PROTEINS: Structure, Function, and Genetics 7:93-98 (1990)
SHORT COMMUNICATION
Computer Analysis of Mutations That Affect Antibody Specificity Jiri Novotny, Robert E. Bruccoleri, and Edgar Haber Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts 02114*
ABSTRACT The mouse hybridoma cell line 40-150 secretes antibodies with high affinity toward the cardiac glycosides digoxin and digitoxin. A spontaneous mutant, 40-150 A2.4, produces an antibody which carries a single residue mutation, Ser + Arg, in its heavy chain (H94) and has an altered specificity. A secondorder mutant, 40-150 A2.4 P.10, produces two antibody molecules, one the same as 40-150 A2.4, the other lacking two residues at the Nterminus of its H chain, and having a specificity profile approaching that of 40-150 antibody.’ The N-terminus and the position H94 are distant from the antigen-binding site of the antibody; thus, the structural basis of the specificity changes was not immediately clear. Approximate structures of the 40-150 antibody and its mutants were constructed in the computer, based on atomic coordinates of the homologous mouse antibody McPC 603. Using the program CONGEN, the torsional space of the polypeptide backbone and side chains around position H94 was uniformly sampled, and the lowest energy conformations were analyzed in detail. The results indicate that when Arg-H94 is substituted for Ser, Arg-H94 can hydrogen bond to side chains of Asp-HlOl, Arg-L46, and Asp-L55. This results in a change in the surface of the combining site which may account for the affinity changes. Deletion of the two N-terminal residues increases solvent accessibility of ArgH94. The solvation may cause a hydrogen bond between Arg-H94 and Asp-H101 to be lost, restoring the structure to one similar to that of 40-150. Key words: protein conformation, CONGEN, immunolgobulin, hydrogen bond, digoxin
site and the “epitope” of the antigen.”4 Amino acid mutations that change antibody specificity are thus expected to occur a t the surface of antibody combining sites, where they directly change the shape of the binding pocket. Paradoxically, some of the reported5-’ single-residue mutations that cause loss of antigen binding were shown to be at positions inaccessible to solvent or antigen.1,9,10 In this communication, we show an application of the method of uniform conformational sampling,” to gain insight into the detailed mechanism of how this “action a t a distance” may be effected among the atoms that constitute an antibody molecule. Our results also suggest an explicit mechanism of how the “framework” parts of antibody structure,’” which are often constant in antibodies of different specificities, are directly involved in, and even essential to, the expression of distinct individual specificities.
MATERIALS AND METHODS
-
The mutation analyzed in this communication is Arg replacement in position 94 (Kabat the Ser numbering, see ref. 12) of the heavy chain of mouse hybridoma 40-150 A2.4.’ This immunoglobulin posseses a high affinity toward cardiac glycosides digoxin = 6.1 kmd and digitoxin (Kas= 5.4nmol),and the mutation alters the relative binding constants of these two compounds. A second-order mutant, 40150 A2.4 P.10, has also been found which possesses Arg-H94, but whose heavy chain is two residues shorter a t the amino terminus. Binding characteristics of the second order mutant are different from that of the 40-150 Arg-H94 variant, and approach those of the parent, i.e., unmutated, 40-150 molecule.’ Examination of the known amino acid se-
INTRODUCTION The accuracy with which antibodies recognize antigens has long been studied as a paradigm of biological specificity. Stable antigen-antibody complexes arise from noncovalent forces acting between the two molecular surfaces, the antibody combining 0 1990 WILEY-LISS, INC
Received June 19, 1989; revision accepted September 6, 1989. Address reprint requests to J. Novotny, Squibb Institute for Medical Research, D4117, Princeton NJ 08543-4000. “The Authors’ present address is The Squibb Institute for Medical Research, P.O. Box 4000, Princeton, N J 08543-4000.
94
J. NOVOTNY ET AL.
Fig. 1. A stereoscopic space filling drawing of the Fv domain of the mouse immunoglobulin McPC 603.” In this view, the ArgH94 and the Asp-H101 are centered, and the antigen combining site is at the top. The color code is as follows: The heavy chain framework atoms are dark gray; the light chain framework is light gray; the first, second, and third heavy chain hypervariable loops
are in lavender, medium green, and yellow, respectively; the first, second, and third light chain hypervariable loops are in purple, brown, and orange, respectively; Arg-H94 is in light blue; AspH101 is in red; the N-terminal two residues of the heavy chain are in dark green; and residues L46 and L55 are in peach.
