`Val. 258, No. 23, Issue of December 10, pp. 14433-14437.1983
`Printed in U.S.A.
`
`Molecular Anatomy of the Antibody Binding Site*
`
`Jiti Novotny, Robert Bruccoleri, John Newell, David Murphy, Edgar Haber, and Martin Karplus
`From the Cellular and Molecular Research Laboratory and Cardiac Computer Center, Massachusetts General Hospital and
`Harvard Medical School. Boston, Massachusetts 021 14 and the Department of Chemistry, Harvard University,
`Cambridge, Massachusetts 02138
`
`(Received for publication, June 10,1983)
`
`The binding region of immunoglobulins, which in-
`cludes the portion of the molecule having the most
`variability in its amino acid sequence, is shown to have
`a surprisingly constant structure that can be charac-
`terized in terms of a simple, well-defined model. The
`binding region is composed of the antigen combining
`site plus its immediate vicinity and arises by noncova-
`lent association of the light and heavy chain variable
`domains (VL and VH, respectively). The antigen com-
`bining site itself consists of six polypeptide chain seg-
`ments (“hypervariable loops”) which comprise some 80
`amino acid residues and are attached to a framework
`of VL and VH @-sheet bilayers. Having analyzed re-
`fined x-ray crystallographic coordinates for three an-
`tigen-binding fragments (Fab KOL (Marquart, M.,
`Deisenhofer, J., and Huber, R. (1980) J. Mol. Biol.
`141, 369-391), MCPC 603 (Segal, D., Padlan, E. A.,
`Cohen, G . H., Rudikoff, S., Potter, M., and Davies, D.
`R. (1974) Proc. Natl. Acad. Sci. U. S. A. 71, 4298-
`4302), and NEW (Saul, F. A., Amzel, L. M., and Poljak,
`R. J. (1978) J. Biol. Chem. 253,585-597)) we use the
`results to introduce a general model for the VL-VH
`interface forming the binding region. The region con-
`sists of two closely packed &sheets, and its geometry
`corresponds to a 9-stranded, cylindrical barrel of av-
`erage radius 0.84 nm with an average angle of -63“
`between its two constituent &sheets. The barrel forms
`the bottom and sides of the antigen combining site. The
`model demonstrates that the structural variability of
`the binding region is considerably less than was
`thought previously. Amino acid residues which are
`part of the domain-domain interface and appear not to
`be accessible to solvent or antigen contribute to anti-
`body specificity.
`
`Immunoglobulins have a modular structure in that both the
`light chains ( M I = 25,000) and the heavy chains (Mr = 50,000)
`(4, 5) are composed of several domains ( 6 4 each of which
`consists of approximately 110 amino acid residues. The NH2-
`terminal variable domains (9, 10) of both chains (VL’ and
`VH) are formed from two layers of @-sheet (1-3, 11-13) and
`have been shown to fold independently of the rest of the
`polypeptide chain (14). Six hypervariable loops of the VL and
`VH domains (Ll, L2, L3, H1, H2, and H3) were predicted by
`
`* The work carried out in the Department of Chemistry was sup-
`ported in part by a grant from the National Institutes of Health. 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 accordance with 18 U.S.C. Section 1734
`solely to indicate this fact.
`’ The abbreviations used are: VL, light chain variable domain; VH,
`heavy chain variable domain.
`
`Wu and Kabat (15,16) to have a role in antigen binding. The
`loops are relatively short segments of primary structure that
`occur in homologous positions in all light and heavy chains,
`but have amino acid sequences and lengths which differ from
`one antibody molecule to another. The x-ray structure shows
`that each of the hypervariable loops forms a connection
`between two antiparallel strands of 8-sheet. By noncovalent
`association of the VL and VH domains, all six loops come
`into close contact and form a contiguous area on the surface
`of the VL-VH dimer from which the binding site
`is con-
`structed (Fig. 1) (2, 11, 17-20).
