Bioconjugate Chem. 1990, 1, 278-284
Kinetics of the Dissociation of Indium-(p-Substituted-benzyl)ethylenediaminetetraaceticAcid Hapten Analogues from the Monoclonal Anti-Hapten Antibody CHA255 Damon L. Meyer,. Mark Fineman, Barbara W. Unger, and James M. Frincke Hybritech Incorporated, P.O. Box 269006, San Diego, California 92126. Received March 16, 1990
Half-lives were measured for the dissociation of a series of 20 indium-benzyl-EDTA derivatives from a monoclonal antibody that binds to them. Most haptens gave expected monoexponential dissociation curves with half-lives ranging from -8 to -100 min at 22 f 1 "C. Precise (*-2.5%) determinations were made using centrifugal ultrafiltration to separate free from bound hapten. A strong pH dependence of the dissociation half-life was found for the two haptens studied. Activation enthalpies were identical (23 f 1 kcal/mol) for the dissociation of four haptens, suggesting that, in contrast to individual rate constants, this parameter is insensitive to hapten modification. The dissociation half-lives provided evidence for the location of a positive charge in the binding site, but gave no clear indication of the role of hydrophobic interactions or of steric requirements in hapten binding. While variations in ionic strength had no effect on the dissociation rate, lowering surface tension with dioxane increased the rate somewhat. Three hapten-antibody complexes showed biexponential dissociation rates. It is postulated that this results from distinct conformations of the complex dissociating at different rates. The dissociation rate constant was found to be an extremely sensitive indicator of the hapten-antibody interaction that can be measured very precisely.
INTRODUCTION
In vivo, anti-hapten monoclonal antibodies can be used to control the pharmacokinetic parameters of haptens and molecules to which a hapten moiety has been chemically attached (1, 2 ) . For small hydrophilic haptens, the presence of anti-hapten antibodies dramatically extends the serum and whole-body lifetime of the hapten. "Hybrid" or "bifunctional" antibodies which have one binding site with high affinity for a hapten and the other binding site with high affinity for a cell marker, such as a tumorassociated antigen, not only prolong the lifetime of the hapten but also cause the hapten to preferentially accumulate at the target in nude mouse tumor models and in humans ( 3 ) . We have begun animal experiments aimed a t exploiting these properties for delivery of radioimaging agents, radiotherapy agents, and chemotherapy agents (2). The biodistribution of a radiolabeled hapten is observed to change if the hapten structure is modified, even though the hapten is complexed to an anti-hapten antibody over 100 times its molecular weight. Some of this variability may be attributed to differences in the binding characteristics of the antibody t o t h e related haptens. Therefore, an understanding of how hapten derivatization will affect antibody binding is useful in designing a hapten-antibody system. Particularly important is the dissociation rate constant, since this parameter should limit the rate at which a hapten leaves the carrier antibody either to diffuse away from the target site or to be ingested into a target cell. Retention time at the target, which may affect therapeutic dose and targetlbackground ratio, is a function of the complex half-life. Information about charge location in, and the size of, the binding site should. emerge from knowledge of the dissociation rate constants for a properly selected group of hapten analogues. Dissociation rates can be used to screen clones for particular applications and to detect changes in binding due to site-specific mutation. The 1043-1802f 9012901-0278$02.50/0
information can also be used to direct the synthesis of new haptens for use in a particular application. Thus, we elected to develop a simple procedure to measure monoclonal antibody dissociation rate constants for a series of chemically related haptens with the intent of applying the information to an in vivo delivery system. For pharmacological prediction and evaluation, rate constant values at 37 "C are of primary importance, but it was found in the course of this study that the dissociation rates in our system can be measured much more precisely at room temperature than a t 37 "C. Consequently, the temperature dependence of the dissociation rate was measured for selected haptens to see whether a relative ordering of dissociation rates at room temperature would extrapolate to body temperature. In addition, a knowledge of the activation parameters for the dissociation provides further insight into the physical nature of the antihapten antibody system for future applications. In fact, it has been concluded ( 4 ) that the activation energy for the dissociation reaction is the primary determinant of the affinity constant. Dissociation and association rate constants have been reported for monoclonal (5-7) and polyclonal(7-10, review: 4 ) antibodies. It is concluded that antibody affinities correlate inversely with dissociation rate Constants and that variations in association rate constants are relatively small (11, 12, 4 ) . Also, Hansch and co-workers (13-16) have developed structure-activity relationships for hapten derivatives binding to polyclonal antibodies. However, no studies have been reported that show quantitatively the sensitivity of the dissociation rate constant to variation in the hapten structure or to variation in solvent conditions for a well-defined anti-hapten monoclonal antibody system. Methods used previously to separate free hapten from complex include adsorption on activated carbon (7)and filtration through cellulose acetate (8,9,4). In this study, we used the Centrifree micropartition apparatus (Ami0 1990 American Chemical Society
Bioconjugate Chem., Vol. 1, No. 4, 1990
Dissociation Kinetics: In-BenryCEDTA’s from CHA255
con Corp.) to separate the unbound hapten from the complex and found this to be a convenient and reproducible method. In this paper, we describe the measurement of the dissociation rate constants for a series of chemically related haptens, all derivatives of the In3+ complex of benzylEDTA, that bind to an anti-hapten monoclonal antibody, designated CHA255. The immunogen from which CHA255 was obtained is a conjugate of (p-thioureidobenzy1)EDTA-In (3) linked to the eamino group of a lysine side chain of keyhole limpet hemocyanin. (p-Isothiocyanatobenzy1)-EDTA and similar (aminobenzy1)-EDTA derivatives (e.g., 4) provide a convenient synthetic route (18) to a wide range of compounds whose indium chelates are all complexed by the antibody. (Benzyl-EDTA chelates of metal ions other than indium are also bound by CHA255 but with significantly lower affinity (19).)Our results indicate that the dissociation rate constant is an extremely sensitive indicator of the antibody-hapten interaction and that it can be measured conveniently with high precision.