quences” shows Arg-H94 to be present in a majority of mouse and human immunoglobulin heavy chains, including the four Fab antibody fragments whose three-dimensional structures are known from X-ray crystallography.13-16 In all the X-ray structures the N-terminus of the heavy chain comes to a contact distance of the Arg-H94 (Fig. 1).In all but one X-ray structure Arg-H94 adopts an identical conformation with its side chain hydrogen bonded to that of AspH101. Both the H94 and HlOl side chains are partially buried and directed away from the antigenbinding cavity (Fig. 1).Indeed, Arg-H94 has been considered to be a “framework” residue, i.e., outside of the hypervariable or complementarity-determining loops12 which comprise the antigen combining site. Thus, a direct impact of position H94, and the heavy chain N-terminus, on antigen binding was unexpected. Conformational invariance of the Arg-H94Asp-H101 motif led us to assume that a similar structural arrangement exists in the 40-150 mutant, which contains Arg-H94 and Asp-H101. We performed computations aimed a t verifying this assumption, as well as a t identifying the possible structural changes associated with the presence of Ser-H94 in the 40-150 heavy chain. The method of uniform conformational sampling, as implemented in the program CONGEN, has been described.” Briefly, the method consists of: (1)the generation of stereochemically acceptable conforma-
tions that avoid significant van der Waals overlaps with the rest of the structure (the chosen torsional degrees of freedom are systematically sampled using a predetermined angular grid). (2) The acceptable conformations are energy ranked, and the lowest energy conformations are chosen. In a n ideal case, the lowest energy conformation should correspond to the natural one. In practice, due to the approximate nature of the empirical energy potential employed, additional criteria also have to be considered in order to identify the correct conformation. For example, we previously considered solvent exposure of calculated conformations as a n aid in finding native-like structures of antibody combining site surface.l7 In this communication, where we deal with residues buried in protein core, criteria other than solvent accessibility have to be considered (see below). Although the conformational sampling method can in principle be applied to macromolecular systems of any size, computer time limitations currently restrict the sampling to small regions of space and short sections of polypeptide backbone. The strategy employed in our particular case was predicated by the backbone structure of that region, a p hairpin with residues H93 and H94 running antiparallel to the segment H101-Hl02, with the antiparallel P-sheet backbone hydrogen-bonding pattern between the two segments (cf. Fig. 2A). First, we constructed models of 40-150 antibody structure in the immediate vicinity of the position
95
COMPUTER ANALYSIS OF MUTATIONS
H94. Beginning with the crystallographic coordinates of a homologous mouse immunoglobulin, McPC 603,15 we replaced all the side chains within a contact distance (4 A) of the Arg-H94 with those that are present in the 40-150 sequence. Second, the conformational space of all the newly introduced side chains, as well as the backbones of residues H93 and H94, was explored with CONGEN. In order to ensure the maximal conformational freedom for the newly generated backbone, residues H95 through HlOO (i.e., the top of the hairpin loop above the sampled residues) were deleted from the system. As an independent test of our sampling protocol, the backbone and side chain torsions of the homologous residues in McPC 603 were sampled using identical conditions, and the results were compared with the crystallographic reality. Third, a similar set of CONGEN runs sampled backbone conformations of residues H101, H102, side chains of H93 and H94, and side chains within the 4 A contact distance of the Asp-H101 residue. These involved light chain residues L46 and L55, which in the sequence of the 40-150 hybridoma light chain are Arg and Asp, respectively. Once again, a n analogous run on McPC 603 gauged the predictive power of our sampling protocol. In the final set of runs, the backbones of residues H94 and HlOl were varied, together with their surrounding side chains. Typically, one such sampling run required in between 2 and 3 days of CPU time on pVAX 11.
RESULTS Results of CONGEN searches, presented in Table I and Figure 2, can be summarized as follows. (1)In all the runs, and particularly in those aimed at reproducing the McPC 603 crystallographic structure, the McPC 603 conformations were closely approximated by the vast majority of the lowest energy conformations (rms shifts from the X-ray structure in between 1.3 and 1.6 A, with backbone atom shifts typically not exceeding 0.5 A, cf. Fig. 2). The fact that test runs on the crystallographic structure of McPC 603 were able to reconstruct quite accurately the main features of this structure is encouraging. It indicates that our search protocols are likely to produce valid results in modeling the 40-150 environments as well. At the same time, constructions were also obtained that had large rms deviations from the McPC 603 crystallographic structure (over 5 A). These, however, were characterized by high potential energy values. (2) As a rule, variants of a single conformational motif dominated the lowest energies, and one or few of those were distinguished by particularly good hydrogen-bonding interactions among electrically charged side chains; such conformations are listed in Table I. Our selection of conformations with the best hydrogen bonding energy is justified by the fact that: (i) the polar side chains varied in our CON-
GEN runs are all buried; and that (ii) polar atoms buried in protein interior always satisfy their hydrogen-bonding potential." In particular, side chains containing formal electric charges usually form hydrogen-bonded clusters in protein interior^.'^,^^ In immunoglobulins, residues involved in such hydrogen-bonding networks belong among the most stringently conserved.21
DISCUSSION Data in the Table I indicate that the Ser + Arg replacement in the 40-150 mutant is not likely to produce a sizable polypeptide backbone rearrangement that would propagate into the parts of antigen binding site. However, large shifts (3-4 8)are seen in positions of side chain atoms of Asp-H101 and Arg-L46, depending on whether Arg or Ser is present in position H94. The serine side chain, due to its small size, is unable to make a hydrogen bond to the Asp-101 side chain and becomes hydrogenbonded to the backbone instead (see Fig. 2B). ArgH94, on the other hand, is hydrogen bonded to AspH101, joining a hydrogen-bonding net which involves the Asp residues HlOl and L55. The Arg residue L46 is central to this net, being hydrogen bonded to both the Asp-H101 and Asp-L55. In particular, the Arg-L46 side chain reorients in response to conformational changes experienced by the side chain of Asp-H101 in the presence of Arg-H94. AspL55 is a complementarity-determining residue and Arg-L46 is a n important domain-domain contacting residue located a t the bottom of the antigen-binding cavity. Rearrangements of these side chains, in response to the Arg or Ser in position H94, are likely to influence directly the shape of antigen-binding surface. Solvent accessibility computations" show that deletion of the two N-terminal residues, as it occurs in the 40-150 A2.4 P10 mutant, increases solvent accessibility of the Arg-H94 by loo%, thus increasing the likelihood that the charged atoms in the Arg side chain would be neutralized by solvation. When this occurs, Arg-H94 can no longer participate in the hydrogen-bonding net, resulting in a situation similar to that found in 40-150 where the residue in this position is Ser. Thus, the side chain shifts affecting the antibody-binding cavity that were postulated to occur in the variant would no longer be seen in the second-order variant.
