An nu. Rev. Bhchem. 1990. 59 :439 73
`
`ANTIBODY-ANTIGEN COMPLEXES 1
`
`David R. Davies and Eduardo A. Padlan
`
`Laboratory of Mole(.;u]ar Biology, National Institute of Diabetes, Digestive and Kid·
`ney Diseases, Bethesda, Maryland 20892
`
`Steven Sheriff
`
`Squibb Institute for Medical Research, Princeton, New Jersey 08543 4000
`
`KEY WORDS: epitopes, complementarity, modeling, three dimensional structure, haptens
`
`CONTENTS
`
`INTRODUCTION .................................................................................... .
`ANTIBODY BINDING TO SMALL MOLECULES ........... ................... .. ....... .. .
`. .............. ..
`McPCfiO.i' <Jfld Phosplwrlioline. .....
`l'lew and Vitamin K 10H .. .................... .. .......... .. .......... .. .......... .. .......... .
`4-4 20 and Fluorescei11
`..... .... .... .... .. .... .... ... .... .. ..... .. ...... ... . .
`.......... .. ...... .. ..................... .
`Other Fabs Bindi1111 to Small Li:;ia11ds
`ANTIBODY ANTIGEN COMPLEXES ......... .. .............................................. .
`The Anti Lys02yme Complexes .. . . . . . . . .. . .. .. .. . ..
`. ............ ..
`T/1e Antit•ody Neurnminidase Complexl'S ..... .............................................. .
`Compllrisr:m of the Ne11rami11idm£' tmd Lysuzyme Comp/e;.es .. .. ..... .................. .
`Amibodies with Specificities for Other L<Jrge Molecules ............ ..
`STRUCTURAi. ASPECTS OF ANTIBODY-ANTIGEN COMPLEXES
`ANTIBODY ENG!NEFRING ......
`................ .. ...... .. ..... ..... .. .... ... .
`CAT AL YT!C ANTIBODIES
`MODELING Of ANTIBODY COMBINING SITES ............. ................. .. . ....... ..
`Template-Based Predictiom . ............................... ... .. ....... ...... ....... ....... .. .
`De Novo Modelin:;i .. . ... .. .. . ... ... .. ..
`. ... .. .... ..... ..
`Future Directions ............................................................................... .
`
`440
`442
`443
`443
`444
`444
`445
`445
`455
`456
`457
`459
`463
`463
`465
`465
`467
`469
`
`1The US Goverrunent has the right to retain a nonexclusive royalty-free licen&e in am! to any
`copyright coyering this paper.
`
`439
`
`1 of 37
`
`BI Exhibit 1112
`
`

`

`440
`
`DAVIES ET AL
`
`INTRODUCTION
`
`Antibodies are made in all vertebrates as part of the immune response to
`antigenic challenge by foreign substances. The diversity of this response is
`impressive: any foreign macromolecule can, under appropriate conditions,
`elicit an immune response. It has been estimated that humans can produce as
`many as 10 million different antibodies in the primary repertoire and that this
`may then be further expanded by several orders of magnitude through the
`effects of somatic mutation (1 ). In addition, the antibody response shows
`remarkable specificity, so that evidence of any significant amount of cross(cid:173)
`reactivity between different antigens is usually taken to indicate close similar(cid:173)
`ity of their structures. The manner in which diversity and specificity operate
`in the antibody molecule at the level of the three-dimensional structure, as
`determined by X-ray diffraction, is the subject of this review.
`Antibodies were the first members of the immunoglobulin superfamily to
`be studied structurally. The domain structure observed in antibodies (2-7) has
`now been seen in many cell-surface proteins that function to control the
`movement and differentiation of many types of vertebrate cells (8, 9). Mem(cid:173)
`bers of this family have characteristic domains of approximately 100 amino
`acids and often contain an internal disulfide loop of 40-70 residues. In those
`proteins where the three-dimensional structure is known, notably antibodies
`and the class I MHC antigens ( 10, 11), the tertiary fold of these domains has
`been very similar. The recently reported structure of the chaperone protein
`from Escherichia coli. PapD (154 ), is of particular interest since both do(cid:173)
`mains of this structure have the immunoglobulin fold despite the absence of
`sequence similarity with the lgG CH2 domains.
`Antibodies (sec Figure I) are multivalent molecules made up of light (L)
`chains of approximately 220 amino acids and of heavy (H) chains of 450-575
`amino acids. The light chains contain two immunoglobulin domains; the
`N-terminal domain is variable, i.e. it varies from antibody to antibody, and
`the C-terminal domain is constant, i.e. it is the same in light chains of the
`same type. The heavy chains are made up of an N-terminal variable domain
`and three or four constant domains. The antibody fragment containing the
`associated variable domains of the light chain (VL) and of the heavy chain
`(VH) is called the Fv; the fragment containing the entire light chain and the
`VH and first constant domain of the heavy chain is called the Fab.
`The combining site of antibodies is formed almost entirely by six
`polypeptide segments, three each from the light and heavy chain variable
`domains. These segments display variability in sequence as well as in number
`of residues, and it is this variability that provides the basis of the diversity in
`the binding characteristics of the different antibodies. These six hypervariable
`
`2 of 37
`
`BI Exhibit 1112
`
`

