Three-Dimensional Structure of an Antigen­
`
`
`Antibody Complex at 2.8 A Resolution
`A. G. AMIT, R. A. MARIUZZA, S. E. v. PHI�LIPS, R. J. POLJAK
`
`antibody combining sites is based on x-ray diffraction studies of
`The 2.8 A resolution three-dimensional
`structure of a
`myeloma immunoglobulins as reviewed (2). These have shown that
`complex between an antigen (lysozyme)
`and the Fab
`the conformation of combining sites is determined by the amino
`acid sequences, unique to each different antibody, of the CDR's.
`
`fragment from a monoclonal antibody against lysozyme
`
`has been determined and refined by x-ray crystallographic
`The strucrures of two complexes of antigen-binding fragments
`changes can be observed
`(Fab) of mycloma immunoglobulins with small ligands have also
`
`techniques. No conformational
`in the tertiary structure of lysozyme compared with that
`been determined (3, 4). Although these $tudies resulted in useful
`in native crystalline forms. The quaternary
`models for ligand-antibody interactions, they are insufficient to
`determined
`structure of Fab is that of an extended conformation.
`The
`establish unequivocally the precise size and shape of antibody
`antibody combining site is a rather flat surface with
`combining sites, the nature and extent of antigen-antibody interac­
`tions, and the occurrence of possible conformational changes (if
`protuberances and depress ions formed by its amino acid
`interface is tightly
`
`side chains. The antigen-antibody
`any) in the antibody after antigen binding. In addition, the precise
`packe� with 16 lysozyme and 17 antibody residues
`structure of antigenic determinants on protein molecules remains to
`be determined (5). Equally imponant are questions concerning the
`
`making close contacts. The antigen contacting residues
`belong to two stretches of the lysozyme polypeptide
`nature of possible conformational changes in the complexed antigen
`
`chain: residues 18 to 27 and 116 to 129. All the comple­
`and the effect of single amino acid substitutions on antigenic
`mentarity-determining regions
`
`and two residues outside
`specificity and antigen recognition by the antibody.
`ble positions of the antibody ma.lee contact with
`hypervaria
`We have recently determined the three-dimensional structure of
`Most of these contacts
`the antigen.
`(10 residues out of 17)
`an antigen-antibody complex, one between lysozyme and the Fab
`
`
`are made by the heavy chain, and in particular by its third
`fragment of a monoclonal antibody to hen egg white lysozyme, at 6
`A resolution (6). We have since extended the resolution of the x-ray
`
`complementarity-determining region. Antigen variability
`and antibody specificity and affini ty are discussed on the
`strucrure detcnn.ination to 2.8 A, and now present a complete
`basis of the determined
`structure.
`description of antigen-antibody interactions in the complex.
`Structure determination. The production of hybrid cell lines
`secreting murinc monoclonal antibody to hen egg white lysozyme,
`and the purification, crystallization (7), and 6 A resolution crystal
`structure determination (6) of the complex between Fab Dl.3 and
`lysozymc have been described. Crystals grown from solutions
`containing 15 to 20 percent polyethylene glycol 8000 at pH 6.0 arc
`monoclinic, space group P2., with a.= 55.6, b = 143.4, c = 49.1
`A,� = 120.5°, and one molecule of complex per asymmetric unit.
`Three heavy atom isomorphous derivatives were prepared with
`(NH.)2PtCI., K3FsU02, and p-hydroxymercuribenzcnesulfonate.
`X-ray intensities were measured to 2.8 A resolution with the use of a
`four-circle automatic diffractometer. Heavy atom sites were refined
`in alternate cycles of phasing and refinement (8); isomorphous
`phases, including anomalous scattering contributions (9), were
`calculated. The mean figure of merit (10) to 2.8 A resolution was
`0.47 for 15592 reflections. The electron density map calculated
`from these data was not readily interpretable, preswnably because of
`lack of isomorphism of the heavy atom derivatives affecting phase
`determination at high resolution. The phases were further refined by
`a density modification technique (11) with a molecular envelope
`traced from the Fab-lysozyme model determined at 6 A resolution
`( 6). The resulting phases depend only on the observed data and the
`overall shape and position of the complex, but are independent of
`the detailed conformation of the previous model (6). The resulting
`
`HE BINDING OF FOREIGN ANTIGENS TO COMPLEMENTARY
`
`T
`
`strucrures on the surface of B and T lymphocytes represents
`the initial step in the sequence of events leading to activation
`of the immune system. The receptor molecule on the surface of B
`lymphocytes responsible for antigen recognition
`is membrane
`immunoglobulin. A mature B cell produces and inserts into its
`plasma membrane only limited amounts of a single kind of immuno­
`globulin. Contact with antigen results in the expansion of B cell
`clones specific for that antigen and in their differentiation into
`plasma cells capable of producing and secreting large amounts of
`antibody of the same specificity (monoclonal antibody).