CONCLUSION In summary, our computations suggest a plausible hypothesis of a detailed atomic mechanism whereby a side chain replacement, via a hydrogen bond "push-pull" mechanism, propagates over a distance of -10 A, and across protein domain interface, to generate a subtle change in the surface of the antigen combining site. The hypothesis is
96
J. NOVOTNY ET AL.
TABLE I. Summary of CONGEN Searches* Conformation number'. 603-37-145 40150-38-153 40150-39-186 603-34-437
40150-35-386
40150-36-410
Sampled residues, backbone + side chains Ala-H93 Arg-H94 Thr-H93 Arg-H94 Thr-H93 Ser-H94 Asp-H101 Val-H102
Asp-H101 Tyr-H102
Asp-HI01 Tyr-H1O2
Sampled residues, side chains only Leu-H4 Phe-H27 Phe-H32 Tyr-H33 Met-H34 Glu-H35 Asp-H101 Val-H102 Leu-H4 Phe-H27 Tyr-H32 Tyr-H33 Met-H34 Ala-H35 Asp-H101 Tyr-H102 Leu-H4 Phe-H27 Tyr-H32 Tyr-H33 Met-H34 Ala-H35 Asp-HI01 Tyr-HI02 Leu-H4 Phe-H27 Phe-H32 Tyr-H33 Met-H34 Glu-H35 Ala-H93 Arg-H94 Leu-L46 Glu-L55 Leu-H4 Phe-H27 Tyr-H32 Tyr-H33 Met-H34 Ala-H35 Thr-H93 Arg-H94 Arg-L46 Asp-L55 Leu-H4 Phe-H27 Tyr-H32 Tyr-H33 Met-H34 Ala-H35 Thr-H93 Ser-H94 Ara-L46 Asp-L55 Leu-H4 Phe-H27 Phe-H32 Tvr-H33 Met-H34 Glu-H35 Aia-H93 Val-H102 Leu-L46 Glu-L55 Leu-H4 Phe-H27 Tyr-H32 Tyr-€I33 Met-H34 Ala-H35 Thr-H93 Tyr-H102 Arg-L46 Asp-L55 Leu-H4 Phe-H27 Tyr-H32 Tyr-H33 Met-H34 Ala-H35 Thr-H93 Arg-H94 Arg-L46 Asp-L55
Number of conformations 308 306 360 507
578
594
Y
603-40-152
40150-41-94
40150-42-83
Conformation number' 603-37-145 40150-38-153 40150-39-186 603-34-437 40150-35-386 40150-36-410 603-40-152 40150-41-94 40150-42-83
Arg-H94 Asp-H101
Arg-H94 Asp-H101
Ser-H93 Asp-H101
In vacuo energy rankrng 2nd 5th 1st 3rd 4th 2nd 1st 13th 2nd
Total energy (kcal) - 125.8 -127.7 -100.6 -132.3 -171.1 -145.6 -146.2 -158.9 - 149.5
Hydrogen bond energy (kcal) -7.1 - 10.3 -8.7 -8.0 - 12.4 - 16.8 -8.2 -15.2 -13.1
181
181
181
rms shift to McPC 603
(A,
1.6 1.4 1.1 1.3 1.4 1.3 1.4 1.5 1.4
"Backbone torsion angles @ and J' ' were sampled using 30" grid maps; side chain torsions were sampled on the three lowest energy points (trans, i gauche) using the van der Waals avoidance option to circumvent close atomic c o n t a ~ t s . Conformations '~ were ranked according to potential energy computed by the CHARMM empirical potential function in V ~ C U O . ' ~Nonbonded interactions were computed within a cut-off distance of 8 A using the E = R electrostatic model as implemented in CONGEN. An explicit geometrydependent term was used to estimate the quality of hydrogen bonds (see ref. 23 for details.). ?The first number in the label, 603 or 40150, indicates whether the run sampled the torsional space of the McPC 603 structure or that of the 40-150 model. The second number is the run serial number, and the third number is the conformation label.
counterintuitive and could be derived only by computations. It is unexpected in that it does not involve backbone rearrangements of the hypervariable loops, and it proposes a n important role for framework, in addition to hypervariable, residues in generating a change in antibody specificity. The hypothesis is open to experimental verification via
site-directed mutagenesis of antibody-coding genes. The computational approach demonstrated here is directly applicable to analysis of single residue mutations of any protein. In principle, uniform conformational sampling should be superior to alternative methods such as interactive molecular graphics, which is subjective, or energy minimization, which
COMPUTER ANALYSIS OF MUTATIONS
ii
J b
x
97
& 5
71 C
Fig. 2. (a) Stereoscopic line diagram of polypeptide backbone and selected side chains in the vicinity of Arg-H94. The light lines trace the backbone of residues H93 through H103 (that is, the third heavy chain hypervariable loop) of the McPC 603 crystallographic structure together with side chains of Arg-H94 (left) and Asp-H101 (right). The heavy lines show positions of the backbone and side chains of residues H94 and HI01 in the computed conformation 603-37-145 (see Table I), aimed at reproducing the crystal conformation drawn in light lines. Although the computed side chain conformations deviate from the crystallographic ones, the hydrogen-bonding between the two side chains has been well reproduced, and backbone conformations match virtually exactly. Compare, for example, position of the free end-point of the computed conformation (C atom of H94, heavy circle) to those found in the X-ray structure (light circle). In this and the following figures, the two conformations compared were matched by least-squares
superposition of the Trp-H103 residue. Side chains are labeled by their one-letter name code and Kabat’’ numbering (H or L stands form the heavy and light chain, respectively). (b)Computed conformation in the 40-150 hybridoma. The light lines show the McPC 603 3rd heavy chain hypervariable loop for reference. The heavy lines trace the backbone and side chains of residues H101 and H I 0 2 in the run 40150-36-410 which analyzed the 40-150 model with Ser side chain in position H94. Two residues appearing to the right are the light chain Arg-L46 and Asp-L55 which participate in hydrogen-bonding with Asp-H101 and Arg-H94. (c) Comparison of computed conformations 401 50-35-386 (i.e., a 40-150 structure model with Arg at position H94, light lines) and 401 50-36-410 (i.e., a model with Ser at position H94, heavy lines). Note the shift in positions of the side chains Arg-L46 and Asp-L55 (on the right side of the picture). The rms difference of Asp-L55 0 6 1 atoms is 4.0 A, that of Arg-L46 Nr12 atoms is 4.3 A.
is known for its tendency to direct atoms into local potential energy wells, different from the global energy minima that distinguish the naturally occurring conformations.