`To obtain a consistent set of refined atomic coordinates
`from the crystallographic data (1-3, 29), we discarded atoms
`which form the constant domains of the Fab fragment and
`energy-minimized the resulting Fv fragments ( i e . VL-VH
`domain dimers, the domains being defined as in Ref. 16) with
`CHARMM, an empirical energy function program designed
`for the study of macromolecules (30). The root-mean-square
`differences between the original and energy-minimized coor-
`dinates were 0.031 nm (0.31 A) for MCPC 603 and 0.021 nm
`for NEW; such small shifts indicate that the original crystal-
`lographic structures were satisfactory and that acceptable
`values for the potential energy can be achieved by small
`adjustments of the coordinates.
`As a first step in deriving the geometric characteristics of
`the binding region, the boundaries between the parts of the
`VL-VH interface forming 8-sheet strands and hypervariable
`loops were determined. The &sheets were defined by their
`interstrand backbone ((24. . . H-N)
`hydrogen-bonding
`pattern. A hydrogen bond list was generated with CHARMM
`(30) for the four polypeptide chain segments which contribute
`to the VL-VH interface (Fig. 2 ) , and amino acids with hydro-
`gen bond energy of -4.2 kJ/mol or less were taken to be parts
`of the @-sheets. By this criterion, portions of the hypervariable
`loops (as defined in Ref. 16) form part of @-sheets (Fig. 2).
`Examination of the four @-sheet segments with adjacent hy-
`pervariable loops by computer graphics showed that they
`consist of nine antiparallel @-strands; the light chain @-sheet
`is 4-stranded, and the heavy chain P-sheet is 5-stranded. The
`atom-packing density (32) of the sheet-sheet interface is on
`the order of that found in interiors of other proteins (33, 34).
`The sheets are strongly curved and wrap counterclockwise
`around a nearly perfect cylindrical surface (Fig. 3). Although
`@-barrels are frequently found among protein structures (35),
`it is unusual to find one formed of sheets contributed by
`different polypeptide chains. Richardson (35) has noted that
`“back-to-back @-barrels that share one wall occur in the
`variable half of immunoglobulin Fab structures.”
`To provide a quantitative assessment of the geometry of
`the binding region, a cylindrical surface was least-squares-
`fitted to the three VL-VH interface barrels (see Fig. 4 for
`details). Radii of the least-squares-fitted cylinders are very
`
`14433
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1061 Page 1 of 5
`
`
`
`14434
`
`Antibody Binding Site
`
`FIG. 1. Antibody binding region. A stereo a-carbon drawing of VL-VH domain dimer of mouse Fab fragment
`MCPC 603 (2). VL domain is at right; VH domain is at left. The interface @-sheets, represented by heavy lines
`(segments S1, S2, S3, and S4 of Fig. 2), are highly curved and their noncovalent contact forms a nearly perfect
`cylindrical barrel the upper surface of which constitutes the antigen combining site. Ca atoms of hypervariable
`residues are represented as heavy dots. The nine @-strands which form the VL-VH interface are numbered 01
`through p9, in correspondence with Fig. 2.
`
`b c 53 5 1
`,60 65
`45 ,50 52a
`35
`30
`,40
`F S S Y A M Y W V R Q A P G K G L E W V A I I W D D G S - - D Q H Y A D S V K G R
` R
`F D D Y Y S T W V R Q P P G R G L E W 1 G Y V F Y H G T S D T D T P L R S " -
`F S D F Y M E W V R Q P P G K R L E W I A A S R N K G N K Y T T B Y S A S V K G R
`I-
`I
`t a l i
`87
`H 2
`
`85
`
`86
`
`105
`9 0 9 5
`l O O a b c d e f g h 101
`G V Y F C A R D G G B G F C S S A S C P G P D Y W G Q G T P V
`
`T
`
`~
`
`88
`
`I
`
`H3
`
`I
`
`89
`
`FIG. 2. Polypeptide-chain segments which constitute the binding region and the VL-VH interface.
`Homologous segments from different structures are aligned to maximize amino acid homology. Boldface letters
`indicate hypervariable loops (as defined in Ref. 16). Boxes denote @-sheet strands as defined in this paper (see
`text). Hypervariable loops as redefined in this paper are denoted by horizontal burs. Segments S1 and S2 are parts
`of light chain variable domain; segments S3 and S4 are parts of heavy chain variable domain. The numbering
`system is that of Ref. 16. Amino acids are given in the one-letter code (31).