(In,
EXPERIMENTAL PROCEDURES Hapten Derivatives. Several of the bifunctional chelates used as hapten derivatives or intermediates were prepared as described previously: 1,2,4 (18);3 (20); 6 and 22 (2). Compound 18 was prepared by treating 17 with glutaric anhydride. Hapten derivatives 7-17, 19, and 20 were prepared by treating 3 with the appropriate amine a t pH -9. Synthesis of the remaining hapten derivatives (5, 21) will be described elsewhere. Purification was conveniently accomplished by ion-exchange chromatography on DEAE Sephadex A25 eluted with a volatile buffer such as triethylammonium formate, followed by lyophilization, or by binding to AG1 ion-exchange resin in 0.5 M formic acid and elution with 8.0 M formic acid. Hapten derivatives were characterized by HPLC retention time with a Hewlett-Packard 10-cm microbore ODS column using a 10-min linear gradient from 50 mM triethylammonium acetate containing 10 mM EDTA to 100% methanol at a flow rate of 0.4 mL/min. Indiumloaded and metal-free chelates were analyzed with this system, giving two characteristic retention times for each hapten. Chelates not containing Ins+ were further characterized by their ‘H and 13C NMR spectra, obtained on a Varian XL-300 Spectrophotometer. These chelates were loaded with In3+ in either ammonium or triethylammonium citrate (18). Antibodies. The antibody used throughout this work was CHA255, a murine antibody of the IgGl type. The immunizations which led to the isolation of CHA255 were performed with a keyhole limpet hemocyanin conjugate of the In3+complex of 3 (17). The “bifunctional” antibody ECHO37 was derived from a fusion of the CHA255 cell line and an anti-CEA monoclonal antibody producing cell line designated CEM231. Kinetic Scheme. It can be shown (21) that isotopeexchange reactions (eq 1) obey first-order kinetics. As Ab-H*
+ H + Ab-H + H*
(1)
applied to antibody-hapten exchange, the general equation is
where x is the fraction of exchange that has occurred and R is the “gross rate of exchange” of labeled and unlabeled hapten (constant for a given reaction mixture). Since,
279
in the case of antibodies and haptens, the affinity constant is very high, the dissociation rate constant is many orders of magnitude smaller than the association rate constant. Consequently, the gross rate of exchange is controlled by the dissociation rate, and if dissociation is a first-order process (eq 3) Ab-H
kl -+
Ab
+H
(3)
and hapten is present in excess, then
R = -k-,([Ab-H]
+ [Ab-H*])
Hence
Experimental conditions have been arranged so that [AbItot is much less than [HItot,so In (1- x ) = - k , t
(4)
If dissociation were a different process, e.g. bimolecular (eq 5) [Ab-H*]
+ [HI k-i
+
[Ab-HI
+ [H*]
(5)
but with other experimental conditions the same, then
R = -k-,[Abl,,[Hl,, and In (1 - x ) = -k-l[H],,t
(6)
Hence, dependence of the observed rate constant on [HI,, distinguishes the dissociation mechanisms. Two simultaneous first-order processes leading to indistinguishable products can be described (22) by equation 7: 1 - x = A [ f exp( -k-,t)
+ (1- f ) exp (-k4)]
(7)
where A is the fraction of isotope originally bound to antibody (approximately l), f is the fraction of complex dissociating with rate constant k-1, and (1- f ) is the fraction of complex dissociating with rate constant k-2. Assay Method. Hapten-antibody dissociation rates were determined by measuring the quantity of ll1In3+chelate released from the antibody-hapten complex as a function of time after addition of a 100-fold excess (over antibody binding sites) of unlabeled hapten. Thus, for each hapten, a set of 700-pL reaction solutions were prepared containing (1)antibody at a concentration of 160 nM, (2) indium-loaded hapten, trace labeled with IllIn (total concentration = 160 nM), (3) normal serum albumin at a total concentration of 1.5 mg/mL, and (4) isotonic phosphate-buffered saline, pH 7.4. At time TOindiumloaded, unlabeled hapten was added to a final concentration of 32 pM. Reactions were stirred by placing the sample containers in a rack mounted on a vortexer head. A t various times after addition of the excess hapten, 400p L portions of the reaction solutions were transferred to Centrifree tubes and centrifuged at lOOg for 8-10 min. The reaction time was taken to be the time between addition of the excess hapten and the time the centrifuge reached
