. the third and fourth fatty acids quench tryptophan emission
(1975) on bovine serum albumin in two respects: 1) the better than the first two fatty acids. Both of these findings are
quenching is greatest for 5 NS and 2) quenching increases consistent with the assignment of the two strongest binding
with Y more rapidly when V = 3 or 4 than when V < 3. The 12 sites to domain III in Browns model (Brown, 1976) and
NS fatty acid is most nearly analogous to cPnA due to the subsequent binding to domain II.
position of its substituent. At all mole ratios, the quenching Quenching of the Emission of Bound cPnA by Nitroxide
by 12 NS is much less than that of eleostearic acid. The Fatty Acids-The occurrence of a negative peak in the in-
quenching by nitroxide fatty acids is isotropic and very effi- duced CD spectrum of cPnA bound to bovine serum albumin
cient for distances less than about 8 A but is inefficient for and human serum albumin has been interpreted in terms of
large distances. The low quenching of tryptophan emission ligand-ligand exciton interaction between bound fatty acids

Downloaded from http://www.jbc.org/ by guest on September 28, 2017

observed for 12 NS therefore indicates that the frst two fatty (Sklar et aE., 1977c). In order to further probe geometric
acids are not very close to the tryptophan and that K does relationships between bound fatty acids, we have examined
not have a vanishingly small value for these ligands. changes in the fluorescence and absorption spectra of cPnA
Our analysis of the cPnA and cEsA tryptophan quenching bound to human serum albumin (V = 1) upon addition of
curves in terms of ligand to tryptophan distances is based on nitroxide-labeled fatty acids. As shown in Fig. 8, 5 NS pro-
the assumption of sequential binding, i.e. discrete binding sites duces sizable quenching of fluorescence when V (5 NS) = 1,
with associated binding constants which are well separated. whereas 10 NS, 12 NS, and oleic acid produce little or no
The cEsA quenching curve of Fig. 4 shows that the second quenching at this level of binding. The possibility that the
and third ligands bind sequentially with respect to the first decreased fluorescence of bound cPnA due to addition of 1
and second ligands. Alternative descriptions of the binding in mol of 5 NS is due to displacement of the cPnA was excluded
which the intrinsic site binding constants are more nearly by observing that the cPnA absorption spectrum does not
equal and ligand-ligand interactions are significant result in shift to shorter wavelength until at least 5 mol of 5 NS have
the conclusion that the distances obtained on the basis of
sequential binding are, in fact, averages over several sites. & ! 1 i I 1
This does not affect our basic conclusion, however, which is m

that the third and fourth fatty acid ligands bind closer to the PnA + OLEATE
tryptophan than the first and second ligands. If the binding is
nonsequential, then the first two sites must be even further
from the tryptophan and the third and fourth sites must be
even closer. The conclusion that the first two fatty acids bind
about equally distant from the tryptophan is supported by
independent experiments which show that these two fatty
acids are close together in the complex.
The energy transfer data presented here may be compared
with the data of Sklar et al. (1977c) in their studies of bovine
serum albumin. Binding of a single oleic acid to bovine serum
albumin reduces the protein fluorescence. This effect is not
seen in human serum albumin until V is greater than 3. When
V = 1 for cEsA, there is virtually no additional quenching of

phanyl
plotted
residue
between

in Fig. 4 and Equations

assumed to be % in each case.
parameters
TABLE

were determined

are given in the text.

