THE OF CHEMISTRY Vol. 268, No. October pp. for in …THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 268, No. 29, Issue of October 15, pp. 21739-21747, 1993 0 1993 by The American Society

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 268, No. 29, Issue of October 15, pp. 21739-21747, 1993 0 1993 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U. S. A.

Heavy Chain Position 50 Is a Determinant of Affinity and Specificity for the Anti-digoxin Antibody 26-10”

(Received for publication, January 25, 1993, and in revised form, June 22, 1993)

Joel F. SchildbachSS, Richard I. Nearn, Robert E. BruccoleriII, Edgar Haber**, Philip D. Jeffrey11 $$, Shi-Chung NgV, Jiri NovotnyII, Steven Sheriffll, and Michael N. MargoliesVfSVV From the $Program on Immumbgy, Harvard University Graduate School of Arts and Sciences, Cambridge, Massachusetts 02138; the Departments of llMedicine and §§Surgev , Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts 021 14; the )I Bristol-Myers Squibb Pharmaceutical Research Institute, Princeton, New Jersey 08543; and the **Cardiovascular Biobgv Laboratow. DeDartment of Biological Sciences, Harvard School of Public Health,

I_ - . ~

Boston, Massachusetts 021 15

Antibody produced by a variant of the murine anti- digoxin hybridoma 26-10 has reduced affinity for digoxin but enhanced recognition of the digoxin 12-hydroxyl due to a Tyr to His substitution at heavy chain position 50 (Schildbach, J. F., Panka, D. J., Parks, D. R., Jager, G. C., Novotny, J., Herzenberg, L. A., Mudg- ett-Hunter, M., Bruccoleri, R. E., Haber, E., and Mar- golies, M. N. (1991) J. Biol. Chem. 266, 4640-4647). Consistent with these data, the 26-10 Fab-digoxin x- ray crystal structure (Jeffrey, P. D., Strong, R. K., Sieker, L. C., Chang, C. Y., Campbell, R. L., Petsko, G. A., Haber, E., Margolies, M. N., and Sheriff, S . (1993) Proc. Natl. Acad. Sci. U. S. A,, in press) reveals that Tyr-50 contacts a region of digoxin that includes the hapten-12 carbon. To determine the effects of other heavy chain position 50 substitutions, mutant antibodies were engineered, and their affinities for digoxin and digoxin analogues were measured. The affinity of the mutant antibodies for digoxin roughly correlates with the size of the position 50 side chain. Substitutions of Trp or Phe have no effect on affinity, whereas substitutions of Asn, His, Leu, Ala, Gly, and Asp confer progressively lower affinities. Although Trp and Phe mutants exhibit wild-type specificity, Asn and Asp mutants have improved affinity for digoxin relative to digitoxin (12-deshydroxydigoxin). Leu, Ala, and Gly mutants have improved affinity for 12-acetyldigoxin relative to digoxin as compared with 26-10. These results indicate that position 50 is a determinant of both antibody affinity and fine specificity for antibody 26-10 and that single-amino acid substitutions can alter antibody fine specificity.

Models of the mutants were computationally constructed, and haptens were docked into the modeled binding sites. The results suggest that 12-acetyldigoxigenin occupies different orientations in the 26-10 and in the Ala mutant binding sites, resulting in altered binding.

~ ~ ~ ~~~~

* This work was supported by National Institutes of Health Grants Pol-HL19259 and R01-HL47415. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accord- ance with 18 U.S.C. Section 1734 solely to indicate this fact.

Supported by a National Science Foundation graduate fellowship, 1986-1989. Present address: Dept. of Biology, Massachusetts Inst. of Technology, Cambridge, MA 02139.

$$ Present address: Dept. of Biophysics and Biochemistry, Johnson Research Foundation, University of Pennsylvania, Philadelphia, PA.

llll To whom correspondence should be addressed: Jackson 1402. Massachusetts General Hospital, Boston, MA 02114. Tel.: 617-726: 8552; Fax: 617-726-4811.

The hallmark of the mammalian humoral immune response is the ability to generate antibodies with exquisite specificity. Early experiments demonstrated that immune sera could discriminate between proteins derived from different species (1) or stereoisomers of dyes or carbohydrates (2, 3). The concept of antigenic specificity is described in biochemical terms as an antibody having a higher affinity for an immunizing antigen than other antigens. Fine specificity refers to the ability to distinguish between antigens that are structur- ally similar. The extent of correlation between affinity and specificity is ill defined, especially at the molecular level.

To examine the molecular basis of antibody specificity, we are studying antibody-hapten interactions using the cardiac glycoside digoxin (digoxigenin tridigitoxoside; see Fig. 1) as a model hapten. A major advantage of using digoxin for studies of antibody specificity is that there are many available digoxin analogues. These analogues differ from digoxin by one or more defined substitutions in the steroid, lactone, or carbo- hydrate moieties and are valuable probes of hapten binding. Furthermore, structures of many of these analogues have been determined by x-ray crystallography.

We initially identified variable region residues affecting hapten recognition of the anti-digoxin monoclonal antibody 26-10 (4) by examining spontaneous variants of the 26-10 hybridoma ( 5 ) that were selected by fluorescence-activated cell sorting. The variant LL2 paradoxically exhibited lower affinity than the parent antibody but an enhanced specificity for digoxin due to a Tyr to His mutation at heavy chain position 50 (H 50).’.’ Here we present a study employing mutagenesis designed to elucidate the role of the amino acid residue at this position. We found an approximate correlation between the affinities for digoxin of the engineered mutant antibodies and the size of the position 50 side chain. In addition, antibodies with smaller side chains display either enhanced or reduced specificity for digoxin relative to other cardiac glycosides. These results indicate that fine specificity and affinity are inexactly related and may be influenced by the identity of the residue a t a single position. Models of the mutants were generated and provide insight into the affinity and specificity differences between 26-10 and the mutant antibodies.

The abbreviations used are: H 50, heavy chain position 50; H, heavy; L, light; ABNR, adopted basis Newton Raphson; CDR, complementarity determining region; dATPotS, deoxyadenosine 5’(ot- 36(S)thio]triphosphate; wt, wild type.

* All variable region position numbers are given according to Kabat numbering (50).

21739

21740 Single Residue Modulates Antibody Fine Specificity

MATERIALS AND METHODS

CellLines and Antibody Purification-The derivation of the murine A/J anti-digoxin hybridoma cell line 26-10 (IgGPa, K ) was described (4). The hybridoma variant LL2 was isolated from a 26-10 subclone by fluorescence-activated cell sorting (5). Cell line 26-10~, a spontaneous H chain-loss variant of 26-10, was subcloned and selected on the basis of H chain isotype loss as determined by isotype enzyme- linked immunosorbent assay.