200
Bioconjugate Chem., Vol. 1, No. 4, 1990
the desired ve1ocity.l Six time points, each in duplicate, were obtained over a range of about two reaction halflives for each rate constant determination. A 200-pL portion from each centrifuged and uncentrifuged solution was then counted for ll1In. The relative concentration of free hapten was taken to be the ratio of counts in the centrifuged solution (filtered) to the uncentrifuged solution at each time point. Times were recorded to the nearest 1 / 1 ~min. Room temperature rate constants were measured a t ambient temperature, which remained a t 22 f 1"C. Rate constants at 4 "C (or 5 "C) were measured in a refrigerated room. The rate constant at 0 "C was measured in an icewater bath in a refrigerated room. Rate constants above room temperature were measured in a water bath controlled to f l "C. Stirring was maintained throughout the reactions. Rate constants and standard deviations for the monoexponential dissociations were obtained from a plot of the logarithm of the relative complex concentration (e.g. 1x in eq 4) vs time. These data were analyzed by using the linear-regression module of the RS/3 program (BBN Software Products Corp.). Biexponential data were fit to eq 7, with uniform weighting of each time point, using the R S / 3 program. (For the purposes of curve-fitting, duplicates were averaged before being plotted.) Control samples were included in each experiment to determine (A) that the initial condition of 100% of labeled hapten bound to antibody was met and (B) that in the absence of antibody, 100% of the counts were able to penetrate the Centrifree membrane. If control A showed that a significant portion of the counts did not bind to antibody initially, the average number of counts in the control duplicates was subtracted from all centrifuged and uncentrifuged sample solutions. If control B showed that not all of the counts penetrated the membrane, each count ratio (i.e., the relative concentration of free hapten) was divided by the average of the fraction penetrating the membrane in the control duplicates. RESULTS
Dissociation rates were calculated from linear firstorder plots (Figure 1) for a series of benzyl-EDTA derivatives para-substituted with nitro, amino, thiourea, or acetamide functional groups (Table I). (Compounds 3 and 4 are intermediates in the synthesis of other haptens, and their dissociation rates were not measured.) The room temperature rate constants show that modification of the hapten at the para position does have a sizable effect on the dissociation rate, and that the method used is capable of distinguishing between very similar haptens. For example, the difference in dissociation rate between benzyl- and phenyl-substituted thiourea derivatives (11 and 12) is considerable. Also, the dissociation rates for the series of alkyl carboxylates 13-15 are strikingly different. Statistical treatment of the errors for the room temperature half-life determinations shows that the average relative error for the method is 2.4 f 1.2%. Precise 1 Since the separation is not instantaneous, the choice of what point in the process to use for the time values is somewhat arbitrary. Fortunately, the slope of the first-order dissociation rate plot (and the t l l 2 ) is dependent only on the interval between measurements so that, as long as the separations all require the same amount of time, any point in the process can serve to identify a measurement time.
Meyer et al. h
0
0.01
: I,
; -. X
Q)
8 c -
,
,
,
,
,
,
\i
Y
Y
-2.0
0
20
40
60
80
100 120
Time (min) F i g u r e 1. First-order plot of the dissociation of compound 17 from CHA255 a t 22 "C.
