THE BINDING OF THE NONSTEROIDAL ANTIINFLAMMATORY DRUGS
BENOXAPROFEN AND FENBUFEN TO HUMAN SERUM ALBUMIN
JEFFREY SCOTT FLEITMAN
A DISSERTATION PRESENTED TO THE
GRADUATE COUNCIL OF THE UNIVERSITY COUNCIL IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE
OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
To my parents, for their unwavering support and encouragement, I
dedicate this with all of my love.
To Dr. John Perrin and Dr. Stephen Schulman for being not only
teachers but friends I will never forget. To Dr. Stephen Curry and
Dr. Federico Vilallonga for always having the time to talk.
TABLE OF CONTENTS
ACKNOWLEDGMENTS... ........................................ ii
I METHODS FOR THE INVESTIGATION OF DRUG-ALBUMIN INTERACTIONS... 1
Introduction..... ....................................... 1
Equilibrium Dialysis ................................... 3
Circular Dichroism....................................... 7
Microcalorimetry ........................................ 11
Fluorescence Spectrometry.......................... ..... 12
Investigations Into the Binding of Phenprocoumon to
Albumin Using Fluorescence Spectroscopy.................. 13
II BENOXAPROFEN, A NEW NONSTEROIDAL ANTIINFLAMMATORY DRUG....... 26
III A RAPID HIGH PRESSURE LIQUID CHROMATOGRAPHIC ASSAY OF
BENOXAPROFEN IN PLASMA.................................... 29
Introduction ............................................ 29
Materials ................................................ 29
Procedures .. ............................................. 31
Results .................................................. 32
Discussion .. ............................................. 32
IV BINDING PARAMETERS FOR BENOXAPROFEN-ALBUMIN COMPLEXES........ 38
Experimental ............................ ................. 38
Binding of Benoxaprofen to Hemodialyzed Plasma............. 45
V CONFORMATIONAL CHANGES IN HUMAN SERUM ALBUMIN................ 48
Basic Structure......................................... 48
Polymeric Forms......................................... 50
N-F Transition........................................... .. 51
N-B Transition........................................... .. 52
VI THE EFFECTS OF pH, CALCIUM AND CHLORIDE IONS ON THE BINDING
OF BENOXAPROFEN TO HUMAN SERUM ALBUMIN.................... 56
Introduction ............................................ 56
Materials and Methods................................... 56
Results and Discussion................................... 57
VII DRUG DISPLACEMENT STUDIES: CHARACTERIZATION OF THE
BENOXAPROFEN BINDING SITES ON HUMAN SERUM ALBUMIN......... 77
Introduction ............................................ 77
Materials and Methods................................ ... .. 78
Results and Discussion................................ ... 79
VIII AN IMPROVED HPLC ASSAY FOR FENBUFEN AND TWO SERUM META-
Introduction .. ........................................... 89
Materials .................................................. 89
Chromatographic Conditions................................ 92
Analytical Procedure................... ................. .... 92
Results and Discussion.............................. ....... 93
IX CLASSIFICATION OF THE FENBUFEN BINDING SITES ON HUMAN SERUM
ALBUMIN: DRUG DISPLACEMENT STUDIES ........................ 99
Introduction ............................................. 99
Materials and Methods............................. ... ..... 99
Results and Discussion........................... .... .. 100
BIOGRAPHICAL SKETCH............................. ........... 116
Abstract of Dissertation Presented to the
Graduate Council of the University of Florida
in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy
THE BINDING OF THE NONSTEROIDAL
ANTIINFLAMMATORY DRUGS BENOXAPROFEN
AND FENBUFEN TO HUMAN SERUM ALBUMIN
JEFFREY SCOTT FLEITMAN
Chairman: John H. Perrin, Ph.D.
Major Department: Pharmacy
Equilibrium dialysis, microcalorimetry and circular dichroism are
used to study the binding of benoxaprofen and fenbufen to human serum
albumin (HSA). The theoretical basis for these techniques is examined
and then they are used as analytical tools in characterizing these drug-
Benoxaprofen has been found to be very highly bound to HSA (>99%)
with an affinity constant for the first binding site (K1) equal to 6.02 x
106 Z/mole. The effects of hemodialysis of benoxaprofen binding in vitro
havealso been studied.
A pH induced conformational change (N-B) in HSA is described and the
effects of the albumin conformation on benoxaprofen binding are elucidated.
It was found that chloride and calcium ions also alter the N-B transition.
Calcium ions alter the albumin conformation such that the binding capacity
for many drugs is increased. However, this study describes an additional
effect in which they compete with the carboxylate residue of benoxaprofen
for binding to the imidazole residue on the amino acid histidine. This
results in a lower binding affinity for benoxaprofen.
Characterization of the binding site on HSA for benoxaprofen is accom-
plished by drug displacement studies. Two high affinity sites of the
binding of anionic drugs are described and the interactions of benoxapro-
fen with drugs specific to these sites is quantitated.
It was found that the binding of fenbufen to HSA was much more diffi-
cult to characterize. Rather than displace drugs common to its binding
site, fenbufen induces a conformational change in the albumin which results
in a cooperativity effect. Fenbufen is described as directly influencing
the N-B transition.
High performance liquid chromatography was used as a method of detec-
tion in the equilibrium dialysis experiments and assays for benoxaprofen
and fenbufen are presented.
A rapid high performance liquid chromatographic assay for the deter-
mination of benoxaprofen in human plasma, is described. Plasma samples
of 1.0 ml, to which benoxaprofen, and warfarin as an internal standard,
had been added, were extracted with ether under acidic conditions. The
samples were analyzed on a MicroPak CN-10 column using 25% acetonitrile
in water (pH 2.5 with phosphoric acid). Detection was made on a variable
wavelength UV absorbance detector at 309 nm.
Samples containing 0.5-10 pg benoxaprofen gave a mean extraction
recovery from control plasma of 90.6 6.8% (n=18). Stability tests
have shown that benoxaprofen in plasma is stable for at least two weeks
A high performance liquid chromatographic assay for the determination
of fenbufen and its two serum metabolites is described. The overall
retention time for the chromatographic procedure is 9 minutes compared
to 22 minutes for the existing method. The compounds were extracted from
a 1.0 ml plasma sample using cyclohexane-ether (7:3) under acidic condi-
tions. Analyses were made on a MicroPak CN-10 column using water-isopro-
panol-acetonitrile-phosphoric acid (84.5:10:5:0.5). Detection was made
on a variable wavelength UV detector at 265 nm. The lower limit of detec-
tion for all compounds was 0.5 ug/ml.
I. METHODS FOR THE INVESTIGATION OF
Techniques for the study of drug-albumin interactions can be broadly
classified into two categories: those which separate the bound from
unbound ligand and those which do not. Methods of separation include gel-
filtration, ultrafiltration, ultracentrifugation and equilibrium dialysis.
Methods which do not involve separation are primarily techniques such as
UV, visible wavelength, fluorescence, NMR, ESR, as well as optical rota-
tory dispersion (ORD) and circular dichroism (CD). Flow microcalorimetry
also does not involve separation.
Choice of technique depends largely upon the physical characteristics
of the drug being studied. If it had poor spectral properties due to
lack of aromaticity or conjugation, the most convenient methods of UV-
visible and/or fluorescence spectroscopy may be useless. Photodegradation
problems may also limit these techniques. In such cases, microcalori-
metry is a useful alternative.
The separation methods permit measurements of free and bound drug
at equilibrium to be made. Once the free drug is separated from the bound
complex it can be analyzed by any one of the standard techniques such as
UV, visible wavelength or fluorescence spectroscopy, high performance
liquid chromatography (HPLC) or gas chromatography (GC). Again, if the
drug has poor spectral properties, GC or HPLC with electrochemical
detection might be used. Using a radiolabelled drug with a liquid scin-
tillation counter as a mode of detection can be a great advantage over
the other analytical procedures. There would be no need for sample
extraction as is usually the case with chromatographic procedures and it
may be more specific than direct spectrophotometric analysis.
The separation techniques allow us to calculate the most important
binding parameters such as the binding constant (K) and the number of
binding sites (n). If the binding study is done at different tempera-
tures, the thermodynamics of the binding reaction can be determined using
the Van't Hoff equation. Spectroscopic techniques can also be used to
determine these parameters if distinctions between the optical properties
of the bound and unbound drug can be made. However, by being able to
directly measure the formation of the drug-albumin complex the spectro-
scopic techniques have the advantage of allowing the researcher to char-
acterize the complex between the drug and albumin. For example, the
observance of an isosbestic point or sigmoidal character of a titration
curve can lead one to make direct inferences as to the number of mole-
cules bound per albumin molecule.
These methods should be thought of as complimentary and not competi-
tive. A disadvantage of the separation methods is that they tend to
involve a perturbation of the protein system. Ultracentrifugation and
ultrafiltration may cause changes in protein concentration during an
experiment which may give rise to inaccurate binding parameters. This
problem tends to be minimized when equilibrium dialysis is used.
Four of the techniques mentioned above, equilibrium dialysis, fluor-
escence spectroscopy, circular dichroism and microcalorimetry have been
found to be most useful in studying drug-albumin interactions in our
laboratory. The theoretical and practical aspects of these methods will
be presented in greater detail.
Basically, the principal objectives of an equilibrium dialysis experi-
ment are to determine the free and bound concentration of a ligand in a
system where it has an affinity for a macromolecule. A typical equili-
brium dialysis experiment involves the separation of a solution of a low
molecular weight ligand by a semipermeable membrane from a solution con-
taining a high molecular weight macromolecule. Eventually, equilibrium
will be reached where part of the ligand is bound to the macromolecule
and the free ligand concentration is equal on both sides of the membrane.
Diffusion of the ligand through the membrane obeys Fick's First Law.
This expression states that the amount of a substance diffusing in time
across a plane of area is directly proportional to the change of concen-
tration with distance travelled. An ideal membrane would therefore allow
maximum diffusion of the ligand and completely retain the macromolecule.
Several authors have used equilibrium dialysis to study the binding
of ligands to macromolecules. Original work was first done by Davis (1943)
and Klotz (1946) with improvements subsequently made on these methods.
Technical innovations which decreased dialysis time and increased preci-
sion have also been accomplished (Weder et al. 1971).
The dialysis apparatus used in all of the experiments in this work
is known as the Dianorm equilibrium dialysis system (Diachema AG, Zurich).
It consists of two teflon dialysis cells separated by a semipermeable
membrane. Teflon cells are chosen as they allow for minimum drug absorp-
tion onto the surface. All experiments were done in cells filled to 4.5
ml. The semipermeable membrane is a hydrated cellulose (Diachema type
10.16) and has a molecular weight cutoff of 10,000. Albumin in the
appropriate buffer is injected into one cell and drug in buffer into
the other cell. Each solution should be identical in pH and ionic
strength. The cells are then rotated in a water bath for 12 hr at 250C.
With this apparatus, equilibrium is reached well within this time
and the free ligand concentration CL(f) is the same on both sides of the
membrane. Since CL(o),the starting initial concentration of ligand is
known, we can determine CL(b), the bound ligand concentration, using the
CL(b) = CL(o)- 2 CL(f) (1)
This relationship might be upset if some ligand binds to the membrane,
in which case
CL(b) = CL(o) 2 CL(f) CL(m) (2)
The concentration or amount bound to the membrane CL(m) is determined in
a series of experiments without any macromolecule. In the concentration
range of ligand which would be used if macromolecule were present in a
binding experiment, equation 2 becomes
CL(m) = CL(o) 2 CL(f) (3)
The only alternative to this approach is to assay both sides of the mem-
brane for ligand. An efficient extraction procedure which will determine
ligand concentration on the bound side is then a necessity.
Another problem which might arise in drug-albumin studies involves
the Gibbs-Donnan effect, which may effect the equilibrium when charged
species are involved in the experiment. This phenomenon can be better
illustrated by the following example where a negatively charged macro-
molecule solution such as albumin (R-) is placed on one side of a semi-
permeable membrane and NaCl is placed in solution on the other side. If
equilibrium is established, the system can be represented by the diagram
in figure 1.
Figure 1. Sodium chloride
of albumin (R~)
in equilibrium in the presence
at physiological pH.