Antibodies were purified by affinity chromatography on columns of ouabain-amine-Sepharose, eluting with 20 mM ouabain (6) or eluting with 0.2 M NHdOH (pH 11) into a neutralizing solution of 1.5 M Tris-HC1 (pH 4.8).

Antibody Mutagenesis-Mutagenesis of the 26-10 H chain variable region gene and expression in 26-10~ cells were performed as described (7). Mutagenic oligonucleotides were synthesized by using an Applied Biosystems 380B DNA synthesizer and had the sequence 5'-GGA- GAAATXXXXCCAATGTAATC-3', where X denotes the nucleo- tides that varied. These oligonucleotides hybridize to the 26-10 H chain variable region (8) from position 46 to 52a, and each oligonu- cleotide encoded 2-base pair differences from the unmutated sequence. Results of mutagenesis were confirmed by dideoxy chain termination nucleotide sequencing using Sequenase (U. S. Biochem- ical Corp.) and dATPaS (Amersham Corp.).

Affinity and Specijicity Deternin~tions-[~H]Digoxin for affinity and specificity measurements was purchased from Du Pont-New England Nuclear. Digoxin and digoxin analogues were obtained as described (5). Binding assays were performed using cell supernatants or, for mutant antibodies H:Tyr50Asp and H:Tyr50Gly, purified antibody.

Affinities were measured by a saturation equilibrium assay using filtration through glass fiber to separate bound and free ligand as described (5), except a Brandel harvester (Life Technologies, Inc.) was used for filtration instead of a Millipore 1225 manifold (Millipore) previously used. Affinity data were analyzed using LIGAND (9). The affinity of LL2 for digoxin reported here (KO = 2.9 X 10' M-'; see Table I) is lower than a previously reported measurement using the glass-fiber filter method with the Millipore manifold (6.4 X 10' M-'; see Ref. 6). This results from the different filtration equipment used (7).

The KI of antibodies for digoxin analogues (KI is KD as determined by a competition assay) was determined using glass-fiber filtration instead of the solid-phase assay utilizing lZ5I-labeled digoxin used previously (5). Briefly, antibody solutions (final concentration, 50.1 K D for digoxin) were added to 5-ml polystyrene tubes (no. 55.476, Sarstead) containing [3H]digoxin and varying concentrations of un- labeled digoxin analogue. To ensure adequate signal to noise ratio, concentrations of [3H]digoxin used ranged from equal to the Kr, of the antibody for highest affinity antibodies to 0.05 KD for the lowest affinity antibody; volumes of incubation ranged from 2 ml for highest affinity antibodies to 0.5 ml for lowest affinity antibodies. Analogue concentrations ranged from 100 PM to 0.1 nM, except for the highest affinity interactions when concentrations of 1 eM to 1 pM were used. Following incubation at 20 "C for 90-120 min, samples were filtered through glass fiber, and the filters were washed with 10 ml of cold (4 "C) PBSA (0.1 M NaC1, 0.01 M KH2P04, 0.02% NaN3, pH 7.4). [3H]Digoxin in the filters was measured by liquid scintillation count- ing. Graphs of counts/min bound uersw log concentration of inhibitor were fitted with a 4-parameter logistic equation relating total and nonspecific binding, 50% inhibitory concentration, and slope of the curve (10) using the FITFUNCTION nonlinear curve-fitting routine of statistical software package RS/1 (BBN Software Products Corp., Billerica, MA). For analogues that inhibited between 50 and 100% of [3H]digoxin binding at maximum inhibitor concentration used (100 mM), nonspecific binding was estimated and fixed during the fit. KI

values were estimated according to the equation of Cheng and Prusoff (11). The KD values used for KI estimation were taken from the K, (KD = I/&) for digoxin as measured by the saturation equilibrium assay. The K, values presented are the averages of a minimum of three assays. Relative KI values were obtained by dividing the average KI value for an analogue by the average K, value for digoxin.

Molecular Modeling-Parameters used for amino acids, including partial atomic charges, were as described (12). The partial atomic charges of the hydroxyl oxygens and hydrogens and the hydroxyl- bearing carbon atoms (-0.65e, 0.4e, and0.25e, respectively) of digoxin

are taken from those for serine (12): For the 0-linked acetyl group of 12-acetyldigoxigenin, the partial atomic charges were C12 = 0.4e, 012 = -0.44e, C24 = 0.35e, 024 = -0.365e, and C25 = 0.055e (see Fig. 1 for numbering). Partial atomic charges for C17 (0.112e), C20 (0.052e), C21 (0.392e), 021 (-0.543e), C22 (-0.284e), C23 (0.873e), and 023 (-0.602e) were calculated from ab initio computations. 3- Isopropyl-2(5H)-furanone, equivalent to carbons 13, 16, 17, 20, 21, 22, and 23 and oxygens 21 and 23 of digoxin, was constructed without charges using CHARMM version 16 parameters (13). The lactone ring oxygen, equivalent to oxygen 21 of digoxin, was treated as a water oxygen. The structure was m-inimized to reduce steric energy. A lattice ofgoints on a grid of 0.8 A was constructed, and all points within 4.8 A of the surface of the molecule were identified by using the program PDGRID (D. E. Williams, Quantum Chemistry Program Exchange, Program 568). The electrostatic potential at each of these points was calculated from the wave function with the program4 Gaussian 90, Revision H using a 6-31GS basis set. Partial charges on each of the atoms were determined by a least squares fit to the electrostatic potential calculated by Gaussian 90 using PDM88 (14).

The values for bond lengths, bond angles, torsion angles, and improper torsion angles of the lactone and hydroxyl atoms were averaged from the x-ray crystal structures of digoxin (15), digoxigenin bisdigitoxoside (16,17), gitoxigenin bisdigitoxoside (16), digitoxigenin bisdigitoxoside (16), gitoxin (la), 5/3-hydroxygitoxigenin (19), ouabain diethanol (20), ouabain (21), oleandrin (22), oleandrigenin monohy- drate (23), dihydrodigoxigenin hydrate (24), digitoxigenin (25), stro- phanthidin (26), A-8,14-anhydrodigitoxigenin (27), methyldigitoxi- genin (28), digoxigenin monodigitoxoside (29), and actodigin (30).

Modeling was based on the crystal structure of the-26-10 Fab- digoxin complex (31) that has been determined to a 2.5-A resolution with an R value of 0.17J and a root mean square deviation from ideal bond lengths of 0.013 A. The modeling procedure is an adaptation of the minimum perturbation approach (32), which combines side-chain conformational searches with energy minimization. The program CONGEN (33) was used." The procedure is as described (7) except that prior to energy minimization of a conformation, the hydrogens were deleted, replaced with the HBUILD command of CONGEN, and energy-minimized (200 steps ABNR minimization; see Ref. 13) while all other atoms were fixed. This established hydrogen bonds for the substituted side chain. As a control, the 26-10 crystal structure used in the figures and calculations was subjected to the same minimization procedure.