determinations were obtained for reactions with halflives as short as 5.5 min and as long as 1400 min. The reaction rate shows no direct dependence on the concentration of excess hapten (Table 11),indicating that the dissociation is a first-order process. The temperature dependence of the dissociation rate for compounds 14, 17, 18, and 20 follows the expected Arrhenius relationship (Table 111) and is experimentally identical for the four compounds. Hapten derivatives for which rate constants are available at only two temperatures (7, Table I) show a similar variation with temperature. This trend suggests that hapten derivative 22, which gave a biexponential dissociation plot a t 22 "C, binds much more tightly than the other hapten derivatives. Only a minor dependence of the dissociation rate of 20 on ionic strength could be detected (Table IV). A strong dependence was expected, given the charge and multiplicity of sites available for hydrogen bonding in this set of hapten derivatives. Decreasing the surface tension by including an organic cosolvent increased the dissociation rate by a factor of 2. In contrast t o the ionic-strength dependence, the dissociation rates of 20 and 13 show dramatic and different pH profiles (Figure 2), which suggest the involvement of several ionizable residues. Compounds 10 and 22 reproducibly exhibited dissociation rate plots fit by biexponential curves at 22 "C (e.g., Figure 3). Compound 10 bound to albumin under the conditions of the standard method but gave a biexponential plot when albumin was omitted from the procedure.2 Compound 7 also exhibited a biexponential plot a t 5 "C. For compounds 7 and 22, the biexponential curvature occurred a t temperatures a t which the dissociation was very slow, and the plots became monoexponential at higher temperature. However, not all haptens gave biexponential plots under conditions of slow dissociation. Also, for compound 10, the biexponential curvature was observed when the dissociation rate was in the range of the other haptens measured. Its behavior was not explored a t higher temperature. DISCUSSION
The dissociation rate constants for para-substituted derivatives in In3+-benzyl-EDTA vary over an unexpectA control experiment was performed in which a mixture of labeled and cold hapten, at the concentration used in the standard assay but in the absence of antibody and albumin, were filtered in the Centrifree apparatus. Approximately 94% of the label penetrated the membrane in multiple experiments, showing that compound 10 does not bind to container surfaces to an extent that interferes with the assay.
Dissociation Kinetics: In-Benzyl-EDTA's from CHA255
Bioconjugate Chem., Vol. 1, No. 4, 1990
281
Table I. Hapten Derivative Dissociation Rate Data no.
structure
2 '3
R1= R2 = H R1= Rz =I 0 R1,R2 = CS
4
X = Br'
lb
5 6
38 "C p-R1R2NC6H4CH2-EDTA-In
22 "C
5 "C
13.0 8.1 (1.5)
XCH2CONHC6H4CH2-EDTA-In
X = S(CH~)~SCH~CONH-~-C~H~CHZ-DTPA X = S(CH2)4S-Co-bleomycin
50.3 (0.7) 51.3 (1.7)
YNHCSNHC6H4CH2-EDTA-In Y=H 9.7 (1.3) 84.2 (2.5) 8 Y = Et 72.6 (1.3) 9 Y = n-Bu 90.6 (2.0) 1 Od Y = n-0ct 11 Y = Ph 78.0 (0.8) 12 Y = CH2CeH4 94.2 (3.0) 13' Y = CHzCOzH 49.7 (2.5) 14 Y = (CH2)&02H 11.6 (0.2) 87.8 (2.6) 1181 (32) 15 Y = (CH2)3COzH 74.3 (1.8) 16 Y = ~-C~HICH&OZH 71.5 (0.8) 17 Y = (CH2)BNHO 5.5 (0.2)f 45.6 (0.8) 611 (16)s 18 Y = (CH2)2NHCO(CH2)3C02H 14.3 (0.3) 98.8 (1.2) 1446 (21) 19 Y = (CH2)4C(NHCOCHs)HCONH2 90.9 (1.2) 20 Y = (CH2)zOH 8.8 (0.2) 59.7 (1.5) 749 (11) 21 Y =~X~H~CHZ-DTPA 22.9 (0.8) 22 Y = N1A-mitomycin C 40.9 (0.8)f Correlation coefficients were between -0.993 and -0.999, except for 2,r = -0.978, and 17,O OC, r = -0,983. Half-life estimated from only two time points. Compounds 3 and 4 are intermediates in the synthesis of the other hapten derivatives; their half-lives were not measured. Compound 10 bound to albumin under standard conditions and gave a nonlinear first-order plot when albumin was omitted. e Values are an average of four determinations. /Actually 37 "C. 8 Actually 4 "C 7
Table 11. 22 OC Dissociation Half-Life of 20 as a Function of Excess Hapten Level excess unlabeled hapten/Ab binding site, mol/mol tllz, min U 20 57.8 1.3 100 56.2 2.1 100 58.9 1.1 lo00 50.9 1.1 Table 111. Activation Parameters for Selected HaDtens hapten H', kcal/mol u S*, eu u -r 14 22.9 0.9 1.2 1.6 0.999 17 22.7 0.8 2.4 1.4 0.999 18 23.5 1.0 2.9 3.6 0.999 20 22.6 0.7 0.4 1.2 0.999 Table IV. 22 "C Dissociation Half-Life of Compound a Function of Solvent relative ionic strength 76 dioxane DH t, /2. min u 1" 0 7.4 59.7 1.5 5 0 7.5 46.2 1.5 10 0 7.5 52.4 1.2 20 7.5 1 24.7 1.9 a 1 = 150 mM NaCl, 15 mM NaP04.