`

`ANTIBODY-ANTIGEN COMPLEXES
`
`441
`
`FAB
`
`Fe
`
`Figure l Line drawing of the alpha carbon trace of a model of hwnan lgG I (E. A. Padlan,
`unpublished). Thick lines show the heavy chain, thin lines the light chain. The various domains
`of the light and heavy chains are labeled (VL, VH, CH I, etc). Th~ antigen binding fragment,
`Fab, and the Fe (consisting of the CH2 and CH3 domains of the two heavy chains) are labeled.
`The hinge is the heavy chain peptide segment joining the Fab and the Fe. It varies considerably in
`size in different antibody types. The hinge is often rich in proline and cysteine residues; the latter
`link the two h1:avy chains through symmetrical disulfide bonds. These interchain disulfide bonds,
`as well as those found in the domain interiors, are shown as filled circles. Carbohydrate has been
`found between the two CH2 domains in hwnan lgG I Fe ( 152), and is probably in a similar
`location in most of the other antibody classes. The two carbohydrate chains are drawn with thin
`lines.
`The molecule was assembled from the Fab of KOL (153), a model of the octapeptide
`Pro Ala Pro Glu Leu-Leu-Gly-Gly, corresponding to residues 230--237 in the hwnan lgG I hinge
`(12), and the human lgGl Fe of Deisenhofer (152).
`
`segments a:re also referred to as the complementarity determining regions or
`CDRs (12).
`The three-dimensional structure of antibodies has been the subject of
`numerous investigations (reviewed in 13-23), and crystal structures for intact
`immunoglobulins and for a variety of fragments are now available. More than
`a dozen Fab structures have been determined crystallographically and at least
`eight of those have been studied with bound ligand. Three Fab-ligand com(cid:173)
`plexes involve small haptens [vitamin K10H (24), phosphocholine (25-27),
`and fluorescein (28)], and five complexes involve proteins [three with hen egg
`white Iysozyme (29-31) and two with influenza virus neuraminidase (32-
`
`3 of 37
`
`BI Exhibit 1112
`
`

`

`442
`
`DA VIES ET AL
`
`34)]. The structures of Fabs with specificities for carbohydrates, DNA, and
`other molecules have also been determined, although at present only in the
`uncomplexed form.
`The three-dimensional structures of antibodies and of antibody fragments
`have been reviewed extensively (l 3-23). Here, we concentrate on the interac(cid:173)
`tion between antibody Fabs and specific ligands. We first consider the binding
`of antibodies to small molecules (haptens) and then to larger molecules (in
`particular, proteins). We then analyze the various antibody-ligand complexes
`of known three-dimensional structure looking at properties that define the
`specificity and strength of the interactions.
`We also discuss briefly how the specificity of antibodies is being utilized
`for a variety of practical applications. For example, it has long been specu(cid:173)
`lated that antibodies could be used as enzymes. Antibodies and enzymes share
`the ability for very specific binding. However, antibodies only bind to
`antigens, while enzymes bind substrates and catalyze their conversion to
`products. It was proposed 20 years ago by Jencks (35) that antibodies raised
`against transition-state analogues could have catalytic activity. This possibil(cid:173)
`ity has now been demonstrated, and antibody enzymes (abzymes) showing
`significant activity have been produced (36, 37).
`Antibodies have many potential uses in diagnosis and in therapy and, with
`the advent of hybridoma technology (38), monoclonal antibodies of almost
`any desired specificity can be produced. However, the monoclonal antibodies
`that are more easily made are of rodent origin, and the long-term use of these
`antibodies in human subjects is blocked by the immune response of the host
`Attempts have been made to circumvent these difficulties by the creation of
`chimaeric antibodies containing the constant domains of the host together
`with rodent variable domains (3H2). This 'humanization' of the more easily
`available rodent monoclonal antibodies has been further extended by the
`grafting of the rodent combining site structures to a human framework
`structure, making feasible the production of essentially human antibodies with
`designer binding properties ( 43-45).
`For many of these applications a knowledge of the three-dimensional
`structures of particular combining sites is very helpful, but a crystallographic
`analysis of each antibody molecule is clearly impractical. The alternative is to
`use the existing structural information to model the new combining sites. The
`progress of these model building studies is discussed.
`
`ANTIBODY BINDING TO SMALL MOLECULES
`
`The first Fabs to be analyzed by X-ray diffraction were prepared from
`myeloma proteins. The antigenic determinants for these antibodies could only
`be inferred from binding studies using small molecules. Large numbers of
`
`4 of 37
`
`BI Exhibit 1112
`
`