ACKNOWLEDGMENTS This work was supported by NIH Grant HL19259-11 and ONR Research Grant 140-86-K-0116.
98
J. NOVOTNY ET AL.
REFERENCES 1 Panka, D. J., Mudgett-Hunter, M., Parks, D. R., Peterson, L.L., Herzenberg, L. A,, Haber, E., Margolies, M. N. Variable region framework differences result in decreased or increased affinity of variant anti-digoxin antibodies. Proc. Natl. Acad. Sci. U.S.A. 85:3080-3084, 1988. 2. Amit A. G., Mariuzza R. A,, Phillips S. E. V., Poljak R. J . Three-dimensional structure of an antigen-antibody complex a t 2.8 A resolution. Science 233:747-752, 1986. 3. Sheriff, S., Silverton, E. W., Padlan, E. A,, Cohen, G. H., Smith-Gill, S., Finzel, B. C., Davies, D. R. Three-dimensional structure of a n antibody-antigen complex. Proc. Natl. Acad. Sci. U.S.A. 84:8075-8079, 1987. 4. Novotny, J., Bruccoleri, R. E., Saul, F. A. On the attribution of binding energy in antigen-antibody complexes McPC 603, D1.3, and HyHEL-5. Biochemistry 28:47354749, 1989. 5. Rudikoff S., Satow S., Padlan E., Davies D. R., Potter, M. Kappa chain structure from a crystallized murine Fab: Role ofjoining segment in hapten binding. Mol. Immunol. 18:705-711, 1981. 6. Azuma T., Igras V., Reilly E. B., Eisen H.. Diversity a t the variable-joining region boundary of A light chains has a pronounced effect on immunoglobulin ligand-binding activity. Proc. Natl. Acad. Sci. U.S.A. 81:6139-6143, 1984. 7. Cook W. D., Rudikoff S., Giusti A. M., Scharff M. D. Somatic mutation in a cultured mouse myeloma cell affects antigen binding. Proc. Natl. Acad. Sci. U.S.A. 79:12401244, 1982. 8. Rudikoff S., Giusti A. M., Cook W. D., Scharff M. D. Single amino acid substitution altering antigen-binding specificity. Proc. Natl. Acad. Sci. U.S.A. 79:1979-1983, 1982. 9. Novotny J., Bruccoleri R. E., Newel1 J., Murphy D., Haber E., Karplus M. Molecular anatomy of the antibody binding site. J. Biol. Chem. 258:14433-14437, 1983. 10. Chothia C., Novotny J., Bruccoleri R.E., Karplus M. Domain association in immunoglobulin molecules; the packing of variable domains. J. Mol. Biol 160:325-342, 1985. 11. Bruccoleri R. E., Karplus M. Prediction of the folding of short polypeptide segments by uniform conformational sampling. Biopolymers 26:137-168, 1987. 12. Kabat E. A,, Wu T. T., Reid-Miller M., Perry H., Gottes-
man K.S. “Sequences of Proteins of Immunological Interest,” 4th ed. U. s. Department of Health and Human Services, N.I.H., Washington D.C., 1987. 13. Saul F. A., Amzel L. M., Poljak R. J . preliminary refinement and structural analysis of the Fab fragment from human immunoglobulin New a t 2.0 A resolution. J . Biol. Chem. 253:585-597, 1978. 14. Marquart, M., Deisenhofer J., Huber R., Palm, W. Crystallographic refinement and atomic models of the intact immunoglobulin molecule Kol and its antigen-binding fragment a t 3.0 A and 1.9 A resolution. J . Mol. Biol. 141: 369-391, 1980. 15. Satow Y., Cohen G. H. Padlan E. A,, Davies D. R. Phosphocholine binding immunoglobulin Fab McPC 603: An X-ray diffraction study at 2.7 A. J. Mol. Biol. 190593-604, 1986. 16. S. W. Suh, M. A. Navia, G. H. Cohen, D. N. Rao, S. Rudikoff, D. R. Davies. The galactan-binding immunoglobulin Fab 5539: An X-ray study a t 2.6 A resolution. Proteins 1:74-80, 1986. 17. Bruccoleri, R. E., Haber, E., Novotny, J . Structure of antibody hypervariable loops reproduced by a conformational search algorithm. Nature (London) 335564-568, and 336: 226, 1988. 18. C. Chothia, The nature of accessible and buried surfaces in proteins. J . Mol. Biol. 105:l-14, 1976. 19. Peeters D, Peeters J. A simple and novel interpretation of the three-dimensional structure of globular proteins based on quantum-mechanical computations on small model molecules. Biopolymers 24:491-508, 1985. 20. Rashin A. A,, Honig B. On the environment of ionizable groups in globular proteins. J. Mol. Biol. 173:515-521, 1984. 21. Novotny, J., Haber, E. Structural invariants of antigen binding: Comparison of immunoglobulin VL-VH and VLVL domain dimers. Proc. Natl. Acad. Sci. U.S.A. 8245924596, 1985. 22. Lee, B. K., Richards F. M. The interpretation of protein structures: Estimation of static accessibility. J. Mol. Biol. 55:379-400 1971. 23. Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States, D. J., Swaminathan, S., Karplus, M. CHARMM: A program for macromolecular energy, minimization and dynamics calculations. J. Comput. Chem. 4:187-217, 1983.