20 as -r 0.997 0.995 0.997 0.978
edly wide range. Determination of these rate constants proved t o be convenient and precise. Consequently, measurement of the dissociation rate constant is a very sensitive method of detecting a change in antibody binding. As such, it may be useful for detecting chemical modification or site-specific mutation of antibodies, for predicting behavior of various in vitro or in vivo systems, or for screening new antibodies or haptens. First-order dissociation rate plots for the exchange reaction could be obtained either from the unimolecular dissociation mechanism of eq 3 or from the bimolecular reaction of eq 5. That eq 5 is not the observed reaction (Le., that dissociation of the antibody-hapten complex is
50 h
.-c E v
40-
c! 307
c
*O 10
L
6.0
7.0
8.0
PH F i g u r e 2. Dissociation half-life of compounds 13, open squares, and 20, filled squares, a t 22 "C as a function of pH. Error bars indicate standard deviation of a single determination, except for 13 a t pH 7.4 (which represents four determinations) and pH 7.5 and 7.7 (which each represent duplicate determinations) and 20 a t p H 6.8 (duplicate determinations) and 7.5 (seven determinations). (The value of the dissociation rate constant a t p H 7.7 for 13 was unexpectedly low. It was repeated, and a value a t p H 7.6 obtained under identical conditions confirmed the trend.)
a unimolecular process) was shown by performing the assay at three levels of unlabeled hapten. Since k o b d is the same in the three cases (Table 111),the reaction is not pseudofirst-order and the data fit the mechanism of eq 3. T h e data reveal several interesting facets of the relationship between the hapten and the antibody binding site. Since the original immunogen was a conjugate of 3 with keyhole limpet hemocyanin, the resultant antibody is expected to have the tightest binding to, and the slowest dissociation from, haptens that contain a thioacyl substituent in the para position. Further, according to Hansch and co-workers (13,14), additional steric bulk in the para position should lead to tighter antibody binding. Clearly, interaction occurs at the para substituent in the
202
Bioconjugate Chem., Vol. 1, No. 4, 1990
Meyer et
al.
Table V. Curve-Fit Parameters for the Biexponential Dissociation Reaction of Eq 7
no.
A 7 b 0.929 (0.005)' 10 0.899 (0.014) 22 0.966 (0.003)
E xz/
f
tip, la
tllz, 2"
parameters
0.10 (0.02) 0.30 (0.02) 0.19 (0.05)
76 (24) 9.5 (1.3) 81 (19)
952 (30) 113 (7) 624 (79)
0.84 0.44 0.81
a Rate constants k-1 and kZ, respectively, expressed as halflives, in minutes. * Compound 7 data were collected at 4 OC;data for compounds 10 and 22 were collected at 22 O C . Numbers in parentheses are standard errors for each parameter. 0.62-
T
0.52
100
200 TIME
300
I
400
(mid
Figure 3. Dissociation plot for Compound 22 at 22 "C overlayed with biexponential curve of formula in eq 7. Compound 7 also exhibited a biexponential plot at 5 "C. For compounds 7 and 22, the biexponentialcurvature occurred at temperatures at which the dissociation was very slow, and the plots became monoexponential at higher temperature. However, not all haptens gave biexponential plots under conditions of slow dissociation. Also, for compound 10, the biexponential curvature was observed when the dissociation rate was in the range of the other haptens measured. Its behavior was not explored at higher temperature. current instance. The para substituents of 1 and 2 (NO2 and NHz, respectively), for example, have different electronic effects and opposite contributions to the binding interactions of the aromatic ring, but both haptens dissociate rapidly relative to other haptens that have either an acyl or a thioacyl substituent on the nitrogen. Hence, the thioacyl substituent present in the immunogen does contribute to the CHA255 binding. Also, the two acetamides tested dissociate faster than almost all the thioureas, suggesting that there is some specificity in the antibody recognition of the thiourea functionality. On the other hand, the antibody binding site apparently does not specifically recognize substituents five carbons removed from the thiourea functionality of haptens 7-22. Compound 19 was prepared to resemble as closely as possible the original CHA255 immunogen. The a-Nacetyl substituent and the amide derivative of the carboxyl group of 19 mimic the backbone of the immunogen carrier protein, yet the dissociation rate for 19 is as fast or faster than several haptens that have much different functionality. For example, compound 9, with only a hydrocarbon substituent, has no functionality available for binding other than hydrophobic interactions but has the same dissociation rate as 19. An explanation for this observation may be that the binding site for the original immunogen surrounds and includes that which is involved with hapten binding but that it does not extend into the backbone of the immunizing carrier protein. Another discernible relationship between hapten and binding site involves charged residues. Comparison of halflives for 14 and 17, which p u t opposite charges a t approximately the same location, suggest that this region of the binding site contains a positive charge. Neutralization of the positive charge on 17 by acylation (18) gives a much slower dissociation rate. Placing the negative charge further from the thiourea, as in compounds 15 and 16, essentially neutralizes the interaction (compound 15 vs compound 8). Placing the negative charge closer to the thiourea (13) may interfere with recognition of the thiourea itself. It is possible that the relatively rapid dissociation of 20 is due to unfavorable interaction of the hydroxyl proton with the positively charged amino acid residue.
