Development of a novel probe for measuring drug binding to the F1*S variant of human alpha 1-acid glycoprotein

Article abstract of DOI:10.1002/jps.1093

A novel probe was developed to measure drug association with the F1*S variant of the human serum protein alpha 1-acid glycoprotein (AGP). The molecule 2-hydroxy-3,5-diiodo-N-[2(diethylamino)ethyl]benzamide (DEDIC) binds to AGP, quenching its native fluorescence. This quenching was fitted to a two-site model giving apparent dissociation constants of 0.049 ± 0.005 and 12 ± 2 μM (mean ± SEM). Quenching of each of the separate variants of AGP by DEDIC was itself described by a two-site model, giving for the F1*S variant K_D(F1*S)¹ = 0.041 ± 0.010 μM and K_D(F1*S)² = 29 ± 7 μM; and for the A variant K_D(A)¹ = 0.31 ± 0.18 μM and K_D(A)² = 8.8 ± 0.7 μM. The utility of DEDIC in probing drug interactions with isolated variants was demonstrated in competition experiments with the model drugs amitriptyline and bupivacaine. In addition, the selectivity of DEDIC for variant F1*S rendered it capable of probing the binding of drugs (including the variant A-selective drug amitriptyline) to F1*S in a mixture of variants, such as occurs naturally in whole AGP. DEDIC is unique as an F1*S variant-selective probe of drug binding to whole AGP that is also sufficiently soluble to serve as a probe of drug binding to the lower affinity sites on isolated A and F1*S variants.

Full text of DOI:10.1002/jps.1093

Development of a Novel Probe for Measuring Drug Binding
to the F1*S Variant of Human Alpha 1-Acid Glycoprotein
LAWRENCE P. COGSWELL III,1,2 DOUGLAS E. RAINES,3 SONALI PAREKH,4 OLIVER JONAS,1 JOHN E. MAGGIO,5
GARY R. STRICHARTZ1,2,6
1Pain Research Center, Department of Anesthesiology, Perioperative and Pain Medicine, Brigham & Women's Hospital,
75 Francis Street, Boston, Massachusetts 02115
2Harvard Biophysics Program, Harvard University, Boston, Massachusetts
3Department of Anesthesia and Critical Care Medicine, Massachusetts General Hospital, Boston, Massachusetts
4Brown University School of Medicine, Providence, Rhode Island
5Department of Pharmacology and Cell Biophysics, University of Cincinnati College of Medicine, Cincinnati, Ohio
6Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, Boston, Massachusetts
Received 31 May 2000; revised 20 February 2001; accepted 5 March 2001
ABSTRACT: A novel probe was developed to measure drug association with the F1*S
variant of the human serum protein alpha 1-acid glycoprotein (AGP). The molecule 2-
hydroxy-3,5-diiodo-N-[2(diethylamino)ethyl]benzamide (DEDIC) binds to AGP, quench-
ing its native ¯uorescence. This quenching was ®tted to a two-site model giving
apparent dissociation constants of 0.049 Æ 0.005 and 12 Æ 2 mM (mean Æ SEM).
Quenching of each of the separate variants of AGP by DEDIC was itself described by
a two-site model, giving for the F1*S variant K1D(F1*S) ˆ 0.041 Æ 0.010 mM and KD2 (F1*S)
ˆ
29 Æ 7 mM; and for the A variant K1D(A) ˆ 0.31 Æ 0.18 mM and K2D(A) ˆ 8.8 Æ 0.7 mM. The
utility of DEDIC in probing drug interactions with isolated variants was demonstrated
in competition experiments with the model drugs amitriptyline and bupivacaine. In
addition, the selectivity of DEDIC for variant F1*S rendered it capable of probing the
binding of drugs (including the variant A-selective drug amitriptyline) to F1*S in a
mixture of variants, such as occurs naturally in whole AGP. DEDIC is unique as an
F1*S variant-selective probe of drug binding to whole AGP that is also suf®ciently
soluble to serve as a probe of drug binding to the lower af®nity sites on isolated A and
F1*S variants. ß 2001 Wiley-Liss, Inc. and the American Pharmaceutical Association J Pharm
Sci 90:1407±1423, 2001
Keywords: alpha 1-acid glycoprotein; drug binding; genetic variants; probes;
methods
INTRODUCTION
glycoprotein (AGP) is a problem of general
pharmacological interest. Many drugs are bound,
principally or in part, by AGP,1 and AGP is the
principal carrier of basic drugs in plasma.2,3 The
extent of this association, at equilibrium, deter-
mines the concentration of unbound drug, and the
traditional thought is that only this fraction is
available for active uptake or diffusion into
surrounding tissues. Clinically, this buffering
effect must be considered in adjusting the dosage
Characterizing the association of blood-borne
drugs with the major serum protein alpha 1-acid
Correspondence to: G. R. Strichartz (Telephone: 617-732-
8797; Fax: 617-730-2801)
Journal of Pharmaceutical Sciences, Vol. 90, 1407±1423 (2001)
ß 2001 Wiley-Liss, Inc. and the American Pharmaceutical Association
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001
1407

1408
COGSWELL ET AL.
of a drug or in assessing the toxicity of a drug
accidentally administered intravenously, a feat
that is often complicated by changes in patients'
AGP concentrations during disease, pregnancy, or
other conditions.2
The heterogeneity of AGP adds an additional
level of complexity. There are three major var-
iants of AGP (F1, S, and A), which differ from one
another in primary structure4 and which may be
separated after desialylation using polyacryla-
mide gel isoelectric focusing.5 This heterogeneity
is important to consider when carrying out
binding studies with AGP, because the A variant
has been reported to bind many drugs with
af®nities different from those of the F1 or S
variant.1,6,7
The genetic polymorphism is the result of the
existence of two different genes that encode for
AGP.8 One of these, denoted AGP-A, encodes the
F1 and S variants, whereas the other, AGP-B/B',
encodes the A variant.9 The relative frequency of
the major phenotypes is ꢀ50% F1*S/A, 35% F1/A,
and 15% S/A, and the relative abundance in
commercial, pooled-source protein of the F1, S,
and A variants has been reported to be 44 Æ 1,
29 Æ 1, 27 Æ 1%, respectively.7 Structurally, the
two genes differ by 22 codons, resulting in 22
amino acid substitutions along the protein of 183
residues. The F1 and S variants, conversely, are
reported to differ from each other by fewer than
®ve residues.8 In addition to the F1, S, and A
variants, there exist a multitude of minor gene
products that arise from the multiple allelic
forms of AGP-A and AGP-B/B'.4 However,
these products, in total, account for < 1% of the
AGP encountered in the general population6 and
are not likely to be signi®cant sources of drug
binding.
Because drugs have been observed to bind
differently to the A variant than to the F1 or S
variants (collectively denoted F1*S),1,6,7 it is clear
that understanding the pharmacological role of
AGP will require an examination of how drugs
interact with each of these variants individually.1
Furthermore, a complete characterization must
include a determination of af®nities for both the
higher and lower af®nity sites on each.
the failure to do so may account for some of the
inconsistent results and fractional binding site
stoichiometries obtained over the years in drug
binding studies with AGP.
We have developed a rapid spectro¯uorometric
method for determining drug af®nities to the high
af®nity site on the F1*S variant of AGP in a
heterogeneous mixture using the novel molecule
2-hydroxy-3,5-diiodo-N[2(diethylamino)ethyl]
benzamide (DEDIC). The method is based on
spectro¯uorometry, which has considerable ad-
vantages over equilibrium dialysis in terms of
speed, convenience, and sensitivity.12
In short, DEDIC binds to and quenches the
native tryptophan ¯uorescence of AGP. Subse-
quent addition of a nonquenching drug that is
known to bind AGP, such as bupivacaine (BUP) or
amitriptyline (AMI), results in ¯uorescence recov-
ery from which the binding constant of the
antagonist may be calculated.
In addition, we demonstrate that DEDIC binds
selectively to the high af®nity site on the F1*S
variant of AGP and can be used to probe the
association of a drug to this site in the naturally
occurring mixture of the two proteins. The se-
lectivity of DEDIC for a single site on one of the
AGP variants, coupled with a solubility suf®cient
to permit its use as a probe of drug binding to
lower af®nity sites on (isolated) single variants,
distinguishes it from other (¯uorescent) probes
currently in use (e.g., 8-anilinonapthalensulfonic
acid, ANS). Among quenchers, DEDIC possesses a
signi®cant advantage over other potentially use-
ful probes. Chlorpromazine (CPZ), for example, is
not variant selective, whereas dipyridamole (DIP)
is variant selective to the same extent as DEDIC
(ꢀ10-fold) but not suf®ciently soluble that it can
be used on isolated variants to probe lower af®nity
binding.
EXPERIMENTAL PROCEDURES
Caution
The following chemicals are hazardous and
should be handled carefully: N, N-diethylethyle-
nediamine, thionyl chloride, and 1,4-dioxane.
Commercially available AGP is derived from
pooled human plasma, and consists of ꢀ70% F1*S
and 30% A variants,5,6,7 re¯ecting the population
average. To accurately determine binding af®-
nities, these variants must ®rst be separated, for
example, according to the chromatographic meth-
Materials
Human alpha 1-acid glycoprotein (AGP, lot
125H9329), dipyridamole (DIP), amitriptyline
(AMI), and bupivacaine (BUP), all > 99% pure,
Â10
ods described by Herve and Halsall.11 Indeed,
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