At equilibrium the sodium chloride concentration (in dilute solutions)
must be the same on both sides of the membrane and the total charges on
both sides of the membrane must be neutral. Equations (4), (5) and (6)
represent these conditions
[Na]o [C1-]o = [Na+]i [Cl-]i (4)
[Na+]o = [Cl-] (5)
[Na+]i = [R-]i + [Cl-]i (6)
Substituting equations (5) and (6) into (4) gives
[CI-]o2 = ([Cl-]i + [R-]i)[Cl-]i
=[Cl- (1]i2 (+ (7)
[Cl]o 1 + [R-]i
0 ( = +[C] ) (8)
When the macromolecule concentration inside the membrane [R-]i is very
large it tends to push the anion [Cl-]i out through the membrane. In
this case, the ratio would be approximately equal to [R-]i. If [Cl']i
is much larger than [R~]i then the ratio in (8) would approach unity and
the salt concentration on both sides of the membrane is the same.
All equilibrium dialysis experiments in this work are done in buffer
solutions of 0.1 M so as to negate the effects of the Donnan Membrane
SWith these considerations realized, we can then use the equilibrium
dialysis apparatus to obtain the drug-albumin binding parameters. The
free fraction f of drug is given as
f L(f) (9)
CL(b) + CL(f)
which in actuality is the concentration of drug analyzed for on the free
side divided by the total drug concentration of the bound side of the
The bound fraction b is
b = 1-f (10)
As the total protein concentration will be known it is convenient to
define an experimentally determinable quantity r, as the average number
of ligand molecules bound per molecule of protein
r = CL(b) ()
Plotting r/CL(f) vs. r will result in a Scatchard plot (Scatchard, 1949)
from which the binding constants Ki, K2, etc. and nl, n2, the average num-
ber of ligands bound to each class of sites (if more than one site exists).
Physiological concentrations of albumin are usually used in dialysis
experiments. The binding parameters obtained from dialysis are lower than
those obtained from the spectrophotometric methods. The lower albumin
concentrations used in fluorescence or CD (sometimes 100X less) tend to
give rise to higher affinity constants. This may be a result of dimeri-
zation of the albumin at high concentrations.
Reviews of methods of handling dialysis data have been done by Weder
et al. (1974), Peters and Pringoud (1979) and Vallner (1974).
Observations about differences in absorption of the components of
circularly polarized light, circular dichroism, were first reported by
Haidinger (1847). Since that time it was used by various authors to study
organic chemical problems but they were very limited by technical diffi-
culties. It wasn't until the 1950's that photoelectric spectropolarime-
ters became available and convenient circular dichroism studies were made
possible (Mitchell, 1957).
Basically, in order to get information from the use of circular
dichroism, one has to have a compound with an optically active chromophore
in the spectral range under examination. This is usually the UV-visible
light absorption region. A molecule may absorb light and not be optically
active so it is necessary to describe what constitutes optical activity.
The Pasteur principle states that in order to be optically active
a molecule must be dissymmetric; the molecule and its mirror image are
not superimposable (Pasteur, 1848). Dissymmetry does not imply the absence
of all elements of symmetry but only the absence of a plane or center of
symmetry. The two structures shown below exhibit optical activity due
to dissymmetry arising from different conditions.
Br- C C1 C3H7 C C2H5
Structure (I) exhibits dissymmetry arising from the differences in elec-
tronegativity of the atoms surrounding the asymmetric carbon atom. The
optical activity of structure II has dissymmetry which arises from the
spatial arrangements of the attached substituents. It is possible for a
symmetric chromophore to exhibit optical activity if it is placed in the
environment of a dissymmetric moiety. Drug-albumin interactions tend to
exhibit optical activity by this phenomenon.
When one wavelength from a beam of light is sent through a polarizer,
the emitted wave of plane polarized light (e) is made up of two vectors.
One vector corresponds to right circularly polarized wave ER and one to
left circularly polarized wave eL. To be optically active, a medium has
to show different indices of refraction for the left and right circularly
polarized light (nL and nR). This is known as circular birefringence, as
the left and right handed wave components will have different speeds due
to the inequality of nL and nR. The lengths of eL and ER remain equal
and plane polarized. If this phenomenon occurs in a region where there
is an optically active chromophore which absorbs light, there will be
differential absorption of EL and ER. The component with the larger index
of refraction or lesser velocity is preferentially absorbed and their
resultant vector e no longer oscillates along the circumference of a
circle but traces out an ellipse. This is called circular dichroism, a
differential absorption between circularly polarized beams. Diagrams
elucidating the waveforms involved along with more detailed elaborations
on the mathematics involvedin these transitions can be found in great
detail in texts by Crabbe' (1965), Djerassi (1960) and Urry (1970).
The circular dichroic signal which one measures on the spectropolari-
meter is reported as the observed ellipticity o in degrees. It can be
reported as the molar ellipticity -, where C is the molar concentration
of the optically active molecule and 1 is the path length in centimeters
required to obtain the molar ellipticity.
Spectra obtained from circular dichroic measurements are characterized
by so called "Cotton" effects of positive or negative maximums, minimums
and points of inflection. Since this work deals with the binding of drugs
to albumin, the factors which give rise to circular dichroic signals in
this system will be examined.
Macromolecules such as albumin have at least three types of dissym-
metry which may lead to optical activity. If the substituents on the a
carbon of an amino acid give rise to an asymmetric configuration,then the
primary structure of the protein is said to be inherently asymmetric and
may lead to optical activity. The secondary structure of albumin which
is helical in nature may give rise to optically active electron transitions.
Finally, the tertiary structure might be influential in that symmetrical
groups such as a drug molecule might be thrust into an asymmetric environ-
ment, as provided by the protein and give rise to a distorted electron
displacement which may lead to an induced optical activity. This case
usually gives a much weaker Cotton effect and occurs at wavelengths char-
acteristic of the drug molecule.
When analytically studying the effects of drugs binding to albumin
the researcher must usually work in the wavelength region above 300 nm
where there are negligible contributions from the albumin and measure the
induced ellipticity, if one arises. Difficulties arise when trying to
study binding at lower wavelengths due to the intrinsic ellipticity of
the albumin molecule. In the wavelength region between 250 nm and 300 nm
a strong CD signal arises from secondary structure, nonbackbone chromo-
phores. Aromatic amino acids, phenylalanine, tyrosine and tryptophan and
cysteine all give a strong signal. Tryptophan is especially influential
at 290 nm. In the region below 250 nm the backbone amino acids have a
very strong influence as well as the aromatic amino acids and histidine
and cysteine (Urry, 1970; Vallner, 1974).
Many small symmetrical molecules give rise to induced circular dichroic
activity when bound to albumin. Examples may be a variety of coumarins
and phenylpropionic acid derivatives (Perrin, 1973). Hydrophobic inter-
actions as well as van der Waal's forces, hydrogen and ionic bonding tend
to give rise to induced optical activity (Klotz, 1973). Structurally,
acidic drugs with large flat hydrophobic areas which are able to form a
rigid complex with the protein are also generally regarded as essential
criteria (Chignell, 1969; Rosen, 1970; Perrin, 1973).
Microcalorimetry has proven to be a useful analytical tool in the
study of drug-albumin binding. It has been used in the study of benzo-
diazepin-albumin interactions (Coassolo et al. 1978a,b), drug-cyclodextrin
interactions (Hardee et al. 1978) and drug-microorganism interactions
Primarily, its use in this study arises from the fact that when drugs
bind to albumin the very small heat evolved or absorbed is proportional
to the amount bound. This allows one to calculate binding parameters
and also serves as a useful analytical tool for drug displacement studies.
It is advantageous in the binding studies of drugs with poor optical pro-
perties and is a non-invasive technique. Unlike the spectroscopic methods,
microcalorimetry does not alter the properties of the system under study
or introduce artifacts during measurements.
An LKB 2107-121 Flow Microcalorimeter (LKB, Bromma, Sweeden) will be
used in this study. It is fitted with two separate gold cells in which
heat reactions can take place. An albumin solution pumped into the calor-
imeter at about 14 ml/min reacts with a drug solution pumped through a
second pump and the binding reaction occurs in a gold mixing vessel. Heat
generated in this reaction flows across a thermocouple and into a heat
sink assembly. If heat is absorbed by the reaction, heat flows from the
heat sink to the mixing vessel.
Heat flowing across the thermocouple generates an electromotive force
proportional to the temperature difference. A second thermocouple acts
as a reference so that when the electromotive force in both thermocouples
is equal the voltage output is zero. A "Wheatstone bridge" circuit is
used to balance the thermocouples so that this zero obtained is actually
the experimental baseline which encompasses heats of dilution, flow effects,
etc. When an exothermic reaction occurs and is measured by the first
thermocouple, the heat in the second thermocouple is subtracted and an
output signal is amplified, fed into a recorder and read in microvolts.
The entire system is isothermal and kept in a Tronac Model 1005 water bath
with a PTC 40 temperature controller (Tronac, Orem, Utah) at 25 0.0020C.
A thorough description of this system has been given by Hardee (1980).
Disadvantages of this system include the large amount and concentra-
tions of reactants which must be used. To get a good measurable signal
in these experiments, albumin in the concentration range of 5.0 x 10-4 M
must be used. Drug concentrations of similar magnitude will also be used
and obviously solubility problems will limit this technique.
Fluorescence spectroscopy has proven to be one of the quickest and
most direct methods of studying drug-albumin interactions. Changes in
drug fluorescence on binding to proteins can be monitored and this tech-
nique has been used by many authors (Chen, 1967; Berde et al. 1979; Geddes
and White, 1979; Otagiri et al. 1979). Fluorescent markers have also
been used to study the displacement of drugs bound to albumin (Chignell,
1969; Sudlow et al. 1975). If a drug being studied does not fluoresce
or if the bound complex has poor fluorescence characteristics,then changes
in the native fluorescence of the albumin may be measured. This fluores-
cence quenching technique in which decreases in tryptophan fluorescence
are monitored has been studied and reviewed by Nieves (1980).
Measuring fluorescence enhancement of drugs upon binding to albumin
is the preferred technique for these studies due to the direct method in
which the binding parameters may be obtained and the low drug to protein
ratios which can be detected. A practical example of this technique will
be illustrated in the following experiment in which the binding of
phenprocoumon to albumin is investigated using fluorescence spectro-
Investigations Into the Binding of Phenprocoumon
to Albumin Using Fluorescence Spectroscopy
The high affinity of the coumarins for human serum albumin (HSA) has
received much attention, particularly since it was realized that their
anticoagulant action was frequently potentiated by the presence of other
drugs in the plasma. More recently, pharmacokinetic and pharmacodynamic
investigations of the enantiomers of warfarin have shown different meta-
bolic fates for the two forms, (Breckenridge et al. 1974) whereas the
enantiomers of phenprocoumon show differences in volumes of distribution
(Jahnchen et al. 1976). The more potent S(-) phenprocoumon is more
strongly bound to HSA than is the R(+) enantiomer. This observation has
been confirmed by a recent limited dialysis investigation (Brown et al.
1977). In the present investigation, the enhanced fluorescence of phen-
procoumon following interaction with HSA is used to investigate the
binding characteristics of the enantiomers to HSA at low drug to protein
ratios and to investigate the possibility of displacement of the drug
from its primary binding site by other commonly prescribed drugs.
Materials and Methods
HSA (lot No. 126C-8070) and BSA (lot No. 17C-8145), crystallized and
lyophilized, and essentially fatty acid free HSA (lot No. 76C-7480) were
obtained from Sigma Chemical Co., St. Louis, MO. Racemic phenprocoumon
(Oraganon, Inc., W. Orange, NJ), S(-) and R(+)-phenprocoumon (Hoffman-
LaRoche, Basel, Switzerland), phenylbutazone and acenocoumarin (Ciba-Geigy
Company, Summit, N J), sodium fenoprofen (Eli Lilly Company, Indiana-
polis, IN), and sodium 2(p-chlorophenoxy)-2-methylpropionate (CPIB) (ICI
LTD., Macclesfield Cheshire, U.K.) were used as supplied by the manufac-
turer. All other materials were reagent grade and all solutions were pre-
pared in deionized water.
All solutions were prepared in 0.1M phosphate buffer of pH 7.4 at
22 1C. HSA concentrations of 0.1 5.0 x 10-5M (M.W. 69,000) were
used. Fluorescence measurements were made using a Perkin-Elmer MPF-44A
fluorescence spectrophotometer (Perkin-Elmer, Norwalk, CT). The fluoro-
metric titrations were carried out as follows: 2.0 ml of the protein solu-
tion of an appropriate concentration in a 10 mm path length cell were
titrated by the successive additions of 2.0 pl volumes of phenprocoumon
solution to give a final drug concentration of 0.1 20 x 106M in the
cell. The fluorescence intensity was measured at 382 nm following exci-
tation at 340 nm. For the experiments in the presence of antagonists, the
fluorescence of phenprocoumon bound to albumin was measured before and
after the addition of the antagonist. The ratio of phenprocoumon to albu-
min was 0.1 and the ratio of competitor to protein ranged from 0.1 to 3.0.