`

`ANTIBODY-ANTIGEN ffiMPLEXES
`
`443
`
`these immunoglobulins with known binding specificities were characterized
`and sequenced. The first two Fabs whose structures were determined, New,
`which was shown to bind to a vitamin K 10H derivative (46), and McPC603,
`which binds to phosphocholine (47), were analyzed in both the complexed
`and uncomplexed forms (5, 7, 24-27, 48-50). Until the advent of hybridoma
`tech11ology, these investigations provided the structural basis for understand(cid:173)
`ing antigen recognition. They are still of considerable interest in showing how
`high bindi111g and specificity can be produced for relatively small molecules.
`Jn addition to New and McPC603, a third small-molecule-antibody com(cid:173)
`plex, fluorescein bound to 4-4-20, has been recently determined in two
`laboratories (Ref. 28; M. Whitlow and K. Hardman, personal communica(cid:173)
`tion). As the complexes of New and of McPC603 have been reviewed
`previously (22), we describe them only briefly here.
`
`McPC603 and Phmphocholine
`McPC603 will precipitate with pneumococcus C polysaccharide (51), which
`contains phosphocholine. This precipitation is inhibited by phosphocholine
`(52), which binds to the McPC603 with a binding constant of 2.0 x 105 per
`mole (53). X-ray studies show that phosphocholine binds in a pocket in the
`McPC603 combining site with the choline buried and the phosphate on the
`surface (25-27). The phosphate group is within hydrogen bond distance of
`H-Arg52NH 1 and H-Tyr330H. There is charge neutralization in that L-Asp91
`is at the back of the pocket together with H-Glu35, which is one layer
`removed from contact with the positively charged choline. At the surface,
`H-Arg52 and H-Lys54 provide positive charges that complement the charge
`on the phosphate. The phosphocholine is in contact with only four of the six
`hypervariable loops: CDRl of the light chain and all three CD Rs of the heavy
`chain. There appears to be no conformational change upon binding phos(cid:173)
`phocholine, but this could be partially due to the presence of a sulfate ion
`from the crystallization medium, which is located in the phosphate-binding
`position (50).
`
`New and Vitamin K10H
`New was shown to bind to, among other ligands, the gamma-hydroxy de(cid:173)
`rivative of vitamin K 1 with a binding con~tant of 1.7 x 105 M 1 (46). X-ray
`studies of the complex of Fab New with vitamin K10H revealed that the
`2-methyl-l .4-naphthoquinone moiety of the vitamin K 1 OH sits in a shallow
`groove between the heavy and light chains of approximately 16 A x 7 A and
`about 6 A deep. The phytyl chain of the vitamin K 10H runs along the surface
`of the antibody-combining site and contacts a number of residues. At the time
`of this investigation, the sequence of the New heavy chain had not been
`determined, so that the interaction between ligand and combining site could
`
`5 of 37
`
`BI Exhibit 1112
`
`