`Antibody molecules of the immunoglobulin G (lgG) class, the
`most abundant in normal serum, are composed of two identical light
`(L) and two identical heavy (H) polypeptide chains. The amino
`terminal regions of the H and L chains, termed V H and V L, are each
`about 110 amino acids long and have variable (and homologous)
`amino acid sequences. The constant (C) half of the L chain, CL, and
`the constant regions CH 1, CH2, and CH3 of the H chain, each about
`100 amino acids long, have homologous sequences that belong to
`one of a few classes (Kand>. for L chains; µ., 8, -y, e, and a for H
`chains). The V H and V L regions each contain three hypervariablc or
`complementarity-determining regions (CDRl, CDR2, and CDR3)
`responsible for antigen recognition. These arc flanked by less
`variable (FRI, FR2, FR3, and FR4) "framework" regions (J).
`Present understanding of the three-dimensional structure of
`
`IS AUGUST 1986
`
`A.G. Amit, R. A. Mariuz.za, and R. J. Poljak are in the DCpartcmcnt
`d'lnunwiologic,
`lnstitut Pasteur, 75724 Paris Ccdcx 15, France. S. E. V. Phillips is in the AstbUry
`of Leeds, Leeds, United Kingdom.
`Dcpanmcnt of Biophysics, University
`
`RESEARCH ARTICLES 74-7
`
`1 of 7
`
`BI Exhibit 1059
`
`

`

`Fig. 1. Stereo diagram of the Ca skeleton of the
`complex. Fab is shown (upper right) with the heavy
`and light chains with thick and thin bonds, respec­
`U\'cly. The lysozyme active site is the cleft containing
`the label HEL. Antibody-antigen interactions are
`most numerous between lysozyme and the heavy
`chain CDR loops.
`
`C81
`
`CIU
`
`electron density map was much improved, and an atomic model was
`fitted to it on an Evans and Sutherland PS300 interactive graphics
`system with the use of the program FRODO (12). The amino acid
`sequence ofFab D 1.3 was derived from the corresponding light and
`heavy chain complementary DNA (cDNA) sequences (13). Of the
`562 amino acid residues in the complex, 24 of those in the constant
`regions could not be located in the initial map. The atomic
`coordinates were submitted to alternate cycles of restrained crystal­
`lographic least-squares refinement (14) and model building. The
`model was checked in the later stages of refinement by sequentially
`omitting segments of the polypeptide chain (up to 20 percent of the
`total) and rebuilding them in maps phased from the remainder of
`the structure in combination with isomorphous replacement data
`(15). All residues have now been located, and the current crvstallo­
`graphic R factor is 0.28 for all data in the 20 to 2.8 A res�lution
`range. (R =I I 1Fol-1Fcl I I IIFol, where Fo, Fe are the observed
`and calculated structure factors of x-ray reflections.) No attempt was
`made to locate solvent molecules. Two isotropic temperature factors
`were used for each residue, one for the main chain atoms, and
`another for the side chain atoms. Stcrcochemical restraints were
`adjusted to give a standard deviation in C-C bonds of ±0.03 A. No
`restraints were applied between residues across the antibody-antigen
`interface. Atomic coordinates will be deposited at Brookhaven Protein
`Data Bank after higher resolution and crystallographic refinement.