The fluorescence intensity of phenprocoumon was enhanced when bound
to crystalline, essentially fatty acid free HSA, and to bovine serum albu-
min (BSA). The fluorescence peak was also shifted from 386 nm to 382 nm
following interaction with the albumin and excitation at 340 nm. Fluoro-
metric titrations were made by altering the drug concentration at a fixed
albumin concentration as shown in figure 2. For curve a, determined at
high protein to drug ratios, all the drug is bound and this is a measure
of the fluorescence of the drug-albumin complex. At lower albumin concen-
trations a curve is obtained as shown in figure 2, curve b. The fractions
of the drug bound, X, can be determined by using the equation:
OF PHENPROCOUMON x 10 (M)
Plots of relative fluorescence intensities as a
function of phenprocoumon concentration for the
X = p (12)
where fp and fo are the fluorescence intensities of a given concentration
of drug in a solution of low albumin concentration and in solution with-
out albumin. Fb is the fluorescence of the same concentration of fully
bound drug. For this treatment to be valid, the fluorescence intensity
of the bound drug must be a linear function of concentration. This is
the case only when the absorbance of the complex at the exciting wave-
length is low. A correction for this absorption can be made by the method
already described (Otagiri et al. 1979) and is made for all the data when
the absorbance at 340 nm is greater than 0.02.
This method of calculating the fraction bound gives the expected
binding constant for 1:1 complexes, and for higher complexes if the quan-
tum yields associated with the various binding sites are identical. Analy-
sis of the data of fig. 2, curve b, by an iterative least squares tech-
nique, assuming 1:1 complex formation were unsuccessful. Apparently the
binding of more than one ligand molecule to the albumin accounts for the
change in fluorescence. This observance is confirmed by the Job plot
shown in figure 3. Although the peak for the two HSA samples is near 0.5,
the value expected for 1:1 complex formation, the tangents at the origins
are not equal in magnitude and opposite in sign as is necessary for 1:1
complex formation (Job, 1928). The quantum yield associated with the lower
affinity sites appears to be much lower than that associated with the
first site. This means that equation (12) does not strictly apply to the
data obtained at high drug to albumin ratios, and the second binding con-
stant obtained by this method is only an estimate of the method. This
problem is shared by other spectroscopic techniques and is not peculiar
to the fluorescence technique (Perrin et al. 1975). It was found that the
0.2 0.4 0.6 0.8 1.0
Figure 3. A Job plot of relative fluorescence intensities
for the phenprocoumon-serum albumin interaction
S: HSA A : FFA : BSA
data fitted a modified Bjerrum model for two independent sites (Naik
et al. 1975) using the following equation:
P + D Z PD (13)
Ki [P] [D]' K2 [PD] [D] (14)
CD [D] KI[D] + 2KiK2[D]2
n = Cp 1 + Ki[D] + KK2[D]2 (15)
and (2 [2 (i 2- )[D] K + KK (16)
where P, D and PD are the concentrations of unbound albumin, unbound drug,
and the complex respectively. CD and Cp are the total concentrations of
drug and albumin and n is the moles of drug bound per mole of albumin.
A least squares analysis of the linear equation 16 gives the binding data
shown in table 1 for the phenprocoumon-HSA and BSA interactions, and in
table 2 for the enantiomers to HSA interactions. Table 1 also shows the
binding parameters calculated by the Scatchard method (Otagiri, M. et al.
1979) assuming two independent binding sites, enough material was not
available to obtain the data points at the high drug to albumin ratios
necessary for reliable estimates of the binding parameters of the enan-
tiomers by the Scatchard method.
The fluorescence intensity of the phenprocoumon-HSA and -BSA com-
plexes increases as the pH is increased from 6-9 (fig. 4) and is decreased
upon the addition of sodium chloride (fig. 5). The fluorescence intensity
of complexes is lowered by phenylbutazone and acenocoumarin, but not sig-
nificantly altered by clofibrate in fig. 6. The fluorescence is slightly
increased by ibuprofen and fenoprofen.
Table 1. Binding parameters for phenprocoumon-serum albumin complexes.
K1 and K2 are binding constants for two binding sites and r
is the correlation coefficient for data points.
Bjerrum pH 6.5 r pH 7.4 r pH 8.5 r
BSA K(n = 1) 0.9 x 106 0.997
K2(n2 = 1) 7.0 x 104
Ki(ni = 1) 1.0 x 106 0.994 1.1 x 106 0.999 1.3 x 106 0.997
K2(n2 = 1) 2.2 x 104 1.6 x 104 2.0 x 104
Ki(ni = 1) 1.0 x 106 0.998 1.0 x 106 0.998 1.4 x 106 0.998
K2(n2 = 1) 1.9 x 104 3.2 x 104 5.2 x 104
Ki(ni = 1) 8.0 x 10 -
K2(n2 1) 1.2 x 105
Ki(ni = 1) 1.0 x 106 9.5 x 105 1.6 x 106 -
K2(n2 = 1) 2.1 x 104 5.0 x 104 1.0 x 105
FFA K(n = 1) 9.8 x 105 1.9 x 106 1.9 x 106 -
K2FFA 1) 7.4 x 3.8 x 2.8 x 10
K2(n2 = 1) 7.4 x 104 3.8 x 104 2.8 x 104
Table 2. Binding parameters
to HSA at pH 7.4.
sites and r is the
of binding of phenprocoumon and its enantiomers
K1 and K2 are binding constants for two binding
correlation coefficient for data points.
Phenprocoumon K1 K2 r
S (-) 1.3 x 106 2.3 x 104 .998
R (+) 0.8 x 106 1.5 x 104 .998
RS (+) 1.1 x 106 1.6 x 104 .997
S--- x 10O
cU--- 0 -0- 0 v 0
I I I I I
Effect of pH on the fluorescence intensities for
phenprocoumon-serum albumin interactions.
0: phenprocoumon alone 9 : phenprocoumon-
HSA system A : phenprocoumon-BSA system
SODIUM CHLORIDE x 10 (M)
Effect of sodium chloride on fluorescence
intensities for phenprocoumon-serum albumin
interactions at pH 7.4
* : HSA 0 : BSA
CONCENTRATION OF DRUG
Figure 6. Effect of various drugs on
of phenprocoumon in the pr
0 : acenocoumarin
x 10 5 (M)
esence of serum albumin
: CPIB /A: fenoprofen
The fluorescence of phenprocoumon is enhanced following the inter-
action with albumins, allowing a quantitative investigation of the binding
phenomena. Although the enhanced fluorescence is mainly the result of the
interaction at a single site, the data cannot be interpreted in terms of
a 1:1 reaction but gave a better fit with a two site model. A similar
situation has earlier been reported for the interaction of warfarin with
HSA, also investigated by this fluorescence technique (Otagiri et al. 1979).
The binding constants given in Table 1 show that the affinity of phenpro-
coumon for albumin increases as the pH is increased. This increased
binding is also reflected in the increased fluorescence of the complex as
the pH is increased. For the binding at the first site there is little
difference between crystalline and essentially fatty acid free HSA. How-
ever, the binding at the second site of crystalline HSA appears to be pH
independent, whereas with essentially fatty acid free HSA, the binding
constant increases with pH. The N B transition in HSA occurs over the
pH region 6-9 (Leonard et al. 1963; Harmsen et al. 1971) and apparently
the binding to the secondary site in the B form is enhanced following the
treatment to remove the fatty acids. It should be noted that the binding
at the first site is the phenomenon of clinical relevance and also that
the affinity of warfarin for HSA also increases as the pH is increased.
These changes as a function of pH must be due to changes in the protein
rather than the changes in the degree on ionization of the coumarins as
the pKa of phenprocoumon is 4.30 and of warfarin is 5.10 (Otagiri et al.
1978). Data of Table 2 show that the affinity for the S(-) enantiomer is
greater than the affinity for the R(+) enantiomer for both the first and
second site on crystalline HSA. The racemic form, the commercial form of
the drug, being of intermediate behavior. These observations are in
agreement with those reported earlier, however, the binding constants
reported here are significantly higher than those reported earlier (Brown
et al. 1977). In the current investigations much more data are available
at low drug to protein ratios, and so, greater reliability can be placed
on the computed binding constants than those of Brown et al. It is inter-
esting to note that although the binding sites seem to be nonspecific in
that a wide range of acidic drugs seem to share the same binding sites,
the sites do show some stereospecificity. Sodium chloride diminishes the
fluorescence of HSA and BSA complexes with phenprocoumon as shown in Fig.
5. Although chloride has a small effect on the N B transition at the
physiological pH (Harmsen et al. 1971), the effect of chloride appears to
be primarily one of displacement as was observed in the case of warfarin
(Wilting et al. 1979). Fenoprofen and ibuprofen slightly increased the
fluorescence of phenprocoumon-HSA complexes at pH 7.4; in an earlier
study ibuprofen had been shown to increase the fluorescence of warfarin-
HSA complexes (Otagiri et al. 1979). When the enhancement caused by these
nonsteroidal antiinflammatory agents is compared to that caused by pH
changes (fig. 4), it is possible to conclude that the two drugs cause an
increase in the B conformation of albumin. Sudlow, Birkett and Wade have
indicated that warfarin and ibuprofen do not share the same primary binding
site on HSA (Sudlow et al. 1976). Clofibrate did not significantly lower
the fluorescence of the phenprocoumon complexes, whereas acenocoumarin and
phenylbutazone did as shown in fig. 6. These observations are in agreement
with the classification of Sudlow et al. Binding constants of 8 x 105 M~1
for acenocoumarin and 1.2 x 105 M-1 for phenylbutazone can be estimated
from these data (Perrin et al. 1975); however, unpublished observations in
these laboratories suggest that phenylbutazone may have a small effect on
the N B transition.
II. BENOXAPROFEN, A NEW NONSTEROIDAL ANTIINFLAMMATORY DRUG
Salicylates such as aspirin have long been used for their antiinflam-
matory effect in the treatment of rheumatoid arthritis. Large doses
required for therapy tend to cause side effects such as gastric distress,
occult bleeding and tinnitus. There is also a widely known interaction
with the coumarins which precludes their use with many geriatric patients.
Researchers have tried to develop drugs which have as good or better
therapeutic efficacy as the salicylates with a lesser degree of side
effects. Phenylpyrazolone derivatives such as phenylbutazone have been
tried but have a high incidence of side effects. Indole derivatives such
as indomethacin also have a high degree of side effects with chronic use.
A newer class of drugs with proven antiinflammatory activity and per-
haps less severe side effects are the phenylpropionic acid derivatives.
Ibuprofen, fenoprofen, naproxen and benoxaprofen are drugs of this class.
It has been hypothesized that the nonsteroidal antiinflammatory drugs
exert their therapeutic effect by inhibiting prostaglandin synthetase
(Vane, 1971; Vane, 1973). Recently, it has been found that other mecha-
nisms which are independent of prostaglandin formation might be respon-
sible for the efficacy of some of these drugs (Crook et al. 1976; Bonta
et al. 1977). Attallah and Lee (1980) have shown that indomethacin and
salicylates interfere with prostaglandin binding to plasma proteins in
vitro. They suggest that this will accelerate the clearance of prosta-
glandin from the blood.
Benoxaprofen appears to be novel in that its mode of action does not
have anything to do with inhibition of prostaglandin synthesis. Rather,
benoxaprofen interferes with the emigration of leucocytes into inflamma-
tory sites (Cashin et al. 1977; Hutchinson et al. 1977) and this alternate
mode of action may result in decreased ulcerogenic potential.
Pharmacokinetic studies with benoxaprofen have shown that after a
single oral dose of 100 mg, the half-life of elimination was 37 hrs and
the volumes of distribution in the central and peripheral compartments
were 6.8 and 3.2 liters respectively (Chatfield et al. 1977). Therapeutic
blood levels after a single dose reach 13 pg/ml while multiple dosing with
100 mg every 12 hrs allows equilibrium concentrations between 35 and 45
jg/ml (Smith et al. 1977). Plasma protein binding studies find benoxa-
profen to be extensively bound to plasma proteins (99.8%) and no displace-
ment of warfarin or salicylate has been noted (Chatfield et al. 1978).
No major metabolic transformation of benoxaprofen was observed, with the
compound excreted mainly as the ester glucuronide (Chatfield and Green,
It has been reported that patients with severe rheumatoid arthritis
have undergone abnormal serum protein alterations including a resultant
decrease in albumin concentration (Weimer et al. 1968; Clarke et al.