The lowest energy models for each substitution were examined. Those in which the H 50 side chain extended into the binding site potentially blocking hapten binding (H:TyrBOTrp, H:Tyr50His; H:TyrSOLeu) were discarded in favor of the model with the next lowest energy side chain conformation. Another H:Tyr50His conformation was discarded because it required the H:Tyr47 0, to accept two hydrogen bonds, which is not characteristic of hydrogen bonds in other proteins (34). Comparison of x-ray crystal structures of 26- 10 with and without bound hapten reveals that the residues in the 26-10 binding site do not shift substantially in position upon complexation with hapten (31). As the effects of side-chain rearrangement on the structure and stability of antibody-hapten complexes are difficult to predict, only models that accommodate hapten in the binding site without substantial rearrangement were selected for further examination.

Following modeling, digoxigenin, ouabagenin, or 12-acetyldigoxigenin was placed in the models in the position of digoxin in the crystal structure. The coordinates of digoxigenin were taken from the 26-10.digoxin complex, and those of ouabagenin were taken from a published crystal structure (21). The 3-position sugars were removed to create the genin forms, and the structures were energy-minimized. During the minimization, the position of the lactone ring relative to the steroid moiety was restrained to that seen in the 26-10 Fab- digoxin complex using harmonic restraints on the hapten dihedral

Parameter sets in CHARMM format (13) are available upon request (to J . F. S.). The programs CONGEN (33) and PLT2 (D. States and R. E. Bruccoleri. unaublished) are also available upon request (to R. E. B.). ' M. J. Frisch, M. Head-Gordon, G. W. Trucks, J. B. Foresman, H.

B. Schlegel, K. Raghavachari, M. Robb, J. S. Binkley, C. Gonzalez, D. J. Defrees, D. J. Fox, R. A. Whiteside, R. Seeger, C. F. Melius, J . Baker, R. L. Martin, L. R. Kahn, J . J. P. Stewart, S. Topiol, and J. A. Pople, Gaussian, Inc., 1990, Pittsburgh, PA.

1 -

Single Residue Modulates Antibody Fine Specificity 21741

angles C13-C17-C2O-C21, C13-C17-C2O-C22, C16-C17-C2O-C21, and C16-C17-C2O-C22. Coordinates of 12-acetyldigoxigenin were constructed from the coordinates of digoxigenin using parameters for the acetyl group taken from the crystal structure of oleandrin (22). After introduction of 12-acetyldigoxigenin into the binding sites, a conformational search using a 30" grid over the Cll-C12-012-C24 and C12- 012-C24-024 dihedral angles was performed, and the lowest energy conformation was selected for the subsequent minimization step. Following introduction of a hapten, the modeled antibody-hapten complex was minimized in a two-step process. First, the H 50 side chain was minimized, whereas all other atoms were fixed (500 steps ABNR). Then, all main chain atoms within 12.5 A of the hapten C8 were restrained (8 kcal/A), all other atoms within 12.5 8, pf the hapten C8 were unrestrained, and all atoms greater than 12.5 A from C8 were fixed during energy minimization (500 steps ABNR). The hapten C8 was arbitrarily chosen as the center of the minimization because it is near the center of the cardenolide structure.

Figures of the 26-10.digoxin structure and the models were made by using the plotting program3 PLTZ. Solvent-accessible areas were calculated according to the algorithm of Lee and Richards (35) using a 1.4-A radius probe.

RESULTS

Affinity for Digoxin-The affinities (K,) of the parent 26- 10, variant, and mutant antibodies for digoxin are listed in Table I. Antibody 26-10wt, produced by expression of the unmutated 26-10 H chain in 26-10~ cells, has an affinity of 9.1 X lo9 M-'. Replacement of the wild type H 50 Tyr with Phe (H.Tyr50Phe) or Trp (HTyr50Trp) results in little change in affinity for digoxin, whereas replacement with other amino acids causes a loss of affinity of up to 4000-fold. With the exception of the Asp mutant, the antibodies exhibit an inexact correlation between affinity and H 50 side-chain volume.

Specificity of Hapten Recognition-The binding of the antibodies to several cardiac glycosides and aglycones was measured in competition assays. Analogues used include those for which HTyr50His (LL2) and 26-10 have different relative affinities (5). The structural differences among the analogues are summarized in Table 11, by reference to Fig. 1. Dissocia- tion constants calculated from the inhibition assays ( K I values) are given in Tables I11 and 1V as relative Kl (values normalized to the Kr for digoxin). The Kt values of the antibodies for digoxin (data not shown) agree well with the K D values obtained from saturation equilibrium assays.

The 26-1Owt antibody binds digoxigenin and digitoxin nearly as well as digoxin, indicating little recognition of the 3-tridigitoxose and 12-OH, respectively, of digoxin (Table 111) (5, 6). Analogues with 0-linked 12-acetyl groups (12-acetyldigoxin, dlgoxigenin-3,12-diacetate) are bound with reduced affinity by 26-10wt, as are analogues with substitutions at the

TABLE I Affinity of antibodies for digoxin

K. for digoxin as measured in a saturation equilibrium assay using filtration through glass fiber to separate bound from free hapten (5, 7). Affinity values given are results from representative assays with estimated standard error as given by the LIGAND program,

Antibody/mutant Affinity

x I 0-7 M I 910 f 100

1200 & 200 670 -C 90 29 -C 6 25 f 4

9.2 f 0.9 3.8 f 0.4

0.29 f 0.10 0.23 -C 0.07

26-1Owt (HTyr50) H:Tyr50Phe H:TyrBOTrp H:TyrBOAsn H:Tyr5OHis (LL2) H:Tyr50Leu H:TyrBOAla H:Tyr50Gly H:Tyr5OAsp

16-position (gitoxin, gitaloxin, 16-acetylgitoxin, oleandrin, oleandrigenin). The affinities for haptens modified at the 16- position decrease concomitantly with an increase in the size of the substituent. The lowered affinity due to a 16-OH (gitoxin) appears to be partially compensated for by a 12-OH (diginatin) although the effect is not much greater than that of a 12-OH in the absence of a 16-OH (compare digoxin and digitoxin). For 16-acetylated compounds, a 3-tridigitoxose (16-acetylgitoxin) confers higher affinity than a 3-oleandrose (oleandrin) or 3-OH (oleandrigenin). For 12-acetylated compounds, a 3-tridigitoxose (12-acetyldigoxin) confers higher affinity than a %acetyl group (digoxigenin-3,12-diacetate). Antibody 26-1Owt has reduced affinity for ouabain and ouabagenin relative to digoxin (Table IV); its high affinity for strophanthidol and acovenoside A suggests that the hapten 11-OH contributes to reduced binding, either alone or in combination with the other hydroxyls of ouabain. Finally, saturation of the C20-C22 bond of the digoxin lactone (dihydrodigoxin, dihydrodigoxigenin; Table IV) causes a greater than 1000-fold loss of affinity.