`Conformation of the complexed antigen and of the Fab. The
`overall structure of the complex at 2.8 A resolution (Fig. 1) confirms
`the results of the 6 A resolution study ( 6). The assignment of the H
`and L polypeptide chains ofFab is unchanged. The closely packed 13
`sheets are seen in Fab as arc the helical and 13-sheer struetures
`surrounding the active site in lysozyme. The Fab appears in an
`almost fully extended conformation, with a definite separation
`between the variable (V) and constant (C) domains. With the
`exception of this difference in quaternary structure, Fab D 1.3
`compares closely to other known Fab's (4, 16), except in the CDR
`loops. Predicted structures for Dl.3 (17) based on other Fab's also
`agree well with the determined structure in the framework 13-sheer
`regions and in some, but not all, of the CD R loops. The relative
`disposition of the variable subunits of the H chain (V H) and of the L
`chain (V L), is unaltered, indicating no change in quaternary struc­
`ture in the V domain resulting from antigen binding. Since the
`crystal structure of the unligandcd Fab Dl.3 has not been deter­
`mined, detailed changes in antibody conformation remain to be
`verified. However, the similarity with other Fab struetures suggests
`
`that possible conformational changes would be small. This observa­
`tion is in agreement with that made by nuclear magnetic resonance
`(NMR) on the unligandcd and hapten-liganded (dinitrophenol)
`mouse myeloma protein MOPC315 (18).
`A least-squares fit of Ca atoms of lysozyme in the complex and
`native lysozyme refined at 1.6 A in its tetragonal crystal form (19) gives
`a root-mean-square (rrns) deviation of 0.64 A between the two (see
`Fig. 2). Since the error in atomic positions in the complex can
`be estimated (20) to be approximately 0.6 A, the difference is not
`significant. Furthermore, the largest changes (up to 1.6 A) occur in
`regions remote from antibody contacts. Similar comparisons of
`native tetragonal lysozyme with other crystal forms gave rms
`deviations of 0.88 A with triclinic lysozyme refined from x-ray and
`neutron diffraction data (21) and 0.46 A for orthorhombic lysozyme
`determined at physiological temperature (22). Some differences in
`side chain conformation are observed between tetragonal and
`complexed lysozyme, but close examination with computer graphics
`revealed these to be similar to differences observed between different
`crystal structures of native lysozyme. Thus, complex formation with
`antibody Dl.3 produces no more distottion of the structure of
`lysozyme than does crystallization.
`The antigen-antibody interface. The interface between antigen
`and antibody extends over a large area with maximum dimensions of
`about 30 by 20 A (Figs. 3 and 4). The antibody combining site
`appears as an irregular, rather fiat surface with protuberances and
`depressions formed by the amino acid side chains of the CD R's of
`V H and V L· In addition, there is a small cleft between the third
`CDR's ofVH and VL, corresponding to the binding site character­
`i.z.ed in hapten-antibody complexes (3, 4). The cleft accepts the side
`chain Gin 121 of lysozyme although this is not the center of the
`antigen-antibody interface (Fig. 3).
`The lysozyme antigenic determinants recogni.z.ed by Dl.3 are
`made up of two stretches of polypeptide chain, comprising residues
`18 to 27 and 116 to 129, distant in the amino acid sequence but
`adjacent on the protein surface. All six CDR's interact with the
`antigen and in all, 16 antigen residues make close contacts with 17
`antibody residues (Tables 1 and 2). Two antibody contacting
`residues, V L Tyr 49 and V H Thr 30, are just outside segments
`commonly defined as CDR's [sequence numbers are as in Kabat et
`aJ. (1) except for VH CDR3; see Tables 2 and 3). VH Thr 30 is a
`constant or nearly constant residue in mouse H chain subgroups I
`and II, as is V L Tyr 49 in mouse kappa chains. While the interaction
`of VL Tyr 49 with antigen is relatively weak (one van der Waals
`
`SCIENCE, VOL. 233
`
`2 of 7
`
`BI Exhibit 1059
`
`

`

`Fig. 2. The Ca skeleton of lysozyme in the com plex
`
`
`(thick trace) superimposed by least squares on that of
`
`
`native lysosyme in the tetragonal crystal form (thin
`trace). The interface to Fab is at the top, and no
`significant conformation change is apparent
`in this
`
`
`region. Greater differences, although still not signifi ­
`
`cant at this resolution, occur at the bottom of the
`molecule.
`
`excludes described above, a tightly packed interface which mostly contact between itS aromatic side chain and Ca of Gly 22 of
`
`
`
`
`
`lysozymc ), there is a strong hydrogen bond between the hydroxyl
`solvent.