1970). McArthur (1979) reports that this disturbance of the plasma pro-
teins causes an increase in the binding of naturally occurring protective
molecules, with rheumatoid arthritis symptoms appearing when the active
free molecule concentration of these protectantss" decreases. These hypoth-
eses implicate the importance of further study of the protein binding
characteristics of the nonsteroidal antiinflammatory drugs.
Changes in the protein binding properties of a drug such as benoxa-
profen are especially interesting because it is so highly bound, with a
small volume of distribution and long half-life. Minor changes in the
bound fraction of the drug can cause large changes in the free fraction
which may have major physiological implications.
Albumin binding characteristics of benoxaprofen will be studied
using equilibrium dialysis, circular dichroism and flow microcalorimetry.
A high performance liquid chromatographic assay is developed and will aid
as an analytical technique in the equilibrium dialysis experiments. Drug
displacement interactions as well as pH induced conformational changes in
albumin which effect benoxaprofen binding will be examined. These in
vitro studies serve to provide a useful foundation for future clinical
III. A RAPID HIGH PRESSURE LIQUID CHROMATOGRAPHIC
ASSAY OF BENOXAPROFEN IN PLASMA
Benoxaprofen (fig. 7), [2-(4-chlorophenyl)-a-methyl-5-benzoxazole-
acetic acid] has been shown to possess notable antiinflammatory activity,
being several times more potent than phenylbutazone (Dunwell et al. 1975;
Cashin et al. 1977). Two gas chromatographic (Ridolfo et al. 1979; Chat-
field and Woodage, 1978) methods have been used to determine benoxaprofen
in biological fluids, but they are time consuming, involving derivatiza-
tion procedures. The high pressure liquid chromatographic determination
described here is rapid in having only a single extraction step and a
retention time for benoxaprofen of 6.0 min. Sodium warfarin (fig. 7), has
been found to be an excellent internal standard for this procedure since
it has a retention time of 5.0 minutes. The use of sodium warfarin as an
internal standard greatly increases the precision and accuracy of the assay.
Sodium benoxaprofen (903.5 mg free acid per gram) and sodium warfarin,
were generously supplied by Eli Lilly (Indianapolis, IN) and Endo Labora-
tories, Inc. (Garden City, NY), respectively. A stock solution of 100 Pg/
ml benoxaprofen in 0.025 M Na2HP04 was prepared. The internal standard
stock solution contained 100 ug/ml warfarin in 0.025 M Na2HP04. Working
standards were prepared by appropriate dilutions with deionized water of
these stock solutions to give 10 ml samples of 2.5, 5.0, 15, 25, 35 and 50
ug/ml benoxaprofen, each containing 25 pg/ml of the internal standard.
Each plasma sample was spiked with 5 to 100 pl volumes of 100 ug/ml
Figure 7. Structures of Benoxaprofen and Warfarin
benoxaprofen stock solution to give final concentrations of 0.5, 1, 3, 5,
7 and 10 ig/ml.
Ether, phosphoric acid, hydrochloric acid and dibasic sodium phos-
phate were all analytical grade and supplied by Mallinckrodt Inc. (St.
Louis, MO). HPLC grade acetonitrile and methanol were obtained from Bur-
dick and Jackson (Muskegon, MI). All solvents including deionized water,
were filtered through 0.45 pm filters (Millipore Corp., Bedford, MA) prior
to use in the liquid chromatograph. Samples analyses were carried out on
a model 5000 liquid chromatograph equipped with a 10 pl manual loop injec-
tor and a Vari-Chrom variable wavelength UV detector (Varian Assoc., Walnut
Creek, CA). An alkylnitrile column, MicroPak CN-10 (Varian Assoc.) ofmedi-
um polarity was used. The mobile phase consisted of 25% acetonitrile in
water (pH 2.5 with H3P04). At ambient operating temperature, the flow rate
was 1.8 ml/min at a pressure of 160 atm. The column effluent was monitored
continuously at 309 nm. All UV spectra of stock solutions were taken on
a Cary model 219 UV spectrophotometer (Varian Assoc.).
Plasma samples of 1.0 ml, to which 5 pg of internal standard had been
added, are acidified with 400 ul of 1N HC1 and mixed thoroughly by hand
shaking. After the addition of 8 ml of ether, the mixture is vortexed for
30 sec. and then placed on a test tube rotator for 10 min. The sample is
centrifuged at 3600 rpm for 10 min. and then the ether layer is transferred
to another test tube and evaporated to dryness under a nitrogen stream.
The residue is reconstituted in 200 pl of methanol and a 10 pl aliquot
injected into the column.
Six point standard curves were obtained from the stock solutions.
Spiked plasma samples were compared to the stock solutions to obtain recovery
percentages. Data were obtained by plotting peak height ratios versus
benoxaprofen concentration and a linear least squares regression program
Figure 8 shows typical chromatograms for control plasma, plasma with
benoxaprofen and plasma with benoxaprofen and warfarin. The retention
times for warfarin and benoxaprofen were 5.0 and 6.0 min respectively.
No interfering peaks in the control plasma were noted.
Table 3 shows the peak height ratios of benoxaprofen to 5 ug warfarin
in the 0.5-10 pg range of benoxaprofen extracted from human plasma. Each
value represents the average of three samples. A linear regression curve
obtained from the data has the equation y=0.257X + 0.041 with r2=0.9996,
where r2 is the coefficient of determination.
Extraction recoveries were determined by injecting known concentrations
of benoxaprofen with internal standard and comparing them to the extracted
plasma samples (Table 4). The extraction recovery was 90.6 6.8% for 18
Stability tests were performed by adding 0.5, 1, 3, 5, 7 and 10 pg
of benoxaprofen to samples of plasma and freezing them. Analyses were
performed at 1, 4, 9 and 14 days, the results of which are listed in table
5. These findings show the samples are stable for this time period.
Two GC methods (Chatfield and Woodage, 1978; Ridolfo et al. 1979)
have been used to determine benoxaprofen levels but no liquid chromato-
graphic method has previously been reported. Radiolabelled [14C] benoxa-
profen has also been used in biological fluid assays (Chatfield and Woodage,
1978; Chatfield et al. 1978; Chatfield and Green, 1978). The extraction
10 3 6
Figure 8. Liquid chromatograms of a) control human plasma, b) human
plasma containing 5 ug benoxaprofen and c) human plasma
containing 5 pg warfarin and 5 pg benoxaprofen.
Table 3. Peak height ratios for benoxaprofen plasma calibration curve.
Peak Height Ratio
Table 4. Benoxaprofen recovery from human plasma. (N=3)
Amount Added Amount Found Recovery, %
(pg) (pg) (Mean SD)
0.5 0.478 0.015 95.5 2.9
1.0 0.942 0.088 94.2 8.8
3.0 2.820 0.120 94.0 4.0
5.0 4.235 0.205 84.7 4.1
7.0 6.069 0.483 86.7 6.9
10.0 8.870 0.270 88.7 2.7
Ave. 90.6 6.8
Table 5. Effect of frozen storage on benoxaprofen stability in human
(pg) 0 1 4 9 14
recovery obtained, 90.6 6.8%, is similar to the extraction efficiency
obtained in the radiolabelled assay (Chatfield et al. 1978).
Though not structurally similar to benoxaprofen, warfarin has proved
to be an excellent internal standard. Retention times were one minute
apart, namely 5.0 min for warfarin and 6.0 min for benoxaprofen. Peak
height ratios exhibited excellent linearity in the concentration range
studied (0.5-10 pg/ml). A variable wavelength UV detector was used to
monitor benoxaprofen levels at 309 nm, its absorbance maximum in the sol-
vent system used. Warfarin has an absorbance maximum of 308 nm in the
same system. The signal-to-noise ratios obtained were on the order of
2-3 times greater than when a 254 nm filter detector was used.
Stability studies have shown that benoxaprofen was stable for at
least two weeks in frozen plasma. It was found, however, that stock
solutions of benoxaprofen were not stable at room temperature during this
time interval. Preliminary studies in this laboratory have shown that
aqueous solutions of benoxaprofen undergo extensive photodegradation.
This is especially apparent when a sample is placed in an intense UV light
source, as in a fluorimeter. New stock solutions were prepared when neces-
sary and concentrations were confirmed by UV spectra.
The lowest level of benoxaprofen detected in this study was 0.5 pg.
Lower levels can be detected by decreasing the reconstituting volume or
using a higher detector sensitivity. However, the concentration range
studied was within the expected therapeutic range (Chatfield and Green,
1978; Ridolfo et al. 1979).
IV. BINDING PARAMETERS FOR BENOXAPROFEN-ALBUMIN COMPLEXES
Equilibrium dialysis and microcalorimetry will be used as techniques
to study the binding parameters of benoxaprofen-albumin complexes. The
stoichiometry and strength of binding (n,K) as well as the thermodynamic
binding parameters (AH,AG,LS) will be determined.
Benoxaprofen has been found to be unstable in the spectrofluorimeter.
The intense light source causes photodegradation to occur, whereas in
aqueous solutions a white precipitate forms instantaneously. This pre-
cludes the use of fluorescence spectroscopy in studying the binding prop-
erties of benoxaprofen.
Materials and Methods
Human serum albumin, fraction V (lot No. 30F-02271), was obtained
from Sigma Chemical Co. (St. Louis, MO). Sodium benoxaprofen (903.5 mg
free acid per gram) was generously supplied by Eli Lilly and Co. (Indiana-
polis, IN). Monobasic, dibasic sodium phosphate and sodium chloride,
analytical reagent grade, were obtained from Mallinckrodt, Inc. (St.Louis,
MO). Dowex 50W-X8 cation exchange resin and 1-X8 anion exchange resin
were supplied by J.T. Baker Chemical Co. (Phillipsburg, NJ). Deionized
water was used in all experiments.
Prior to all experiments, human serum albumin (HSA) is deionized by
running a solution through a column containing cationic and anionic exchange
resins. The resins are prepared by adjusting the pH of the cationic and
anionic forms to 1 and 12, respectively. The pH is then adjusted to 4.5
for the cationic form and 7.5 for the anionic form by rinsing with de-
ionized water. Anionic and cationic resin are then added to a column
where they are again rinsed with water. An albumin solution is run through
the column and the final pH is adjusted to 7.4 in 0.1 M phosphate buffer.
Microcalorimetry is used to obtain a Job's plot for the determina-
tion of the stoichiometry of the benoxaprofen-HSA complex. HSA solutions
of 0-5.0 x 10-4 M are pumped into one half of the mixing cell while benoxa-
profen solutions of 5.0 x 10-4 -0 M are pumped into the other side. After
1:1 dilution in the mixing cell, the total final concentration for the
benoxaprofen-HSA complex is 2.5 x 10-4 M. The heats of binding in pV vs.
mole fraction benoxaprofen are then recorded.
A titration of HSA with benoxaprofen is done calorimetrically. HSA
at a constant concentration of 2.56 x 10-4 M is pumped into one side of
the mixing cell while benoxaprofen in concentrations of 0-5.12 x 10-4 M
is pumped into the other side. The binding affinity for the first site
can be obtained from this method.
Equilibrium dialysis is used to obtain a Scatchard plot from which n
and K can be determined. HSA at a concentration of 2.0 x 10-4 M is dia-
lyzed vs. benoxaprofen in concentrations of 2.0 x 10-5 8.0 x 10-4 M.
The system is rotated at 4 rpm in a water bath at 250C for 12 hr. Both
sides are analyzed for benoxaprofen using the previously described HPLC
Results and Discussion
Rather than use the charcoal treatment devised by Chen (1967) to de-
ionize albumin, a less harsh resin deionization is used. Perrin and
Vallner (1975) found that major differences in albumin preparations might
exist due to impurities arising from commercial manufacturing techniques.
Hardee et al. (1979) found that different heats of binding for the
salicylic acid-albumin interaction would be measured depending upon the
amount of ionic impurities present. It was later found by Chen and Koes-
ter (1980) that charcoal treatment causes major changes in the fluores-
cence characteristics of albumins and even after this treatment, major
changes in albumin samples may exist. In this experiment, albumin is
passively deionized to remove ionic impurities such as chloride, citrate,
lactate and various cations and fatty acids. Minimal perturbation is
made to the albumin. For microcalorimetric work, this has proven to be
an acceptable technique (Hardee et al. 1979).
A Job's plot (Job 1928) for the benoxaprofen-HSA interaction is shown
in fig. 9. An inflection point between 0.75 and 0.80 is a strong indica-
tion that at least three moles of benoxaprofen bind per mole of albumin.