`
`6 0
`
`
`
`
`
`
`
`27a 5 5
`K O L S 1
`NEW S 1
`
`25
`
`,
`
`3 0
`
`b c
`
`
`
`2 8
`
`I
`
`L1
`
`83
`
`I-
`
`,
`
`81
`
`I-
`
`82
`
`L2 -1
`
`H N N A R
`
`I
`
`95 97a b 90 100
`
`L3 - I
`
`84
`
`K O L S2
`NEW S2
`MCPC603 S2
`
`K O L S 3
`NEW S3
`M C P C 6 0 3 S 3
`
`K O L S 4
`NEW S4
`MCPC603 S4
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1061 Page 2 of 5
`
`
`
`Binding
`
`
`
`Antibody
`
`.
`
`I
`
` I
`
`'
`
`
`
`FIG. 3. Topology of the binding site barrel. Four hairpin
`segments, S1 to S4, primary structures of which are given in Fig. 2,
`organize themselves along a cylindrical surface. Segments S1 and S2
`form the light chain 0-sheet (on the right, counterclockwise, p2, pl,
`p3, and p4) as well as the three light chain hypervariable loops L1,
`L2 and L3. Segments S4 and S3 form the heavy chain 0-sheet, consist
`of five @-strands (on the left, clockwise, p9, p8, p5, p6, and p7), and
`the three heavy chain hypervariable loops H1, H2, and H3. Geneti-
`cally, the strand 64 corresponds to the light chain J gene (21,22); 09
`corresponds to the heavy chain J gene (23-26); and the H3 loop is
`coded for by the D gene segment (27, 28). The shapes of the hyper-
`variable loops are indicated schematically, and only their most general
`features are given. In all three structures studied, the domain-domain
`interface is closely packed, leaving no free space between the two p-
`sheets.
`
`similar (0.82 nm for KOL, 0.86 nm for MCPC 603, and 0.84
`nm for NEW), and the root-mean-square deviation of back-
`bone atoms from the fitted surface (0.25 nm for KOL, 0.22
`nm for MCPC 603, and 0.21 nm for NEW) is small considering
`that the fit includes parts of the hypervariable segments. For
`all three binding-site cylinders, we also computed angles
`between the backbone atoms of adjacent 0-strands; the back-
`bone atoms were represented by straight, least-squares-fitted
`lines (axes of inertia). The average twist angle between adja-
`
`cent strands in the same 0-sheet is -20", a value correspond-
`ing to that found for sheets in other proteins (36-38). If the
`binding site barrels were perfect cylinders, strands which form
`sheet-sheet junctions would be expected to subtend the same
`angle (i.e. -20"); the angles are, in fact, significantly larger
`(the @4-@7 angle is -57" and the @2-/39 angle is 33").
`Interacting @-sheets of a single domain normally pack with
`the relative orientation of the sheet axes equal to either -30"
`or -90" (36-38); the sheet-sheet angle of the VL-VH interface
`by contrast is -52" in KOL, -54" in MCPC 603, and -52" in
`NEW.
`The KOL, MCPC 603, and NEW interface cylinders closely
`resemble each other
`(Fig. 4), and their interface @-sheet
`segments which are adjacent to hypervariable
`loops have
`virtually identical primary structures. The same, or very sim-
`ilar, amino acid sequences occur in corresponding segments
`of all immunoglobulins (16). This suggests that the geometry
`of the VL-VH interface is an important feature of the anti-
`body binding region and that structurally similar 9-stranded
`
`cylinders form the essential element of all the binding regions;
`the antigen combining site is the concave part of the "upper"
`surface of the cylinder (Fig. 3).
`To ascertain the role which hypervariable residues play in
`domain-domain association
`(17, 39, 40), we computed the
`total contact surface between
`VL and VH domains. The
`
`Site
`14435
`interdomain contact surfaces were obtained as differences
`between total solvent-accessible surface (41, 42) of isolated
`domains and of the domains involved in the dimer. The VL-
`VH contact surfaces, obtained as sums of the individual VL
`and VH interfaces, are
`17.56 nm2 in KOL, 19.39 nm2 in
`MCPC, and 603 and 17.09 nm' in NEW. Thus, the average
`interface area of a single variable domain is 9 nm' (900 A'),
`36% of which consists of polar residues. The values are on
`the order of those reported for other protein-protein interfaces
`(e.g., CY@ oxyhemoglobin dimer) (34). The hypervariable loops
`contribute about half of the total contact
`surface (47% in
`KOL, 53% in MCPC 603, and 42% in NEW).