`.
`group of V H Thr 30 and the carbonyl oxygen of Lys 116 of
`Although the antigen-antibody interface involves all six CDR's of
`
`
`
`antibody involving an invariant lysozyme. This specific interaction
`
`the Fab, there are more
`
`interactions with VH than with VL CDR's,
`and with VH CDR3 in particular (Tables 1 to 3). The geometrical
`
`residue demonstrates that the functional distinction between
`
`
`
`
`"framework" (FR) and CDR residues, although largely maintained,
`The interacting
`is n0t absolute.
`
`surfaces arc complementary, with
`
`
`protruding side chains of one lying in depressions of the other (Fig.
`
`3) in commo n with other known protein-protein interactions (23).
`There arc many van der Waals interactions interspersed with
`
`hydrogen bonds. This is most striking for the side chain ofGln 121,
`
`
`which penetrates deeply into the Fab, surrounded by three aromatic
`side chains, V L Tyr 32 and Trp 92 and V H Tyr 101 (Figs. 3, 4, and
`
`5). Its amide nitrogen forms a strong, buried hydrogen bond to the
`main chain carbonyl oxygen ofV L Phe 91 (Fig. 5 and Table 3). The
`V H Tyr I 0 I extends to the surface of lysozyme, its terminal
`adjacent
`
`hydroxyl group forming hydrogen bonds to the main chain nitro­
`gens of Val 120 and Gln 121, and to 081 of Asp 119. Many
`hydrogen bonds occur between the side chains of the antigen and
`
`the main polypeptide chain of the antibody, and vice versa (Table
`3). Hydrogen bonds between main polypeptide chain atoms, similar
`to those in J3-sheet strucrures
`, occur between Lys 116 of lysozyme
`and VH Gly 31, and between Gly 117 and VH Gly 53, where the
`
`lack of side chains allows close approach. There are many side chain­
`form ing, together with the ones
`side chain close interactions
`
`Fig. 3. Space filling representation of Fab D 1.3 and lysozyme. (A) Antigcn­
`
`
`
`antibody complex strucrure as determined in this work. The: antibody H
`chain is shown in blue, the L chain in yellow, lysozymc: in grttn, and
`
`Gin 121 in red. (B) The Fab and lysozyme models have been pulled apart to
`
`
`
`
`indicate protuberances and depressions of each fit in complementary surface
`features of the other. c:ornparc with (A) above. At the top of the interface,
`
`protru ding V L residues His 30 and Tyr 32 fit into a depression in lysozyme,
`between residues Ile 124 and Leu 129 (sec also Table 1). Below the Gin 121,
`
`
`in red, a protuberance of lysozyme
`
`
`consisting of residues around Thr 118 fits
`
`
`into a surface depression by V H residues of CDRl and CDR2 (V H
`formed
`
`Trp 52 can be seen at the bottom of this depression). views of
`(C) End-on
`
`
`the antibody combining site (left) and the antigenic markers of lysozyme
`recognized by antibody Dl.3, formed from (B) above, by rotating each of
`
`
`
`
`approximately 90° about a vertical axis. Contacting residues on
`the molecules
`the antigen and antibody are shown in red, except for Gin 121 shown in light
`
`
`purple. L chain residues that contact the antigen are labeled 1 (His 30), 2
`(Tyr 32), 3 (Tyr 49), 4 (Tyr 50), 5 (Phe 91), 6 (Trp 92), and 7 (Ser 93). H
`
`chain residues that contact the antigen are labeled 8 (Thr 30), 9 (Gly 31), 10
`
`(Tyr 32), 11(Trp52), 12 (Gly 53), 13 (Asp 54), 14 (Arg99), 15 (Asp 100),
`16 (Tyr 101), and 17 (Arg 102); sec Table l. Lysozymc residues that contact
`the antibody are labeled 1 (Asp 18), 2 (Asn 19), 3 (Arg 21), 4 (Gly 22), 5
`(Tyr 23), 6 (Ser 24), 7 (Leu 25), 8 (Asn 27), 9 (Lys 116), 10 (Gly 117), 11
`(Thr 118), 12 (Asp 119), 13 (Val 120), 14 (Gin 121), 15 (Ile 124), and 16
`
`(Leu 129). Gin 121 fits into the antibody surface pocket surrounded by V L
`
`and VH residues 2, 5, 6, 7, and 16 (Table 1).