Figure 10 illustrates a Scatchard plot (Scatchard, 1949) for the beno-
xaprofen-HSA interaction. Each data point represents the average of two
determinations from an equilibrium dialysis experiment. Both the bound
and free sides are analyzed since preliminary studies have shown that
5-6% of the drug is bound to the membrane. Computer fitting the data to
a three site model utilizes equation 17
nz KI[D] n2 K2[D] n3 K3[D]
1 + KI[D] 1 + K2T] 1 + K3[D] (17)
where r is the number of moles of benoxaprofen bound per mole of albumin,
[D] is the free benoxaprofen concentration and Ki and ni are the binding
constants and number of sites of the ith class. The binding constant
obtained for the first site (K1 = 6.02 x 106 L/mole) is high but not out
of the order for other drugs in its class as Ki for ibuprofen was reported
to be 2.73 x 106 L/mole (Whitham et al. 1979). A good graphical estimate
of n2 and n3 is not obtained since solubility problems with benoxaprofen
preclude getting higher values of r. Using lower HSA concentrations might
).4 Q6 0.8
MDLE FRACTION OF BENOXAPROFEN
Job's plot for the benoxaprofen-HSA interaction.
Final total concentrations = 2.50 x 104 M.
25 K1 = 6.02 x 106 L/Mole
N1 = 1.0
K2 = 6.93 x 104
N2 = 2.0
K3 = 1.08 x 104
20 N3 = 3.0
0,5 1.0 1,5 2,0 2,5 3,0 3,5
Figure Scatchard plot of the benoxaprofen-HSA interaction.
Equilibrium dialysis experiment.
0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 10. Scatchard plot of the benoxaprofen-HSA interaction.
Equilibrium dialysis experiment.
be a way of getting around this problem but a concentration of 2.0 x 10-4M
HSA is about the minimum necessary to achieve accuracy and precision with
The association constant for the binding site of physiological signi-
ficance (Kj) obtained from the Scatchard plot is confirmed by a titration
of HSA with benoxaprofen done on the ricrocalorimeter. Figure 1 illus-
trates this titration curve where the heat flux generated is proportional
to the amount of benoxaprofen bound to albumin. More than one site con-
tributes to the heat flux, as is indicated by the linearity. If only one
site of high affinity was involved, the curve would be expected to plateau
soon after that site was saturated. Solubility problems with benoxaprofen
and the multiple number of binding sites limit attainment of a high enough
concentration to saturate the binding sites. The first binding constant
can therefore be calculated by mathematical manipulation of the data.
If it is assumed that at drug to protein ratios below 0.5 all of the
benoxaprofen is bound to the first site, the slope of a line drawn tan-
gent to these points and through the origin is represented by the dashed
line in fig. 11 and will give the value of the maximum heat produced by
the formation of one mole of bound complex. This heat is equivalent to
the plateau value if one could be obtained for a 1:1 complex and its units
are microvolts.mole-l.liter. A linear regression program is used to find
the value of 'Vmax and K1 is calculated from equation 18 where a and b are
the initial concentrations of reactants
1 = (a-c)(b-c) (18)
and C is given by
representing the concentration of the 1:1 complex formed.
(An) Xnd IV3-
Using this method a value of 5.05 x 106 L/mole for Ki is obtained,
which is in excellent agreement with the value 6.02 x 106 L/mole obtained
for KI from the Scatchard plot. Equation 20 allows the heat of the reac-
tion in joules/mole to be calculated,
AH (joules/mole) = uvmax x .057 x total flow rate (20)
where 0.057 is the conversion factor for this microcalorimeter in uv/pw.
A value of -21,000 joules/mole is obtained for the heat of formation of
one mole of complex at 250C. From equations 21 and 22, AG and AS can
also be calculated and their values are -38.242 joules/mole and 57.9 joules/
mole K, respectively.
AG = -RT In K (21)
AS = AH AG (22)
Binding of Benoxaprofen to Hemodialyzed Plasma
Studies by Shoeman and Azarnoff(1972) and Boobis (1977) have suggested
that a change in the composition of albumin is a major factor in the
decrease of drugs binding to the plasma of uremic patients. Dirnhuber
and Shutz (1948) have found that in urea solutions above pH 6 cyanate is
in equilibrium with urea. A recent study by Erill et al. (1980) has
shown that cyanate can carbamylate plasma proteins and this carbamylation
results in decreased binding of acidic drugs.
Other factors such as hypoalbuminemia and increased levels of fatty
acids may also play a role in decreased acidic drug binding. It is there-
fore of interest to study the binding of benoxaprofen to uremic plasma
and to note the effects hemodialysis may have on the binding.
Two middle aged male uremic patients have 10 ml of whole blood drawn
at 0, 2 and 4 hrs of hemodialysis. The blood is immediately centrifuged
for 10 min and 4 ml of plasma is obtained. The plasma is spiked to give
a 40 pg/ml concentration of benoxaprofen. An equilibrium dialysis experi-
ment is devised so that 1.8 ml of plasma is dialyzed vs. 1.8 ml Sorenson
buffer (Schumacher, 1966) at pH 7.4 for 14 hr at 370C. Two samples from
each patient at each time are run. Normal human plasma is treated under
the same conditions for use as a blank. The previously described HPLC
procedure is used to assay the free concentration of benoxaprofen.
Results and Discussion
Table 6 gives the free concentrations of benoxaprofen after equilib-
riur has been reached. Due to the very high degree of binding of benoxa-
profen, the differences between the free concentrations in normal plasma
as compared to that in uremic plasma can at best be a result of inter-
Of significance is the decrease in free concentration of the drug in
uremic patients after 4 hrs of hemodialysis. There is not much difference
in the free concentrations between 0 and 2 hrs but after 4 hrs, major
decreases in the free concentrations are apparent. Again, interindividual
differences arising from the different types and levels of impurities in
each patient are probably responsible for different free benoxaprofen con-
centrations reported for the two patients. However, the trend is clear
in that the increase in the amount of bound benoxaprofen is a result of
the removal of competitive impurities from the plasma by hemodialysis.
Removal of urea might also lead to a decrease in the formulation of the
carbamylated products previously described. In any case, the effects of
hemodialysis upon patients receiving benoxaprofen should be carefully monitored.
Table 6. Effects of Hemodialysis on Benoxaprofen Binding to Human Plasma.
Patient Time (hr) Free Concentration (pg/ml) X
A 0 0.206 0.237
2 0.284 0.243
4 0.134 0.102
B 0 0.121 0.101
2 0.099 0.114
4 0.030 0.0204
Normal Plasma 0.130
V. CONFORMATIONAL CHANGES IN HUMAN SERUM ALBUMIN
Proteins are high molecular weight copolymers of about twenty amino
acids linked together to form a polypeptide chain. Their length, amino
acid composition and sequence are genetically determined. Diversity is
introduced by residues attached to the a-carbon of each monomeric amino
acid unit. The amino acid residues are classified according to the pre-
sence of polar and/or apolar groups. There are six apolar aliphatic resi-
dues (glycine, alanine, valine, leucine, isoleucine and methionine); four
apolar heterocyclic and aromatic residues prolinee, phenylalanine, trypto-
phan and tyrosine); five polar aliphatic residues serinee, threonine, cys-
teine, asparagine and glutamine); and five ionizable residues (aspartic
acid, glutamic acid, lysine, arginine and histidine). As will be evidenced,
the last five ionizable amino acids (fig. 12) will be most relevant to the
conformational changes discussed later.
In an aqueous environment, hydrophobic residues entropically interact
with each other to reduce water contact while ionizable and polar residues
interact with water resulting in an enthalpy change that reduces AG. Thus
apolar residues tend to be on the inside of a protein molecule while charged
residues exist on the outer surfaces. The protein chain will assume the con-
formation which best satisfies the requirements of minimization of exposure
of hydrophobic residues to the aqueous environment, forming what is described
as a hydrophobic core. This leads to maximization of intramolecular inter-
actions such as hydrogen bonding, Van der Waals forces and ion pair forma-
tion. There is also a maximization of interactions between charged and
NH3- CH- CH2- CH2- CH2- C- oo
NH2- Cj- NH- CH2- CH2- CH2- q- C00
C- CH2- C- COO0
HN NH NHI
CHM.- C- CO0-
LYSINE PKA 0.53
ARGININE PKAl 12.48
HISTIDINE PKA- 6.00
ASPARTIC ACID PKAZ 3.86
C- CH2- CP2- C- COo
GLUTAMIC ACID PKA- 4.25
Figure 12. Ionizable Amino Acids.
polar residues with the solvent molecule described as "polar shell" forma-
tion. The resulting compact structure is characterized by the presence or
absence of chain segments in helical, B-sheet or "random coil" configura-
tions as well as B-turns.
Three-dimensional structure of a protein is a reflection of the bal-
ance of enthalpic and entropic forces of interaction between the amino
acid monomeric units (backbone peptide units and side chain residues) with
one another and with the solvent molecules. This results in a minimum AG.
The structure is stabilized by cooperative noncovalent interactions
(Brandts, 1969; Tanford, 1970). Small local changes in structure can be
readily transmitted by a "chain reaction" to the rest of the molecule
leading to the loss of native structure of the entire molecule.
The charge state of the protein is determined by the total number of
ionizable residues, accessibility to the solvent, pKa values, and solution
pH. pKa's are affected by inductive effects, temperature, chemical nature
of the solvent dielectricc constant), and ionic strength of the solvent.
A major source of heterogeneity in serum albumin involves the presence
of dimer and even higher oligomeric forms. The presence of dimers was
first indicated by ultracentrifuge work and the more advanced work done by
Pedersen (Pedersen, 1962) with Sephadex chromatography clearly resolved the
monomer, dimer, trimer and tetramer forms.
Dimers and trimers are the result of the manufacturing process and
are not in the bloodstream (Andersson, 1966). Williams and Foster found
dimerization to be most probable at a pH of 3.3 and result from the forma-
tion of disulfide linkages through oxidative reactions (Williams and Fos-
ter, 1960). Cu+2 ion was also found to increase dimerization; and in fact
it is theorized that Cu+2 actually acts as a catalyst in the dimerization.
Thiol reagents are usually used to reduce the dimeric forms of albumin.
Oxidation with ferricyanide in 4 M guanidine hydrochloride at pH 8 is
used to artificially prepare the dimer.
Dimers usually have less affinity for drug binding than monomeric
albumin. A commercial albumin preparation has about 6% dimer content.
Optical rotary dispersion studies have shown that there is a large
increase in the hydrodynamic volume of the albumin molecule at low pH.
This change occurs from coulombic repulsion incurred by the large positive
charge. These studies have also shown that the change occurring is a
helix-coil transition with a reduction of apparent helical structure from
51% to 35%. Thus, the pH expansion can be viewed as moving apart of the
disulfide loops due to electrostatic forces, and the accompanying break-
down of the helical structure (Sogami and Foster, 1968).
The two forms of albumin in equilibrium at low pH are N for native
and F for fast migrating. Physically, the N form is extremely soluble,
while the F form is insoluble in 3 M KC1 (Rachinsky and Foster, 1957).
This property can be attributed to the hydrophobic regions. This hypothe-
sis is supported by a study done by Wishnia and Pinder (Wishnia and Pinder,
1964), which showed that the N form binds substantial amounts of alkanes
such as pentane; this ability is lost in the F form.
Optical rotation has been a useful method of studying the N-F tran-
sition. At 233 nm, the albumin spectrum shows a deep trough in rotation
which is common to other proteins of high helical content. A smooth
decrease in intensity is displayed until a pH of 3.7 where a slight pla-
teau forms, then a decrease again until a plateau is reached at a pH of
2.9. These two steps are now commonly referred to as the N-F' and F-F-
The N-F transition can be described as the opening of the hydrophilic
interface, while the second step involves the opening of the hydrophobic
regions. Absorption spectroscopy has shown that there is a slight blue
shift in the spectrum at low pH (Williams and Foster, 1959). Perhaps this
is due to tyrosyl residues which are in a hydrophobic environment in the
N form, but exposed to an aqueous environment at the low pH's. This
results in an increase in polarizability, indicative of tryptophyl resi-
dues which have shown fluorescence intensity to diminish with decreasing
pH (Chen, 1966; Halfman and Nishida, 1971; Rudolph et al. 1975). A slight
blue shift is noted with the decrease in intensity. Sogami et al. have
concluded that this shift and intensity change accompany the N F tran-
sition. The decrease in fluorescence intensity can be due to quenching of
chromophores from protonated carboxyl groups (Sogami et al. 1973).