Comparison of the relative Kr values for H:TyrSOPhe and H:Tyr50Trp to those of 26-1Owt indicates that the fine specificities of these antibodies are similar. Antibody H:Tyr50Trp has a 3-fold higher affinity for ouabain than does 26-10wt, and H:Tyr50Trp and H:Tyr50Phe may have slightly higher affinities for 12-acetyldigoxin than 26-1Owt.

Antibody H:Tyr50Asn differs from 26-1Owt in recognition of the hapten 12-OH, i.e. H:Tyr50Asn binds digoxin with 32- fold higher affinity than digitoxin, whereas 26-1Owt binds the two haptens with similar affinities. This specificity difference is revealed also by reduced relative affinities for haptens lacking a 12-OH, including those that are modified at the 16- position (Table 111) and strophanthidol and acovenoside A (Table IV). Antibody H:Tyr50Asn has relative KI values for ouabain, ouabagenin, and 12-acetyldigoxin similar to those of 26-1Owt although the relative Kt value for digoxigenin-3,12- diacetate is reduced. The binding specificity of H:Tyr50His (LL2) resembles that of H:Tyr50Asn, except the relative K, value for digoxigenin-3J2-diacetate does not differ greatly from 26-10wt, and the increased relative affinity for analogues with a 12-OH is not as pronounced.

While mutants H:Tyr50Leu, H:Tyr50Ala, and H:Tyr50Gly exhibited little recognition of the 12-OH (relative KI for digitoxin of 4 or less), compared with 26-1Owt they have a decreased relative KI for 12-acetylated haptens (Table 111). These mutants also have decreased relative KI for haptens with a saturated lactone ring (Table IV). The difference between relative KI values for dihydrodigoxin of 26-1Owt and H:Tyr50Leu or H:Tyr50Ala is small (3-7-fold) but is 80-fold between 26-1Owt and H:Tyr50Gly. In addition, the mutant H:Tyr50Leu has slightly higher relative K, values for ouabain, ouabagenin, and 16-acetylgitoxin than 26-1Owt. H:Tyr5OGly has a lower relative affinity for acovenoside A and a slightly higher relative affinity for ouabain.

The specificity of mutant H:Tyr50Asp most closely resembles that of H:Tyr5OAsn and H:Tyr50His with lowered affinities for digitoxin, strophanthidol, and acovenoside A, relative to digoxin. However, antibody H:Tyr50Asp differs by having equal affinities for digoxin and ouabain. The specificities of H.Tyr50Gly and H:Tyr50Asp could not be thoroughly examined, however, because their low affinities for many analogues (KI < 10 g ~ ) preclude measurement by the method used.

Molecular Modeling of the Binding Sites-The x-ray crystal structure of 26-10 Fab complexed with digoxin (31) reveals that the Tyr at H 50 contacts digoxin in the region of the


TABLE I1 Structural characteristics of digoxin analogues

Analogue Substitutions at steroid positions"

18 38 58 l l a 128 168 19 C20-C22 bond

Tridigitoxose -OH Unsaturated Digoxin Digoxigenin Digitoxin 12-Acetyldigoxin Digoxigenin-3,12-diacetate Diginatin Gitoxin Gitaloxin 16-Acetylgitoxin Oleandrin Oleandrigenin Ouabain Ouabagenin Strophanthidol Acovenoside A

-OH - Tridigitoxose Tridigitoxose

Tridigitoxose Tridigitoxose Tridigitoxose Tridigitoxose Oleandrose

-OH L-rhamnose

-0COCHs

-OH

-OH -OH -OH

-OH 6-Deoxy-3-O-methyl-~-talose Tridigitoxose

-OH -OH -OH -OH -OH

-OH

-0COCH3 -0COCH3 -OH -OH

-OH -0CHO -0COCH3 -0COCHs -0COCH3

-OH -OH -OH

Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Saturated -OH Dihydrodigoxin

Dihydrodigoxigenin -OH ~ -OH Saturated a Numbering given in Fig. 1.

FIG. 1. Diagrams of the cardenolide numbering system, digoxin (digoxigenin tridigitoxoside), ouabain, and 12-acetyldigoxin, including the numbering for the 12-acetyl moiety.

2

3

4 6

DIGOXIN

0

I W W -

hapten 12-position (Fig. 2). The Cg, C,, Caz, Cr2, and 0, atoms of the 26-10 Fab-digoxin complex as a starting point. The of H:Tyr-50 are in contact (as defined by Sheriff et al. (36)) models of H:Tyr50Tyr (modeled 26-10), H:Tyr50Asn, with atoms C2, C12, C15, and C16 of the hapten.' Models of H:Tyr50His, and H:Tyr50Leu are shown in Fig. 3 in a view all but the Gly mutant were constructed by a side-chain depicting the relationship between the modeled position-50 conformational search procedure using the crystal structure side chain, H:Tyr-47, and bound digoxigenin. The

H:Tyr5OTyr model places the H 50 side chain in a position P. D. Jeffrey, unpublished observation. similar to that of H:Tyr50 in the crystal structure although it

Single Residue Modulates Antibody Fine Specificity

TABLE 111 Binding of 26-10 H 50 mutants to digoxin and gitoxin analogues

21743

Relative K, values as measured using a solution-phase competition assay (see “Materials and Methods”). Kr values were normalized to the KI of digoxin for a given antibody. Values for those haptens that at a concentration of 100 ~ L M inhibited less than 50% of t3H]digoxin binding are denoted as greater than (>) the lowest measurable value. Antibodies are listed from left to right in order of decreasing affinity for digoxin. The KD of the antibodies for digoxin, relative to the KD of 26-1Owt for digoxin, are given in parentheses in the first row.