`IS AUGUST 1986
`
`RESEAlt.CH ARTICLES 7+9 .
`
`3 of 7
`
`BI Exhibit 1059
`
`

`

`Fig. 4 . Stereo diagram of the antibody-antigen interface in
`a similar orientation to Fig. 1 . All atoms are shown for
`those residues involved in the interaction. Heavy and light
`main chains are indicated by thick and thin bonds, respec­
`tively, and hydrogen bonds by dotted lines. Lysozymc
`residues broadly lie below the diagonal from top left to
`lower right of the diagram.
`
`center of the surface lies near V H CDR3, and is occupied by the side
`chain of V H Asp l 00, which forms H bonds to the side chains of
`Ser 24 and Asn 27 of lysozyme. Of the antibody hypervariable
`regions, VL CDR2 contributes the least to antigen binding. A large
`number of antibody side chains in the interface (9 out of 15 if we
`exclude Gly residues) are aromatic, thus presenting large areas of
`hydrophobic surface to the antigen; in addition, some of them such
`as V L Tyr 50 and V H Tyr l 0 l participate in hydrogen bonding with
`the antigen via their polar atoms. In all, 748 A2 or about 11 percent
`of the solvent-accessible surface (24) of lysozyme is buried on
`complex formation, together with 690 A2 for the antibody.
`Antigen variability and antibody specificity. The fine specificity
`of monoclonal antibody Dl.3 for other avian lysozymes shows its
`ability to distinguish a single amino acid change in the antigen, at
`position 121. Fab Dl.3 binds hen e� white lysozyme with an
`equilibrium affinity constant of 4.5 x 10 M-1 (25). Bobwhite;: quail
`lysozyme, with four amino acid sequence differences (26) from hen
`lysozyme but none in the interface with Fab D 1.3, binds with
`similar affinity (25). The binding of antibody Dl.3 to the lysozymes
`of partridge [three amino acid differences (26) ], California quail
`[four amino acid differences (26)], Japanese quail [six amino acid
`differences (27)], turkey [seven amino acid differences (28)], and
`pheasant and guinea fowl [ten amino acid differences each (29)] is
`undetectable (KA < 1 x 105�1) with the enzyme-linked immu­
`noabsorption assay used in our laboratoty. These lysozymes differ
`from hen lysozyme in the amino acid residue at position 121, which
`makes dose contacts with the antibody. Except for Japanese quail
`and pheasant lysozymes, all have Gin replaced by His.
`
`Table 1 . Antibody residues involved in contact with lysozyrne. Sequence
`positions arc numbered as in Kabat a Ri. ( l) except for V H CD R3, where the
`numbers of Kabat a Ri. (1) are given in parentheses.
`Antibody residues
`Lysozyme residues in contact
`Light chain
`CDRl
`FR2
`CDR2
`CDR3
`
`His 30
`Tyr 32
`Tyr49
`Tyr 50
`Phe 91
`Trp92
`Scr93
`
`Leu 129
`Leu 25, Gin 121 , Ile 124
`Gly22
`Asp 18, Asn 19, Leu 25
`Gin 121
`Gin 121, Ile 124
`Gin 121
`
`Hcavychain
`FRI
`CDRl
`CDR2
`
`CDR3
`
`750
`
`Thr30
`Gly 31
`Tyr 32
`Trp52
`Gly53
`Asp54
`(96)
`Arg99
`Asp 100 (97)
`Tyr 101 (98)
`Arg 102 (99)
`
`Lys 116, Gly 117
`Lys 116, Gly 117
`Lys 116, Gly 117
`Gly 117, Thr 118, Asp 119
`Gly 117
`Gly 117
`Arg 2 1, Gly 22, Tyr 23
`Gly 22, Tyr 23, Ser 24, Asn 27
`Thr 118, Asp 119, Val 120, Gin 121
`Asn 19, Gly 22
`
`A computer graphics analysis indicates that a His residue could be
`placed in the interface, in the space occupied by Gin 121, with small
`displacements of the contacting antibody side chains, maintaining
`the H bonds made by Gin 121. Conformational energy calculations
`(30) con1irm this possibility, the total energy being little changed on
`substitution of His fur Gin 121. The buried hydrogen bond is
`maintained with good geometry, and only vety small shifts of neigh­
`boring groups are necessary to accommodate the mutation.