`Single amino acid exchanges in hypervariable positions at
`the bottom of the binding site and inaccessible to solvent or
`antigen have a dramatic effect on antibody specificity. Partic-
`ularly, side chains of residues Leu L96, Glu H35, and Asp
`H101, known to be indispensable for the specificity of phos-
`phorylcholine-binding myelomas (43-45), are either totally or
`partially buried in the VL-VH interface. Other hypervariable
`residues (COOH-terminal parts of L1 and H1, NH,-terminal
`parts of J gene-coded segments, i.e. L3 and H3) also form part
`of the VL-VH contact
`surface. Thus the domain-domain
`contact surface, even though it is not accessible in the free
`
`antibody structure, is directly involved in antibody specificity.
`This suggests that, due to the close packing at the interface,
`changes in volumes of buried side chains may induce changes
`in the bottom surface of the binding site. Such a result would
`be consistent with the structural variability found in hemo-
`globin (46) and the core of immunoglobulin domains (47).
`Alternatively, structural changes induced by antigen binding
`(48-58) could expose some of these buried residues.
`Visual inspection of the Fv fragments (Figs. 1 and 3) shows
`that some of the 80 amino acid residues which constitute the
`hypervariable loops are distant from the binding site surface
`and may not be directly involved in antigen binding. Others,
`as indicated in Fig. 2, participate in the well-defined p-sheet
`secondary structures, so that their conformational freedom is
`rather limited. Those which are distant from the site include
`the whole L2 loop (it has been observed in the past that L2
`is not directly involved in antigen binding (2, 59, 60)), two
`NH,-terminal residues of L1, and five COOH-terminal resi-
`dues of H2. The NH, and COOH termini of L1, L3, H3, and
`the middle of H2 are involved in backbone-backbone hydrogen
`bonds and have an extended conformation
`with backbone
`torsion angles (9 and q ) in the @-sheet range. Of the 80 amino
`acids in the hypervariable loops (16), only 47 are intimately
`involved in the binding site, having at the same time confor-
`mations which vary in such a way that they cannot be inferred
`by analogy to known structures. This greatly reduces the
`problem of determining the binding site geometry of immu-
`noglobulins with unknown structure, although, due to varia-
`tions in the lengths of hypervariable loops, the exact number
`of structurally variable positions is expected to vary somewhat
`in different molecules.
`of a
`The model proposed here represents the first part
`project aimed at developing a method for the analysis of the
`antigen binding region including prediction of its structure
`and evaluation of the interactions between the antibodies and
`the antigen.
`
`thank Drs. F. Saul, R. Poljak (Pasteur
`Acknowledgments-We
`Institute, Paris), E. Padlan (Johns Hopkins University, Baltimore),
`D. Davies (National Institutes of Health, Bethesda), M. Marquart,
`R. Huber (Max-Planck Institut, Munich), and E. Abola (Brookhaven
`Data Bank) for crystallographic coordinates, Drs. F. M. Richards and
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1061 Page 3 of 5
`
`
`
`14436
`
`Site
`
`
`
`Binding
`
`
`
`Antibody
`
`n U
`
`n
`Y-
`
`n
`0
`
`n n
`
`n
`0
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1061 Page 4 of 5
`
`
`
`Site
`
`
`
`
`
`Antibody B linding
`
`14437
`
`J. Matthews (Yale University) for computer programs, and R. Ladner
`and J. Ladner (Harvard University) for computer graphics.
`
`(1981) J .
`
`Brice, M. D., Rodgers, J. R., Kennard, O., Shimanouchi, T.,
`and Tasumi, M. (1977) J. Mol. Biol. 112, 535-542
`30. Brooks, B., Bruccoleri, R. E., Olafson, B. D., States, D. J.,
`Swaminathan, S., and Karplus, M. (1983) J. Comput. Chem. 4,
`187-217
`31. IUPAC-IUB Tentative Rules (1968) Eur. J. Biochem. 5,151-153
`32. Richards, F. M. (1974) J. Mol. Biol. 82, 1-14
`33. Chothia, C. (1974) Nature ( L o n d . ) 248, 338-339
`34. Chothia, C., and Janin, J. (1975) Nature (Lond.) 256, 705-708
`35. Richardson, J. S. (1981) Ado. Protein Chem. 34, 167-339
`36. Chothia, C., Levitt, M., and Richardson, D. (1977) Proc. Natl.