Relative K ,

26-1Owt HTyr50Phe H:TyrBOTrp H:Tyr50Asn H:Tyr5OHis HTyr50Leu H.Tyr5OAla H:Tyr50Gly H:Tyr50Asp

Digoxin l(1) l(0.7) l(1.4) 1 (31) 1 (36) 1 (99) 1 (240) 1 (3100) 1 (4000) Dieoxkenin 2 2 2 1 2 2 2 0.9 1

Analogue

DiiitoGn 2 12-Acetyldigoxin 330 Digoxigenin-3J2-diacetate 1,400 Digination 2 Gitoxin 6 Gitaloxin 32 16-Acetylgitoxin 250 Oleandrin 920 Oleandrigenin 9,900

1 120 850

2 5

31 300

1,100 14,000

2 160

1,500 4 5

35 240

1,500 16,000

32 220 340

12 86

410 1,500

>2,800 >2,800

5 210

1,200 4

22 140 460

2,100 >2,700

2 2 20 6 35 14 0.9 1 8 9

32 37 76 76

730 490 X 4 0 >470

4 6

28 3

12 27

>54 >54 >54

11 >30 >30

4 >30 >30 >30 >30 >30

TABLE IV Binding of 26-10 H 50 mutants to digoxin and ouabain analogues

Relative KI values as measured using a solution-phase competition assay (see “Materials and Methods”). KI values were normalized to the KI of digoxin for a given antibody. Values for those haptens that at a concentration of 100 ~ L M did not inhibit 50% of [3H]digoxin binding are denoted as greater than (>) the lowest measurable value. Antibodies are listed from left to right in order of decreasing affinity for digoxin. The KD of the antibodies for digoxin, relative to the KD of 26-1Owt for digoxin are given in parentheses in the first row.

Analogues Relative KI

26-1Owt H:Tyr5OPhe H:Tyr50Trp H:Tyr5OAsn H:Tyr50His H:Tyr50Leu H:Tyr5OAla H:TyrBOGly H:Tyr5OAsp

Digoxin 1 (1) 1 (0.7) l(1.4) l(31) 1 (36) l(99) l (240) l(3100) l(4000) Ouabain 42 64 15 24 28 31 12 15 1 Ouabagenin 40 35 22 55 20 24 9 19 5 Strophanthidol 1 2 2 15 7 1 2 2 6 Acovenoside A 2 3 3 59 13 9 4 8 18 Dihvdrodigoxin 1300 1300 1700 820 810 410 200 16 >30 Dihydrodigoxigenin 2400 2000 2500 770 1500 680 210 24 >30

is shifted slightly (root mean square difference between position of the modeled H:Tyr50 side chain atoms and those of the minimized crystal structure is 1.1 A): The H 50 side chain of H:Tyr50Phe (not shown) occupies the same position as the H:Tyr50 in the crystal structure (root mean square difference of shared H 50 atoms is 0.4 A), maintaining all the contacts except that involving the Tyr 0,. The H:Tyr50Trp model (not shown) suggests that the Trp side chain provides contacts similar to the H:Tyr50 of 26-10 by occupying a similar position.

The H 50 side chain of the H:Tyr50His model (Fig. 3) lies within the same plane as the HTyr50 of 26-10 and appears capable of providing many of the contacts of H:Tyr50. The Cb, C,, and C6z of Tyr and His are in similar positions as are the C I p of Tyr and the Ne2 of His, but the substitution of N I P for CeZ may affect the strength of the interaction due to its smaller van der Waals radius. The side chains of H:Tyr50Asn (Fig. 3) and H:Tyr50Asp (not shown) cannot provide a number of the contacts of H:Tyr5O, but the Co and C , atoms of the Asn, Asp, and Tyr side chains are in similar positions. The Asp and Asn side chains are stabilized by hydrogen bonding to the hydroxyl of H chain Tyr-47. The Cg, C,, and C62 atoms of the Leu of H:Tyr50Leu (Fig. 3) are nearly superimposable upon the corresponding atoms of the H:Tyr50 of the crystal structure, mimicking some of the hapten con-

’ The second lowest energy conformation, however, is nearly iden- tical to the minimized crystal structure (root mean square difference in H:Tyr5O position is 0.1 A). The reason for the energy difference in the two conformations is not clear, although the lowest energy structure does have a slightly larger solvent-exposed area, which is unfavorable and not directly accounted for in the energy function used.

tacts. The position of the Cb atom in the H:Tyr50Ala model (not shown) is similar to that of the 26-10 H:Tyr50, indicating that a contact to hapten through this atom is possible. A model of H:Tyr50Gly was not constructed.

The mutants H:Tyr50Leu, H:Tyr50Ala, and H:Tyr50Gly demonstrate higher affinity for 12-acetylated analogues, relative to digoxin, than does 26-10 (Table 111). To examine this phenomenon, 12-acetyldigoxigenin was introduced (see “Ma- terials and Methods”) into the binding site of the H:Tyr50Ala model and into the 26-10 crystal structure (Fig. 4). The acetyl group occupies a position in the 26-10 binding site that causes unfavorable close contacts with both the LPro-96 and H:Tyr50 (as calculated by CONGEN; data not shown). Anal- ogous results were obtained when docking 12-acetyldigoxigenin into the modeled binding site of H:Tyr5OAsn (not shown). In the binding site of H:Tyr50Ala, however, the entire hapten is shifted relative to its position in the 26-10 binding site, with the acetyl group now in the space occupied by the H:Tyr50 of 26-10. The unfavorable contacts between the acetyl group and L:Pro96 occurring in the 26-10. hapten are thereby relieved without causing unfavorable contacts with Ala at H 50.

Antibody 26-10 has a 42-fold greater relative KI for digoxin, apparently due to the ouabain 11-OH, alone or in combination with other ouabain hydroxyls. However, no unfavorable contacts involving the 11-OH are obvious in a modeled complex of 26-10 and ouabagenin (not shown). The preference for digoxin over ouabain is shared by most of the mutant antibodies. The exception is H:Tyr50Asp, which has approximately equal affinities for ouabain and digoxin, although a modeled complex of ouabagenin and H:TyrBOAsp (not shown)


L CDR2 L CDR2

L CDR3 H CDR2

97 a C D R 3

L-

FIG. 2. Stereo view of the CDRs and hapten from the x-ray crystal structure of the 26-10 Fab-digoxin complex (31). Upper panel, a view of the 26-10 CDRs and digoxigenin from solvent into the binding site. The side chains of H:TyrBO (in heavy chain CDR2 (Hz)) and LPro96 (in light chain CDR3 (L3)) are shown with the CDR main chains. Lower panel, a view of digoxigenin and its orientation to H:Tyr47 (within the second framework region), H:TyrBO, and L:Pro96. The view is as in the upper panel except rotated 110 degrees about the vertical axis. The atoms of H:Tyr50 and hapten that are in contact are shown as solid circles. Hydrogens are shown only for the Tyr phenolic and the digoxigenin hydroxyl groups.

suggests that the Asp side chain is unable to form hydrogen bonds with any of the ouabain hydroxyls. Examination of the modeled 26-10.oubagenin complex indicates that the lp, 5p, 11-a, and 19-OH are partially or fully solvent accessible (as measured according to Lee and Richards (35); data not shown).