NEW PROBE TO MEASURE DRUG BINDING TO F1*S OF HUMAN ALPHA 1-ACID GLYCOPROTEIN
1409
were obtained from Sigma Chemical Company
(St. Louis, MO). N,N-Diethylethylenediamine
(DED, 99%), 3,5-diiodosalicylic acid (DIS, 99%),
1,4-dioxane (99.8%, anhydrous), thionyl chloride
(SOCl2, 97%), and diethyl ether (99 ‡ %, anhy-
drous) were obtained from Aldrich Chemical
Company (Milwaukee, WI). Ethanol (EtOH,
100%, anhydrous) in glass bottles was obtained
from Pharmco Products (Brook®eld, CT). All other
chemicals were ACS reagent grade obtained from
Aldrich. Water was doubly deionized to > 18.2 MO
with a Milli-Q ultra pure water system (Millipore
Corp., Woburn, MA) and was free of ¯uorescent
contaminants. Iminodiacetic acid (immobilized on
Sepharose 6B Fast Flow Gel) and copper(II)
chloride were obtained from Sigma.
Structure Determination
Nuclear magnetic resonance (NMR) measure-
ments were performed on a Bruker 500 MHz
instrument at 208C in either deuterated chloro-
form or a 1:2 v/v mixture of deuterated water
and methanol. Isotopic purity for all solvents was
> 99.9%. pH was adjusted with sodium deuter-
oxide or deuterium chloride from 2 to 13, as
measured by spotting a drop of solution onto
pHydrion-brand pH paper (Microessential Labo-
ratories, Brooklyn, NY). The precision of the
paper is Æ 0.5 pH unit. Low-resolution fast atom
bombardment mass spectroscopy (FABMS) was
performed in a glycerol matrix in the positive ion
mode.
Elemental Analysis
Synthesis of DEDIC
Standard C, H, N analysis was performed, as
well as I and Cl analysis, using ion-selective
electrodes.
First, 30 g (0.077 mol) of DIS were added to
120 mL of SOCl2 and the suspension was re¯uxed
under dry nitrogen at 1208C for 40 min. SOCl2
was evaporated at 1308C over a steady nitrogen
stream until a clear, light yellow oil remained.
Remaining SOCl2 was evaporated at 2 Torr for
2 min, yielding a viscous yellow-brown oil. This
residue solidi®ed on cooling to 208C in vacuo. The
residue was removed from vacuum and dissolved
in 150 mL of dioxane, heating as necessary, to give
a clear, light yellow solution. Aliquots (3 mL) of
this solution were added at 1-min intervals to 50 g
(0.44 mol) neat DED with gentle stirring. A
voluminous yellow precipitate formed in the
recipient ¯ask before addition was complete, and
several milliliters of additional dioxane were
added to facilitate stirring. After standing for 1 h
at 208C, the mixture was vacuum ®ltered, and the
resinous yellow solid was washed thrice with
25 mL of 1.0 M HCl to remove excess amine salts.
The solid was then agitated with a spatula and
washed liberally with deionized water until a
powdery solid remained. The product was washed
four times with 10 mL of ethanol and three times
with 25 mL of ether, and allowed to dry on the
funnel. Sixteen grams of crude product were
recovered and then dissolved in a minimum
(ꢀ100 mL) of boiling methanol. This solution
was gravity ®ltered and then decanted into a
crystallizing dish at room temperature. After 24 h
at 208C, white crystals were collected on a glass
frit and washed with a minimum of cold methanol
and then with 5 mL ether. The product was dried
overnight in a vacuum oven at 20 Torr and 708C.
The yield was 8.0 g (0.017 mol; 22%).
Chemical Stability
Aqueous DEDIC solutions at pH 3, 7.4, and 12
were stored at either room temperature (19±
208C)/room light, room temperature/dark, or 48C/
dark. Hydrolysis was measured as free DIS, and
photo-deiodination as aqueous iodine, which were
easily detected spectrophotometrically.
Preparation of Biochemical Solutions
Except where noted, all solutions were prepared
in standard aqueous medium (SAM), containing
0.15 M NaCl and 5 mM MOPS buffer adjusted to
pH 7.40. Glassware was triply rinsed with double-
deionized water and oven dried prior to use.
For ¯uorescence experiments, glassware was
initially cleaned with 7 M HNO3 to degrade any
adsorbed ¯uorescent contaminants; plastic con-
tainers were avoided. Determinations of mass
were made on a Mettler AE100 electronic balance
and were accurate to Æ 0.0002 g as determined
with standard weights. Adjustable volumetric
pipettes (Gibson or Eppendorf) were used routi-
nely and ®tted with polypropylene tips that were
observed not to contaminate solutions with ¯uor-
escent plasticizers. Solutions were covered with
aluminum foil, Te¯on tape, or glass stoppers (no
grease).
Solutions of whole AGP were prepared
gravimetrically, using a molecular weight of
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

1410
COGSWELL ET AL.
40,800 g/mol and water content of 13% (Sigma
Chemical Co., personal communication). The
molar absorptivity of whole AGP was determined
the excitation wavelength was 278 nm, except
as noted. The emission spectrum of AGP was
essentially independent of excitation wavelength
over the range 265±295 nm (data not shown),
indicating that tyrosine residues did not contri-
bute signi®cantly to the observed ¯uorescence.
Therefore, the protein was excited at its wave-
length of maximum absorbance (278 nm) to
give maximal ¯uorescence. Quartz cuvettes were
soaked daily in 7 N HNO3 for 30 min. Cuvettes
were ®tted with copper handles and not touched
with gloves or paper tissues. Between scans,
cuvettes were cleaned with ethanol dispensed
from glass bottles and dried over air ®ltered
through a 0.2-mM ®lter (Gelman PTFE ACRO-
disk).
to be 3.24 Â 104
M
1cm 1, and this value was
used to determine the concentration of the isolat-
ed F1*S and A variants. In so doing, we assumed
that absorbance by the tryptophan chromophores
was unaffected by the substitution of other
residues. This assumption is supported by the
observation that scaled absorbance spectra of
AGP, F1*S, and A variant are identical (data
not shown), a situation that could arise in practice
only if amino acid substitution did not affect the
chromophores.
DEDIC solutions were prepared by adding the
solid material to SAM and stirring vigorously to
form a suspension. Aqueous HCl (1 M) was then
added in a dropwise manner until the solid
dissolved. The pH of the solution was immediately
readjusted to pH 7.40 with NaOH. Ultraviolet±
visible (UV±vis) spectroscopy con®rmed that
DEDIC was not hydrolyzed by this procedure.
DEDIC was soluble in SAM at pH 7.40 to
ꢀ250 mM. Although it is possible to form nomin-
ally more concentrated solutions, in our experi-
ence these are, in fact, supersaturated, and slow
aggregation may occur over a period of hours to
days.
Analysis of Steady-State Fluorescence Data
Raw ¯uorescence data (Fraw) were ®rst corrected
for inner ®lter effects13 and second for spurious
¯uorescence (e.g., from Raman scatter.) Third,
data were corrected for the ¯uorescence contribu-
tion of the drug, if any. For ligands used in
this study, this correction was minimal and
accounted for < 10% of the total signal. The cor-
rected ¯uorescence, normalized to unity (F), was
then plotted versus nominal [DEDIC] and ®tted
directly to modi®ed Stern±Volmer equations
allowing for either one (eq. 13) or two (eq. 26)
classes of binding sites either on the same or
different protein molecules (see Appendix). The
derivations of these equations is complicated by
the inapplicability of common simplifying as-
sumptions in our experiments. Most signi®cantly,
because our ligand concentrations were compar-
able to those of the protein, we were unable to
approximate the concentration of free ligand at
equilibrium by its nominal (total) concentration.
Equations 13 and 26 (Appendix) are, therefore,
exact descriptions of the isotherms.
In some experiments, a titration method was
used instead, wherein successive aliquots of drug
were added in situ to a cuvette containing 2.5 mL
of protein solution. To minimize the possibility of
degradation associated with repeated interroga-
tion of the sample, single wavelength measure-
ments (lem ˆ 345 nm) were collected instead
of complete emission scans, and the excitation
shutter was closed between measurements. Fluo-
rescence data were ®rst corrected for inner-®lter
effects13 and then corrected for dilution by multi-
plying each datum by the appropriate dilution
factor. Under the conditions used in this study
DIP solutions were similarly prepared by ini-
tial acidi®cation to speed dissolution. The solubi-
lity limit was ꢀ5 mM in SAM.
Optical Titration
Spectrophotometric titrations were performed at
228C with a Beckman DUS model 600 single-
beam spectrophotometer with matched far UV
quartz cuvettes. Spectro¯uorometric titration
was performed with an Aminco SPF 500 spectro-
¯uorometer. The pKas were determined either by
plotting the observed lmax versus pH or by ®rst
computing the ratio of absorbances (¯uores-
cences) at two wavelengths corresponding to the
l
maxs for the conjugate acid and base, and then
plotting versus pH. pH was measured with a
Corning combination electrode with an Orion
digital pH meter (model 611).
Fluorescence Assays
Emission scans were performed in ratio mode at
228C with an SLM-Aminco SPF-500 spectro¯uo-
rometer using rhodamine B (3 g/L in ethylene
glycol) as reference. Slits were 7.5 nm, and
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