`

`444
`
`DAVIES ET AL
`
`not be described in detail. Amzel et al (24) estimate that the contact site
`consists of at least 10-12 residues and at least four hypervariable loops
`(L-CDRI, L-CDR3, H-CDR2, H-CDR3). The New~vitamin K 10H coordin(cid:173)
`ates are not available from the Protein Data Bank
`
`4-4-20 and Fluorescein
`Anti-fluorescein antibodies have been found with affinities for fluorescein
`ranging from about 105 to 1010 M 1
`. The 4-4-20 antibody binds fluorescein
`with an affinity at the high end of the range. Fluorescein consists of a
`xanthonyl moiety with enolic oxygens on each exterior ring and a phenyl
`carboxylate, which is attached through the ortho position to the central ring of
`the xanthonyl. The antibody is quite selective in that it does not bind the
`closely related rhodamine molecules, which have amino or diethylamino
`groups substituted for the enolic oxygens.
`The fluorescein is bound with the planar xanthonyl ring at the bottom of a
`relatively deep slot formed by a network of tryptophan and tyrosine side
`chains. The phenylcarboxyl group is partially exposed to the solvent. The
`xanthonyl group is flanked by L-Tyr37 andH-Trp33. L-TrplOI makes up the
`floor of the combining site. The phenyl ring of the hapten is close to
`H-Tyrl 03; H-Tyrl 02 lines part of the boundary of the slot. The fluorescein is
`more than 90% buried by interaction with the 4-4-20 antibody and interacts
`with five of the hypervariable loops (L-CDR 1, L-CDR3, H-CDR I, H-CDR2,
`H-CDR3) (28). The large xanthonyl moiety contributes more than 60% of the
`interactions with the antibody, which is consistent with that portion being
`'immunodominant.'
`One enolic oxygen on the xanthonyl moiety forms a hydrogen bond with
`L-His3 l NE2, and the other forms a salt link with L-Arg39NH2 and a hydro(cid:173)
`gen bond to L-Ser960G. The negative charge on the carboxyl group that is
`part of the phenylcarboxylate is not neutralized, but faces towards solvent,
`and one of the oxygens hydrogen bonds to L-Tyr370H. lt is easy to un(cid:173)
`derstand why rhodamine analogues of fluorescein do not bind. First, the
`positive charge of the amino groups would create unfavorable interactions
`with L-Arg39. Second, in the case of diethylamino groups, there would be
`considerable steric problems in packing the ethyl moieties into the complex.
`Animals are immunized to fluorescein by attaching it to a lysine group on
`keyhole limpet hemocyanin through an isothiocyanate group to the para
`position of the phenyl moiety. From this para position there is a channel in the
`complex through which the lysine from keyhole limpet hemocyanin could
`extend.
`
`Other F abs Binding to Small Ligands
`A number of model antigens have provided experimental probes of antibody
`diversity, specificity, and genetic control of the immune response. One of the
`
`6 of 37
`
`BI Exhibit 1112
`
`

`

`ANTIBODY-ANTIGEN COMPLEXES
`
`445
`
`the hapten, p(cid:173)
`most extensively studied model antigens has been
`azobenzenearsonate. Recently the structure of Fab RI9.9, which is specific
`for p-azobenzenearsonate, has been determined (54 ). This structure shows
`that nine aromatic residues, mostly tyrosines, in the antibody combining site
`are solvent exposed and could provide interactions with hapten.
`Very recently two new hapten-antibody complexes have been determined.
`R. Fox, D. Leahy, and A. Brunger (personal communication) have de(cid:173)
`termined lhe structure of AN02 Fab with the immunizing hapten 2,2,6,6-
`Tetrameth_yl-l-piperidinyloxy-dinitrophenyl bound. R. L. Stanfield, I. A.
`Wilson, and coworkers (personal communication) have determined the struc(cid:173)
`ture of an Fab both uncomplexed and complexed with a peptide that has the
`sequence of the C helix of myohemerythrin (55). Thus there will soon be an
`expanded view of hapten-antibody complexes, including the first examination
`of whether a peptide, which is helical in the native protein (56), maintains that
`structure when bound to antibody.
`Wilson and coworkers have crystallized a number of other antibody-hapten
`complexes, which are suitable for structure determination. They have crystal(cid:173)
`lized 17/9 Fab with and without the immunizing peptide, which is sequential(cid:173)
`ly identical to residues I 00-- J 08 of subunit I of the influenza virus hemagglu(cid:173)
`tinin. This sequence comes from the 'top' of the hemagglutinin, which is the
`immunodominant region (57). They have also crystallized
`the anti(cid:173)
`progesterone DB3 Fab' both in the presence and absence of steroids, includ(cid:173)
`ing proges1terone, pregnanedione, and aetiocholanolane (58, 59). The space
`group and unit cell parameters are unaltered by the presence of the various
`steroids, a~though there is a change in crystal morphology.
`
`ANTIBODY-ANTIGEN COMPLEXES
`
`During the last three years there have heen several structural investigations of
`monoclonal antibodies complexed with protein antigens. The structures an(cid:173)
`alyzed include three complexes with lysozyme (Dl.3, HyHEL-5, HyHEL-10)
`(Figure 2) and two complexes with the neuraminidase from influenza virus
`(32-34). ln spite of their recent publication, these structures have already
`been extensively reviewed (33, 60 64), although there has been no com(cid:173)
`prehensive review of the results for all five complexes. In this section we first
`describe the individual results for each complex, and then discuss some of the
`common properties of the group.
`
`The Anti-Lysozyme Complexes
`These three structures form one subgroup among the complexes. The antigen,
`lysozyme, is a small 14.6-kd, rugged (65) glycosidase, with specificity for
`hexasaccharide having alternating beta-1,4-linked N-acetyl glucosamine and
`N-acetylmuramic acid residues (66). The crystal structure of tetragonal chick-
`
`7 of 37
`
`BI Exhibit 1112
`
`