DISCUSSION

Altered Affinity of the Mutants-The affinities and specificities of the 26-10 H 50 mutants clearly indicate that the residue at H 50 contributes to hapten recognition by the 26- 10 antibody. The K, values for digoxin of 26-10 and the mutant antibodies (Table I) increase concomitant with an increase in size of the H 50 side chain. Examination of the crystal structure of the 26-10 Fab-digoxigenin complex (31) and models of the mutants (Fig. 3) suggest that the reduced affinity of some of the mutants is due in part to reduction of the number of contacts and the surface area of the antibody- hapten interface. Although the affinity of 26-10 for digoxin derives largely from shape complementarity and hydrophobic effect (31), reduction in the contact area between hapten and the mutant antibodies may not entirely account for the re-

duction in affinity. The free energies of complexation (calculated according to AG = R T In KO, where R is the gas constant, T is 293 K, and KD is the experimentally determined dissocia- tion constant) of digoxin with 26-10 and with H:Tyr50Ala differ by 3.2 kcal/mol. The difference in surface areas of these two antibody-hapten interfaces is only 24 A' (calculated from the model of H:Tyr50Ala and the structure of 26-10 using the algorithm of Lee and Richards (35)). If the reduced affinity were due solely to lost contact area, the ratio of energy to area is approximately 130 cal/A2/mol, much higher than expected (25-47 cal/b'/rnol; see Refs. 37 and 38) even if one accounts for the hydrophobic nature of the interaction (39, 40). This discrepancy may reflect an underestimation of the difference in surface areas resulting from inaccuracies in the H:Tyr50Ala model, an underestimation of the ratio of the energy of complexation to contact area, or an overestimation of the differences in the measured KD for digoxin of these antibodies. Alternatively, the substitution of the H:Tyr50 of 26-10 may affect the positioning or stability of other contact residues, reducing antibody affinity. For example, the Tyr side chain may stabilize other contact residues; substitution of Ala for Tyr may result in increased mobility of these residues. This

Single Residue Modulates Antibody Fine Specificity 21745

T y r 4 5

Tyr5$

Asn5&5

H i

-1

s5$

L0u5$

T y r 4 7 -I

TyrP Asn5>+&5

-1 His50 @ Leu5$

FIG. 3. Stereo views of models of H:TyrBOTyr (modeled 26- lo), H:TyrSOAsn, H:TyrSOHis (LL2). and H:TyrBOLeu. The view depicts the relationship of the H 50 amino acid to digoxigenin in the binding site. The H:Tyr47 residue is included to illustrate the hydrogen bonding between this residue and the modeled Asn H 50 side chain. Models are shown in heauy lines, and H:Tyr50 from the crystal structure is shown in thin lines. Hydrogen bonds are depicted as dashed lines. Hydrogens are shown explicitly only on potential hydrogen bond donors. The differences between models in the loca- tion of the 12-OH hydrogen of digoxigenin are the result of the hapten introduction procedure, during which the Cll-Cl2-012-Hl2 dihedral angle was varied, the complex minimized, and the lowest energy conformation selected.

would increase the entropic penalty incurred if the motion of these contact residues is restricted by bound hapten. To distinguish between these possibilities, it may be necessary to determine the structure of H:TyrBOAla, both uncomplexed and with bound digoxin.

The reduction in affinities of the H 50 mutants reflects the hydrophilicity of the substituted H 50 side chain in addition

to changes in hapten contact area. The contact between the H:Tyr50 of 26-10 and digoxin (Fig. 2) involves apolar atoms. Desolvation of these atoms, as required during hapten binding, should be energetically favorable. Desolvating a more polar side chain would be less favorable, resulting in a reduced affinity for digoxin. This effect is most clearly seen when comparing the affinities of H:Tyr5OAsn and H:TyrSOAsp for digoxin (Table I). The affinity of H:Tyr50Asp is 100-fold lower than that of H:Tyr50Asn although, as suggested by modeling experiments, the H 50 residues of these antibodies may occupy similar positions and provide similar hapten contacts. The energetic penalty is greater when the environ- ment of the negatively charged Asp side chain changes from solvent-exposed (when uncomplexed) to apolar (when complexed with hapten) as compared with the neutral Asn side chain. The complex of H:Tyr50Asp and digoxin is thus less stable than that of H:Tyr50Asn and digoxin, and the affinity of the H:Tyr50Asp antibody is lower.

Altered Fine Specificity of the Mutunts-H.Tyr50Phe, H:Tyr50Trp, and 26-1Owt have similar fine specificities, including an apparent lack of recognition of the digoxin 12-OH. The digoxin 12-OH and the side chains of H:Tyr50Asn, H:Tyr50His (LLZ), and H:Tyr50Asp interact favorably, as evidenced by the higher affinities of these antibodies for digoxin than for digitoxin. The basis for this enhanced recognition of the 12-OH is unclear; models of these mutants suggest that their H 50 side chains do not contact the digoxin 12-OH and that the distances between the 12-OH and the H 50 side chains and the angles of interaction make formation of hydrogen bonds unlikely. This fine specificity is likely related to the polar nature of the H 50 side chains because only mutants having these hydrophilic residues exhibit recognition of the 12-OH.

The higher relative affinity for ouabain of H:Tyr50Asp than of 26-1Owt and the other H 50 mutants may also be explained by the polar nature of side chains and hapten. H:Tyr50Asp binds digoxin and ouabain with equal affinities, presumably due to an interaction between the H 50 side chain and one of the hydroxyls and/or the sugar moiety of ouabain. The other H 50 mutant antibodies and 26-1Owt bind ouabain with at least a 12-fold lower affinity than digoxin, despite partial or

FIG. 4. Stereo view of the orientation of 12-acetyldigoxigenin in the binding sites of (upper panel) 26-10 and (lower panel) H:TyrSOAla. In both panels, 12-acetyldigoxigenin is in the center with the 12-acetyl group extending downward. The residue on the left is L:Pro96, and the amino acid on the right is the H 50 residue. Also included is the main chain of the amino acids N-terminal and C-terminal to LPro96 and H 50. Hydrogens are shown explicitly only on potential hydrogen bond donors.

2 1746 Single Residue Modulates Antibody Fine Specificity

total solvent accessibility of all of the ouabain hydroxyls in a modeled complex of 26-10 and ouabagenin. Because the improved affinity for ouabain relative to digoxin is observed for H:Tyr50Asp but not H:Tyr50Asn, the negative charge of the Asp side chain is most probably responsible for this fine specificity shift.