SHORT COMMUNICATION
Computer Analysis of Mutations That Affect Antibody Specificity Jiri Novotny, Robert E. Bruccoleri, and Edgar Haber Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts 02114*
ABSTRACT The mouse hybridoma cell line 40-150 secretes antibodies with high affinity toward the cardiac glycosides digoxin and digitoxin. A spontaneous mutant, 40-150 A2.4, produces an antibody which carries a single residue mutation, Ser + Arg, in its heavy chain (H94) and has an altered specificity. A secondorder mutant, 40-150 A2.4 P.10, produces two antibody molecules, one the same as 40-150 A2.4, the other lacking two residues at the Nterminus of its H chain, and having a specificity profile approaching that of 40-150 antibody.’ The N-terminus and the position H94 are distant from the antigen-binding site of the antibody; thus, the structural basis of the specificity changes was not immediately clear. Approximate structures of the 40-150 antibody and its mutants were constructed in the computer, based on atomic coordinates of the homologous mouse antibody McPC 603. Using the program CONGEN, the torsional space of the polypeptide backbone and side chains around position H94 was uniformly sampled, and the lowest energy conformations were analyzed in detail. The results indicate that when Arg-H94 is substituted for Ser, Arg-H94 can hydrogen bond to side chains of Asp-HlOl, Arg-L46, and Asp-L55. This results in a change in the surface of the combining site which may account for the affinity changes. Deletion of the two N-terminal residues increases solvent accessibility of ArgH94. The solvation may cause a hydrogen bond between Arg-H94 and Asp-H101 to be lost, restoring the structure to one similar to that of 40-150. Key words: protein conformation, CONGEN, immunolgobulin, hydrogen bond, digoxin
site and the “epitope” of the antigen.”4 Amino acid mutations that change antibody specificity are thus expected to occur a t the surface of antibody combining sites, where they directly change the shape of the binding pocket. Paradoxically, some of the reported5-’ single-residue mutations that cause loss of antigen binding were shown to be at positions inaccessible to solvent or antigen.1,9,10 In this communication, we show an application of the method of uniform conformational sampling,” to gain insight into the detailed mechanism of how this “action a t a distance” may be effected among the atoms that constitute an antibody molecule. Our results also suggest an explicit mechanism of how the “framework” parts of antibody structure,’” which are often constant in antibodies of different specificities, are directly involved in, and even essential to, the expression of distinct individual specificities.
MATERIALS AND METHODS
-
The mutation analyzed in this communication is Arg replacement in position 94 (Kabat the Ser numbering, see ref. 12) of the heavy chain of mouse hybridoma 40-150 A2.4.’ This immunoglobulin posseses a high affinity toward cardiac glycosides digoxin = 6.1 kmd and digitoxin (Kas= 5.4nmol),and the mutation alters the relative binding constants of these two compounds. A second-order mutant, 40150 A2.4 P.10, has also been found which possesses Arg-H94, but whose heavy chain is two residues shorter a t the amino terminus. Binding characteristics of the second order mutant are different from that of the 40-150 Arg-H94 variant, and approach those of the parent, i.e., unmutated, 40-150 molecule.’ Examination of the known amino acid se-
INTRODUCTION The accuracy with which antibodies recognize antigens has long been studied as a paradigm of biological specificity. Stable antigen-antibody complexes arise from noncovalent forces acting between the two molecular surfaces, the antibody combining 0 1990 WILEY-LISS, INC
Received June 19, 1989; revision accepted September 6, 1989. Address reprint requests to J. Novotny, Squibb Institute for Medical Research, D4117, Princeton NJ 08543-4000. “The Authors’ present address is The Squibb Institute for Medical Research, P.O. Box 4000, Princeton, N J 08543-4000.
94
J. NOVOTNY ET AL.
Fig. 1. A stereoscopic space filling drawing of the Fv domain of the mouse immunoglobulin McPC 603.” In this view, the ArgH94 and the Asp-H101 are centered, and the antigen combining site is at the top. The color code is as follows: The heavy chain framework atoms are dark gray; the light chain framework is light gray; the first, second, and third heavy chain hypervariable loops
are in lavender, medium green, and yellow, respectively; the first, second, and third light chain hypervariable loops are in purple, brown, and orange, respectively; Arg-H94 is in light blue; AspH101 is in red; the N-terminal two residues of the heavy chain are in dark green; and residues L46 and L55 are in peach.