`

`446
`
`DA VIES ET AL
`
`HyHEL-5
`
`01. 3
`
`HyHEL-10
`
`Figure 2 A composite of the Fvs of the three anti lysozyme antibodies: DI .3 (29), HyHEL 5
`(30), HyHEL IO (31) binding to lysozyme (75). Forth is picture the Fvs have been separated from
`lysozyme. C alpha representations are used for 1he Fvs and for lysozyme, and a dot representation
`is used for their interacting surfaces. It should be noted that the three epitopes are nearly
`non overlapping with only a small overlap between HyHEL IO and DI.3.
`
`en lysozyme was first determined in 1965 (67). Subsequently, the structures
`of several different crystal forms of chicken lysozyme have been determined
`(68, 69), as well as several closely related lysozymes from different sources
`(70, 71). The three-dimensional structure of lysozyme contains a mix of beta
`sheet and alpha helix. Across its middle it has an extensive cleft containing
`the catalytic residues, into which the hexasaccharide substrate binds. Lyso(cid:173)
`zyme has been extensively used as a model antigen, both for antibodies and
`for T cell activation (72-75).
`
`THE D 1. 3 COMPLEX The crystal structure of the antibody D I. 3 complexed
`to lysozyme has been reported at 2.8 A resolution (29). The interface between
`the antibody and antigen is extensive, involving 748 A2 of the solvent(cid:173)
`accesible surface of the lysozyme and 680 A2 of the antibody (29). For the
`molecular surface the corresponding areas are 690 A 2 and 680 A 2
`, respective(cid:173)
`ly (see Table 4, in a later section). The interface extends over maximum
`dimensions of 30 x 20 A. The fit between the two surfaces is very close with
`only one water molecule remaining in the interface.
`
`8 of 37
`
`BI Exhibit 1112
`
`