NEW PROBE TO MEASURE DRUG BINDING TO F1*S OF HUMAN ALPHA 1-ACID GLYCOPROTEIN
1411
(where the concentration of stock titrant exceeded
the KD or KI by a factor of at least 50), the effect
of volume dilution on the free fraction of protein
was assumed to be negligible. This assumption
was con®rmed for quenching by DEDIC both
theoretically by simulations conducted using
Mathematica1 software (Wolfram Research) and
experimentally by direct comparison of data col-
lected using both the titration method and the
method of individual aliquots.
The relative yield of each component was deter-
mined using absorbance data at 280 nm and our
previously determined molar absorptivity for
1
whole AGP of 3.24 Â 104
M
cm 1. Fractions
from the F1*S and A variant components were
pooled separately. Because they were eluted in
the same buffer to be used in subsequent ¯uo-
rescence measurements and the concentration of
each variant (as determined spectrophotometri-
cally) was suf®ciently large, concentration and
dialysis steps were unnecessary.
Assessing the Signi®cance of Dynamic Quenching
Con®rmation of the Identity of the Variants
The contribution of dynamic quenching was
deemed to be negligible for the concentrations of
protein and quencher used in this study (i.e.,
[AGP] < 5 mM, and [Drug] < 50 mM) according to
the following argument. Assuming a maximum,
The identi®cation of the separated variants was
made by comparing our relative yield of each
variant with that reported by Halsall.15 In ad-
dition, the F1*S variant-selective drug DIP was
added to equal concentrations (0.39 mM) of F1*S
variant, A variant, and whole AGP. From the re-
sulting quench curves, the af®nity of DIP for each
variant was determined. The relative af®nities of
the drugs were compared with those previously
reported using a different method and different
(phosphate) buffer.1
diffusion-limited, collision rate of drug and pro-
1
tein equal to 1010
M
s
1, and a quencher
5
concentration of 5 Â 10 M, the rate of collision
of quencher with AGP is 5 Â 105 s 1. The average
time between collisions on any given AGP mole-
cule is thus ꢀ2 ms. The ¯uorescence lifetimes of
the tryptophans in AGP have been reported to be
on the order of 1±10 ns,14 which is 200±2000 less
than the average time between collisions, effec-
tively obviating a role for dynamic quenching.
Competition between DEDIC and AMI on Whole
AGP and Its Isolated Variants
The model drug AMI was added to partially
quenched solutions of DEDIC and protein. From
the resulting ¯uorescence recovery curves, the
af®nities of these drugs were determined by ®t-
ting the data to a model of competitive antagon-
ism at a single site (see Appendix, eq. 36)
The concentration of DEDIC was chosen so that
only a single high af®nity site would be occupied
and competition at lower af®nity sites was neg-
ligible. Where experiments were repeated at
several [DEDIC], the data were ®t simultaneously
to eq. 36.
Chromatographic Separation of Variants
AGP was separated chromatographically into the
major variants F1*S and A using Halsall's
10
modi®cation15 of Herve's method.
Brie¯y, a
Â
20-cm column (i.d. ˆ 1 cm) containing ꢀ40 mL of
iminodiacetic acid (immobilized on Sepharose 6B
Fast Flow Gel) was allowed to equilibrate with
several volumes of 0.2 M copper(II) chloride/
150 mM NaCl and then rinsed with several
volumes of pH 3.8 100 mM acetic acid/150 mM
NaCl, followed by several volumes of the standard
pH 7 buffer, SAM (150 mM NaCl/5 mM MOPS).
Whole AGP, 25 mg in 1 mL SAM, was applied to
the column and eluted at a constant ¯ow rate of
6 mL/min. Elution was monitored spectrophoto-
metrically by scanning over the wavelength range
210 to 450 nm, and fractions of 5 to 20 mL each
were collected. After elution of the adsorbing
component (F1*S), the pH of the elution buffer
was stepped down to 4.6 to elute the A variant.
The column was reconditioned between runs by
removing chelated copper with 50 mM EDTA,
pH 7, rinsing with several volumes of deionized
water, and repeating the equilibration procedure.
RESULTS
Structure of DEDIC
The consensus structure of the novel molecule
DEDIC is shown in Figure 1A. From the 1H NMR
spectra in 2:1 v/v CD3OD/D2O (Figure 1B), peak
assignments were made as reported in Table I.
No impurities were detected. Low-resolution
FABSMS gave an M‡1 ion peak at m/z ˆ 489
consistent with the calculated molar mass of
DEDIC of 488 g/mol.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

1412
COGSWELL ET AL.
48.68% I (theory 52.0%). Furthermore, the analy-
sis was entirely consistent with DEDIC sesquihy-
drate (DEDICÁ 1.5 H2O) to within Æ 0.4% for C, H,
and N, and Æ 1% for I. Chloride was not detected
by trace analysis (limit of detection ˆ 0.05%),
indicating that DEDIC had not crystallized as a
hydrochloride salt.
Physical Properties of DEDIC
The molar absorptivity of DEDIC in SAM (pH
7.40 and 22 8C) at lmax ˆ 356 nm was 5.02
( Æ 0.05) Â 103
M
1cm 1, determined from solu-
tions at < 20 mM [DEDIC]. The aqueous solubility
of DEDIC at pH 7.4 and room temperature was
ꢀ250 mM.
Spectrophotometric titration of DEDIC was
characterized by changes in absorbance intensity
and peak wavelength. Over the pH range 4±6,
lmax (DEDIC) shifted from 328 to 355 nm and
peak absorbance rose, with an isosbestic point at
331 nm (Figure 2A, inset). pH-dependent shifts
in lmax as a function of pH gave pK1a ˆ 5.0 Æ 0.2
(Figure 2A). From pH 7 to 10, lmax shifted
from 355 to 351 nm, and peak absorbance
remained unchanged (Figure 2B, inset). The
calculated pK2a of 9.5 Æ 0.2 (Figure 2B) was
determined by pH-dependent changes in the
absorbance ratio A355/A351 that resulted from the
wavelength shift.
Because of the relatively small peak shift
observed spectrophotometrically for the second
protonation event, pK2a was also determined by
spectro¯uorometric titration. The ¯uorescence
emission spectrum of DEDIC was considerably
more sensitive than absorbance to the second
protonation event, as evidenced by a red shift of
almost 10 nm (Figure 2C, inset). The ®tted pH
dependence of this shift yielded a pK2a of 9.8 Æ 0.2
(Figure 2C), which is consistent with that derived
from absorbance changes (vide supra).
Figure 1. (A) Structure of the novel molecule
DEDIC. Chemically distinct hydrogen nuclei are iden-
ti®ed by bold letters. (B) 1H NMR spectra of DEDIC (in
2:1 v/v D2O/CD3OD, room temperature) at pH 2 (top) ,
pH 7 (middle) and pH 13 (bottom). (See Figure 1 for an
assignment of resonances.) From pH 2 to pH 7 (top to
middle traces) the aromatic resonance f shifts up®eld as
the phenol moiety of DEDIC is deprotonated, and from
pH 7 to pH 13 (middle to bottom traces) the resonances
of the aliphatic protons (a±d) shift up®eld in response
to deprotonation of the tertiary amine. Isotopic impu-
rities in the solvents cause the strong resonances at
d ˆ 3.3 and 4.9 ppm. The internal reference was sodium-
3-(trimethylsilyl)propionate-2,2,3,3-d4.
Elemental analysis (C, H, N, I) agreed well
with theory, considering the potential of phenolic
compounds to form hydrates and the relatively
large uncertainty in determination of iodine
( Æ 2%): 30.35% C (theory 31.96%), 3.84% H
(theory 3.68%), 5.23% N (theory 5.73%), and
The NMR spectra in 2:1 v/v CD3OD/D2O
(shown in Figure 1B with 1H assignments
provided in Table I) showed pH-dependent shifts
consistent with two protonation events. On tra-
versing pK1a (i.e., increasing pH from 2 to 7), the
resonance of one of the aromatic hydrogens (f)
shifted signi®cantly up®eld, whereas the shifts of
the aliphatic hydrogens (i.e., a±d) were relatively
unaffected. Conversely, on traversing pK2a, the
opposite response was observed, with the reso-
nances of the aliphatic hydrogens shifting sig-
ni®cantly up®eld and those of the aromatic
hydrogens remaining approximately the same.
Table I. Chemical Shiftsa of Hydrogen Nuclei of
DEDIC at Varying pH Values
1
d (ppm) H
pH 2
pH 7
pH 13
a
b
c
d
e
F
1.32
3.32
3.40
3.79
8.07
8.30
1.31
3.38
3.27
3.76
8.04
8.05
1.06
2.63
2.72
3.51
8.01
8.04
aDetermined in 2:1(v/v) CD3OD/D2O (see Figure 1A for 1
labels and 1B for raw spectra).
H
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