Several H 50 mutants, especially H:Tyr50Ala and H:Tyr50Gly, display increases of varying magnitude in relative affinity for dihydrodigoxin and dihydrodigoxigenin (Table IV). While there are several contacts between 26-10 and the lactone ring (31), H:Tyr50 provides none. It is therefore unexpected that smaller residues at H 50 might increase relative affinity for a hapten with a saturated lactone. The lactone of dihydrodigoxigenin is non-planar, and the positions of many of its constituent atoms are shifted in relation to the steroid moiety of the molecule (24). Modeling of the interaction between 26-10 and dihydr~digoxigenin~ suggests that a small rotation of the hapten steroid moiety is required to accommodate the saturated lactone in the binding site, resulting in unfavorable van der Waals contact with the H:Tyr50 Cc2. Substitution of a smaller residue at H 50 would permit the hapten shift without the unfavorable contact with the H 50 side chain, thus improving the relative affinity for dihydrodigoxigenin.

The results of docking experiments (Fig. 4) suggest that the higher relative affinity for 12-acetyldigoxin of H:Tyr50Ala compared with 26-1Owt is due to the absence of an unfavorable contact in the H:Tyr50Ala binding site between the hapten acetyl group and L:Pro96. A role for L:Pro96 in the lowered affinity of 26-10 for 12-acetylated haptens was not antici- pated, and its identification illustrates the usefulness of modeling. For example, we can now suggest that engineered mutations of L:Pro96 might modulate the affinity of antibodies for 12-acetyldigoxin relative to digoxin.

The H 50 mutants described here exhibit a variety of additional fine specificity changes compared with 26-10wt. Antibody H:TyrSOAsn is more specific for digoxin due to an increased recognition of the hapten 12-OH. Digoxin and digoxigenin are bound equally by H:Tyr50Asn, but all other analogues are bound with affinities at least 10-fold lower than digoxin. In contrast, 26-1Owt binds digoxin and several analogues with similar high affinities. Antibody H:Tyr50Ala appears to be less specific for digoxin than 26-1Owt based on its improved relative affinities for 12-acetylated haptens and those haptens with saturated lactones. Antibody H:Tyr50Asp may be considered more specific for digoxin because of its increased recognition of the 12-OH but less specific given its equal affinities for digoxin and ouabain. In each case, however, the altered fine specificity occurs at the cost of significantly reduced affinity, relative to 26-10wt, for all haptens. None of the antibodies mutated at H 50 demonstrated substantially improved affinity for any tested digoxin analogue above that

We have shown here that whereas the 26-10 antibody does not distinguish between digoxin, the immunizing hapten, and digitoxin, a substitution at H 50 could result in antibodies with up to a 32-fold higher affinity for digoxin relative to digitoxin. This observation is somewhat unexpected. Previous studies have suggested that an increase in or acquisition of specificity of an antibody for an antigen is concomitant with an increase in affinity for that antigen (41-44). The specificity of the H:Tyr50Asn antibody demonstrates that an increase

’ J. F. Schildbach, S.-Y. Shaw, R. I. Near, R. E. Bruccoleri, E. Haber, L. A. Herzenberg, P. D. Jeffrey, D. J. Panka, David R. Parks, J. Novotny, and M. N. Margolies, submitted for publication; J. F. Schildbach, unpublished observation.

of 26-10.

in specificity for a given hapten can also occur with a loss of affinity for that hapten. It is not surprising, however, that the “less specific” 26-10 antibody, rather than H:Tyr50Asn, was isolated following immunization with a digoxin-protein con- jugate. If the sole basis of in vivo selection were affinity for antigen, the 26-10 antibody would have a clear advantage.

Single variable region mutations have been shown to be sufficient to alter antibody fine specificity (43, 45-48). Of these single mutations causing specificity shifts, two are known from crystallographic data to involve contact residues (45, 46). These mutations removed hydrogen bonds between antibody and hapten. The contribution of the 26-10 H:Tyr50 residue to digoxin binding is derived from shape complementarity rather than through hydrogen bond formation (31). The binding displayed by the H 50 mutants is consistent with the conclusion drawn from the crystal structure of the 26-10. digoxin complex (31) that shape complementarity alone can significantly contribute to specificity and affinity of an antibody. The observation that a mutant antibody (H:Tyr50Asn) may exhibit both lowered affinity and enhanced fine specificity is unusual and may be unique to an antibody deriving binding energy largely from shape complementarity. Other substitutions at H 50 and substitutions at H 3~5,~ however, caused lower affinity for digoxin with reduced fine specificity. These observations are analogous to results obtained using antibodies directed against other antigens (45, 46).

The results reported here for the H 50 residue, as well as similar studies at other 26-10 H chain CDR residues7 (7) emphasize that antibody affinity and specificity may be mod- ulated in a variety of ways by different amino acid substitutions at a single site and thus have implications for antibody engineering experiments. It would be desirable to engineer antibodies with high affinities that readily distinguish struc- turally similar compounds. However, for the limited case in which mutations were made at H 50 of antibody 26-10, improved specificity for digoxin is possible but at the cost of affinity. To improve both specificity and affinity of 26-10 for digoxin, mutations at other positions or at multiple positions in the binding site would be required. Even extensive mutagenesis of the 26-10 binding site, however, may not improve the affinity for digoxin. While engineered antibodies with improved affinities have been reported, these experiments involve antibodies with affinities considerably lower than that of 26-10 (41, 49). It is likely that 26-10, as a high affinity secondary response antibody, is the result of somatic mutation of lower affinity precursors and antigen-driven selection (affinity maturation). Thus the complementarity between hapten and antibody is extensive and may not be easily improved by mutation, particularly at single sites.

Acknowledgments-We thank Terry R. Stouch (Bristol-Myers Squibb) for assistance with the electrostatic charge calculation and Rou-fun Kwong and Lii Suen (Massachusetts General Hospital) for technical assistance.

REFERENCES 1. Ehrlich,.P. (1900) Proc. R. Soc. Lond. Ser. B Biol. Sci. 66,424-448 2. Landstemer, K., and van der Scheer, J. (1928) J. Exp. Med. 48, 315-320 3. Landsteiner, K., and van der Scheer, J. (1929) J. Exp. Med. 50,407-417 4. Mudeett-Hunter. M.. Maraolies. M. N., Ju, A., and Haber, E. (1982) J .