`This seems to rule out steric hindrance in explaining the absence of
`complex formation when His occurs at position 121. Other possible
`explanations for the effect of this amino acid substitution include the
`following. (i) His 121 could be charged, and consequently unstable
`in the hydrophobic pocket occupied by Gin 121; (ii) its side chain
`may have a different orientation from that of Gin, forming, for
`example, a salt bridge with Asp 119; and (iii) substirution of His for
`Gin at position 121 may induce a local change of conformation in
`the polypeptide backbone making the antigenic determinant unrec­
`ognizable by the antibody. Not enough information is available to
`decide on the relative importance of these factors. Nevertheless, the
`
`Table 2. Lysozymc residues in contact with antibody.
`
`Lysozyrne
`residues
`Asp 18
`Asn 19
`Arg21
`Gly22
`Tyr23
`Ser 24
`Lcu25
`Asn27
`
`Antibody residues
`in contact (No.)
`I L chain
`2 H,L
`lH
`4 H(3), L
`2H
`lH
`1 L
`lH
`
`Lysozyme
`residues
`Lys 116
`Gly 117
`Thr ll8
`Asp 119
`Val 120
`Gin 121
`Ile 124
`Leu 129
`
`Antibody residues
`in contact (No.)
`3H
`6H
`2H
`2H
`lH
`5 H( l), L(4)
`2L
`lL
`
`Table 3. Hydrogen bonded interactions between antibody and lysozymc.
`Sequence positions are numbered as in Kabat a Ri. ( l) except for V H CD R3
`where the numbers of Kabat a td. (1) are given in parentheses.
`Antibody residue Lysozyrnc residue
`
`Light chain
`
`Heavy chain
`
`Ne2 His30
`0,, Tyr50
`0
`Phe91
`
`Oyl Thr30
`N Gly 31
`N Gly 53
`N11l Arg 99
`(96)
`081 Asp 100 (97)
`0&2 Asp 100 (97)
`0,, Tyr 101 (98)
`0,, Tyr 101 (98)
`0,, Tyr 101 (98)
`*Denotes the: closest interactions (distances s; 2.5 A).
`
`0
`Leu 129
`081 Asp 18
`Ne2 Gin 121*
`
`Lys 116*
`0
`0 Lys 116
`0 Gly 117*
`0 Gly22
`N82 Asn27
`Oy Ser 24 *
`N Val 120
`N Gin 121
`rnn Asp 119
`
`SCIENCE, VOL. 233
`
`4 of 7
`
`BI Exhibit 1059
`
`

`

`Fig. 5. Stereo view of the environment of Gin 121
`with (above) atoms drawn with their van der Waals
`radii, showing the close packing of the three antibody
`aromatic rings around the antigen side chain. The
`dotted line indicates the hydrogen bond from NE.2 of
`Gin to the main chain carbonyl oxygen ofV L Phc 91.
`
`fact that the His residue is not induced to fit into the interface
`position occupied by Gln 121 is in agreement with a "lock and key"
`model (see below) of complex formation between conformationally
`stable antigen and antibody structures.
`Japanese quail lysozyme has an Asn at position 121 and the
`additional differences Asn 19 --+ Lys and Arg 21 --+ Gin in the
`antigen-antibody interface. Asn 121 would be unable to form
`hydrogen bonds as strong as those of Gln 121. In addition,
`replacement of Asn 19 by Lys causes the loss of weak interactions
`between 081 of Asn 19 and N of VH Arg 102. The positively
`charged Lys 19 side chain would be repelled by V H Arg 102 and
`would probably remain outside the interface, further reducing
`packing efficiency. Only main chain atoms of Arg 21 make contact
`with Fab and the side chain is external; therefore cl)anges at this
`position in the antigen surface are probably not detrimental to
`complex formation.
`The equilibriwn affinity constant of Fab Dl.3 binding of hen
`lysozyme is 4.5 x 107M-1• Other monoclonal antibodies to lyso­
`zyme that we (25) and others (31) have obtained and characterized
`show similar affinity constants for the homologous lysozyme anti­
`gen. Moreover, the determined equilibrium constants of protein
`antif,ens with their specific antibodies range from 1<>5M-1 to
`101 '.M-1 (32). Thus, Dl.3 is a typical antibody of the monoclonal
`response in BALB/c mice and one of an about average affinity
`constant in immune responses to protein antigens in general.