The nature of the charge environment in the CHA255 binding site is not clarified by the pH dependence of the dissociation half-life (Figure 2). The effect of pH over the relatively narrow range from 6.2 to 8.1 is strong and unexpectedly complex. It suggests that at least three amino acid residues, which change ionization state in the pH range 6-8, are significantly involved in binding 13. (Compounds 13 and 20 do not change ionization state in this pH range.) The importance of hydrophobic interactions and steric crowding in hapten-CHA255 binding are difficult to identify. Comparison of the half-lives for 11 and 12 suggests that steric repulsion in the region close to the thiourea has a noticeable effect on the rate, but both 21 and 22 are bulky in this region yet exhibit very slow dissociation at room temperature (see below). It is likely that steric effects that are present are manifest weakly, if at all, in dissociation rates, since once a ligand is bound, steric interactions could either hasten or hinder its dissociation if conformational changes in the binding site must be reversed for dissociation to occur. Nonetheless, 22 appears to exhibit both the slowest dissociation and the highest affinity (data not shown) even though it is highly sterically crowded in the area close to the thiourea, which is clearly involved in binding. In an effort to determine which forces ( 4 ) are primarily responsible for the slow dissociation (high affinity) of CHA255 and indium-benzyl-EDTA's, half-life measurements were performed for compound 20 a t three levels of ionic strength and in the presence of a solvent which lowers surface tension (Table IV). That ionic strength variation causes no variation in dissociation rate is surprising since the pH profile implicates charged residues in binding. The involvement of van der Waal's forces in binding is reflected in the decrease in dissociation halflife in the presence of dioxane. However, this does not provide a complete explanation for the affinity (19)of the antibody for this set of haptens. Measurement of the dissociation rate constants for selected haptens a t different temperatures (Tables IV and V) shows that even though the rates themselves vary significantly (as a function of hapten structure), the temperature dependence of the rate constants is remarkably similar. The activation enthalpies and entropies of the four compounds are experimentally identical, despite opposite para-substituent charges (14, 17) and vastly different room temperature rate constants (14, 18, 20).3 The temperature dependence of the rate constants for 7 appears to parallel 18, 14, 20, and 17. The temperature dependence of antibody dissociation rate constants is infrequently explicitly measured. The The fact that the rate constants are clearly distinguishable while the activation parametersare experimentally identical points out the precision with which rate constantscan be measured. The interpretation must be that the activation enthalpiesand entropies are not identical, but the differences are too small to be easily measurable. This is clearly shown in a comparison of the standard error for the rate constants and the activation parameters.
Dissociation Kinetics: In-Benzyl-EDTA’s from CHA255
Bioconjugate Chem., Vol. 1, No. 4, 1990 283
activation enthalpies measured in this system (-23 kcal/ mol) are higher than those reported by Pontarotti et al. (- 16.1 and 18.7 kcal/mol; ref 7) for a series of antivinca antibodies by about 5 kcal/mol. The negligible activation entropy (Table 111) may result from a combination of large opposite terms or may simply suggest that the constraint in the transition state resembles that in the fully bound complex. It does, however, argue against the requirement for major antibody conformational changes in the dissociation reaction. Nonlinear first-order dissociation rate plots for certain haptens were disconcerting when first encountered, despite a report of a similar finding in kinetic investigation of antibody binding to cellular antigens ( 6 ) . Prolonged preincubation of the complex did not produce linear firstorder plots, eliminating the possibility that the “secondary binding” phenomenon ( 4 ) reported for polyclonal interactions was responsible. We also increased the buffer ionic strength to test the possibility that nonspecific hydrophobic interactions were responsible, but we saw no change. Eliminating albumin from the assay made no change, and using a larger concentration of excess unlabeled hapten also did not produce linear first-order dissociation rate plots. Curve fits were attempted with several different functions. The best fits were of biexponential form (eq 7). The best fit parameters with their standard errors, for the three observed cases, are listed in Table V. Fits that were of statistically equivalent quality were obtained with a monoexponential function of the form of eq 8. However,
+
1- x = Ae-kf B
(8)