NEW PROBE TO MEASURE DRUG BINDING TO F1*S OF HUMAN ALPHA 1-ACID GLYCOPROTEIN
1413
Because changes in pH around pK1a resulted in
shifts in the resonances of one of the aromatic
protons (f) of DEDIC, this protonation event was
assigned to the dissociation of hydrogen from the
phenolate moiety. Likewise, the pH-dependent
1
shifts of the aliphatic H (a±d) resonances near
pK2a were attributed to protonation of the tertiary
amine (see Figure 1B).
Consistent with this assignment, the optical
absorbance spectrum of DEDIC changed on
traversing the pH region near pK1a (Figure 2A),
re¯ecting the perturbation of the electronic
structure of the aromatic ring associated with
dissociation of the phenol hydrogen. Notably,
DEDIC is a zwitterion at physiologic pH.
Stability
Solutions are stable in the dark for at least 1 week
at 228C and for at least 1 month at 48C. In room
light at 228C, solutions are stable for ꢀ2 days
before traces of free iodine could be detected
spectrophotometrically. DEDIC hydrolysis was
negligible between pH 3 and pH 11. DEDIC was
stable in the spectro¯uorometer during continu-
ous irradiation at 278 nm for at least 15 min
(300 mW Hg lamp, 15 nm slit widths). Continuous
excitation near the absorption maximum (356 nm)
resulted in accelerated photo-deiodination, which
was easily detected by yellow-brown coloration of
the solution after ꢀ15 min. The absorbance
spectrum and odor of the solution were consistent
with that of aqueous molecular iodine.
Separation of Protein Variants
To investigate the identities of the two apparent
binding sites observed for DEDIC binding to
whole protein, 25 mg of AGP was separated chro-
matographically into its variants (cf. Methods).
Recovery of the protein was quantitative, with
81% of the material (as monitored spectrophoto-
metrically) eluting in the ®rst fraction and 19%
in the second fraction (Figure 3); these propor-
tions were similar to those reported by Halsall.15
Furthermore, the absorbance spectra of these
fractions were identical in shape to that of pure
AGP. Because we had simply repeated Halsall's
procedure, with changes only in the identity
(MOPS rather than phosphate) and ionic strength
of the eluting buffer, we therefore equated
the ®rst and second elution fraction with the
F1*S and A variants, respectively, identi®ed by
Halsall.
Figure 2. (A) Spectrophotometric titration of the
acid moiety of DEDIC. lmax is plotted against pH (open
circles). The data were ®tted to the Henderson±
Hasselbalch equation (solid line), giving a pK1a of
5.0 Æ 0.2. Inset: lmax shifts from 328 to 355 nm over
the pH range 4±6, with an isosbestic point at 331 nm.
(B) Spectrophotometric titration of the basic moiety of
DEDIC. The ratio A355/A351 (open circles) is plotted
against pH. These data were ®tted to the Henderson±
Hasselbalch equation (solid line), giving a pK2a of
9.5 Æ 0.2. Inset: lmax shifts bathochromically from 355
to 351 nm over the pH range 7 to 10. (C) Spectro-
¯uorometric titration of the basic moiety of DEDIC.
lmax (open circles) is plotted against pH. These data
were ®tted to the Henderson±Hasselbalch equation
(solid line), giving a pKa2 of 9.8 Æ 0.2. Inset: Emission
spectra of DEDIC (F 345/lem) over the pH range 7.3±
11.6.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

1414
COGSWELL ET AL.
Figure 3. Chromatographic separation of AGP into
the A and the F1S variant. Twenty-®ve milligrams of
whole AGP was applied to the copper-iminodiacetate
column. The absorbance at 280 nm (solid line) shows
that the ®rst fraction (F1*S variant) was eluted at
pH 7.0 and the second fraction (A variant) was eluted
after stepping the pH down to 4.6. The cumulative
protein recovery (in milligrams) is also shown (dashed
line). Total protein recovery was ꢀ99%, with the A and
the F1*S variants comprising 19 and 81%, respectively,
of the material.
Figure 4. Fluorescence quenching of 0.39 mM AGP
with added DIP. Quench data for variant A (squares),
variant F1S (circles), and whole AGP (triangles) were
®tted by eq. 13 to yield KD(A/DIP) ˆ 0.75 Æ 0.12 mM,
KD(F1*S/DIP) ˆ 0.07 Æ 0.02 mM, and KD(AGP/DIP) ˆ 0.84 Æ
0.14 mM. The amplitudes of these processes were
0.48 Æ 0.03, 0.69 Æ 0.01, and 0.63 Æ 0.01, respectively.
Con®rmation of Variants Using the F1*S-Selective
Drug DIP
appears to be largely independent of emission
wavelength between 325 and 400 nm (Figure 5B),
so any analytical wavelength in this range would
be suitable. Outside this range, the apparent
wavelength dependence is dif®cult to interpret
because it may be an artifact of a decreasing
signal-to-noise ratio. By convention, we have
chosen a near maximal analytical wavelength
(345 nm). The emission spectra of the F1*S and A
variants are identical to each other and to that of
whole AGP as well (data not shown).
The ¯uorescence intensity (ex. 278/em. 345,
corrected for inner ®lter effects13) of 0.36 mM
whole AGP, variant A, and variant F1*S, is shown
in Figure 6 as a function of nominal [DEDIC]. For
whole AGP (Figure 6, triangles), the concentra-
tion dependence could not be ®tted satisfactorily
by a one-binding-site model (eq. 13), but was ®t
well by a two-site model (eq. 26), using the ana-
lysis presented in the Appendix (vide infra). From
nonlinear least squares regression of these data,
we obtained apparent equilibrium dissociation
constants for DEDIC binding to whole AGP of
K1D(native) ˆ 0.049 Æ 0.005 mM, n1 ˆ 0.63 Æ 0.04 mM,
and K2D(native) ˆ 12 Æ 2 mM, with n2 indeterminate.
The amplitudes of the two processes were
A1 ˆ 0.59 Æ 0.03 and A2 ˆ 0.22 Æ 0.02 with an
unquenched fraction having Aunq ˆ 0.19 Æ 0.03.
Con®rmation of variants was made by determin-
ing the af®nities of the F1*S variant-selective
drug DIP to the isolated variants and comparing
these to previously reported values.
The DIP quench data (Figure 4) were ®t by a
modi®ed one-site Stern±Volmer equation (eq. 13)
to give af®nities of KD(A/DIP) ˆ 0.75 Æ 0.12 mM and
KD(F1*S/DIP) ˆ 0.07 Æ 0.02 mM, with an nA of 0.73
and an indeterminate nF1*S. The amplitudes of
these processes were 0.48 Æ 0.03 and 0.69 Æ 0.01,
respectively. Unfortunately, solubility limitations
prevented us from investigating the binding of
DIP to lower af®nity sites on AGP, as were
7
Â
reported by Herve.
Binding of DEDIC to AGP and to the Isolated
Variants
The spectral changes in ¯uorescence during
quenching of 0.44 mM whole AGP by increasing
concentrations of DEDIC are depicted in
Figure 5A. The emission spectrum of AGP results
principally from the three tryptophan residues of
the protein, and in SAM is maximal near 340 nm.
Note that the relative degree of quenching
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

NEW PROBE TO MEASURE DRUG BINDING TO F1*S OF HUMAN ALPHA 1-ACID GLYCOPROTEIN
1415
Figure 6. The corrected, normalized ¯uorescence (F)
was plotted as a function of nominal [DEDIC], and the
resulting quench curves were ®tted to a two binding site
model (eq. 26). Parameters obtained from replicate
titrations (at least two) were averaged, yielding for
F1*S (circles), K1D(F1*S) ˆ 0.041 Æ 0.010 mM, n1
ˆ
(F1*S)
0.69 Æ 0.02, K2D(F1*S) ˆ 29 Æ 7 mM, and n2
was
(F1*S)
indeterminate. For the A variant (squares), KD1 (A)
ˆ
0.31 Æ 0.18 mM, and K2D(A) ˆ 8.8 Æ 0.7 mM, with both
Figure 5. (A) Addition of increasing [DEDIC] to 0.44
mM AGP quenches the native tryptophan ¯uorescence
of the protein. The ordinate is the corrected, normalized
¯uorescence (F) with excitation at 278 nm. Nominal
[DEDIC] are as shown. (B) Relative quenching (de®ned
as the quotient F divided by F in the absence of DEDIC)
is invariant over a range of emission wavelengths.
n1
and n2
indeterminate; and for whole AGP
(A)
(A)
(triangles), K1D(native) ˆ 0.049 Æ 0.005 mM, n1 ˆ 0.63 Æ
0.04 mM, and KD2 (native) ˆ 12 Æ 2 mM, with n2 indetermi-
nate. Note also that for both the isolated variants and
whole AGP, the fraction of ¯uorescence unquenched at
limiting nominal [DEDIC] is 0.21 Æ 0.02.
The isolated protein variants F1*S and A were
quenched similarly by DEDIC (Figure 6, circles
and squares, respectively). Both data sets suggest
the existence of two classes of binding sites, and
neither could be ®t satisfactorily to a one-site
model. The data for the F1*S and A variants also
were ®tted by eq. 26, which assumes two sites per
protein molecule. This treatment gave for the
F1*S variant K1D(F1*S/DEDIC) ˆ 0.041 Æ 0.010 mM
nominal concentration (given on the abscissa of
Figure 6), because of high af®nity binding. The KD
(e.g., 0.049 mM) is therefore well below the
apparent high af®nity IC50 (0.21 mM). (In general,
for a one-site model, the KD is smaller than the
IC50 by an amount equal to one-half the nominal
protein concentration.)
In summary, in whole AGP, DEDIC is selective
both for the F1*S high af®nity site over the F1*S
low af®nity site and for the F1*S high af®nity site
over the A high af®nity site by factors of 700 and
8, respectively.
with n1
ˆ 0.69 Æ 0.02 mM, and KD2 (F1*S/DED-
(F1*S)
ˆ 29 Æ 7 mM with n2
indeterminate. The
(F1*S)
normalized amplitudes of the two processes were
IC)
A1
ˆ 0.62 Æ 0.03 and A2
ˆ 0.20 Æ 0.02
(F1*S)
(F1*S)
with an unquenched fraction having A3
ˆ
ˆ
(F1*S)
0.18 Æ 0.02. For the A variant, KD1 (A/DEDIC)
Proof of Method
0.31 Æ 0.10 mM and K2D(A/DEDIC) ˆ 8.8 Æ 0.7 mM
Using DEDIC to Determine Inhibition Constants
for Drug Binding to the Isolated Variants
with both n1
and n2
indeterminate. The
(A)
(A)
normalized amplitudes of the two processes were
A1 ˆ 0.31 Æ 0.02 and A2 ˆ 0.44 Æ 0.02 with an
AMI and BUP are members of a large class of
ligands that produce no appreciable alteration in
the ¯uorescence of a protein when they bind.
Indeed, AMI and BUP were chosen to illustrate
(A)
(A)
unquenched fraction having Aunq ˆ 0.25 Æ 0.02.
Note that in these experiments the free con-
centration of DEDIC is much smaller than the
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