`

`ANTIHODY-ANTIGEN COMPLEXES
`
`447
`
`Table I Antibcxly Dl.3 residues in contact with lysozyme
`
`Antibcxly
`
`Lysazyme
`
`Light chain
`CDRI
`
`FR2
`CDR2
`CDR3
`
`Heav~ chain
`FRI
`CDRJ
`
`CDRJ
`
`CDR3
`
`HisJO
`Tyr32
`Tyr49
`Tyr50
`Phe91
`Ttp92
`Ser93
`
`Thr30
`Gly3l
`Tyr32
`Trp52
`Gly5::l
`Asp54
`Arg99
`Asp JOO
`TyrJOJ
`Arg102
`
`Leu 129
`Leu25, Glnl2 I, !lel24
`Gly22
`Asp!!!, Asnl9, Leu25
`Glnl21
`Glnl21, Tlel24
`Ginl21
`
`Lysl 16, Glyl 17
`Lysll6, Glyll7
`Lysl 16, Glyl 17
`Giyl 17, Thrll8, Asp119
`Glyll7
`Glyl 17
`Arg2l, Gly22, Tyr23
`Gly22, Tyr2], Ser24, Asn27
`Thcl 18, Aspll9, Va1120 . Glnl21
`Asnl9, Gly22
`
`The epitope (Figure 3a) consists of two stretches of polypeptide chain,
`comprising residues 18 to 27 and 116 to 129, distant in sequence but adjacent
`on the protein surface (Table 1). The side chain of Glnl 21 projects outwards
`from the lysozymc and penetrates deeply into a pocket on the antibody, where
`the amide nitrogen makes a hydrogen bond with a main-chain carbonyl
`oxygen of L-Phc91. The pocket is lined by aromatic side chains of L-Tyr32,
`L-Trp92, and H-TyrlO I. Replacement of Gin 121 by His in several other
`avian lysozymes effectively abolishes the binding to the antibody.
`All six CDRs of the antibody make contact with the lysozyme. In addition,
`there are t1.vo framework residues that are involved in contact, L-Tyr49 and
`H-Thr30. Of the 17 contacting residues, nine are aromatic, including five
`tyrosines.
`No significant change in the conformation of the lysozyme is observed in
`the complex. In a recent report the structure of the unliganded Dl.3 Fab is
`briefly des,;ribed (76). This structure, currently under refinement, shows no
`significant change in the relative orientation of the VL and VH domains from
`the Fab complexed to the lysozyme, nor is there any change in the tertiary
`structure of the CD Rs. There is a difference in the elbow bend (18) of the Fab,
`which is 172Q in the complex vs 138° alone_
`Also reported is the structure determination of a complex of the DI. 3 Fab
`
`9 of 37
`
`BI Exhibit 1112
`
`

`

`t
`
`00
`
`0
`
`~
`
`~
`~
`
`Figure 3a Stereo drawing of the lysozyme epitope for D l .J defined in Table I (coordinates courtesy of Dr. R. Poljak).
`The contacting residues are labeled.
`
`10 of 37
`
`BI Exhibit 1112
`
`

`

`70
`
`ASP 48
`
`47
`
`Figure 3b Stereo drawing of the lysozyme epitope for HyHEL-5 defined in Table 2. The contacting residues are
`labeled.
`
`~ @
`§
`-<
`
`~
`
`~ n
`~
`£
`
`11 of 37
`
`BI Exhibit 1112
`
`

`

`c
`
`~
`VI
`0
`
`ti >
`~
`f;l
`gj
`
`> "
`
`Figure 3c Stereo drawing of the lysozyme epitope for HyHEL-10 defined in Table 3. The contacting residues are
`labeled.
`
`12 of 37
`
`BI Exhibit 1112
`
`

`

`ANTIBODY-ANTIGEN ffiMPLEXES
`
`451
`
`with its anti-idiotypc, E225, Fab (76). Herc too the structure is in the course
`of refinement, but the principal structural features can be described. Again
`there is no indication of significant conformational change in either the
`relative orientations of the VL and VH domains, or in the tertiary structure of
`the CDRs of Dl.3. The elbow bend is 140° in this complex, almost in(cid:173)
`distinguishable from that of the uncomplcxed D 1.3. The idiotypc recognized
`by E225 is largely centered on the VL domain, but with contributions from
`CDR3 and, to a lesser extent, CDR2 of VH.
`
`THE HyHEL-5 COMPLEX The monoclonal antibody, HyHEL-5, complexed
`to lysozymc, has been determined in two different crystal forms (30). It binds
`to the opposite end of the lysozyme molecule from Dl .3 (Figure 2). The
`epitope (Figure 3b) is made up of three segments of the polypeptide chain:
`residues 41-53, 67-70, and 84 (Table 2). In addition, several surrounding
`residues, while not in direct contact with the antibody, are partly buried by the
`interaction. This epitope includes Arg68, which, together with Arg45 forms a
`ridge on the surface of the molecule. Arg68 had been demonstrated by epitope
`mapping to be a 'critical' residue, replacement of which by a lysine led to a
`1000-fold decrease in binding to HyHEL-5 (77).
`
`Table 2 Antibody HyHEL 5 residues
`lysozyme
`
`in contact with
`
`Antibody
`
`Lysozymc
`
`Li,sht chain
`
`CDRI
`
`CDR2
`CDR3
`
`Heavy chain
`
`CORI
`
`FR2
`CDR2
`
`CDR3
`
`Asp31
`Tyr32
`Asp50
`Trp91
`Gly92
`AJE93
`Pro95
`
`Trp33
`Glu35
`Trp47
`GluSO
`Ser55
`Ser57
`Thr58
`Asn59
`Gly95
`Tyr97
`
`Asp48
`Pro70
`Pro70
`Arg45, Gly49, Arg68
`Arg45, Asn46, Thr47
`Arg45, Asn46, Thr47
`Arg45
`
`Tyr53, Arg68
`Arg68
`Arg45
`Arg45, Arg68
`Gln41, Leu84
`Gln41, Thr43
`Thr43
`Thr43, Asn44
`Arg68
`Gly67, Arg68, Thr69, Pro70
`
`13 of 37
`
`BI Exhibit 1112
`
`