the values of A and B required by this fit imply that close to 50% of the hapten becomes irreversibly bound to the antibody. While a covalent reaction may be conceivable for 22, it is certainly unlikely for 10 and 7. Furthermore, if such a reaction did occur, it would be very difficult to explain the return to simple first-order dissociation at higher temperature (for 7 and 22). While the relative standard error for the curve-fit parameters is large, each case suggests a minor fast-dissociating component and a major slow-dissociating component. The ratio of the rate constants of fast-/slow-dissociating components is constant within experimental error, for the three cases. The fact that these dissociations were fit by biexponential curves offers the explanation that two simultaneous first-order dissociations contribute to the observed dissociation (see the Experimental Procedures). This circumstance would occur if (1)the hapten were impure, so that two different labeled haptens dissociated with different rate constants (eq 9), (2) two slightly different Ab-H
2 Ab + H; Ab-H’
ki
Ab + H’
(9)
forms of antibody binding site released the hapten with different rate constants (eq lo), or (3) two conformations
ki
Ab-H
Ab + H; AM-H
kz
Ab‘ + H
(10)
of the antibody-hapten complex dissociated at different rates (eq 11). The first of these possibilities is excluded
Ab-H
ki
Ab + H; (Ab-H)’
2 Ab + H
(11)
by analytical data. The second possibility cannot be eliminated but seems unlikely since such a small percentage of the conditions tested (hapten derivatives X temperature)
exhibited the phenomenon. The most likely explanation, that two forms of complex exist for these particular hapten derivatives, also offers a potential explanation for the fact that the dissociations become first-order at higher temperature: if the two conformations are in equilibrium (eq 121, with the proper kinetic barrier to interconver(Ab-H)
(Ab-H)’
(12)
sion, then at low temperature independent dissociation of the two forms should be observed, whereas a t higher temperature, rapid equilibration between the two forms should lead to dissociation of an “average”form. However, there is no direct evidence for the two-conformationmodel, and there is no clear rationalization for the fact that only these particular hapten derivatives bind in more than one conformation. A hybrid bifunctional antibody has been prepared that recognizes the benzyl-EDTA hapten with one binding site and the tumor-associated antigen CEA with the other binding site. We have found that this antibody causes hapten derivatives to accumulate at CEA-bearing tumor sites (3). T o show whether the hapten binding site of the bifunctional antibody is different from that of its parent, CHA255, the dissociation rate constant for compound 17 was measured with the bifunctional antibody. The room temperature complex dissociation half-life with the bifunctional antibody is 49.6 f 1.3 min, which is close to the value with the parent antibody (45.6 f 0.8 min). This result confirms that the binding sites are identical. Thus, measurement of a dissociation rate constant with the parent antibody appears to yield a value that is representative of the bifunctional antibody. We have found that the dissociation rate con%;ant is useful for comparing small changes in the interaction between a monoclonal antibody and a hapten. Although the rate constants cannot be directly related to affinity constants, the convenience and precision with which rate constants can be measured, as described in this paper, makes them attractive indicators of the interaction between haptens and antibodies. Also, when considering the behavior of a formed hapten-antibody complex under conditions of infinite dilution (e.g., in vivo), the dissociation rate constant assumes importance independent of affinity. Interpreting these relative half-lives based on the generalization ( 4 ) that dissociation rates are inversely correlated with affinities, we conclude that there is a positive charge in a certain location of the binding site, and that van der Waal’s forces are more important than ionic forces in binding hapten and antibody. In addition, questions have emerged about the observed pH dependence and the small amount of binding energy that could be accounted for by tests for ionic and van der Waals’s contribution and by the biexponential rate plots for particular hapten derivatives. To test these conclusions and answer these questions, the structure of the antibody binding site is being investigated through X-ray crystalstructure analysis of the complex of CHA255 F(ab)’ with selected hapten derivatives. ACKNOWLEDGMENT
We would like to acknowledge contributions to this work made by S. Christensen, who performed several of the rate determinations and helped with the development of the procedure, and T. Battersby and L. Anderson, who prepared many of the haptens and performed some of the radiolabeling. Also, M. Pellegrino, C. Ahlem, P. Sundstrom, and W. Butler provided helpful discussion and criticism of the manuscript.