1416
COGSWELL ET AL.
Figure 8. Addition of the nonquenching drug BUP to
a mixture of 1 mM DEDIC (nominal) and 0.39 mM of the
isolated variant results in ¯uorescence recovery.
(Squares and circles represent the A and F1*S variants,
respectively.) This recovery was successfully ®tted to a
model of competitive inhibition (eq. 36), yielding
KI(BUP/A) ˆ 1.9 Æ 0.4 mM and KI(BUP/F1*S) ˆ 1.7 Æ
0.1 mM.
Figure 7. (A) Fluorescence recovery following addi-
tion of AMI to a pre-equilibrated solution of 0.39 mM A
variant and 1 mM DEDIC (squares), 2.5 mM DEDIC
(circles), or 10 mM DEDIC (triangles). All concentra-
tions are nominal. The recovery curves were plotted
consistent with a model of competitive antagon-
ism. Moreover, the measured inhibition constants
were comparable to those determined by Herve
Â1
simultaneously to a model of competitive inhibition
using equilibrium dialysis: KD(A/AMI) ˆ 0.30 Æ
0.02 mM and KD(F1*S/AMI) ˆ 13.5 Æ 13.2 mM.
In competition with DEDIC, BUP also reversed
the quenching of both variants, A and F1*S
(Figure 8, squares and circles, respectively). This
recovery was successfully ®tted to a model of com-
petitive inhibition, yielding KI(BUP/A) ˆ 1.9 Æ
0.4 mM and KI(BUP/F1*S) ˆ 1.7 Æ 0.1 mM. BUP
therefore binds with equal af®nity to both var-
iants. Although the apparent af®nity of BUP to
native AGP has been reported (0.9 mM),16 its
af®nity to the isolated variants has not. However,
given that BUP binds to both with equal af®nities,
its apparent KD for binding to whole AGP is
ꢀ2 mM, which is in near agreement with the
previously reported KD of 0.9 mM (standard error
not reported), as determined by BUP competition
experiments on a mixture of AGP using the
¯uorescent probe 8-ANS.
AMI
I(A)
(eq. 36) to yield K
ˆ 0.10 Æ 0.01 mM. (B) Fluore-
scence recovery following addition of AMI to a pre-
equilibrated solution of 0.39 mM F1*S variant and
0.2 mM DEDIC (squares), 0.4 mM DEDIC (circles), or
1 mM DEDIC (triangles). Fitting the data to eq. 36 gave
AMI
I(F1*S)
K
ˆ 4.6 Æ 0.3 mM.
the utility of DEDIC in the spectro¯uorometric
determination of ligand af®nity to the isolated
variants.
The equilibrium ¯uorescence recovery that is
reached following addition of AMI to DEDIC-
quenched A and F1*S variants is shown in
Figures 7A and 7B, respectively. Experiments
were repeated several times at different nominal
[DEDIC]. Using a model of competitive antagon-
ism, eq. 36 (see Appendix) and the parameters
already reported for DEDIC binding, KI(AMI) for
the A and F1*S variants were determined to be
0.10 Æ 0.01 mM (A) and 4.6 Æ 0.3 mM (F1*S).
For both variants, the simultaneous ®t of the
several curves by single values of KI(AMI) was
Using DEDIC to Determine Drug Af®nity for the
F1*S Variant in Whole (Native) AGP
The ¯uorescence recovery following addition of a
competitor (AMI and BUP) to a pre-equilibrated
mixture of 1 mM DEDIC and 0.39 mM native AGP
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

NEW PROBE TO MEASURE DRUG BINDING TO F1*S OF HUMAN ALPHA 1-ACID GLYCOPROTEIN
1417
Potential of DEDIC as a More General Probe of
Drug±Protein Interactions
Preliminary experiments in this lab have shown
that DEDIC quenches the native tryptophan
¯uorescence of human serum albumin (HSA) in
a concentration-dependent manner. Fitting the
data (not shown) to a single-binding-site model
gave a KD of 3.3 Æ 0.5 mM. We also observe that
DEDIC interacts with chain A of the lipocalin
bovine beta lactoglobulin (BLG-A), a structural
homologue of AGP, to produce changes in ¯uor-
escence (data not shown). With BLG, however, the
concentration dependence is biphasic. At low
(substoichiometric) concentrations of DEDIC,
the ¯uorescence is enhanced, but further eleva-
tion of [DEDIC] results in ¯uorescence quench-
ing. Experiments are ongoing to investigate both
the identity of the DEDIC binding sites on these
proteins and to examine the displacement of
DEDIC by added drugs.
DISCUSSION
Figure 9. Drug inhibition constants for the F1*S
variant were determined using DEDIC and whole
(unfractionated) AGP. The addition of the nonquench-
ing drugs BUP and AMI to a mixture of 1 mM DEDIC
(nominal) and 0.39 mM native AGP resulted in ¯uores-
cence recovery. The curves were ®tted to eq. 36 to yield
Design and Synthesis of DEDIC
DEDIC was the product of an effort to character-
ize the binding of local anesthetics (LAs) to AGP.
The LA BUP, for example, had been reported to be
heavily (95%) bound to AGP in human serum.17
However, we observed that many LAs (including
BUP and lidocaine) produced little or no change in
either the absorbance or ¯uorescence of AGP on
addition.
AMI
BUP
inhibition constants for K
and K
of 3.0 Æ 0.5
D(F1*S)
D(F1*S)
and 1.9 Æ0.1 mM, for AMI and BUP, respectively.
Thus, we based the structure of DEDIC on
that of an archetypal LA, having a hydrophobic
aromatic moiety connected to a relatively hydro-
philic tertiary amine via a polar (ester or amide)
linker. For DEDIC, an amide bond was selected
because of its greater resistance both to hydro-
lysis and degradation by adventitious pseudo-
acetylcholinesterases in biological samples. To
produce a likely quencher, we chose to append two
iodine atoms to the hydrophobic ring. Speci®cally,
a salicylate moiety was chosen because of its
directing effects during iodination, although the
®nal synthesis of DEDIC was achieved without
having to perform the iodination.
Although it is plausible that the appended
iodines of DEDIC are indeed responsible for
its quenching of AGP, it is equally plausible
that quenching occurs via ¯uorescence resonance
energy transfer between an excited tryp-
tophyl donor (lmax(em) ˆ 340 nm and DEDIC
(nominal concentrations) is shown in Figure 9.
Given the ꢀ10-fold selectivity of DEDIC for
binding to the high af®nity site on the F1*S
variant over both the sites on the A variant and
the lower af®nity site on F1*S, it was possible to
select a nominal DEDIC concentration (1 mM with
[AGP] ˆ 0.39 mM) for which DEDIC occupied only
high af®nity F1*S sites. Therefore, in the compe-
tition experiment, one observes only the displace-
ment of DEDIC from F1*S, regardless of whether
or not the A variant is present in solution. In this
manner, KI(AMI/F1*S) was determined to be
3.0 Æ 0.5 mM, which is in agreement with the KI
already determined with the isolated F1*S var-
iant (i.e., 4.6 Æ 0.3 mM). Likewise, KI(BUP/F1*S)
was determined to be 1.9 Æ 0.1 mM, which is in
agreement with that previously determined using
the isolated F1*S variant (i.e., 1.7 Æ 0.1 mM).
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