II
Distances between the tryptophanyl residue of human serum
albumin and polyene fatty acid chromophores as determined by

Apparent
energy transfer
distances polyene fatty acids and the trypto-
using the quenching
1 to 4. The orientation
Values of the relevant
efficiencies
factor I? was
spectroscopic
2
FIG. 8. Quenching
2

by nitroxide
M
MOLE
ESNS
HSA
4

fatty acids of the fluorescence of

v PnA EsA cPnA bound to human serum albumin. A solution of human serum
albumin (1 x 10m6 M) was prepared and cPnA was added in equimolar
4
ratio. Excitation of cPnA fluorescence was performed at 335 nm (3
1 25 24 nm slit) and emission was collected at 420 nm (5 nm slit). Nitroxide
2 23 23 fatty acids in ethanol were added, and changes in fluorescence inten-
3 19 sity were recorded. As a control, a similar titration was performed
4 18 with oleic acid.
 Ligand Binding to Albumin

been added. This kind of experiment can be used to estimate TABLE III
the relative binding affinities of these ligands for human serum Binding of combinations of nitroxide fatty acids to human serum
albumin. Absorption and fluorescence data in cases in which albumin
V ranges from 6 to 9 for oleic acid or nitroxide fatty acids Nitroxide fatty acids were added to a human serum albumin
suggest the following order of binding affinities: oleic acid zz solution and percentages of bound and unbound probe were deter-
mined by double integration and subtraction of EPR spectra using a
cPnA > 5 NS 2 10 NS >> 12 NS. PDP8E computer, and values were checked by hand calculation of
The polyene chromophore of cPnA is located toward the peak heights and comparison with standards. Calculations are derived
methyl end of the chain, and the best quenching occurs when from the suectra shown in Fig. 9.
the doxyl moiety is nearer to the COOH terminus. This Nitroxide fatty acid Unbound nitroxide fatty acid
suggests that for the fist two fatty acids bound, the arrange- %
ment is roughly antiparallel side by side as shown in Fig. 15. 5NS CO.2
Electron Paramagnetic Resonance Spectroscopy of Nitrox- 10 NS 0.8
ide-labeled Fatty Acids Bound to Human Serum 12 NS 4.0
Albumin-The EPR spectra of nitroxide-labeled fatty acids lONS+5NS 2.9
change dramatically upon transfer from buffer to binding sites lONS+ 12NS 7.2
IONS+ 10NS 12.9
on human serum albumin. Fig. 9A shows spectra of three 5NS+12NS 6.0
nitroxide fatty acids bound to human serum albumin (V 5 1). 12NS+12NS 9.4
The general features of these spectra are similar to those 5NS+5NS 0.3
observed by Morrisett et al. (1975) in their study of bovine a Each fatty acid listed is present in an equimolar ratio to human
serum albumin. The spectral lines are broadened, reflecting serum albumin at 5 X 10-S M.
the relative immobilization of the nitroxide moieties upon
binding to the protein. Superimposed on these broad lines are
much smaller sharp components due to the presence of un-