Another important aspect of the N-F transition concerns the hydrogen
binding properties of the protein. It has been hypothesized that up to
40% of the carboxylate groups are "masked" in the N state so that they
will not protonate (Stroupe and Foster, 1973). Though it is not likely
that charged carboxylate groups cannot be isolated from solvent molecules
without neutralization from positive charges, ion pairing may occur. The
partners may be protonated amino groups, imidazolium groups, or guanidin-
ium groups. The N-F transition involves the rupture of a number of these
Though much work has been done on the N-F transition, its relevance
to physiological conditions is doubtful. However, there is another tran-
sition which takes place much closer to the pH range encountered in blood.
Leonard and Foster (Leonard et al. 1963), using optical rotation studies
at 313 nm, first discovered a transition in the pH range of 7-9. This
wavelength was chosen because it monitors change in tertiary structure,
perhaps primarily alterations in aromatic chromophores. They noted that
the magnitude of this transition was smaller than that for the N-F transi-
tion and hypothesized that there was even less loss of helical structure
in the higher pH transition. This transition is now commonly referred to
as the N-B or neutral transition.
Histidine, whose imidazole residues have a pKa of 6.4 to 7.0, seems
to be involvedin this transition, and it is hypothesized than ten imida-
zolium residues are "hidden" in the N form and become available in the B
form (Harmsen et al. 1971). At physiological pH, albumin will exist in
two forms. Organs such as the liver and certain inflamed tissues have a
slightly lower pH; therefore, changes in protein conformation at these
sites resulting in differences in drug binding must be taken into account.
It is interesting to note that calcium ions in physiological concen-
trations favor the N-B transition. This has been confirmed by optical
rotation (Zurawski and Foster, 1974) and spectrophotometric studies (Harm-
sen et al. 1971; Wilting et al. 1979). With physiological levels of cal-
cium ion present there are nearly equal amounts of the two conformers since
the transition is pushed to the N form at physiological pH. Calcium ion
binding can be thought of as pH dependent (Katz and Klotz, 1953) and it
is possible that calcium ions compete with imidazolium residues for car-
boxylate binding sites (Pedersen, 1972a) supporting one reason for differ-
ences in binding constants for various drugs.
It has been found that calcium ions bind to albumin with increasing
affinity in the pH range 5-9 (Pedersen, 1972b). These studies have also
shown that calcium ions compete reversibly with twelve of the sixteen
imidazole groups on HSA with the calcium binding increasing over this pH
range. In a phosphate buffer at a pH of 6.9, the presence of calcium ions
results in 50% of the albumin existing in the B form. When no calcium
ions are present, 50% of the albumin is in the B form at a pH of 7.4.
Circular dichroism studies have shown that 80% of human albumin is in the
B form when calcium ions are present, but only 50% in the absence, at a
pH of 7.4.
Work done by Wilting et al. (Wilting et al. 1980a) has shown that high
concentrations of chloride ion (0.1 M) also alter the N-B transition.
He has shown that in warfarin binding experiments in the presence of cal-
cium and/or chloride ions, the N B transition is altered. Albumin is
"pushed" to the B form in the presence of these ions and in this form it
has a higher affinity for warfarin.
Physiologically the importance of this phenomenon is very important
as is the case with the anticoagulant drug which is mainly metabolized in
the liver. The pericellular pH in the liver is slightly lower than in the
blood, and the free fractions of warfarin available for metabolism and phar-
macological action are not accurately estimated from plasma sample deter-
minations. Slight changes in blood pH can be expected to alter free war-
farin concentrations dramatically. A shift in pH from 7.4 to 7.0 almost
doubles the free warfarin concentration in the presence of calcium ions.
These changes can be expected when a patient has diabetes mellitus or
severe renal failure.
The mechanics of the N-B transition are not as clearly classified as
with the N-F transition. The amount of helical structure involved in the
N-B transition is small and is almost unchanged by temperature (Wallevik,
1972). The apparent net loss of a-helix structure in the unfolding between
pH 7 and 9 is 2.5%, in the N-F transformation it is 8%. It could also be
concluded that the N-B transition causes a pK shift of imidazole groups
with protons released in the neutral regions. The highest pK was found
in the low pH conformation suggesting that several histidyl residues are
involved in salt bridges. In the presence of calcium ions, the affinity
of albumin for protons decreases, resulting in a shift to the neutral state.
VI. THE EFFECTS OF pH, CALCIUM AND CHLORIDE IONS ON THE
BINDING OF BENOXAPROFEN TO HUMAN SERUM ALBUMIN
Most of the studies of the pH induced N-B transitions in HSA have been
done with the coumarins. This class of drugs usually has good fluores-
cent properties and also gives a strong induced CD signal. Wilting et al.
(1979;1980a;1980b) and Otagiri et al. (1978) found that warfarin binds with
increasing affinity to HSA in the pH range from 6 to 9 and it was shown
that calcium and chloride effect this binding. Similar results were
obtained with phenprocoumon in Chapter I of this work.
Experiments in this chapter were designed to study the effects of pH,
calcium and chloride ions on the binding of benoxaprofen to HSA. Equilib-
rium dialysis and circular dichorism are the techniques used to perform
Materials and Methods
Human serum albumin, fraction V (lot No. 30F-02271) was obtained from
Sigma Chemical Co. (St. Louis, MO). Sodium benoxaprofen (903.5 mg free
acid per gram) was supplied by Eli Lilly (Indianapolis, IN). Monobasic,
dibasic sodium phosphate and sodium chloride, analytical reagent grade,
were obtained from Mallinckrodt, Inc. (St. Louis, MO). Calcium chloride
dihydrate, A.C.S. grade, was obtained from Fisher Scientific Co. (FairLawn,
NJ). Dowex 50W-X8 cation exchange resin and 1-X8 anion exchange resin
were supplied by J.T. Baker Chemical Co. (Phillipsburg, NJ).
Albumin was deionized as previously described and its molecular weight
was taken to be 66,500 g/mole. Concentrations were determined by UV
spectroscopy (e = 4.123 x 104 Z/mole-cm). All solutions were prepared
using deionized water.
Free concentrations of benoxaprofen were obtained by means of equilib-
rium dialysis using a Dianorm equilibrium dialyzer (Diachema, A.G.)
Rischlikon, Switzerland) with teflon cells of 10 ml total volume. Dialy-
sis membranes of hydrated cellulose were used with a molecular weight cut-
off of 10,000 (Diachema type 10.16). The samples were dialyzed at 4 rpm
for 12 hr at 25C. Benoxaprofen was analyzed using the previously described
Circular dichroic (C.D.) measurements were made using a Jasco model
J-500 spectropolarimeter (Jasco International Co. LTD, Tokyo, Japan).
The extrinsic benoxaprofen-albumin signal was measured at a slit width
of 1 nm, wavelength expansion of 5 nm cm 1, chart speed 1 cm min-1, time
constant 32 sec and a sensitivity of 0.5 or 1.0 m cm-1. Pathlengths of
2, 5 and 10 mm were used. Observed ellipticities (eobs) are the actual
C.D. spectraof the benoxaprofen-albumin complex while the induced ellip-
ticity is the observed ellipticity of the complex minus the ellipticity
of albumin alone at the corresponding wavelength. The dynode voltage was
kept below 0.5 in all experiments.
Results and Discussion
Prior to beginning the study on the pH dependence of benoxaprofen
binding, it was necessary to determine if a constant ionic strength be
maintained when workingin the region of the N-B transition. Mixing dif-
ferent ratios of 0.1 M monobasic and dibasic phosphate buffer would be
used to adjust the pH but this would cause the ionic strength to vary.
Equilibrium dialysis was used to study the binding of a 1 to 1 ratio of
benoxaprofen to albumin in phosphate buffer concentrations of 0.05, 0.10,
0.20 and 0.30 M. From table 7 it can be shown that there would only be a
Table 7. Dialysis data. Effect of ionic strength on the binding of
benoxaprofen to HSA.
Peak Height Ratios (Benoxaprofen/Warfarin)
Phosphate Buffer Conc. (M) X S.D. (N=5)
0.05 1.50 0.09
0.10 1.34 0.11
0.20 1.31 0.03
0.30 1.26 0.07
[Benoxaprofen] = 2.5 x 10-4 M
[HSA] = 2.5 x 10-4 M
notable ionic strength contribution if the phosphate buffer concentration
went below 0.1 M. Therefore, adjusting the pH with 0.1 M phosphate buffer
should not effect benoxaprofen binding.
Benoxaprofen, when bound to HSA,gives rise to an extrinsic Cotton
effect with a positive maximum at 303 nm and a negative minimum at 333 nm.
These wavelengths can shift as is shown by the titration at pH 7.4 in fig.
13. The isosbestic point at 328 nm is maintained throughout the titration
at low drug to protein ratios, suggestive of a single binding site of high
affinity. When the drug to protein ratios exceed 1 to 1 the isosbestic
point is lost and at 3 to 1 the positive maximum shifts to shorter wave-
lengths and decreases in magnitude. This is easier to see when the same
data is plotted as induced ellipticities in fig. 14.
The sigmoidal character of this curve at low drug to protein ratios is
indicative of a secondary binding site of higherinducedellipticity or a
cooperative effect between the primary and secondary sites (Chignell, 1969).
Drug displacement studies done later in this work will tend to support the
first hypothesis. At a drug to protein ratio greater than 3 to 1 the ellip-
ticity decreases. This evidence of at least three binding sites correlates
with the datain the Scatchard and Job's plots (figs. 9 and 10) previously
described. Changes in the sign of the ellipticity upon increasing the drug
to protein ratio have also been observed with indomethacin (Ekman et al.
1980). Here, it was found that the first two binding sites give rise to a
negative ellipticity and the third site has a positive ellipticity.
The binding of various aromatic propionic derivatives has been studied
by Perrin (1973) and as with fenoprofen, it appears the extrinsic Cotton
effects of the benoxaprofen-albumin complex arise as a result of hydropho-
bic binding of the aromatic rings and hydrogen bonding of the carbonyl and
2 300 310 320 330 30 350 3
Figure 13. Observed ellipticity of benoxaprofen-HSA complex
titration at pH 7.4. 0.5 cm. cell.
3.0 4.0 5.0
Induced ellipticity of benoxaprofen-HSA complex
titration at pH 7.4 in 0.1 M phosphate buffer.
[HSA] = 5.0 x 10-5 M. 0.5 cm. cell at 300 nm.
ether oxygen to albumin. No Cotton effect is generated by benoxaprofen
itself. Studies done on a molecule similar to benoxaprofen, [2-(4-chloro-
phenyl)-4-oxazole]acetic acid, indicated that the molecule failed to
exhibit an extrinsic Cotton effect with albumin. It was postulated that
the absorption maximum of this compound was at too low a wavelength (281
nm) to distinguish the extrinsic ellipticity from the intrinsic ellipticity
of the albumin (Mitchell and Rosen, 1978). Benoxaprofen has one more
phenyl ring than this compound which gives it an absorption maximum of
309 nm. This, plus its added planar rigidity, probably gives rise to the
observable Cotton effect of benoxaprofen.
Albumin was titrated with benoxaprofen at pH 6.5 (fig. 15). It is
in the N form at the pH and the magnitude of the observed ellipticity is
greater than at pH 7.4. Since the pKa of benoxaprofen is between 3 and 4,
the changes in the nature of the binding are due to changes in the albumin
and not benoxaprofen. There also seems to be a loss in the distinction of
the isosbestic point in this conformation. Figure 16 shows the data at pH
6.5 plotted as induced ellipticities and there does seem to be a loss in
distinction between the primary and secondary sites as is noted by decreased
sigmoidal character of the curve at low drug to protein ratios. At ratios
higher than 3 to 1 the decrease in ellipticity is still observed.
At pH 9.2 the titration had to be done in a 0.2 cm cell as there was
too high a dynode voltage which was caused by the increased ionic strength
of the phosphate buffer (fig. 17). The induced ellipticities of these data
in fig. 18 exhibit the interesting observation that at drug to protein
ratios greater than 3 to 1 the ellipticity keeps increasing rather than
decreasing as was seen at pH 6.5 and 7.4. Much more sigmoidal character
is noted in this curve. It is clear from these observations that in the
B form, the character of albumin changes so that there is a greater
Observed ellipticity of benoxaprofen-HSA complex
titration at pH 6.5. 0.5 cm. cell.