1418
COGSWELL ET AL.
(lmax(abs) ˆ 355 nm). Signi®cantly, two molecules
that do not quench protein ¯uorescence on
binding to high af®nity sites, BUP and AMI, are
nonabsorbing at wavelengths > 280 and 300 nm,
respectively. Moreover, the drug DIP is a quen-
cher; although lacking iodines, it absorbs in the
region of maximum AGP ¯uorescence, with
separate peaks centered near 280 and 420 nm.
We did not attempt to differentiate between these
mechanisms in this report.
Finally, a DED moiety (as occurs on the LA
procainamide) was employed as a hydrophilic
head after an attempt with the less hydrophilic
dibutyl analog resulted in a probe having a
solubility too low to be useful.
Con®rming our assignment of chromato-
graphic fractions to the AGP variants, the af®nity
of the A and F1*S variants for DIP (0.75 Æ 0.12
and 0.07 Æ 0.02 mM, respectively) agree to within a
factor of three with those previously reported by
Â7
Herve using equilibrium dialysis: 2.9 Æ 0.04 and
0.31 Æ 0.02 mM, respectively. Moreover, our ratio
of the af®nities of DIP for the F1*S and A variants
Â
(10.7) is comparable to Herve's (9.4). The small
(factor of three) differences between the absolute
af®nities are not unexpected given the differences
in both the ionic strength and composition of the
buffer. Furthermore, we noted that DIP could
form supersaturated solutions, which decayed
to their equilibrium solubilities over a period of
Â
The synthesis of DEDIC as described herein
was convenient for making large quantities of the
compound from readily available, inexpensive
materials. Our previous attempts had involved
iodinating 2-hydroxy-N-[2(diethylamino)ethyl]-
benzamide, but this procedure yielded a mixture
of mono- and diiodinated products that were
dif®cult to isolate on a preparative scale. The suc-
cess of the current method is particularly de-
pendent on the speci®c conditions described
because they minimized polymerization of the
diiodosalicyl chloride intermediate and also ef®-
ciently removed principal contaminants from the
product.
hours to days. Herve reported using concentra-
tions of DIP in her buffer solutions that were up to
10-fold more concentrated than we could achieve
in ours, which also might account for the small
discrepancy.
Utility of DEDIC as a Probe of Interactions
Between Drugs and Isolated AGP Variants
Addition to a pre-equilibrated solution of AGP
and DEDIC of a drug that minimally (0±10%)
quenches, such as BUP or AMI, results in
¯uorescence recovery from which the equilibrium
inhibition constants can be calculated without
radiolabelling or drug modi®cations of any kind.
Equilibrium interactions are achieved within a
few seconds, and one sample of protein can be
examined with increasing concentrations of the
drug in question.
For both variants, the simultaneous ®t of the
several curves by single values of KI(AMI) was
consistent with a model of competitive antagon-
ism. Moreover, the measured inhibition con-
stants, 0.10 Æ 0.01 mM (A) and 4.6 Æ 0.3 mM
(F1*S), were comparable to those determined by
Noteworthy Physical Properties
With pKas of 5.0 and 9.8, DEDIC is a zwitterion at
physiologic pH (7.4). The acidity of the phenolic
hydrogen of DEDIC was initially surprising, but
may be explained, in part, by the inductive effects
of the ortho and para iodines. Additionally, if
zwitterionic DEDIC can participate in either
intra- or intermolecular hydrogen bonding, then
this bonding would also act to lower the observed
pK1a. In agreement with these hypothetical inter-
actions, it has been reported that the structurally
analogous drug raclopride,18 a 3,5-dichlorosalicy-
lamide, has a pK1a near 4 and pK2a near 9.5.
Â1
Herve using equilibrium dialysis: KD(A/AMI) ˆ
0.30 Æ 0.02 mM and KD(F1*S/AMI) ˆ 13.5 Æ
13.2 mM. As was observed with DIP, the calcu-
lated inhibition constants for AMI were each
roughly three-fold lower than those previously
determined using equilibrium dialysis.1 As with
DIP, our calculated selectivity of AMI for the F1S
variant (46 Æ 5) agrees with that reported by
Separation of AGP Into the Major Variants
A and F1*S
Â
The chromatographically separated variants A
and F1*S possessed absorbance and ¯uorescence
spectra of the same shape as that of whole AGP.
Their relative abundances (81% F1*S and 19% A)
were similar to those reported by Halsall15 (75
and 25%, respectively).
Herve (i.e., 45 Æ 44).
Of further interest here is the ®nding that BUP
binds to the two variants with comparable af-
®nities. Whereas a KD for BUP binding to whole
AGP has been previously reported (0.9 mM16),
these are the ®rst reports of the af®nities of BUP
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

NEW PROBE TO MEASURE DRUG BINDING TO F1*S OF HUMAN ALPHA 1-ACID GLYCOPROTEIN
1419
for the individual variants. As reported by
between the drug and a ¯uorescent probe
1
Â
Herve, the structurally and functionally related
whereby the displacement of probe by the drug
is observed as a decrease in the ¯uorescence of the
probe. Then, given the af®nity of the probe for
AGP, the af®nity of the drug may be calculated.
The probes ANS21,22,23 and auramine O are com-
monly used for this purpose.
compound lidocaine also displays little variant
selectivity.
Utility of DEDIC as a Probe of Drug Interactions
With the F1*S Variant in Whole (Native) AGP
Here, however, we report the use of the com-
plementary method: the use of a reversibly bound
probe (DEDIC) as a quencher of the intrinsic
¯uorescence of AGP. Although the drugs CPZ and
DIP have been reported to quench AGP,19 they
have not been used as spectro¯uorometric probes
of drug binding. Even so, DEDIC would possess a
signi®cant advantage over them in terms of its
ability to probe variant-speci®c binding. CPZ does
not exhibit variant-selective binding as indicated
by its almost equal af®nities to bind the F1*S
and A variants (0.37 Æ 0.03 and 0.74 Æ 0.09,7
respectively). Furthermore, DEDIC is a more
versatile probe than DIP. Indeed, while posses-
sing a selectivity for the F1*S variant comparable
to that of DIP, DEDIC not only binds F1*S
with almost a 10-fold higher af®nity (0.041 Æ
0.010 mM) than does DIP (0.31 Æ 0.02 mM7), but
also has 40-fold greater solubility, a property
that permits its additional use as a general probe
of drug binding to lower af®nity sites on either
variant.
Because DEDIC binds selectively to the high
af®nity site on the F1*S variant, it can be used to
probe the association of drugs to F1*S when both
variants are present, as in whole AGP. The
agreement between the inhibition constants cal-
culated for AMI and BUP, KIs (F1*S), obtained in
whole AGP and those obtained in the isolated
variant con®rms the utility of DEDIC as a probe of
binding to the F1*S variant in the naturally
occurring mixture. Moreover, the use of the
strongly A variant-selective drug AMI provided
a particularly stringent test of the ability of
DEDIC to selectively probe binding to the F1*S
variant in native AGP. In short, DEDIC offers a
convenient method for the accurate determina-
tion of ligand af®nities without the need to ®rst
separate the variants.
Comparing DEDIC With Other Probes of
Drug±AGP Binding
Spectro¯uorometric measurements of drug±pro-
tein binding may be of two types; they are, direct
and indirect. Direct spectro¯uorometric determi-
nation of drug af®nities requires that either the
drug or the protein exhibit binding-dependent
changes in ¯uorescence. Clearly, this require-
ment constrains the utility of the method because
the binding of many drugs produces no detectable
changes in optical signals; that is, a drug may not
possess an appreciable ¯uorescence or, if it does,
the spectrum is not altered by protein binding.
Notwithstanding, this method has been applied to
examine the binding of both non¯uorescent,
quenching drugs (e.g., DIP and CPZ19) and
¯uorescent drugs whose intensity increases on
binding (e.g., warfarin20). We used this approach
in our direct determination of the KD for DIP
binding to the F1*S variant, 0.13 Æ 0.03 mM,
which approximates the previously reported
value of 0.38 Æ 0.01 mM as determined by equili-
brium dialysis in 0.067 M phosphate buffer at
pH 7.40.1
For drugs that produce a relatively small
¯uorescence change (e.g., < 20%), DEDIC pro-
vides a means of amplifying the signal change
that occurs on binding because limiting concen-
trations of DEDIC quench roughly 80% of the
native protein ¯uorescence. In addition, because
DEDIC binds selectively to the F1*S variant of
AGP, the F1*S af®nities of partial quenchers may
be determined in a mixture of variants, such as
occurs naturally in AGP, thereby eliminating the
need to separate them chromatographically. For
example, limiting concentrations of the tricyclic
antidepressant AMI quench ꢀ8% of the ¯uores-
cence of AGP (data not shown), but this small
signal change did not permit us to directly
determine its precise af®nity. Using DEDIC as a
probe of AMI binding, however, we were able to
determine af®nities of the drug for both variants.
In addition, DEDIC allowed us to determine the
F1*S af®nity of this strongly A variant-selective
drug in native AGP.
Our work suggests that the binding of many
drugs produces little, if any, change in protein
¯uorescence. Indeed, only minimal (0±8%) que-
nching of the ¯uorescence of AGP is produced by
In cases where the direct observation of bind-
ing is not possible, indirect methods have been
used. Traditionally, these rely on antagonism
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