Downloaded from http://www.jbc.org/ by guest on September 28, 2017

bound probe. We note that more fatty acids remains unbound
in our experiment than in that of Morrisett et al. (1975)
because we work at a protein concentration which is approx-
imately an order of magnitude lower (5 X 10e5 M). It is
apparent from these spectra that 12 NS, which has its spin
label near the methyl terminus, binds less strongly than 10
NS or 5 NS.
The model for human serum albumin proposed by Brown 2
300 350 400 450 500
(1976) and the fluorescence data presented above suggest that
two fatty acids may bind to a single domain. The fluorescence WAVELENGTH (nm)
energy transfer measurements indicate that the first two FIG. 10. Spectral overlap of bilirubin absorption with human se-
bound fatty acids are equally distant from the tryptophan. rum albumin (HSA) and cPnA fluorescence. The clashed and dot-
The CD measurements and the efficient quenching of bound dashed curues are the emission spectra of cPnA and human serum
albumin, respectively. The solid curve is the absorption spectrum of
cPn4 by bound 5 NS indicates that these two species bind bilirubin when bound to human serum albumin. The shaded areas
close together. The positional dependence of this quenching represent regions of spectral overlap. It is apparent that cPnA and
indicates an antiparallel side by side arrangement. In an bilirubin have extremely strong spectral overlap, and that human
attempt to further examine this hypothesis, we recorded EPR serum albumin (i.e. tryptophan) and bihrubin have moderately strong
spectra of pairs of different nitroxide fatty acids bound to spectral overlap.
human serum albumin. In each case, 1 mol of each type was nitroxide fatty acid depends on the position in the chain of
added per mol of human serum albumin. The spectra of these the nitroxide group on the other fatty acid bound. An inter-
combinations are shown in Fig. 9B. Close approach of two pretation for this finding can be given which is consistent with
nitroxide moieties (about 10 A) would be expected to produce the geometrical information provided by the previously men-
spin-spin broadening of the spectra; no obvious broadening tioned studies of nitroxide quenching of cPnA. The fatty acid
was observed. Evidence for ligand-ligand interaction can binding sites on human serum albumin appear to have rea-
nevertheless be obtained by examination of sharp components sonably strict steric requirements. While there is only a slight
in the spectra, which are due to unbound probe. As shown in dependence of the binding affinity on the position of a single
Fig. 9B and Table III, the amount of unbound probe and nitroxide moiety into a binding site, it appears that there is
hence the binding affinity when V = 2 depends on the partic- not enough room for two nitroxide groups to approach each
ular pair of nitroxide fatty acids used, and not merely on their other closely without a substantial reduction in binding aftin-
individual affinities when I, = 1. For example, there is very ity. The extent of increase in the amount of unbound fatty
little unbound probe when V = 1 for either 5 NS or 10 NS acid for a given pair of nitroxide fatty acids with nitroxide
(~0.2% unbound 5 NS and 0.8% unbound 10 NS, respectively), groups at the jth and kth positions in their respective chains
while there is relatively weaker binding of 12 NS (4.0% un- is probably related to the proximity of the jth carbon on a
bound 12 NS). When 5, equals 2 for 5 NS, the percentage of first chain to the kth carbon in a second chain. As depicted in
unbound probe increases to 0.3%, as would be expected for Fig. 15, the proposed antiparallel side by side binding arrange-
sequentially decreasing binding constants for successive sites. ment will result in relatively large steric hindrance for the pair
Similarly, when V = 2 for 12 NS, then the amount of unbound 10 NS + 10 NS, and relatively less for the pair 5 NS + 5 NS,
probe increases to 9.4%. The surprising finding is that when as is observed.
t, = 2 for 10 NS the amount of unbound probe is 12.9%, greater Energy Transfer between Human Serum Albumin and
than that observed for 12 NS, despite the fact that 10 NS is Bilirubin-The absorption spectrum of bilirubin is plotted
much more strongly bound than 12 NS when b = 1. Moreover, along with the emission spectrum of human serum albumin in
the amount of unbound 10 NS when V = 2 is greater than the Fig. 10. There is sizable spectral overlap, permitting efficient
amount of unbound probe when the pair 10 NS + 12 NS is Forster excitation transfer. Plots of the fluorescence of human
studied. We conclude that the binding affinity of a particular serum albumin in the presence of successive aliquots of bili-
 Ligand Binding to Albumin 397
___-__ ----__---__
robin are shown in Fig. 11, and relative emission intensities at / ---_
-..
346 nm are plotted in Fig. 12. This quenching proftie is quite / \
similar to that observed by Chen (1973). It can be seen that I / \\
\ \
bilirubin quenching is quite efficient (T = 66% when V = 1). l
From the data shown in Fig. 10, the value of the spectral \
overlap integral J is 1.74 X lo-l4 and Ro~ is 31 A. Using the I \
measured transfer efficiencies when V = 1 or 2, we calculate
the apparent distances from the tryptophanyl residue to the
first and second bilirubin chromophores to be 27 A and 24 A,
respectively (see Table IV). As noted by Chen (1973), fluores-
cence quenching at different protein concentrations can be ,,---
/
used to obtain binding constants. We have repeated his mea- ,
TRYPTOPHAN
surements and studied binding at 5C, 23C and 37C. Bind-
II
ing data were analyzed by a stepwise multiple equilibrium
model and are shown in Fig. 13 and Table V. Results at 23C
compare well with those obtained by Chen (1973) and Levine
(1977), and by other methods (Jacobsen, 1969) using a Scat-
chard model. Results over this range of temperatures suggest
moderate temperature dependence of binding.
Energy Transfer between cPnA and Bilirubin-The ab-
sorption spectrum of bilirubin also overlaps the emission