5. Schildbach, J. F., Panka, D. J., Parks, D. R., Jager, G . C., Novotny, J., Imkunol. 129’, 1165-1172

Herzenberg, L. A,, Mudgett-Hunter, M., Bruccoleri, R. E., Haber, E., and Margolies, M. N. (1991) J. Ed. Chem. 266,4640-4647

6. Mudaett-Hunter. M., Anderson, W., Haber, E., and Margolies, M. N. (1985)

7. Schildbach, J. F., Near, R. I., Bruccoieri, R. E., Haber, E., Jeffrey, p. D., Mol. Immunol.’ 22,477-488

Novotny, J., Sheriff, S., and Margolles, M. N. (1993) Protein S c m m 2 ,

8. Near, R. I., Ng, S. C., Mudgett-Hunter, M., Hudson, N. W., Margolies, M. N., Seidman, J. G., Haber, E., and Jacobson, M. A. (1990) Mol. Imrnunol. 27,901-909

206-214

9. Munson, P. J. (1983) Methods Enzyrnol. 92,543-576

Single Residue Modulates Antibody Fine Specificity 21747 10.

11. 12.

13.

14. 15.

16.

17. 18.

19.

20. 21.

23. 22.

24.

25.

26.

27.

28.

29. 30.

DeLean, A., Munson, P. J., and Rodbard, D. (1978) Am. J. Physwl. 235,

Cheng, Y., and Prusoff, W. H. (1973) Biochem. Pharmacol. 22,3099-3108 Novotny, J., Bruccoleri, R. E., and Saul, F. A. (1989) Biochemistry 28,

Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States, D. J., Swaminathan,

Williams, D. E. (1988) J. Comput. Chem. 9, 745-751 Go, K., Kartha, G., and Chen, J. P. (1980) Acta Crystallogr. Sect. B Struct.

Go, K., and Bhandary, K. K. (1989) Acta Crystallogr. Sect. B Struct. Sei.

Go, K., and Kartha, G. (1982) Cryst. Struct. Comm. 11, 285-290 Go, K., and Kartha, G. (1980) Acta Crystallogr. Sect. B Struct. Sci. 36,

Przybylska, M., and Ahmed, F. R. (1979) Acta Crystallogr. Sect. B Struct.

Go, K., and Kartha, G. (1981) Cryst. Struct. Comm. 10,1329-1334 Messerschmidt, A. (1980) Cryst. Struct. Comm. 9, 1185-1194

Go, K., and Kartha, G. (1984) Acta Crystallogr. Sect. C Struct. Commun. Kartha, G., and Go, K. (1981) Cryst. Struct. Comm. 10, 1323-1327

Mostad, A. (1982) Acta Chem. S c a d . Ser. B Og. Chem. Biochem. 36,635-

Karle, I. L., and Karle, J. (1969) Acta Crystallogr. Sect. C Cryst. Struct.

Gilardi, R. D., and Karle, I. L. (1973) Acta Crystallogr. Sect. B Struct. Sci.

Gilardi, R. D., and Flippen, J. (1970) Acta Crystalbgr. Sect. B Struct. Sci.

Prasad, L., and Gabe, E. J. (1983) Acta Crystallogr. Sect. C Cryst. S trut .

E97-El02

4735-4749

S., and Karplus, M. (1983) J. Comput. Chem. 4, 187-217

Sei. 36,1811-1819

45,306-312

3034-3040

Sei. 36,2436-2440

40,1866-1869

639

Commun. 40,1866-1869

29,1842-1848

26,207-218

&mmrm 39. 372-215

Petsko, G. A., Haber, E., Margolies, M. N., and Sheriff, S. (1993) Proc.

32. Shih, H. H.-L., Brady, J., and Karplus, M. (1985) Proc. Natl. Acad. Sci. Natl. Acad. Sci. U. S . A,, in press

U. S. A. 82, 1697-1700 33. Bruccoleri, R. E., and Karplus, M. (1987) Biopolymers 26,137-168 34. Baker, E. N., and Hubbard, R. E. (1984) Prog. Biophys. Mol. Bid. 44,97-

179 35. Lee;B. K., and Richards, F. M. (1971) J. Mol. Biol. 55 379-400 36. Sheriff, S., Hendrickson, W. A,, and Smith, J. L. (1987; J . Mol. Biol. 197,

37. Chothia, C. (1974) Nature 248,338-339 38. Sharp, K. A., Nicholls, A., Fine, R. F., and Honig, B. (1991) Science 262,

39. EiE-nberg, D., Wesson, M., and Yamashita, M. (1989) Chem. Scr. 29,217-

273-296

106-109

40. 41. 42.

43.

44.

45.

46.

47.

48.

49.

Horton, N., and Lewis, M. (1992) Protein Science 1,169-181 Roberts, S., Cheetham, J. C., and Rees, A. R. (1987) Nature 328, 731-734 Ne arstek Y. Andre-Schwartz, J., Manser, T., Wysocki, L. J., Breitman, E, Stolfar, B. D., Gefter, M., and Schwartz, R. S. (1986) J. Exp. Med.

164,614-626 Diamond, B., and Scharff, M. D. (1984) Proc. Natl. Acad. Sci. U. S. A. 81,

5841-5844 Barbas, C. F., Bain J. D., Hoekstra D. M., and Lerner, R. A. (1992) Proc.

Natl. Acad. Sci. h. S. A. 89,4457'4461 Glockshuber, R., Stadlmuller, J., and Pluckthun, A. (1991) Biochemistry

30,3049-3054 Denzin, L. K., Whitlow, M., and Voss, E. W., Jr. (1991) J. Biol. Chem.

266,14095-14103 Lavoie, T. B.. Drohan, W. N., and Smith-Gill, S. J. (1992) J. Immunol.

148,503-513 Near, R. I., Bruccoleri, R. R., Novotny, J., Hudson, N. W., White, A., and

Mudgett-Hunter, M. (1991) J. immunol. 146,627-633 RiEtkrnann, L., Weill, M., and Cavanagh, J. (1992) J. Mol. Biol. 224,913-

221

. . . . . . . -. . . - - , - . - - . - Go, K., and Kartha, G. (1982) Cryst. Struct. Comm. 11,279-284 Fullerton, D. S., Yoshioka, K., Rohrer, D. C., From, A. H. L., and Ahmed,

50. Kabat, E. A., Wu, T. T., Perry, H. M., Gottesman, K. S., and Foeller, C. (1991) Sequences of Proteins ofImmunologica1 Interest, U. S. Department

31. Jeffrey, P. D., Strong, R. K., Sieker, L. C., Chang, C. Y., Campbell, R. L., of Health and Human Services, U. S. Government Printing Office, Wash- ington, D. C.

Y l O

K. (1980) Mol. P h a r m o l . 17,43-51

Documents

THE OF CHEMISTRY Vol. 268, No. October pp. for in …THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 268, No. 29, Issue of October 15, pp. 21739-21747, 1993 0 1993 by The American Society