1420
COGSWELL ET AL.
representative drugs from three major classes:
the local anesthetic BUP, the steroid progester-
one, and the tricyclic antidepressant AMI.
For drugs producing larger ¯uorescence
changes (e.g., > 20%) on binding, it is obviously
not necessary to use a probe such as DEDIC to
obtain a suitably large signal. Even with these
quenching drugs, however, DEDIC could still be
useful as a probe of drug binding to the F1*S
variant in native AGP.
ACKNOWLEDGMENTS
This research was supported in part by NIH grant
GM35647 (to GRS) and a graduate stipend to LPC
from the BWH Anesthesia Foundation.
APPENDIX
Derivation of the One-Site Binding Isotherm
Interest in AGP has been growing amid reports
that it may play a role in mediating in¯amma-
tion,24,25 so we hope that the methods presented
here will prove useful in studying drug±AGP
binding and perhaps contribute to our knowledge
of the structure and function of this important
protein.
The drug D is assumed to bind reversibly to the
protein P to form a complex DP with an
equilibrium dissociation constant of KD. The
nominal [D] and [P] are denoted DT and PT,
respectively, and the ¯uorescence of P and DP is
given by the product of their concentrations with
their molar emissivity f. No ¯uorescence arises
from free drug.
From the equilibrium condition and mass
conservation we obtain the following:
NOMENCLATURE
DEDIC
2-hydroxy-3,5-diiodo-N[2(diethyla-
mino)ethyl] benzamide
alpha 1-acid glycoprotein
human serum albumin
bovine beta lactoglobulin, Chain A
imipramine
‰DŠ‰PŠ
KD
ˆ
ꢁ1†
‰DPŠ
AGP
HSA
BLG-A
IMI
DT ˆ ‰DŠ ‡ ‰DPŠ
ꢁ2†
ꢁ3†
PT ˆ ‰PŠ ‡ ‰DPŠ
From eqs. 2 and 3, it follows that
‰DPŠ ˆ DT ‰DŠ
DIP
dipyridamole
bupivacaine
BUP
8-ANS
AMI
F1*S
AGP-A
8-anilinonaphthalenesulfonic acid
amitriptyline
F1 and S variants of AGP
gene encoding for the F1 and S
variants of AGP
ꢁ4†
ꢁ5†
‰PŠ ˆ ‰PTŠ ‰DPŠ ˆ PT ꢁDT ‰DŠ†
AGP-B/B' gene encoding for the A variant of
Substituting eqs. 4 and 5 into eq. 1 and re-
arranging yields a quadratic equation:
AGP
DED
DIS
N,N-diethylethylenediamine
3,5-diiodosalicylic acid
2
‰DŠ ‡ ‰DŠꢁPT DT ‡ KD† KDDT ˆ 0
ꢁ6†
SAM
standard aqueous medium, contain-
ing 0.15 M NaCl and 5 mM MOPS
buffer adjusted to pH 7.40
for which the only physically meaningful root is
ꢀ
1
‰DŠ ˆ
DT PTr KD
Fraw
F
raw ¯uorescence
2
ꢁ
ꢁ7†
q
¯uorescence corrected for inner ®lter
and Raman errors and then normal-
ized to unity
‡
KD2 ‡ 2KDPT ‡ 2KDDT ‡ P2T 2DTPT ‡ D2T
F(lex/lem
)
corrected, normalized ¯uorescence
with excitation and emission at lex
and lem
Neglecting the effects of dynamic quenching
(which we already argued to be negligible), the
observed ¯uorescence (corrected for inner-®lter
effects and background; e.g., Raman scatter from
water) is given by
Frel
quotient of the corrected, normalized
¯uorescence emission spectrum in
the presence of a quencher divided
by that obtained in its absence
ratio of binding site concentration to
nominal protein concentration
F ˆ fP‰PŠ ‡ fDP‰DPŠ
ꢁ8†
n
because the molar ¯uorescence of the drug is zero.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

NEW PROBE TO MEASURE DRUG BINDING TO F1*S OF HUMAN ALPHA 1-ACID GLYCOPROTEIN
1421
Substituting eqs. 4 and 5 into eq. 8 yields
‰P1Š ˆ PT1 ‰DPŠ ˆ P1T ꢁDT ‰DŠ†
ꢂ
ꢃ
ꢂ
ꢃ
ꢁ16†
KD
‰DŠ
‰P2Š ˆ PT2 ‰DPŠ ˆ P2T ꢁDT ‰DŠ†
F ˆ fP
PT ‡ fDP
PT
ꢁ9†
KD ‡ ‰DŠ
KD ‡ ‰DŠ
where [D] is given by eq. 7.
If the ¯uorescence data are normalized to unity
from drug±free protein, then
Substituting eqs. 16 into eqs. 14 and solving for
[DP] yield the familiar relations
ꢂ
ꢃ
‰DŠ
‰DP1Š ˆ
‰DP2Š ˆ
P1T
P2T
F ꢂ 1 ˆ fPPT
ꢁ10†
KD1 ‡ ‰DŠ
‰DŠ
ꢂ
ꢃ
ꢁ17†
KD2 ‡ ‰DŠ
Using eqs. 3 and 10, substituting into eq. 8, and
rearranging yields
F ˆ 1 ꢁfP fDP†PT ‡ ‰DPŠ
ꢁ11†
Now, from mass conservation of drug
D ˆ DT ‰DPŠ ‰DP2Š
ꢁ18†
1
De®ning the amplitude A of the change due to
and substituting eqs. 17, the following relation-
ship is formed:
drug-induced quenching as follows:
A ꢂ ꢁfP fDP†PT
and substituting eq. 4 into eq. 11 and rearranging
gives
ꢁ12†
ꢂ
ꢃ
ꢂ
ꢃ
‰DŠ
‰DŠ
D ˆ DT
P1T
PT2 ꢁ19†
KD1 ‡ ‰DŠ
KD2 ‡ ‰DŠ
Equation 19 is a cubic equation that may be
solved27 for [D]:
ꢂ
ꢃ
‰DŠ
F ˆ 1
A
ꢁ13†
‰DŠ ‡ KD
p
Á
a
2
3
ꢀ
‰DŠ DT; PT1 ; P2T; KD1 ; KD2
ˆ
‡
a2 3b Á cos
where [D] is not identical to the nominal drug
concentration, but is again given by eq. 7.
3
3
ꢁ20†
where
Derivation Of the Two-Site Binding Isotherm
a ˆ KD1 ‡ KD2 ‡ PT1 ‡ PT2 ‡ DT
b ˆ KD1 KD2 ‡ P1T ꢁKD2 KD1 † DT ꢁKD1 ‡ KD2 †
c ˆ KD1 KD2 DT
The drug D is assumed to bind reversibly to
two noninteracting sites, P1 and P2, to form
complexes DP1 and DP2, with respective equili-
brium dissociation constants K1D and KD2 .
The nominal concentrations of the sites are
denoted PT1 and PT2. The species P1, P2, DP1, and
DP2 each contributes to the measured ¯uores-
cence F by an amount equal to the product of
their concentrations and their respective molar
¯uorescences, f. No ¯uorescence arises from the
drug.
2
3
2a3 ‡ 9ab 27c
2
6
4
7
5
q
ꢀ ˆ arccos
3
ꢁa2 3b†
Substitution of eq. 20 into eqs. 16 and 17 gives
expressions for the equilibrium values of the
species [P1], [P2], [DP1], and [DP2].
The observed ¯uorescence is given by
From equilibrium and mass conservation, we
get
F ˆ fP ‰P1Š ‡ fP ‰P2Š ‡ fDP ‰DPŠ ‡ fDP ‰DP2Š ꢁ21†
1
2
1
2
1
‰DŠ‰P1Š
KD1
KD2
ˆ
ˆ
Substituting eqs. 16 and 17 into eq. 21 gives
‰DP1Š
‰DŠ‰P2Š
‰DP2Š
ꢂ
ꢃ
ꢂ
ꢃ
KD1
KD2
ꢁ14†
ꢁ15†
F ˆ fP
PT1 ‡ fP
P2T
1
2
KD1 ‡ ‰DŠ
KD2 ‡ ‰DŠ
ꢂ
ꢃ
ꢂ
ꢃ
‰DŠ
‰DŠ
‡ fDP
PT1 ‡ fDP
P2T
1
2
KD1 ‡ ‰DŠ
KD2 ‡ ‰DŠ
PT1 ˆ ‰P1Š ‡ ‰DP1Š
PT2 ˆ ‰P2Š ‡ ‰DP2Š
ꢁ22†
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