Downloaded from http://www.jbc.org/ by guest on September 28, 2017

FIG. 15. Proximity relationships and the partial domain model for
human serum albumin. The figure is a schematic representation of
human serum albumin showing three domains. Adapted from Brown
(1976).

TABLE V
Binding constants for bilirubin and human serum albumin
Binding was quantified by the use of fluorescence quenching titra-
tions at different protein concentrations as shown in Fig. 14. It is
assumed that at concentrations greater than 6 X 1Om6 M the fust 2
mol are essentially all bound. As before, the nonlinear least squares
routine was used, and three binding constants were required to
produce adequate fits. The third binding constants were 8 x 105, 7 x
105, and 6 X lo5 at 5C, 23C, and 37C, respectively. These constants
are not reliable due to the absence of sufficient data points for Y > 2.
I. C.
The RMS error is described in the legend to Table I. The enthalpy
change calculated for the first binding constant from the 23C and
MOLES BILIRUBIN 37C data is roughly 8.2 kcaI/mol and the entropy change is roughly
MOLE HSA 9 cal/mol . degree.
FIG. 12. Energy transfer from human serum albumin (HSA) to 5C 23C 37C
bilirubin. Relative emission intensities at 340 nm are recorded from liters/m01
the experiment (described in Fig. 11) in which bihrubin is added to
human serum albumin (6 X 1Om6 M, filled circles). Fluorescence K1 1.3 x lo* 1.2 x lo* 6.4 x 10
intensities were corrected for inner and outer filtering, scaled relative K2 2.4 x lo6 4.9 x lo6 2.6 x lo6
to the initial value, and plotted versus the bihrubin to human serum RMS error 0.119 0.154 0.079
albumin molar ratio. The open circles and squares show the quench-
ing efficiency of bihrubin at human serum albumin concentrations of spectrum of cPnA as shown in Fig. 10. In this case the spectral
3.5 X 10e7 M and 8 X 10-s M, respectively. The data shown are for
quenching at 23C; similar curves were obtained for 5C and 37C. overlap is extremely large, and the spectral overlap integral is
1.031 X 10-13. The value of Ro213 is calculated to be 41 A. Fig.
TABLE IV 14 shows the quenching of the fluorescence of cPnA bound to
Distances from bound bilirubin to polyene fatty acids and the human serum albumin (Y = 1) as aliquots of bilirubin are
tryptophanyl residue of human serum albumin added. A similar quenching profile is observed when 2 mol of
Apparent distances between chromophores are calculated using cPnA are bound. As noted under Materials and Methods,
the transfer efficiencies derived from Figs. 12 to 14, similar curves for the quantum yields of the first 2 bound cPnA mol are identical,
the case in which V = 2 for PnA (data not shown), and Equations 1 to and this allows calculation of distances between both bound
4. Values of the pertinent spectroscopic parameters are given in the cPnA chromophores and the first two bound bilirubin chro-
text, and K* is set equal to % in each case. mophores. As shown in Table IV, in each case, the fatty acid-
Donor-acceptor p/3