284
Bioconjugate Chem., Vol. 1, No. 4, 1990
LITERATURE CITED (1) Frincke, J. M., Unger, B. W., Burnett, K., Hersh, E., Rosen-
blum, M., and Gutterman, G. (1984) Antibody Compositions of Therapeutic Agents and Extended Serum Half-life, U.S.Pat. Ser. No. 732, 154. (2) Frincke, J. M., David, G. S., Bartholomew, R. M., and Meyer, D. L. (1986) Antibody Complexes of Hapten-Modified Diagnostic or Therapeutic Agents, U S . Pat. Ser. No. 906,728. (3) Anderson, L. D., Meyer, D. L., Battersby, T. R., Frincke, J. M., Mackensen, D., Lowe, S., a n d Connolly, P. (1988) Optimization of Ligand Structure for Imaging with a Bifunctional Antibody. J. Nucl. Med. 29, 835. (4) Absolom, D. R., and van Oss,C. J. (1987) The Nature of the Antigen-Antibody Bond a n d t h e Factors Affecting its Association and Dissociation. CRC Crit. Rev. Immunol. 6,l-46. (5) Birnbaum, D., and Kourilsky, F. M. (1981) Differences in the Cell Binding Affinity of a Cross-Reactive Monoclonal AntiIa Alloantibody in Mice of Different H-2Haplotypes. Eur. J. Immunol. 11,734-738. (6) Mason, D. W., and Williams, A. F. (1980) The Kinetics of Antibody Binding to Membrane Antigens in Solution and at the Cell Surface. Biochem. J. 187, 1-20. (7) Pontarotti, P. A., Rahmani, R., Martin, M., and Barbet, J. (1985) Monoclonal Antibodies to Antitumor Vinca Alkaloids: Thermodynamics and Kinetics. Mol. Immunol. 22,277-284. ( 8 ) Barbet, J., Rougon-Rapuzzi, G., Cupo, A., and Delaage, M. A. (1981) Structural Requirements for Recognition of Vasopressin by Antibody; Thermodynamics and Kinetic Characteristics of the Interaction. Mol. Immunol. 18,439-446. (9) Denizot, F. C., Hirn, M. H., and Delaage, M. A. (1979) Relationship Between Affinity and Kinetics in a HaptenAntibody Reaction. Studies on Anti-Cyclic ATP Antibodies. Mol. Immunol. 16, 509-513. (10)Hornick, C. L., and Karush, F. (1972) Antibody Affinity111. The Role of Multivalence. Immunochemistry 9,325-340. (11) Eisen, H. A. (1974) Immunology. An Introduction to the Molecular and Cellular Principles of the Immune Response, pp 365,366, Harper and Row Publishers, Incorporated, Hagerstown, MD.
Meyer et al. (12) Froese, A. (1968) Immunochemistry 5, 253-264. (13) Kutter, E., and Hansch, C. (1969) The Use of Substituent Constants in the Quantitative Treatment of Hapten-Antibody Interaction. Arch. Biochem. Biophys. 135, 126-135. (14) Hansch, C., and Moser, P. (1978) Structure-Activity Relationships in Immunochemistry-The Interaction of Phenylsuccinanates and Pyridine Haptens with Anti-3-Azopyridine Antibody. Immunochemistry 15,535-540. (15) Hansch, C., Yoshimoto, M., and Doll, M. H. (1976)Structure Activity Relationships in Immunochemistry. 4. Inhibition of Complement by Benzylpyridinium Ions. On the Predictive Value of Correlation Equations. J. Med. Chem. 19,108!+1093. (16) Yoshimoto, M., Hansch, C., and Jow, P. Y. C. (1975) Structure Activity Relationships in Immunochemistry. 111. Inhibition of Complement by Benzyl Pyridinium Ions. Chem. Pharm. Bull. 23,437-444. (17) Goodwin, D. A,, Meares, C. F., David, G. F., McTigue, M., McCall, M. J., Frincke, J. M., Stone, M. R., Bartholomew, R. M., and Leung, J. P. (1986) Monoclonal Antibodies as Reversible Equilibrium Carriers of Radiopharmaceuticals. Nucl. Med. Biol. 13, 383-391. (18) Meares, C. F., McCall, M. J., Reardan, D. J., Goodwin, D. A., Diamanti, C. I., and McTigue, M. (1984) Conjugation of Antibodies with Bifunctional Chelating Agents: Isothiocyanate and Bromoacetamide Reagents, Methods of Analysis and Subsequent Addition of Metal Ions. Anal. Biochem. 142,68-78. (19) Reardan, D. J., Meares, C. F., Goodwin, D. A,, McTigue, M., David, G. S., Stone, M. R., Leung, J. P., Bartholomew, R. M., and Frincke, J. M. (1985) Antibodies Against Metal Chelates. Nature 316, 265-267. (20) DeRiemer, L. H., Meares, C. F., Goodwin, D. A., and Diamanti, C. I. (1981) Bledta 11: Synthesis of a New Tumor Visualizing Derivative of Co(II1)-Bleomycin. J. Labelled Compd. Radiopharm. 18,1517-1534. (21) Moore, J. W., and Pearson, R. G. (1981) Kinetics and Mechanism, 3rd ed., pp 311-313, John Wiley and Sons, New York. (22) See ref 21, pp 286-288.