`Comparison of evolutionarily related proteins has been used to
`identify antigenic sites in proteins such as lysozyme (5, 25, 33). The
`detection of antigenic determinants by these fine specificity studies
`is biased coward the recognition of evolutionarily variable residues,
`such as Gln 121 by antibody Dl.3. As our results show, such
`analyses are limited in defining antigenic determinants and, in
`particular, in defining the area of the antigen-antibody interaction,
`or even its center. Antigenic determinants have also been localized
`by measuring the reactivity of natural or synthetic peptides corre­
`sponding to different parts of the sequence of the protein with
`antibodies to the protein. This method cannot identify, or it can
`identify only partially, noncontinuous determinants such as those
`recognized by Dl.3. Furthermore, given the large size of an
`
`IS AUGUST 1986
`
`antibody combining site (about 690 A 2 of accessible surface area in
`our study) plus the fact that only a small portion of the surface of a
`globular protein is made up of linear arrays of residues, the
`probability that all of the antigen residues contacted by a given
`antibody come from the same continuous segment of polypeptide
`chain is very low (34). Thus, most antibodies to native protein
`molecules probably recognize noncontinuous determinants.
`In the three-dimensional structure of the antigen-Fab complex
`presented in this article, the axes of the V and C domains of Fab
`make an angle close to 180°. This gives an extended conformation,
`with the V and C domains further apart than they would be if that
`angle were smaller. This observation is not in agreement with
`hypotheses (35) in which the liganded antibody molecule is postu­
`lated to asswne a more rigid conformation with an "elbow bending"
`angle (between the axes of the V and C domains) close to 120°. In
`fact, the angle observed in the lysozyme-Fab D 1.3 complex corre­
`sponds to that postulated to occur in unliganded Fab's. Thus, the
`allosteric model of antibodies (35, 36) in which antigen binding
`induces changes in quatema.ry structure resulting in closer contacts
`across V and C domains is not consistent with the structure of this
`complex.
`No other change of conformation in the antibody or antigen can
`be established by the present analysis. The classical "lock and key"
`metaphor (37) is an adequate simplification to describe the interac­
`tion of lysozyme and antibody D 1.3. It implies that somatic
`recombination of the gennline gene repertoire provides all the
`complementary antibody templates necessary to bind all possible
`antigens. These combining site templates preexist and are basically
`unaltered in binding their specific antigens. The lysozyme-D 1.3
`binding is accomplished by van der Waals and hydrogen bonding
`interactions, and the number of contacts is of the order of that seen
`in other protein-protein systems, with similar implications for the
`specificity and the energetics of the interacting molecules (23).
`Although the fit of the antigen-antibody contacting surface is
`remarkably good, there are some imperfections in the form of holes.
`One of these holes, between VH residues 52 and 100 and lysozyme
`residues 24 and 118, is probably filled by a water molecule,
`hydrogen-bonded co the N of lysozyme Gly 117, as suggested by an
`
`RESEARCH ARTICLES 7SI
`
`5 of 7
`
`BI Exhibit 1059
`
`

`

`electron density peak at that location. Other holes appear not to be
`filled by water molecules. This observation might explain the
`occurrence of heteroclitic antibodies, that is, antibodies that have
`higher affinities for heterologous, closely related antigens, which
`would fill those holes and provide a tighter association. It might also
`explain an improved fit between antigen and antibody by somatic
`mutations in the antibody genes. For example, in antibody Dl.3,
`further increases in affinity could be achieved by amino acid changes
`which would permit salt links and hydrogen bonding to lysozyme
`residues, such as Arg 21 and Thr 118, whose polar parts are
`unbonded in the interface.