bilirubin distance is large, ranging between 35 and 40 A. This
A finding is consistent with the observation (Woolley and
Tryptophan-bihrubini 27 Hunter, 1970; Berde et aL4) that the first two fatty acids
Tryptophan-bihrubinz 24 bound to human serum albumin do not cause measurable
PnAi-bihrubinl 40
PnAl-bibrubinn 37 displacement of bilirubin.
PnAz-bihrubim 40 4 C. B. Berde, F. Rasmussen, W. Benitz, J. A. Kerner, J. D. Johnson,
PnAz-bihrubinz 35
and R. P. Wennburg, in preparation.
398 Ligand Binding to Albumin
CONCLUSIONS Arvidsson, E. O., Green, F. A., and Laurell, S. (1971) J. Biol. Chem.
Binding and proximity relationships for fatty acids and 246,5373-5379
Ashbrook, J. D., Spector, A. A., Santos, E. C., and Fletcher, J. E.
bilirubin bound to human serum albumin were studied using
(1975) J. Biol. Chem. 250,2333-2338
absorption, circular dichroism, fluorescence enhancement, and Behrens, P. Q., Spiekerman, A. M., and Brown, J. R. (1975) Fed.
fluorescence energy transfer techniques with conjugated linear Proc. 34,591
polyene fatty acids. Electron paramagnetic resonance spec- Brown, J. R. (1976) Fed. Proc. 35,2141-2144
troscopy of nitroxide fatty acids and fluorescence quenching Brown, J. R., Low, T., Behrens, P., Sepulveda, P., Parker, M., and
by nitroxide fatty acids were also used to determine proximity Blakeney, E. (1971) Fed. Proc. 30, 1241
Burstein, E. A., Vedenkins, N. S., and Ivkova, M. N. (1973) Photo-
relationships. Binding and proximity relationships for biliru-
them. Photobiol. 18, 263-279
bin were monitored by fluorescence quenching. The data Chen, R. F. (1967) J. Biol. Chem. 242, 173-181
presented here generate the following picture of ligand binding Chen, R. F. (1973) in Fluorescence Techniques in Cell Biology
td albumin. There are roughtly six strong binding sites for (Thaer, A. A., and Sernetz, M., ed) pp. 273-282, Springer-Verlag,
fatty acids, with the fist two fatty acids binding more strongly New York
than the next four. The binding is very weakly temperature- Fisher, P. A., and Kern, D. (1977) J. Biol. Chem. 252,6528-6535
dependent, indicating that the major driving force for binding Fletcher, J. E., Spector, A. A., and Ashbrook, J. D. (1970) Biochem-
istry 9,4580-4587
is a positive entropy contribution. The first two fatty acids Forster, T. (1948) Ann. Phys. 437, 55-75
bind close together and further from the tryptophanyl residue Goodman, D. S. (1958) J. Am. Chem. Sot. 80,3892-3898
than the third and fourth fatty acids. Taken together with the Hubbell, W. L., and McConnell, H. M. (1971) J. Am. Chem. Sot. 93.
bovine serum albumin results of Sklar et al. (1977c), this 314-326
supports the view that the first two binding sites are located Hudson, B. S., and KohIer, B. E. (1974) Annu. Rev. Phys. Chem. 25,
in domain III, and the third and fourth fatty acids bind in 437-460
Jacobsen, J. (1969) FEBS Lett. 5, 112-114
domain II, which contain the tryptophan. Fluorescence Klotz, I. M., and Urquhart, J. M. (1949) J. Am. Chem. Sot. 71,
quenching, EPR spectroscopy, and CD spectroscopy provide 847-851