quences” shows Arg-H94 to be present in a majority of mouse and human immunoglobulin heavy chains, including the four Fab antibody fragments whose three-dimensional structures are known from X-ray crystallography.13-16 In all the X-ray structures the N-terminus of the heavy chain comes to a contact distance of the Arg-H94 (Fig. 1).In all but one X-ray structure Arg-H94 adopts an identical conformation with its side chain hydrogen bonded to that of AspH101. Both the H94 and HlOl side chains are partially buried and directed away from the antigenbinding cavity (Fig. 1).Indeed, Arg-H94 has been considered to be a “framework” residue, i.e., outside of the hypervariable or complementarity-determining loops12 which comprise the antigen combining site. Thus, a direct impact of position H94, and the heavy chain N-terminus, on antigen binding was unexpected. Conformational invariance of the Arg-H94Asp-H101 motif led us to assume that a similar structural arrangement exists in the 40-150 mutant, which contains Arg-H94 and Asp-H101. We performed computations aimed a t verifying this assumption, as well as a t identifying the possible structural changes associated with the presence of Ser-H94 in the 40-150 heavy chain. The method of uniform conformational sampling, as implemented in the program CONGEN, has been described.” Briefly, the method consists of: (1)the generation of stereochemically acceptable conforma-
tions that avoid significant van der Waals overlaps with the rest of the structure (the chosen torsional degrees of freedom are systematically sampled using a predetermined angular grid). (2) The acceptable conformations are energy ranked, and the lowest energy conformations are chosen. In a n ideal case, the lowest energy conformation should correspond to the natural one. In practice, due to the approximate nature of the empirical energy potential employed, additional criteria also have to be considered in order to identify the correct conformation. For example, we previously considered solvent exposure of calculated conformations as a n aid in finding native-like structures of antibody combining site surface.l7 In this communication, where we deal with residues buried in protein core, criteria other than solvent accessibility have to be considered (see below). Although the conformational sampling method can in principle be applied to macromolecular systems of any size, computer time limitations currently restrict the sampling to small regions of space and short sections of polypeptide backbone. The strategy employed in our particular case was predicated by the backbone structure of that region, a p hairpin with residues H93 and H94 running antiparallel to the segment H101-Hl02, with the antiparallel P-sheet backbone hydrogen-bonding pattern between the two segments (cf. Fig. 2A). First, we constructed models of 40-150 antibody structure in the immediate vicinity of the position
95
COMPUTER ANALYSIS OF MUTATIONS
H94. Beginning with the crystallographic coordinates of a homologous mouse immunoglobulin, McPC 603,15 we replaced all the side chains within a contact distance (4 A) of the Arg-H94 with those that are present in the 40-150 sequence. Second, the conformational space of all the newly introduced side chains, as well as the backbones of residues H93 and H94, was explored with CONGEN. In order to ensure the maximal conformational freedom for the newly generated backbone, residues H95 through HlOO (i.e., the top of the hairpin loop above the sampled residues) were deleted from the system. As an independent test of our sampling protocol, the backbone and side chain torsions of the homologous residues in McPC 603 were sampled using identical conditions, and the results were compared with the crystallographic reality. Third, a similar set of CONGEN runs sampled backbone conformations of residues H101, H102, side chains of H93 and H94, and side chains within the 4 A contact distance of the Asp-H101 residue. These involved light chain residues L46 and L55, which in the sequence of the 40-150 hybridoma light chain are Arg and Asp, respectively. Once again, a n analogous run on McPC 603 gauged the predictive power of our sampling protocol. In the final set of runs, the backbones of residues H94 and HlOl were varied, together with their surrounding side chains. Typically, one such sampling run required in between 2 and 3 days of CPU time on pVAX 11.
RESULTS Results of CONGEN searches, presented in Table I and Figure 2, can be summarized as follows. (1)In all the runs, and particularly in those aimed at reproducing the McPC 603 crystallographic structure, the McPC 603 conformations were closely approximated by the vast majority of the lowest energy conformations (rms shifts from the X-ray structure in between 1.3 and 1.6 A, with backbone atom shifts typically not exceeding 0.5 A, cf. Fig. 2). The fact that test runs on the crystallographic structure of McPC 603 were able to reconstruct quite accurately the main features of this structure is encouraging. It indicates that our search protocols are likely to produce valid results in modeling the 40-150 environments as well. At the same time, constructions were also obtained that had large rms deviations from the McPC 603 crystallographic structure (over 5 A). These, however, were characterized by high potential energy values. (2) As a rule, variants of a single conformational motif dominated the lowest energies, and one or few of those were distinguished by particularly good hydrogen-bonding interactions among electrically charged side chains; such conformations are listed in Table I. Our selection of conformations with the best hydrogen bonding energy is justified by the fact that: (i) the polar side chains varied in our CON-
GEN runs are all buried; and that (ii) polar atoms buried in protein interior always satisfy their hydrogen-bonding potential." In particular, side chains containing formal electric charges usually form hydrogen-bonded clusters in protein interior^.'^,^^ In immunoglobulins, residues involved in such hydrogen-bonding networks belong among the most stringently conserved.21
DISCUSSION Data in the Table I indicate that the Ser + Arg replacement in the 40-150 mutant is not likely to produce a sizable polypeptide backbone rearrangement that would propagate into the parts of antigen binding site. However, large shifts (3-4 8)are seen in positions of side chain atoms of Asp-H101 and Arg-L46, depending on whether Arg or Ser is present in position H94. The serine side chain, due to its small size, is unable to make a hydrogen bond to the Asp-101 side chain and becomes hydrogenbonded to the backbone instead (see Fig. 2B). ArgH94, on the other hand, is hydrogen bonded to AspH101, joining a hydrogen-bonding net which involves the Asp residues HlOl and L55. The Arg residue L46 is central to this net, being hydrogen bonded to both the Asp-H101 and Asp-L55. In particular, the Arg-L46 side chain reorients in response to conformational changes experienced by the side chain of Asp-H101 in the presence of Arg-H94. AspL55 is a complementarity-determining residue and Arg-L46 is a n important domain-domain contacting residue located a t the bottom of the antigen-binding cavity. Rearrangements of these side chains, in response to the Arg or Ser in position H94, are likely to influence directly the shape of antigen-binding surface. Solvent accessibility computations" show that deletion of the two N-terminal residues, as it occurs in the 40-150 A2.4 P10 mutant, increases solvent accessibility of the Arg-H94 by loo%, thus increasing the likelihood that the charged atoms in the Arg side chain would be neutralized by solvation. When this occurs, Arg-H94 can no longer participate in the hydrogen-bonding net, resulting in a situation similar to that found in 40-150 where the residue in this position is Ser. Thus, the side chain shifts affecting the antibody-binding cavity that were postulated to occur in the variant would no longer be seen in the second-order variant.