`Complete immunoglobulin V region genes are generated by
`somatic recombination of gene segments during B lymphocyte
`differentiation (38). The VH polypeptide chain is encoded by three
`gene segments, VH, D (diversity), and 1H (joining). A complete H
`chain V gene is generated by V.,D and D-JH joinings (39). The
`core portion of V H CDR3, composed of 1 to 13 residues, is
`e::icoded by the D segment. As the V.,D and D-JH boundaries fall
`with the V H CDR3, the combinatorial joining contributes to the
`high sequence variability of this CDR3, higher indeed than that of
`other CDR's. This variability suggests an essential role for VH
`CDR3 in antigen recognition (40), which we now confirm for
`antibody Dl.3. Specifically, our three-dimensional model shows
`
`that (i) the physical center of the lysozyme-antibody interface is at
`V H CDR3; (ii) this CDR makes a proportionately greater contribu­
`tion to the fonnation of the complementary antigen-binding surface
`of the combining site than the other CDR's, as measured by the
`number of van der Waals contaets and hydrogen bonds (6 out of 12;
`see Table 3) it makes with lysozyme. All four VH CDR3 residues
`(99 to 102) in contact with antigen are encoded by the D segment,
`thus illustrating its critical role in the generation of functionally
`different combining sites.
`Our structural model permits us to evaluate the contribution of
`the imprecise joining of gene segments at the V L -JL, V .,DH, and
`D.,JH junctions to the antigen contaets at the binding site. In
`mouse kappa chains, the somatic recombination process between V L
`and 1L generates sequence diversity at position 96 (41). The
`interactions ofV L CDR3 with antigen are illustrated in Figs. 3 to 6.
`Only atoms of residues 91 to 93 are within 4 A of neighboring
`lysozyme atoms; the V-J junction at Arg 96 is relatively distant (>6
`A) from the interface. Thus neither residue 96 nor JK·cncoded
`residues participate directly in contaets with antigen in antibody
`Dl.3. The interactions ofVH CDR3 with lysozyme along with the
`locations of the V-D and D-J boundaries are shown in Fig. 6. The
`side chains ofVH Arg 99, Asp 100, and Tyr 101 fonn a number of
`hydrogen bonds with several lysozyme residues (Table 3), while
`
`Fig. 6. (A) Stereo diagram of V H CDR3 with
`
`
`interacting residues from lysozyme shown below and
`
`
`
`to the right. Thick bonds indicate residues coded by
`the D gene segment through which all CDR3 con­
`Similar view to (A)
`tacts to the antigen are made.,(B)
`for the VL CDR3. The V-J boundary is indicated,
`
`
`showing lack of involvement in the interaction of
`coded by the JL gene segment.
`residues
`
`752
`
`SCIENCE, VOL. 233
`
`6 of 7
`
`BI Exhibit 1059
`
`

`

`guanidino group atoms of Arg 102 are in van der Waals contact
`with lysozyme Asn 19 and Gly 22. These are all D-segment encoded
`residues; thus, neither J 8 residues nor those that might arise from
`imprecise joining at the D-J junction (42) contribute directly to
`antigen contacts made by antibody Dl.3.
`It is noteworthy that most lysozyme residues at the interface
`contact only one of the antibody chains: Asp 18, Leu 25, Ile 124,
`and Leu 129 contact only the L chain, while Arg 21, Tyr 23, Ser 24,
`Asn 27, Lys 116, Gly 117, Thr 118, Asp 119, and Val 120 contact
`only the H chain (Table 2). Thus, pairing of an L or H chain with
`different H or L chains could generate antibodies that will bind antigenic
`variants ar positions contacting only one of the antibody chains.
`The three-dimensional model obtained from the antigen-antibody
`complex presented here differs in important respects from that
`obtained in the study of the three-dimensional srructure of hapten­
`antibody complexes (3, 4). In these latter srudies no conformational
`change in the Fab's was detected after hapten binding. The lyso­
`zyme-Fab complex suggests a similar result except in the relative
`disposition of the V and C domains, as discussed above. However,
`in the hapten-Fab studies and, even more, in those of the Jess specific
`ligand binding to an L-chain dimer (43), a cavity or pocket
`surrounded by the CDR's of V8 and VL was the most relevant
`structural feature of the antibody combining site. In the antigen-Fab
`model presented here the combining site appears as a large, irregu­
`lar, and rather flat surface with protrusions and depressions formed
`by the amino acid side chains of the CDR's of the antibody. The
`complementarity of shape between antigen and antibody in the
`interface is striking: protruding side chains of one lie in depressions
`of the other. In addition, two antibody FR residues also closely
`con