`

`452
`
`DAVIES ET AL
`
`The antibody combining site contributes residues from all six of the CDRs
`to the contact with lysozyme (Table 2). There is also at least one framework
`residue, H-Trp4 7, that makes contact with lysozyme. In addition to many van
`der Waals' contacts, there are several salt bridges between the guanidinium
`groups of Arg45 and Arg68 and the carboxyl groups of H-Glu35 and H(cid:173)
`Glu50. There are also 10 hydrogen bonds between uncharged groups of
`atoms. The contacting residues from the antibody include two tyrosines and
`three tryptophans.
`The surface areas of the antibody and lysozyme in contact are about 750 A 2
`each. The interaction between the two surfaces shows good shape com(cid:173)
`plementarity; there are only two water molecules remaining within the in(cid:173)
`terface. The ridge on the lysozyme surface formed by the two arginine side
`chains fits into a groove on the antibody.
`There does not appear to be any gross conformational change in the
`lysozyme as a result of binding; superposition of the bound and unbound
`lysozymes shows a root mean square displacement of the backbone atoms of
`0.48 A. The biggest C-alpha displacement is 1.7 A for Pro70. The largest
`side-chain change occurs forTrp63 within the substrate binding groove where
`the indole ring system has rotated by about 180° about the CB-CG bond, even
`though it is not in contact with the Fab.
`
`THE HyHEL-10 COMPLEX The monoclonal antibody HyHEL-10 binds to yet
`another part of the lysozyme surface (31) (Figure 2). The epitope is centered
`on the alpha helix running from residues 89 to 99, and can be regarded as
`containing this helix together with some surrounding residues. There are
`antibody contacts with Trp63 within the lysozyme groove and with residues
`on the other side of the groove from the helix_ The area of lysozyme in contact
`with the antibody is 774 A2
`• In Tabk 3 the contacting residues between the
`antibody and the lysozyme are shown_ In addition to 114 pairwise van der
`Waals' contacts, thereare 14 hydrogen bonds. There is only one salt bridge, a
`weak hydrogen bond between the oppositely charged side chains of H-Asp32
`and Lys97.
`The surface of the HyHEL-10 that interacts with lysozyme is unusual in
`that it is not noticeably concave and contains no pronounced grooves or
`cavities. Instead it has a large protrusion consisting largely of the side chains
`of H-Tyr33, and H-Tyr53 that fit into the substrate-binding groove of lyso(cid:173)
`zyme. The shape complementarity between the antibody and antigen is again
`remarkable, with no water molecules that can be detected within the interface.
`The interacting surface of the antibody is rich in aromatic residues, particu(cid:173)
`larly tyrosines, which make up 6 out of 19 contacting residues (Table 3,
`Figure 4). These tyrosines point outward from the combining site to make
`contact with the lysozyme_
`
`14 of 37
`
`BI Exhibit 1112
`
`