CONCLUSION In summary, our computations suggest a plausible hypothesis of a detailed atomic mechanism whereby a side chain replacement, via a hydrogen bond "push-pull" mechanism, propagates over a distance of -10 A, and across protein domain interface, to generate a subtle change in the surface of the antigen combining site. The hypothesis is
96
J. NOVOTNY ET AL.
TABLE I. Summary of CONGEN Searches* Conformation number'. 603-37-145 40150-38-153 40150-39-186 603-34-437
40150-35-386
40150-36-410
Sampled residues, backbone + side chains Ala-H93 Arg-H94 Thr-H93 Arg-H94 Thr-H93 Ser-H94 Asp-H101 Val-H102
Asp-H101 Tyr-H102
Asp-HI01 Tyr-H1O2
Sampled residues, side chains only Leu-H4 Phe-H27 Phe-H32 Tyr-H33 Met-H34 Glu-H35 Asp-H101 Val-H102 Leu-H4 Phe-H27 Tyr-H32 Tyr-H33 Met-H34 Ala-H35 Asp-H101 Tyr-H102 Leu-H4 Phe-H27 Tyr-H32 Tyr-H33 Met-H34 Ala-H35 Asp-HI01 Tyr-HI02 Leu-H4 Phe-H27 Phe-H32 Tyr-H33 Met-H34 Glu-H35 Ala-H93 Arg-H94 Leu-L46 Glu-L55 Leu-H4 Phe-H27 Tyr-H32 Tyr-H33 Met-H34 Ala-H35 Thr-H93 Arg-H94 Arg-L46 Asp-L55 Leu-H4 Phe-H27 Tyr-H32 Tyr-H33 Met-H34 Ala-H35 Thr-H93 Ser-H94 Ara-L46 Asp-L55 Leu-H4 Phe-H27 Phe-H32 Tvr-H33 Met-H34 Glu-H35 Aia-H93 Val-H102 Leu-L46 Glu-L55 Leu-H4 Phe-H27 Tyr-H32 Tyr-€I33 Met-H34 Ala-H35 Thr-H93 Tyr-H102 Arg-L46 Asp-L55 Leu-H4 Phe-H27 Tyr-H32 Tyr-H33 Met-H34 Ala-H35 Thr-H93 Arg-H94 Arg-L46 Asp-L55
Number of conformations 308 306 360 507
578
594
Y
603-40-152
40150-41-94
40150-42-83
Conformation number' 603-37-145 40150-38-153 40150-39-186 603-34-437 40150-35-386 40150-36-410 603-40-152 40150-41-94 40150-42-83
Arg-H94 Asp-H101
Arg-H94 Asp-H101
Ser-H93 Asp-H101
In vacuo energy rankrng 2nd 5th 1st 3rd 4th 2nd 1st 13th 2nd
Total energy (kcal) - 125.8 -127.7 -100.6 -132.3 -171.1 -145.6 -146.2 -158.9 - 149.5
Hydrogen bond energy (kcal) -7.1 - 10.3 -8.7 -8.0 - 12.4 - 16.8 -8.2 -15.2 -13.1
181
181
181
rms shift to McPC 603
(A,
1.6 1.4 1.1 1.3 1.4 1.3 1.4 1.5 1.4
"Backbone torsion angles @ and J' ' were sampled using 30" grid maps; side chain torsions were sampled on the three lowest energy points (trans, i gauche) using the van der Waals avoidance option to circumvent close atomic c o n t a ~ t s . Conformations '~ were ranked according to potential energy computed by the CHARMM empirical potential function in V ~ C U O . ' ~Nonbonded interactions were computed within a cut-off distance of 8 A using the E = R electrostatic model as implemented in CONGEN. An explicit geometrydependent term was used to estimate the quality of hydrogen bonds (see ref. 23 for details.). ?The first number in the label, 603 or 40150, indicates whether the run sampled the torsional space of the McPC 603 structure or that of the 40-150 model. The second number is the run serial number, and the third number is the conformation label.
counterintuitive and could be derived only by computations. It is unexpected in that it does not involve backbone rearrangements of the hypervariable loops, and it proposes a n important role for framework, in addition to hypervariable, residues in generating a change in antibody specificity. The hypothesis is open to experimental verification via
site-directed mutagenesis of antibody-coding genes. The computational approach demonstrated here is directly applicable to analysis of single residue mutations of any protein. In principle, uniform conformational sampling should be superior to alternative methods such as interactive molecular graphics, which is subjective, or energy minimization, which
COMPUTER ANALYSIS OF MUTATIONS
ii
J b
x
97
& 5
71 C
Fig. 2. (a) Stereoscopic line diagram of polypeptide backbone and selected side chains in the vicinity of Arg-H94. The light lines trace the backbone of residues H93 through H103 (that is, the third heavy chain hypervariable loop) of the McPC 603 crystallographic structure together with side chains of Arg-H94 (left) and Asp-H101 (right). The heavy lines show positions of the backbone and side chains of residues H94 and HI01 in the computed conformation 603-37-145 (see Table I), aimed at reproducing the crystal conformation drawn in light lines. Although the computed side chain conformations deviate from the crystallographic ones, the hydrogen-bonding between the two side chains has been well reproduced, and backbone conformations match virtually exactly. Compare, for example, position of the free end-point of the computed conformation (C atom of H94, heavy circle) to those found in the X-ray structure (light circle). In this and the following figures, the two conformations compared were matched by least-squares
superposition of the Trp-H103 residue. Side chains are labeled by their one-letter name code and Kabat’’ numbering (H or L stands form the heavy and light chain, respectively). (b)Computed conformation in the 40-150 hybridoma. The light lines show the McPC 603 3rd heavy chain hypervariable loop for reference. The heavy lines trace the backbone and side chains of residues H101 and H I 0 2 in the run 40150-36-410 which analyzed the 40-150 model with Ser side chain in position H94. Two residues appearing to the right are the light chain Arg-L46 and Asp-L55 which participate in hydrogen-bonding with Asp-H101 and Arg-H94. (c) Comparison of computed conformations 401 50-35-386 (i.e., a 40-150 structure model with Arg at position H94, light lines) and 401 50-36-410 (i.e., a model with Ser at position H94, heavy lines). Note the shift in positions of the side chains Arg-L46 and Asp-L55 (on the right side of the picture). The rms difference of Asp-L55 0 6 1 atoms is 4.0 A, that of Arg-L46 Nr12 atoms is 4.3 A.
is known for its tendency to direct atoms into local potential energy wells, different from the global energy minima that distinguish the naturally occurring conformations.