Downloaded from http://www.jbc.org/ by guest on September 28, 2017

evidence that the first two fatty acids bind antiparallel and Levine, R. (1977) Clin. Chem. 53,2292-2301
side by side. The first 2 mol of bilirubin bind a very large Morrisett, J. D., Pownall, H. J., and Gotto, A. M. (1975) J. Biol.
distance from the first two fatty acids. Fluorescence quenching Chem. 250,2487-2494
data (not shown) and competition studies with linolenic acid4 Mukerjee, P. (1965) J. Phys. Chem. 69,2821-2827
Ostrow, J. D., Hammaker, L., and Schmid, R. (1961) J. Clin. Incest.
indicate that the third fatty acid binds closer to the first two 40, 1442-1452
bilirubin sites, and that there is some competition between Pederson, D. M., and Foster, J. F. (1969) Biochemistry 8, 2357-2365
these fatty acids and bilirubin. Quenching studies indicate Reed, R. G., Feldhoff, R. C., Clute, 0. L., and Peters, T., Jr. (1975)
that the first two bilirubin sites are roughly equidistant from Biochemistry 14,4578-4583
the first two fatty acid sites and from the tryptophanyl residue. Scatchard, G. (1949) Ann. N. Y. Acad. Sci. 51, 660-672
Fluorescence quenching by bilirubin supports the notion that SkIar. L. A.. Hudson. B. S.. and Simoni. R. D. (1975) Proc. Natl.
Acid. Sci: U. S. A. ?2,1649-1653
there are at least two strong binding sites whose affinities SkIar, L. A., Hudson, B. S., Petersen, M., and Diamond, J. (1977a)
differ by an order of magnitude. Studies at different temper- Biochemistry 16,813-819
atures indicate moderate temperature dependence of binding Sk&, L. A., Hudson, B. S., and Simoni, R. D. (1977b) Biochemistry
affinity. 16,819-828
The relationships described here provide a set of approxi- SkIar. L. A.. Hudson. B. S.. and Simoni. R. D. (1977c) Biochemistry
mate distances between ligands which follow the features of 16;5100-5108
Smith, R., and Tanford, C. (1973) Proc. Natl. Acad. Ski. U. S. A. 70,
Browns model and data on proteolytic fragments presented 289-293
by Pederson and Foster (1969), Reed et al. (1975), and other Spector, A. A., Fletcher, J. E., and Ashbrook, J. D. (1971) Biochem-
workers. While the measurements depend on simplifying as- istry 10,3229-3232
sumptions and the model for domain structures may be a Spector, A. A., John, K., and Fletcher, J. E. (1969) J. L&id Res. 10,
simplification of the real state of affairs, a quantitative picture 56-57
emerges which may be of heuristic value and also of clinical Tanford, C. (1971) Physical Chemistry of Macromolecules, pp.
526-546, John Wiley & Sons, New York
utility in describing drug-ligand interactions. Tanford. C. (1973) The Hydroohobic
- _ Effect:
. Formation of Micelles
Acknorulealgments-We have benefitted from helpful discussions and Biological Membranes, pp. 12-21, John Wiley & Sons, New
with Drs. James R. Brown, Richard P. Wennburg, and Alonzo Ross. York
We thank Dr. Harden McConnell for use of his EPR spectrometer. Teale, F. W. J. (1960) Biochem. J. 76,381-388
Van Holde, K. E. (1971) Physical Biochemistry, pp. 57-65, Prentice
REFERENCES Hall, Englewood Cliffs, N. J.
Anderson, S. R., and Weber, G. (1969) Biochemistry 8,371-377 Woohey, P. V., III, and Hunter, M. J. (1970) Arch. Biochem. Biophys.
Andrews, J. R., and Hudson, B. S. (1978) Chem. Phys. Lett. 57, 140, 197-209
600-604
 Downloaded from http://www.jbc.org/ by guest on September 28, 2017
399
to Albumin
Binding
Ligand
 Ligand Binding to Albumin

& 4.0

z
z
Y 2.0
cl
::
k!
2

Downloaded from http://www.jbc.org/ by guest on September 28, 2017

2 320 340 360 380 400 420
WAVELENGTH IN NANOMETERS
 Human serum albumin. Spectroscopic studies of binding and proximity
relationships for fatty acids and bilirubin.
C B Berde, B S Hudson, R D Simoni and L A Sklar
J. Biol. Chem. 1979, 254:391-400.

Access the most updated version of this article at http://www.jbc.org/content/254/2/391

Alerts:
When this article is cited
When a correction for this article is posted

Click here to choose from all of JBC's e-mail alerts

Downloaded from http://www.jbc.org/ by guest on September 28, 2017

This article cites 0 references, 0 of which can be accessed free at
http://www.jbc.org/content/254/2/391.full.html#ref-list-1