1.0 2.0 3.0 4,0 5,0 6.0
Figure 16. Induced ellipticity of benoxaprofen-HSA
pH 6.5 in 0.1 M phosphate buffer. [HSA]
0.5 cm. cell at 300 nm.
complex titration at
= 5.0 x 10 5 M.
? +1 -
SI I I I II
290 300 310 320 330 340 350 360
Figure 17. Observed ellipticity of benoxaprofen-HSA complex
titration at pH 9.2. 0.2 cm. cell.
1.0 2,0 3,0 4.0
Induced ellipticity of benoxaprofen-HSA complex titration
at pH 9.2 in 0.1 M phosphate buffer. [HSA] = 5.0 x 105 M.
0.2 cm. cell at 300 nm.
distinction between the primary and secondary sites for benoxaprofen.
The unfolding of the albumin in this conformation (Wallevik, 1972) and the
change in the ionization states of the imidazole groups (Harmsen et al.
1971) might favor a more rigid fit of the benoxaprofen at its third binding
site resulting in the change of sign to a positive ellipticity.
The effects of chloride and calcium ions on the N-B transition of
benoxaprofen-HSA complexes are studied with circular dichroism and equi-
librium dialysis. From fig. 19 it can be seen that the observed ellipti-
city of the benoxaprofen-HSA complex in 0.1 M phosphate buffer decreases
with increasing pH. This is contrary to the effects observed with warfarin
where the ellipticity increased with increasing pH. Under the conditions
in which this experiment was performed, there is no change in HSA with pH
so the observed signals arise strictly from the benoxaprofen-albumin complex.
The fact that the benoxaprofen-albumin and warfarin-albumin complexes have
ellipticities of opposite sign in the N-B transition is perhaps related
to the nature of their respective binding sites. Later, it will be proven
that they actually bind to two physically separate sites on HSA and the
different environments of these sites can be the reason for the opposite
ellipticities in the N-B transition.
Figure 20illustrates the effect sodium chloride has on decreasing the
ellipticity of the benoxaprofen-albumin complex. The mechanism involved
may be one of direct displacement of benoxaprofen, as the molar ratio of
chloride to benoxaprofen is 2000 to 1. It is also possible that chloride
"pushes" albumin to the B form, which when completed with benoxaprofen, has
a greater negative ellipticity. Calcium ion also has the same effect (fig.
21) but of a greater magnitude than with chloride ion.
The different effects of phosphate, chloride and calcium ions on the
N-B transition can be seen more clearly when they are plotted as induced
1o 3 3 0
Observed ellipticity vs. pH in the presence of 0.1 M
phosphate buffer. [HSA] = 5.0 x 10 5 M = [benoxaprofen].
1.0 cm. cell.
290 30 30 30
I / / "" 9.65----
290 3C0 310 320 330
3O0 350 360 370
Observed ellipticity vs. pH in the presence of 0.1 M
sodium chloride. [HSA] = 5.0 x 10 5 M = [benoxaprofen].
1.0 cm. cell (pH adjusted with 0.1 N sodium hydroxide).
Observed ellipticity vs. pH in the presence of 2.50 x 10 3
calcium chloride dihydrate. [HSA] = 5.0 x 10-5 M = [Benoxa-
profen]. 1.0 cm. cell pH adjusted with 0.1 N sodium
(ow) e nDonNI
Figure 23. Dialysis data. 2.0 x 10-4 M HSA.
2.5 x 10-3 M CaC12-2H20 0.1
A 0.1 1M Phosphate buffer.
2.0 x 10~4 M Benoxaprofen.
M NaCl, M 0.1 M NaC1,
ellipticities in fig. 22. When albumin is in the N form, the effects of
chloride and calcium are emphasized, indicating that in this conformation
albumin is most sensitive to the effects of these ions. Table 8 lists
the parameters which are used to describe this type of curve. The pHso
is defined as the point at which the change from the neutral to basic
form is 50% completed. The range over which the change in the induced
ellipticity is from 10 to 90% completed is defined as the ApHo1,90o
HSA in 0.1 M phosphate buffer has a pHso at 7.4. The pH50 for the
benoxaprofen-albumin complex is 7.7, indicating that benoxaprofen itself
may "push" HSA to the basic conformation. Chloride ion does not have much
effect on the transition as the pH50 of 7.65 is almost identical to that
in phosphate buffer. Calcium ion has a major effect on the transition as
is seen by the pH50 of 6.90. As previously noted (Wilting et al. 1980),
this is a result of calcium ion inducing albumin into due B conformation.
Though the effect of chloride is mainly one of displacement, calcium seems
to both displace benoxaprofen and push the transition to the B conformation.
Calcium ions are reported to compete with carboxylate residues for imida-
zole binding sites (Pedersen,1972a,b)and this might be the displacement
Correlation of the C.D. data with the physical changes in the benoxa-
profen-HSA system has been done with equilibrium dialysis experiments as
shown in fig. 23. In phosphate buffer, the concentration of benoxaprofen
decreases from 0.200 to about 0.065 pg/ml as the pH is increased. When
chloride ions are present, the free concentration of benoxaprofen increases
as it is displaced by chloride throughout the N-B transition. At pH 7.4
the free benoxaprofen concentration is almost double in the presence of
0.1 M chloride than in phosphate buffer. When calcium ions are present
at a concentration of 2.50 x 10-3 M (with 0.1 M NaCl added to eliminate
Table 8. Parameters for benoxaprofen N-B curves.
0.1 M phosphate buffer 7.70 1.60
0.1 M chloride ion 7.65 2.30
2.5 x 10-3 M calcium ion 6.90 4.10
Table 9. Dialysis data. Variation of binding constants with pH.
[HSA] = 2.0 x 10-4 M [Benoxaprofen] = 2.0 x 10-4 M
0.1 M Phosphate Buffer
6.00 1.38 x 106
6.50 1.46 x 106
7.00 1.70 x 106
7.50 2.66 x 106
8.00 3.82 x 106
8.50 3.88 x 106
9.00 3.34 x 106
0.1 M NaCl
6.35 6.75 x 105
6.85 6.29 x 105
7.30 6.19 x 105
7.80 8.81 x 105
8.40 1.07 x 105
9.10 1.33 x 105
0.1 M NaCI
& 2.5 x 10-3 M
6.30 3.72 x 105
6.59 4.17 x 105
6.81 4.98 x 105
7.35 6.08 x 105
7.74 6.16 x 105
8.12 6.07 x 105
9.13 5.61 x 105
Donnan effects) there is an even greater displacement of benoxaprofen.
The flatness of the calcium curve at pH values greater than 7.0 confirms
that it does induce albumin into the B conformation. Solubility problems
encountered at low pH values made the plateau region for the N form inexact
but the general trend is still clear.
Table 9 lists the binding constants calculated from dialysis data
for 1:1 benoxaprofen-HSA complexes. In all cases, K1 increases as the
pH is increased. Displacement of benxoaprofen by chloride and calcium
ions is characterized by the smaller magnitude of the primary binding con-
stants in these cases as compared to phosphate buffer. As benoxaprofen
is very highly albumin bound (>99%), the bound fraction does not change
greatly but the free fraction can more than triple when pH, calcium or
chloride ion concentrations are changed. The free concentration of a drug
is responsible for its therapeutic or toxic effect.
Future protein binding studies with benoxaprofen should take careful
note of buffer composition and pH as this alone can significantly alter
its binding characteristics.
VII. DRUG DISPLACEMENT STUDIES: CHARACTERIZATION OF THE
BENOXAPROFEN BINDING SITES ON HUMAN SERUM ALBUMIN
It is generally accepted that most drugs which are bound to albumin
are bound via only a few high affinity sites and a number of low affinity
binding sites (MUller and Wollert, 1979). Evidence has been increasing
which shows these high affinity sites possess a degree of specificity
for distinct classes of drugs such as coumarins and nonsteroidal anti-
inflammatory agents (Sudlow et al. 1975, 1976). Changes in the fluores-
cence intensity of probes which had separate and distinct binding sites
were monitored as competitors were added to the system. Using equilib-
rium dialysis as a confirming technique, they found that the relative
affinity of drugs for these binding sites was measured by their ability
to displace probes specific to these sites.
Two distinct binding sites of high affinity have been described and
drugs binding to these sites classified. Site II drugs tend to be aro-
matic carboxylic acids which are ionized at physiological pH. Examples
of drugs which bind to this site are ibuprofen, naproxen, ethacrynic acid,
clofibric acid and flufenamic acid. These drugs are characterized as having
an extended configuration with a negatively charged carboxylate group on
an aliphatic residue extended from a nonpolar aromatic region. Site I
drugs are usually aromatic acids which have a delocalized negative charge.
Examples of Site I drugs are warfarin, phenprocoumon, acenocoumarin, phenyl-
butazone, tolbutamide and sulfonamides. Tests performed with diluted
adult sera correlate with studies which used crystalline serum albumin.
Other authors have studied the specificity of the high affinity
sites on HSA and a wide range of drugs have been classified as to their
binding sites (Sjoholm et al. 1979). A third discrete site which binds
digitoxin has also been reported in these studies.
Much attention has been given the warfarin (site I) binding site as
displacement of the coumarins can have severe pharmacological consequen-
ces (McElnay and D'Arcy, 1980; Veronich et al. 1979; Wosilait and Ryan,
1979; Minn and Zinny, 1980). The nature of this site has been described
as involving the lone tryptophan residue on HSA (Fehske et al. 1979).
Modification of this site with different reagents has led to decreased
Less information is known about the nature of the diazepam site
(site II) but studies in this laboratory found that some nonsteroidal
antiinflammatory drugs binding to this site may have a tendency to alter
the conformation of HSA as well as cause displacement interactions. Evi-
dence for anionic molecules inducing changes in albumin conformation has
been given in the binding of fatty acids (Birkett et al. 1977; Ashbrook
et al. 1974; Muller, 1976).
The nature of benoxaprofen's binding site will be described in this
study and the conformational adaptability of this site will be investi-
Materials and Methods
Human serum albumin, fraction V (lot No. 30F-02271) was obtained from
Sigma Chemical Co. (St. Louis, MO). Monobasic and dibasic sodium phos-
phate, analytical reagent grade, were obtained from Mallinckrodt, Inc.
(St. Louis, MO). The following drugs were used as supplied by the manu-
facturers: sodium benoxaprofen (903.5 mg free acid per gram), Eli Lilly
and Co. (Indianapolis, IN); fenbufen, Lederle Laboratories (Pearl River,
NY); flufenamic acid, Aldrich Chemical Co. (Milwaukee, WI); sodium
clofibrate, Imperial Chemical Industries, LTD. (Great Britain); oxphenyl-
butazone and phenylbutazone, Ciba Pharmaceutical Co. (Summit, NJ); ibu-
profen, Upjohn Co. (Kalamazoo, MI); sodium warfarin, Endo Labs, Inc.
(Garden City, NY), phenprocoumon, Organon, Inc. (W. Orange, NJ).
Albumin was deionized as previously described and its concentration
was determined by UV spectroscopy (M.W. = 66,500; e = 3.946 x 104 /mole.cm).
All solutions were prepared in 0.1 M phosphate buffer at pH 7.4. Free
concentrations of benoxaprofen were obtained from equilibrium dialysis
experiments and detected by HPLC using techniques described previously.
Circular dichroic measurements were also done as described in Chapter VI.
Displacement studies using the microcalorimetric technique were per-
formed by pumping a 5.0 x 10-4 M albumin solution into one half of the
mixing cell and 5.0 x 10-4 M drug solutions into the other side. The
heats of reaction for drug 1-HSA, drug 2-HSA and then drugs 1,2-HSA were
obtained. At all times the drug to protein ratios were kept at 1:1.
Equilibrium dialysis experiments were done with 2.50 x 10-4 M HSA
solutions dialyzed vs. 1.25 x 10-4 M drug solutions. Benoxaprofen is
dialyzed vs. HSA and then benoxaprofen and competitor are dialyzed vs.
HSA. Drug to protein ratios were maintained at 0.5.
Circular dichroic displacement studies were performed by recording
the ellipticity of the benoxaprofen-HSA complex and then recording the
change in ellipticity as the concentration of competitor was increased.
Separate solutions were prepared for each run and all work was done at
Results and Discussion
Some experience was gained by studying the much more thoroughly
investigated binding of warfarin and this was used to ascertain whether
benoxaprofen was a site I or site II drug. As can be seen from table 10,
at a drug to protein ratio of 0.5 there is 1.00 Ig/ml free warfarin.