`Acad. Sci. U. S. A. 74,4130-4134
`37. Chothia, C., and Janin, J. (1981) Proc. Natl. Acad. Sci. U. S. A.
`78,4146-4150
`38. Cohen, F. E., Sternberg, M. J. E., and Taylor, W. R.
`Mol. Biol. 148, 253-272
`39. Stevens, F. J., Westholm, F. A., Solomon, A., and Schiffer, M.
`(1980) Proc. Natl. Acad. Sci. U. S. A. 77, 1144-1148
`40. Horne, C., Klein, M., Polidoulis, I., and Dorrington, K. J. (1982)
`J. Immunol. 129,660-664
`41. Lee, B., and Richards, F. M. (1971) J. Mol. Biol. 55,379-400
`42. Richmond, T. J., and Richards, F. M. (1978) J. Mol. Biol. 119,
`537-555
`43. Rudikoff, S., Satow, Y., Padlan, E., Davies, D., and Potter, M.
`(1981) Mol. Immunol. 18, 705-711
`44. Cook, W. D., Rudikoff, S., Giusti, A. M., and Scharff, M. D.
`(1982) Proc. Natl. Acad. Sci. U. S. A. 79, 1240-1244
`45. Rudikoff, S., Giusti, A. M., Cook, W. D., and Scharff, M. D.
`(1982) Proc. Natl. Acad. Sci. U. S. A. 7 9 , 1979-1983
`46. Lesk, M. A,, and Chothia, C. (1980) J. Mol. Biol. 136, 225-270
`47. Lesk, M. A., and Chothia, C. (1982) J. Mol. Biol. 160,325-342
`48. Roberts, S., and Grabar, P. (1957) Ann. Inst. Pasteur (Paris) 9 2 ,
`56-61
`49. Warner, C., and Shumaker, V. N. (1970) Biochemistry 9 , 451-
`458
`50. Holowka, D. A., Strosberg, A. D., Kimball, J. W., Haber, E., and
`Cathou, R. E. (1972) Proc. Natl. Acad. Sci. U. S. A. 69, 3399-
`3403
`51. Liberti, P. A,, Stylos, W. A., Maurer, P. H., and Callahan, H. J.
`(1972) Biochemistry 11, 3321-3326
`52. Tumerman, L. A., Nezlin, R. S., and Zagyansky, Y. A. (1972)
`FEES (Fed. Eur. Biochem. Soc.) Lett. 19,290-292
`53. Pilz, I., Kratky, O., Licht, A., and Sela, M. (1973) Biochemistry
`12,4998-5005
`54. Levison, S. A., Hicks, A. N., Portman, A. J., and Dandliker, W.
`B. (1975) Biochemistry 14,3778-3786
`55. Schlessinger, J., Steinberg, I. Z., Givol, D., Hochman, J., and
`Pecht, I. (1975) Proc. Natl. Acad. Sci. U. S. A. 72, 2775-2779
`56. Kaivarainen, A. I., and Nezlin, R. S. (1976) Biochem. Biophys.
`Res. Commun. 68, 270-274
`57. Lancet, D., and Pecht, I. (1976) Proc. Natl. Acad. Sci. U. S. A.
`73,3549-3553
`58. Cser, L., Franek, F., Gladkikh, I. A., Nezlin, R. S., Novotny, J.,
`and Ostanevich, Y. M. (1977) FEES (Fed. Eur. Biochem. Soc.)
`Lett. 80,329-331
`59. Margolies, M. N., Cannon, L. E., Strosberg, A. D., and Haber, E.
`(1975) Proc. Natl. Acad. Sci. U. S. A. 72, 2180-2184
`60. Novotny, J., and Margolies, M. N. (1983) Biochemistry 22,1153-
`1158
`
`I
`
`
`
`’
`
`-
`
`REFERENCES
`1. Marauart. M.. Deisenhofer, J., and Huber. R. (1980) J. Mol. Biol.