ACKNOWLEDGMENTS This work was supported by NIH Grant HL19259-11 and ONR Research Grant 140-86-K-0116.
98
J. NOVOTNY ET AL.
REFERENCES 1 Panka, D. J., Mudgett-Hunter, M., Parks, D. R., Peterson, L.L., Herzenberg, L. A,, Haber, E., Margolies, M. N. Variable region framework differences result in decreased or increased affinity of variant anti-digoxin antibodies. Proc. Natl. Acad. Sci. U.S.A. 85:3080-3084, 1988. 2. Amit A. G., Mariuzza R. A,, Phillips S. E. V., Poljak R. J . Three-dimensional structure of an antigen-antibody complex a t 2.8 A resolution. Science 233:747-752, 1986. 3. Sheriff, S., Silverton, E. W., Padlan, E. A,, Cohen, G. H., Smith-Gill, S., Finzel, B. C., Davies, D. R. Three-dimensional structure of a n antibody-antigen complex. Proc. Natl. Acad. Sci. U.S.A. 84:8075-8079, 1987. 4. Novotny, J., Bruccoleri, R. E., Saul, F. A. On the attribution of binding energy in antigen-antibody complexes McPC 603, D1.3, and HyHEL-5. Biochemistry 28:47354749, 1989. 5. Rudikoff S., Satow S., Padlan E., Davies D. R., Potter, M. Kappa chain structure from a crystallized murine Fab: Role ofjoining segment in hapten binding. Mol. Immunol. 18:705-711, 1981. 6. Azuma T., Igras V., Reilly E. B., Eisen H.. Diversity a t the variable-joining region boundary of A light chains has a pronounced effect on immunoglobulin ligand-binding activity. Proc. Natl. Acad. Sci. U.S.A. 81:6139-6143, 1984. 7. Cook W. D., Rudikoff S., Giusti A. M., Scharff M. D. Somatic mutation in a cultured mouse myeloma cell affects antigen binding. Proc. Natl. Acad. Sci. U.S.A. 79:12401244, 1982. 8. Rudikoff S., Giusti A. M., Cook W. D., Scharff M. D. Single amino acid substitution altering antigen-binding specificity. Proc. Natl. Acad. Sci. U.S.A. 79:1979-1983, 1982. 9. Novotny J., Bruccoleri R. E., Newel1 J., Murphy D., Haber E., Karplus M. Molecular anatomy of the antibody binding site. J. Biol. Chem. 258:14433-14437, 1983. 10. Chothia C., Novotny J., Bruccoleri R.E., Karplus M. Domain association in immunoglobulin molecules; the packing of variable domains. J. Mol. Biol 160:325-342, 1985. 11. Bruccoleri R. E., Karplus M. Prediction of the folding of short polypeptide segments by uniform conformational sampling. Biopolymers 26:137-168, 1987. 12. Kabat E. A,, Wu T. T., Reid-Miller M., Perry H., Gottes-
man K.S. “Sequences of Proteins of Immunological Interest,” 4th ed. U. s. Department of Health and Human Services, N.I.H., Washington D.C., 1987. 13. Saul F. A., Amzel L. M., Poljak R. J . preliminary refinement and structural analysis of the Fab fragment from human immunoglobulin New a t 2.0 A resolution. J . Biol. Chem. 253:585-597, 1978. 14. Marquart, M., Deisenhofer J., Huber R., Palm, W. Crystallographic refinement and atomic models of the intact immunoglobulin molecule Kol and its antigen-binding fragment a t 3.0 A and 1.9 A resolution. J . Mol. Biol. 141: 369-391, 1980. 15. Satow Y., Cohen G. H. Padlan E. A,, Davies D. R. Phosphocholine binding immunoglobulin Fab McPC 603: An X-ray diffraction study at 2.7 A. J. Mol. Biol. 190593-604, 1986. 16. S. W. Suh, M. A. Navia, G. H. Cohen, D. N. Rao, S. Rudikoff, D. R. Davies. The galactan-binding immunoglobulin Fab 5539: An X-ray study a t 2.6 A resolution. Proteins 1:74-80, 1986. 17. Bruccoleri, R. E., Haber, E., Novotny, J . Structure of antibody hypervariable loops reproduced by a conformational search algorithm. Nature (London) 335564-568, and 336: 226, 1988. 18. C. Chothia, The nature of accessible and buried surfaces in proteins. J . Mol. Biol. 105:l-14, 1976. 19. Peeters D, Peeters J. A simple and novel interpretation of the three-dimensional structure of globular proteins based on quantum-mechanical computations on small model molecules. Biopolymers 24:491-508, 1985. 20. Rashin A. A,, Honig B. On the environment of ionizable groups in globular proteins. J. Mol. Biol. 173:515-521, 1984. 21. Novotny, J., Haber, E. Structural invariants of antigen binding: Comparison of immunoglobulin VL-VH and VLVL domain dimers. Proc. Natl. Acad. Sci. U.S.A. 8245924596, 1985. 22. Lee, B. K., Richards F. M. The interpretation of protein structures: Estimation of static accessibility. J. Mol. Biol. 55:379-400 1971. 23. Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States, D. J., Swaminathan, S., Karplus, M. CHARMM: A program for macromolecular energy, minimization and dynamics calculations. J. Comput. Chem. 4:187-217, 1983.