`

`ANTIBODY-ANTIGEN COMPLEXES
`
`453
`
`1rable 3 Antibody HyHEL IO residues in contact with lysozyme
`
`Antibody
`
`Lysozyme
`
`!jght chain
`
`CDRl
`
`CDR2
`
`CDR3
`
`HeaV;)' chain
`
`FRI
`CDRl
`
`CDR2
`
`CDR3
`
`Gly30
`Asn31
`Asn32
`Tyr50
`Gln53
`Ser91
`Asn92
`Tyr96
`
`Thr30
`Ser31
`Asp32
`Tyr33
`Tyr50
`Ser52
`Tyr53
`Ser54
`Ser56
`Tyr58
`Trp98
`
`Glyl6
`His 15, Gly 16, Lys96
`Gly 16, Tyr20
`Asn93, Lys96
`Thr89, Asn93
`Tyr20
`Tyr20, Arg21
`Arg21
`
`Arg73
`Arg73, Leu75
`Lys97
`Trp63, Lys97, lle98, SerIOO, AspIOI
`Arg21, Ser!OO
`Asp IOI
`Trp63, Leu75, AspIOI
`Asp IOI
`AspIOI, Glyl02
`Arg21, SerIOO, Glyl02
`Arg21, Lys97, Ser!OO
`
`No major conformational change takes place in the lysozyme upon binding.
`A comparison of the bound lysozyme coordinates with those of uncomplexed
`lysozyme shows a root mean square displacement of 0.47 A for the alpha
`carbons, with significant deviations at residues 4 7, 10 l, and 102 of 1.44,
`1.80, and 2.13 A, respectively. Larger deviations are found in the side chains,
`most notably that of Trp62 where the aromatic ring has been rotated by 150°
`about the CB-CG bond, presumably to avoid close contact with a tyrosine side
`chain of the antibody.
`Although no structure is available for the uncomplexed antibody, Padlan et
`al (31) believe that no significant conformational change can have taken place
`because of the overall similarity of the HyHEL-10 domain structures to those
`of other Fabs and the similar ways in which the domains associate.
`
`SUMMARY OF RESULTS FOR THE LYSOZYME COMPLEXES The three anti(cid:173)
`lysozyme complexes described above have been reviewed (62). They have
`epitopes that are almost entirely non-overlapping; there is a small overlap
`between Dl.3 and HyHEL-10 (Figures 2, 3), but Smith-Gill (private com(cid:173)
`munication) has observed little or no functional competition. Together, these
`
`15 of 37
`
`BI Exhibit 1112
`
`

`

`454
`
`DA VIES ET AL
`
`Figure 4 A line drawing of the HyHEL-10 Fv. The structure shown is that of the Fv in the
`complex, with the lysozyme removed. The combining site surface is at the top of the figure.
`Residues that contact the antigen are displayed with side chains; only the backbone is shown for
`the rest of the structure. VH is on the right, VL is on the left. Note the many aromatic side chains
`that protrude out to contact the antigen.
`
`epitopes cover more than 40% of the lysozyme surface area. A comparison of
`the 'B' factor for lysozyme at different parts of the sequence with the
`positions of the epitope showed no particular correlation. This, together with
`the large total area covered by the three epitopes, was interpreted to mean that
`any part of the surface that is accessible to the antibody combining site is
`potentially antigenic, in agreement with Benjamin et al (74). The role of
`mobility in defining the antigenicity of regions of the protein surf ace has been
`extensively discussed recently by van Regenmortel (78) and by Getzoff et al
`(79, 80). The extent to which a correlation with mobility occurs has been
`deduced from experiments where the epitopes are defined largely by peptide
`mapping. It seems reasonable that when the antibody and antigen interact,
`some adjustment of the positions of the interacting residues, particularly the
`
`16 of 37
`
`BI Exhibit 1112
`
`

`

`ANTIBODY-ANTIGEN COMPLEXES
`
`455
`
`side chains, will be observed, corresponding to induced fit. The regions most
`likely to show these changes will be those with the greatest mobility. Howev(cid:173)
`er, any ad,,antage gained in fitting these regions to the antibody will be at least
`partially offset by the entropic cost of immobilization upon forming the
`complex.
`The conformational changes observed here for the backbone atoms of the
`lysozyme in the three anti-lysozyme complexes vary from very small for D 1.3
`(29), to a few angstroms for HyHEL-5 (30) and HyHEL-10 (31 ). The changes
`observed in the side chains are larger, and sometimes involve rotations of
`aromatic ring systems (HyHEL-5, HyHEl-10), supporting the contention that
`rearrangements of side chains sometimes occur in myohemerythrin (79, 80),
`with the effect of exposing otherwise buried groups to the antibody.
`A striking feature of the interaction of these three antibodies with lysozyme
`is the shape complementarity of the two surfaces, as revealed by the almost
`total exclusion of water from the interface over an area of about 700 A 2 for
`both the antibody and antigen. This is a common feature of multisubunit
`protein interfaces as shown by X-ray analysis (81), and suggests that the water
`exclusion must play a significant role in stabilizing the c