`1 4 i , 369-391
`2. Segal, D., Padlan, E. A., Cohen, G. H., Rudikoff, S., Potter, M.,
`and Davies, D. R. (1974) Proc. Natl. Acad. Sci. U. S. A. 71,
`4298-4302
`3. Saul. F. A,. Amzel. L. M.. and Poliak. R. J. (1978) J. Biol. Chem.
`253,585-597
`4. Porter, R. R. (1973) Science (Wash. D. C.) 180, 713-716
`5. Edelman, G. M. (1973) Science (Wash. D. C.) 180,830-840
`6. Singer, S. J., and Doolittle, R. F. (1966) Science (Wash. D. C.)
`153, 13-25
`7. Hill, R. L., Delaney, R., Fellows, R. E., and Lebovitz, H. E. (1966)
`Proc. Natl. Acad. Sci. U. S. A. 5 6 , 1762-1769
`8. Edelman, G. M. (1970) Biochemistry 9,3197-3205
`9. Hilschmann, N., and Craig, L. C. (1965) Proc. Natl. Acad. Sci. U.
`S. A. 53, 1403-1409
`10. Titani, K., Whitley, E., Avogadro, L., and Putnam, F. M. (1965)
`Science (Wash. D. C.) 149, 713-716
`11. Amzel, L. M., Poljak, R. J., Saul, F., Varga, J. M., and Richards,
`F. F. (1974) Proc. Natl. h a d . Sci. U. S. A. 7 1 , 1427-1430
`12. Schiffer, M., Girling, R. L., Ely, K. R., and Edmundson, A. B.
`(1973) Biochemistry 12, 1620-1631
`13. Epp, O., Colman, P., Fehlhammer, H., Bode, W., Schiffer, M.,
`and Huber, R. (1974) Eur. J. Biochem. 45, 513-524
`14. Hochman, J., Inbar, D., and Givol, D. (1973) Biochemistry 12,
`1130-1135
`15. Wu, T. T., and Kabat, E. A. (1970) J. Exp. Med. 132, 211-250
`16. Kabat, E. A., Wu, T. T., and Bilofsky, H. (1983) Sequences of
`Proteins of Immunological Interest, United States Public Health
`Service, National Institutes of Health, Bethesda
`17. Davies, D. R., Padlan, E. A., and Segal, D. (1975) Annu. Reu.
`Biochem. 44,639-667
`18. Padlan, E. A., Davies, D. R., Rudikoff, S., and Potter, M. (1976)
`Immunochemistry 13,945-949
`19. Padlan, E. A. (1977) Quart. Rev. Biophys. 10.35-65
`20. Amzel, L. M., and Poljak, R. J. (1979) Annu. Reu. Biochem. 48,
`961-997
`21. Sakano, H., Hiippi, K., Heinrich, G., and Tonegawa, S. (1979)
`Nature (Lond.) 2 8 0 , 288-294
`22. Max, E. E., Seidman, J. G., and Leder, P. (1979) Proc. Natl. Acad.
`Sci. U. S. A. 76,3450-3454
`23. Bernard, O., and Gough, N. M. (1980) Proc. Natl. Acad. Sci. U.
`S. A. 77,3630-3634
`24. Early, P., Huang, H., Davis, M., Calame, K., and Hood, L. (1980)
`Cell 19,981-992
`25. Newell, N., Richards, J. E., Tucker, P. W., and Blattner, F. R.
`(1980) Science 209, 1128-1132
`26. Sakano, H., Maki, R., Kurosawa, Y., Roeder, W., and Tonegawa,
`S. (1980) Nature (Lond.) 286,676-683
`27. Sakano, H., Kurosawa, Y., Weigert, M., and Tonegawa, S. (1981)
`Nature ( L o n d . ) 290,562-565
`28. Siebenlist, U., Ravetch, J. V., Korsmeyer, S., Waldmann, T., and
`Leder, P. (1981) Nature (Lond.) 294,631-635
`29. Bernstein, F. C., Koetzle, T. F., Williams, G. J. B., Meyer, E. F.,
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1061 Page 5 of 5