1422
COGSWELL ET AL.
which rearranges to
and
2
3
ꢂ
ꢃ
ꢂ
ꢃ
fP K1 ‡ fDP ‰DŠ
fP K2 ‡ fDP ‰DŠ
2a3 ‡ 9ab 27c
2
1
1
2
2
PT1 ‡
P2T
D
D
6
4
7
5
F ˆ
q
y ˆ arccos
:
KD1 ‡ ‰DŠ
KD2 ‡ ‰DŠ
3
ꢁa2 3b†
ꢁ23†
The observed ¯uorescence is given by
F ˆ fP‰PŠ ‡ fPD ‰PD1Š ‡ fPD ‰PD2Š
By mass conservation of protein,
ꢁ29†
ꢁ30†
In addition, if the data are normalized, then
1
2
F ꢂ 1 ˆ fP PT1 ‡ fP PT2
ꢁ24†
1
2
P ˆ PT ‰PD1Š ‰PD2Š
Substituting eq. 30 into eq. 29 and rearranging
yields the analog of eq. 25:
Substituting into eq. 21 making use of eqs. 15 and
24 gives
F ˆ 1 ꢁfP
fDP †PT1 ꢁfP
fDP †PT2
ꢁ25†
1
1
2
2
F ˆ fPPT ꢁfP fPD †DT1 ꢁfP fPD †DT2 ꢁ31†
1
2
Finally, substituting for DP1 and DP2 using eqs.
17 gives
Substituting the analogs of eq. 17
ꢂ
ꢃ
ꢂ
ꢃ
ꢂ
ꢃ
‰PŠ
‰DŠ
‰DŠ
‰PD1Š ˆ
‰PD2Š ˆ
D1T
D2T
F ˆ 1 A1
A2
ꢁ26†
KD1 ‡ ‰PŠ
‰PŠ
KD1 ‡ ‰DŠ
KD2 ‡ ‰DŠ
ꢂ
ꢃ
ꢁ32†
where the amplitudes A1 and A2 are de®ned as
KD2 ‡ ‰PŠ
A1 ꢂ ꢁfP
A2 ꢂ ꢁfP
fDP †PT1
1
2
1
ꢁ27†
gives
fDP †PT2
2
"
#
"
#
‰PŠ
‰PŠ
FˆfPPT A1 Á
A2 Á
ꢁ33†
K1D‡‰PŠ
K2D ‡ ‰PŠ
Equations 23 and 27 are also applicable to the
case of both independent protein binding sites
being on the same molecule.
where the amplitudes A1 and A2 are de®ned as
A1 ꢂ ꢁfP fPD †DT1
1
ꢁ34†
A2 ꢂ ꢁfP fPD †DT2
2
Derivation of the One-Site Competition
Equation
If the data are normalized to unity, then
F ꢂ 1 ˆ fP Á PT
ꢁ35†
ꢁ36†
The situation here (two drugs D1 and D2 compet-
ing for a single site on the protein) is mathema-
tically analogous to that of two proteins
competing for a single drug, which has been
described in detail in the last section. Thus by
transposing the ``drug'' and ``protein'' terms in eq.
20, one obtains an expression for the concentra-
tion of [P] as a function of the total drug
concentrations.
and
"
#
"
#
‰PŠ
‰PŠ
F ˆ 1 A1
A2
K1D‡‰PŠ
K2D‡‰PŠ
REFERENCES
Â
1. Herve F, Caron G, Duche JC, Gaillard P, Abd RN,
Tsantili-Kakoulidou A, Carrupt PA, d'Athis P,
Tillement JP, Testa B. 1998. Ligand speci®city of
the genetic variants of human alpha1-acid glyco-
protein: generation of a three-dimensional quanti-
tative structure-activity relationship model for
drug binding to the A variant. Mol Pharmacol
54:129±138.
p
Á
a
2
3
y
3
ꢁ28†
‰PŠ PT; DT1 ; D2T; KD1 ; KD2
ˆ
‡
a2 3b Á cos
3
where
a ˆ KD1 ‡ KD2 ‡ DT1 ‡ DT2 ‡ PT;
b ˆ KD1 KD2 ‡ D1TꢁKD2 KD1 † PTꢁKD1 ‡ KD2 †;
c ˆ KD1 KD2 PT;
2. Kremer JM, Wilting J, Janssen LH. 1988. Drug
binding to human alpha-1-acid glycoprotein in
health and disease. Pharmacol Rev 40:1±47.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001

NEW PROBE TO MEASURE DRUG BINDING TO F1*S OF HUMAN ALPHA 1-ACID GLYCOPROTEIN
1423
3. Routledge PA. 1986. The plasma protein binding of
basic drugs. Br J Clin Pharmacol 22:499±506.
4. Eap CB, Baumann P. 1989. The genetic polymorph-
ism of human alpha 1-acid glycoprotein. Prog Clin
Biol Res 300:111±125.
15. Sun H. (H. Brian Halsall, Advisor) Characteriza-
tion of the ligand-binding properties of human
orosomucoid gene products I and Ii (lysine, histi-
dine, warfarin, propranolol). Thesis, University Of
Cincinnati. Cincinnati, OH; pp. 1±214.
Â
5. Herve F, Duche JC, Barre J, Millot MC, Tillement
16. Essassi D, Zini R, Tillement JP. 1989. ANS-drug
interactions to AAG and HSA ¯uorescence and
equilibrium dialysis studies. Prog Clin Biol Res
300:423±426.
17. Tucker GT, Mather LE. 1975. Pharmacology of
local anaesthetic agents. Pharmacokinetics of local
anaesthetic agents. Br J Anaesth 47: suppl.213±
224.
JP. 1992. pH titration curves of the desialylated
human alpha 1-acid glycoprotein variants by
combined isoelectrofocusing-electrophoresis: Utili-
zation in the development of a fractionation method
for the protein variants by chromatography on
immobilized metal af®nity adsorbent. J Chroma-
togr 577:43±59.
Â
6. Herve F, Gomas E, Duche JC, Tillement JP. 1993.
18. Pagliara A, Carrupt PA, Caron G, Gaillard P, Testa
B. 1997. Lipophilicity pro®les of ampholytes.
Chem.Rev (Washington, D.C.) 97:3385±3400.
19. Carrupt PA, Tsai RS, El Tayar N, Testa B, De
Paulis T, Hoegberg T. 1991. Ionization behavior
and ionization-dependent conformation of raclo-
pride, a dopamine D2 receptor antagonist. Helv
Chim Acta 74:956±968.
20. Friedman ML, Schlueter KT, Kirley TL, Halsall
HB. 1985. Fluorescence quenching of human
orosomucoid. Accessibility to drugs and small
quenching agents. Biochem J 232:863±867.
21. Zini R, Copigneaux C, Tillement JP. 1989. 1-
Anilino-naphthalene 8-sulfonic acid (ANS) as a
probe for the binding of antidepressant drugs to
human alpha 1-acid glycoprotein (AAG). Prog Clin
Biol Res 300:417±421.
Evidence for differences in the binding of drugs to
the two main genetic variants of human alpha 1-
acid glycoprotein. Br J Clin Pharmacol 36:241±249.
Â
7. Herve F, Duche JC, d'Athis P, Marche C, Barre J,
Tillement JP. 1996. Binding of disopyramide,
methadone, dipyridamole, chlorpromazine, ligno-
caine and progesterone to the two main genetic
variants of human alpha 1-acid glycoprotein:
Evidence for drug-binding differences between the
variants and for the presence of two separate drug-
binding sites on alpha 1-acid glycoprotein. Phar-
macogenetics 6:403±415.
8. Dente L, Pizza MG, Metspalu A, Cortese R. 1987.
Structure and expression of the genes coding for
human alpha 1-acid glycoprotein. EMBO J 6:2289±
2296.
9. Tomei L, Eap CB, Baumann P, Dente L. 1989. Use
of transgenic mice for the characterization of
human alpha 1-acid glycoprotein (orosomucoid)
variants. Hum Genet 84:89±91.
22. Epps DE, Raub TJ, Kezdy FJ. 1995. A general,
wide-rage spectro¯uorometric method for measur-
ing the site- speci®c af®nities of drugs toward
human serum albumin. Anal Biochem 227:342±
350.
Â
10. Herve F, Gomas E, Duche JC, Tillement JP. 1993.
Fractionation of the genetic variants of human
alpha 1-acid glycoprotein in the native form by chro-
matography on an immobilized copper(II) af®nity
adsorbent. Heterogeneity of the separate variants
by isoelectrofocusing and by concanavalin A af®nity
chromatography. J Chromatogr 615:47±57.
11. Sun H, Dage JL, Halsall HB. 1996. Separation of
human orosomucoid major gene products using
immobilized copper af®nity chromatography and
identi®cation of the metal-interactive residues.
Chromatographia 46:430±436.
23. Maruyama T, Furuie MA, Hibino S, Otagiri M.
1992. Comparative study of interaction mode of
diazepines with human serum albumin and alpha
1-acid glycoprotein. J Pharm Sci 81:16±20.
24. Johansen AK, Willassen NP, Sager G. 1992.
Fluorescence studies of beta-adrenergic ligand
binding to alpha 1-acid glycoprotein with 1-ani-
lino-8-naphthalene sulfonate, isoprenaline, adre-
naline and propranolol. Biochem Pharmacol
43:725±729.
25. van Dijk W, Brinkman-van der Linden EC,
Havenaar EC. 1998. Occurrence and possible
function of in¯ammation-induced expression of
sialyl Lewis-x on acute-phase proteins. Adv Exp
Med Biol 435:145±150.
12. Oravcova J, Bohs B, Lindner W. 1996. Drug-protein
binding sites. New trends in analytical and experi-
mental methodology. J Chromatogr B Biomed Appl
677:1±28.
13. Lakowicz JR. 1999. In Principles of Fluorescence
Spectroscopy, 2nd Edition. New York: Kluwer
Academic/Plenum Publishers.
14. Hof M, Vajda S, Fidler V, Karpenko V. 1996.
Picosecond Tryptophan Fluorescence of Human
Blood Serum Orosomucoid. Collect Czech Chem
Commun 61:808±818.
26. Shiyan SD, Bovin NV. 1997. Carbohydrate compo-
sition and immunomodulatory activity of different
glycoforms of alpha1-acid glycoprotein. Glycoconj J
14:631±638.
27. Wang ZX, Jiang RF. 1996. A novel two-site binding
equation presented in terms of the total ligand
concentration. FEBS Lett 392:245±249.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 90, NO. 9, SEPTEMBER 2001