In the presence of phenylbutazone and oxyphenylbutazone which binds to
the same primary site, the free concentrations of warfarin increases to
1.77 and 1.66 ug/ml, respectively. Other nonsteroidal antiinflammatory
drugs, especially those with aliphatic carboxylate residues in an extended
linear configuration, do not bind to the same site (Sudlow et al. 1976;
Sjiholm et al. 1979; Otagiri et al. 1979) and evidence a smaller degree
of warfarin displacement. Benoxaprofen does not significantly displace
warfarin and appears to be a site II drug.
The drugs in table 10 are ranked in order of the ability to displace
warfarin by a significance test on the difference between two means (AX).
Mean values of free warfarin concentration are compared with and without
competitor and the 90% confidence limit of the difference between these
two means is calculated. If AX >C.L., the difference between the means
is due to drug displacement and not random scatter. If AX C.L.,the sam-
ples are not significantly different or there is an enhancement on binding
in the presence of a competitor. The magnitude of AX-C.L. is used to
rank the samples in terms of warfarin displacement. Ibuprofen gives a
negative value for AX-C.L. and this may signify an enhancement of warfarin
binding at this low drug to protein ratio. This observation has been made
by other authors (Sudlow et al. 1976; Otagiri et al. 1979) and previously
in Chapter I where the effects of ibuprofen on phenprocoumon binding were
noted. It has been hypothesized in the above studies that ibuprofen may
"push" albumin into the B conformation, which has a greater affinity for
Table 11 lists dialysis data for the interactions between benoxapro-
fen and competitors. It is displaced by flufenamic acid, clofibrate
Table 10. Dialysis data.
[HSA] = 2.5 x 10-4 M [Warfarin] =
(X S.D. ig/ml)
1.25 x 104 M = [competitor]
C.L. (I = 0.05)*
90% confidence limit for the difference between two means
**Largest value = greatest displacement
and oxyphenylbutazone. This data, plus the fact that it didn't displace
warfarin, confirms benoxaprofen as a site II drug. It is unusual that
ibuprofen does not displace benoxaprofen and oxyphenylbutazone causes a
slight displacement. These drugs may be altering the conformation of
albumin or causing some type of cooperative effect such that the distinc-
tion between site I and site II may not be as exact when benoxaprofen is
monitored. With phenylbutazone, this hypothesis has been advanced by
Maes et al. (1979) and Madsen and Tearne (1979).
A clearer explanation as to what type of mechanisms might be involved
in these site II displacements might be reached if this phenomenon is
examined with different experimental techniques. The heat flux of binding
for the mixture of a drug and competitor (AHAB) is compared to the sum of
the values obtained with each drug acting alone with HSA (AHA + AHg). If
AHAB is equal to AHA + AHB no interaction between drugs A and B has occurred.
If the value of AHAB is less than the sum of AHA + AHB, then a mutual dis-
placement is probable. If AHAB is greater than the sum, an interaction
causing enhancement of the binding has taken place (Hardee, 1980).
Table 12 lists the results for different displacement experiments
at pH 7.4. As expected, the heat flux generated for a mixture of two
site I drugs, warfarin and phenprocoumon, gave a heat which was less than
the additive heat when each was measured separately. This indication of
a displacement is in agreement with the dialysis data in table 10. Within
experimental error, there does not appear to be competition between benoxa-
profen and warfarin or phenprocoumon. There is a displacement of benoxa-
profen in the presence of flufenamic acid and clofibrate and this again
is confirmed by the dialysis data in table 10. In the presence of ibupro-
fen and fenbufen AHAB was greater than AHA and AHB indicating some type
Binding site specificity and drug
(p j-oule sec-1)
(AHA + AHB
r I. 1
(+) = greater than additive heat
(-) = less than additive heat
(o) = additive heat = observed heat
of cooperative phenomenon or conformational change resulting in the
enhanced binding of benoxaprofen may be taking place.
All of the drugs listed in table 12 reacted exothermally with HSA.
Phenylbutazone was to be included in this table but it was found that at
drug to protein ratio less than 1.0 an endothermic reaction occurred.
The value obtained for AH at this ratio was approximately +4.0 Kcal/mole
which is close to that seen by Maes et al. (1979). It was not clear
whether adding heats of different signs would be applicable to the model
described, so this problem was examined further by circular dichroism.
Induced ellipticities of benoxaprofen-HSA complexes in the absence
and presence of increasing amounts of phenylbutazone are shown in fig.
24. In the wavelength region above 315 nm the ellipticities overlap and
without the aid of a computer, are extremely difficult to interpret. For
clarity, the wavelength region of maximum change in ellipticity is pre-
sented and the scale has been expanded. The effects of flufenamic acid
and clofibrate have also been studied in this way but they have an induced
C.D. signal of their own in this region whereas phenylbutazone does not.
At drug to protein ratios greater than 0.5 of these competitors, the
ellipticity of benoxaprofen is "masked" so these spectra will not be pre-
sented. However, at 0.5 drug to protein ratios where the induced benoxa-
profen signal is not "masked" an increase in ellipticity of the benoxapro-
fen-HSA complex is observed. This is similar to the effect of phenylbuta-
Figure14 illustrates that as HSA is titrated with benoxaprofen, a
secondary binding site of higher ellipticity exists. Equilibrium dialysis
and microcalorimetry have shown that flufenamic acid and clofibrate both
displace benoxaprofen. The increase in induced ellipticity as these com-
petitors are added can best be explained by a displacement from its primary
Induced C.D. curves for interaction of benoxaprofen-
HSA complex with phenylbutazone. [HSA] = 2.0 x 10-5
= [benoxaprofen] 0-3.0 D/P phenylbutazone
29 295 305 310 315 32
Induced C.D. curves for
complex with ibuprofen.
0-4.0 D/P ibuprofen.
interaction of benoxaprofen-HSA
[HSA] = 2.0 x 10"5 M = [benoxaprofen
I-- I V-- I -- I -
290 295 300 305 310 315 320
Figure 26. Induced C.D. curves for interaction of benoxaprofen-HSA
complex with fenbufen. [HSA] = 2.0 x 10-5 = [benoxaprofE
0-4.0 D/P fenbufen.
binding site to this secondary site of higher ellipticity. A similar
effect is seen with indomethacin, which also has a secondary binding site
of higher ellipticity (Ekman et al. 19801. Phenylbutazone does not bind
to the same primary site as benoxaprofen so the increase in the induced
ellipticity might be due to the phenylbutazone causing a conformational
change in the albumin which gives rise to the increased positive ellipti-
city. Figure 19 illustrates that as the pH is decreased and albumin is
induced into the N conformation, the ellipticity of the benoxaprofen-HSA
complex increases. As the ellipticity of the complex increases in the
presence of phenylbutazone, it too can possibly be inducing albumin into
the N conformation or it may be causing some other type of cooperative
effect. Studies in Chapter IX will confirm that phenylbutazone binding
to HSA involves a cooperative process rather than a B to N conformational
Figures 21 and 26 illustrate the induced CD signals of benoxaprofen
in the presence of ibuprofen and fenbufen. In both cases, the elliptici-
ties decrease as competitor is added. Microcalorimetric and dialysis data
(tables 12 and 11) have indicated that these drugs might enhance benoxa-
profen binding. Figure 19 shows that the ellipticity of the benoxaprofen-
HSA complex decreases as HSA is induced into the B conformation. It can
be hypothesized that both ibuprofen and fenbufen induce HSA into the B
conformation, resulting in the decreased ellipticities in figs. 25 and 26.
To confirm whether this is a correct hypothesis, the effects of these
competitors on site II binding must be studied at different pH values.
If they are involved in drug induced conformational changes as described
above, we would expect different competitive effects at pH 7.4 when HSA
is in both the N and B conformation as compared to pH 9 when HSA exists
solely in the B conformation. These experiments are presented in Chapter IX.
VIII. AN IMPROVED HPLC ASSAY FOR FENBUFEN
AND TWO SERUM METABOLITES
Fenbufen, 3-(4-biphenylcarbonyl)propionic acid (fig. 27, structure 1),
a nonsteroidal antiinflammatory drug, has been found to possess the same
spectrum of activity as aspirin, phenylbutazone and indomethacin. Its
two major serum metabolites, 3-(4-biphenylhydroxymethyl)propionic acid
(structure II) and 4-biphenylacetic acid (structure III) have a similar
degree of activity (DeSalcedo, 1975; Sunshine, 1975; Vergara et al. 1979;
Coutinho et al. 1976; Child et al. 1977).
Previous GC (Cuisinaud et al. 1978) and HPLC (Van Lear et al. 1978)
methods have been developed for the determination of fenbufen and its
metabolites in biological fluids. The GC method is very time consuming,
involving double extraction, derivatization and TLC separation before GC
injection. The established HPLC method has retention times of up to 22
minutes for the compounds of interest. A column heated to 450 and a 2 ml
plasma sample are also necessary.
In the new method the chromatographic procedure was improved to keep
the total retention time for all compounds under 9 minutes (table 13).
A plasma sample of only 1 ml is necessary and the column is operated at
Fenbufen, its metabolites and an internal standard, 4'-hydroxy-4-
biphenylacetic acid (structure IV), were kindly supplied by Lederle
/ / CH2CH2CH I
H/\O CH2COOH IV
Figure 27. Structures of fenbufen (I) and major metabolites.
Table 13. Retention times for fenbufen and serum metabolites.
Compound Retention Time (min)
Laboratories (Pearl River, NY). HPLC grade acetonitrile and isopropanol
were obtained from Burdick and Jackson (Muskegon, MI). Analytical grade
phosphoric acid, hydrochloric acid, sodium hydroxide and methanol were
supplied by Mallinckrodt, Inc. (St. Louis, MO). All chromatographic sol-
vents including deionized water, were filtered through 0.45micron filters
(Millipore Corp., Bedford, MA).
Sample analyses were carried out on a Varian model 5000 liquid
chromatograph equipped with a 10 pl manual loop injector and a column
heater (Varian Assoc., Walnut Creek, CA). An alkylnitrile column, Micro-
Pak CN-10 (Varian Assoc.) of medium polarity was used. Detection was made
on a Vari-Chrom variable wavelength UV detector.
During the extraction procedure, metabolite II forms a lactone (struc-
ture V) (Van Lear et al. 1978), as will be described later. The lactone
was prepared as follows: Met'abolite II was saturated in water at 400.
An excess of concentrated HC1 was added and the mixture was allowed to
stand for 5 minutes. The white precipitate was extracted into cyclohexane-
ether (7:3). The organic layer was evaporated under a nitrogen stream and
thelactone crystals were collected. Purity was determined chromatographi-
cally to be greater than 99%.
The mobile phase consisted of water-isopropanol-acetonitrile-phosphoric
acid (84.5:10:5:0.5). The flow rate was 2.5 ml/min and the column was kept
at 25. Detection was made at 265 nm at 0.02 aufs. A 10 vl sample was
injected onto the column.
The extraction procedure is similar to the one previously described
(Van Lear et al. 1978). Stock solution containing 200 ug/ml of fenbufen,
metabolites and internal standard are prepared in methanol. Accurate
volumes are pipetted into test tubes which will give final concentrations
of 0.5, 1.0, 3.0, 5.0, 10.0 and 25.0 ug/ml of plasma. The internal stan-
dard (2 pg) is then added. The methanol is evaporated in a stream of dry
nitrogen. Samples are reconstituted in 125 pl of 0.1 N NaOH and 1 ml of
plasma is added. Limited solubilities of fenbufen and its analogs neces-
sitate this procedure rather than direct spiking of plasma samples from an
aqueous stock solution.
After the contents of the tube are thoroughly mixed, 1 ml of concen-
trated HC1 is added and the samples are vortexed for 15 seconds. The sam-
ples are allowed to stand for 2 minutes and then 8 ml of cyclohexane-ether
(7:3 v/v) are added and the mixture is vortexed for 30 seconds. The sam-
ples are centrifuged at 3600 rpm for 12 minutes and the organic layer is
transferred to another tube and evaporated under a stream of dry nitrogen.
Samples were reconstituted in 100-200 pl of methanol and 10 il were injected
onto the column.
For absolute recovery experiments, spiked plasma samples were compared
to unextracted stock solutions. Peak height ratios were calculated and
amounts found were compared.
Precision determinations were performed by comparing the peak height
ratios of five extractions at each concentration.
Results and Discussion
Figure 28 shows a chromatogram of plasma containing fenbufen (I), two
metabolites (II) and (III) and internal standard (IV). The retention times
are listed in table 13.
Table 14 lists recovery data for compounds I, II and III. In the 0.5
to 25 Pg/ml range, the average recovery for I was 88.9 9.2%, II, 82.4
9.0% and III, 100.4 2.6%.