Ann. Rev. ImmunoL 1983. 1:87-1I7
`
`
`
`
`Copyright © 1983 by Annual Reviews Inc. All rights reserved
`
`STRUCTURAL BASIS
`
`OF ANTIBODY FUNCTION
`
`David R. Davies and Henry Metzger
`
`National Institute of Arthritis, Diabetes, Digestive, and Kidney Diseases,
`
`
`
`
`
`
`
`
`
`National Institutes of Health, Bethesda, Maryland
`20205
`
`INTRODUCTION
`
`It is less than 20 years since the general architecture of antibodies was
`
`
`
`
`
`
`
`elucidated and even less time since the explicit molecular basis of antibody
`specif
`about the immune system, two basic principles have emerged: (a) Antibod­
`
`
`
`
`
`
`ies remain the only known structures whose diversity is sufficient to explain
`
`
`
`the fne specificity exhibited by the immune response; and, (b) antibody
`
`
`
`
`function is mediated by a molecule whose structure consists of two distinct
`
`
`
`
`
`regions-- ne that carries a recognition site for antigenic determinants, and
`
`
`
`
`a second by which the antibody reacts with receptors of a variety of effector
`systems.
`In this review we examine the current information on the structure of
`
`
`
`
`
`
`
`
`antibodies. We do not describe again the basic four-chain structure of
`
`
`
`
`
`immunoglobulins nor the division into variable and constant regions, which
`
`
`
`are by now well known (e.g. 4, 90, 1 18). We instead concentrate on the
`
`
`
`
`
`higher resolution data, much of which is still in the course of refinement.
`
`
`
`We discuss the Fabs, in particular with reference to the combining site and
`the specif
`binding site of protein A of Staphylococcus aureus and of Clq; and the
`
`
`
`
`
`
`structure of the hinge with reference to its possible role in separating Fab
`and Fe.
`Whereas the characterization of the structures of individual proteins and
`
`
`
`
`
`
`
`
`
`of their interactions with small molecules can now be carried out with some .
`
`
`
`0732-0582183/0410-0087$02.00
`
`87
`
`Further
`
`ANNUAL
`REVIEWS
`Quick links to online content
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1058 Page 1 of 31
`
`

`
`88 DAVIES & METZGER
`
`sophistication, the interaction of two or more macromolecules still presents
`
`
`
`
`
`
`
`considerable difficulties. It is not surprising therefore that our under­
`
`
`
`
`
`standing of how antibodies interact with the macromolecular receptors on
`
`
`
`
`
`effector systems (e.g. Clq in the complement pathway, Fc receptors on cell
`
`
`
`membranes) is much less advanced. Our review of this aspect of antibody
`
`
`
`
`structure and function therefore involves more questions than answers.
`
`IMMUNOGLOBULIN STRUCTURE
`
`General Comments
`Our knowledge of the three-dimensional structure of antibodies at atomic
`
`
`
`
`
`
`
`resolution rests mainly on X-ray diffraction investigations of fragments.
`
`
`
`
`
`Intact proteins for which X-ray analyses have been carried out consist of
`
`
`
`
`
`Kol (106), an IgGl(A) human myeloma protein, the protein Dob (161), an
`
`
`
`IgGl(K) human cryoglobulin, and recently the human myeloma IgGl(A)
`
`
`
`
`protein, Mcg (129). In Kol, and also apparently in another immunoglobu­
`
`
`
`
`
`
`lin, Zie (58a), the crystal structure contains an unusual feature: The Fc
`
`
`
`
`occupies a number of different positions in the crystal that are not crystallo­
`
`
`
`
`
`
`graphically related, with the result that no significant electron density oc­
`
`
`curs in this region of the crystal, there being an abrupt drop in density at
`
`the end of the hinge. In both Dob and Mcg there is a 15-amino-acid-residue
`
`
`
`
`deletion in the hinge (63, 169), thus, presumably, reducing the flexibility of
`
`
`
`
`the molecule and enabling the Fc to be located in the electron density,
`
`
`
`although in the case of Dob the crystals are disordered and do not diffract
`to high resolution.
`The structures of three Fabs have been published, Newm (148), Kol
`
`
`
`
`
`
`(106), and McPC603 (152), as well as the structures of a number of VL
`
`dimers and L chain dimers (4). The structure of human IgG Fc has been
`
`
`
`
`determined, and also its complex with protein A of Staphylococcus aureus
`by Amzel & Poljak
`
`
`(43). These structures have been reviewed most recently
`
`
`(4) and are not covered comprehensively in this review.
`Because of the deficiencies in the crystals of the intact molecules, our
`
`
`
`
`
`
`
`
`
`knowledge of the whole antibody molecule has to be a sum of its parts. The
`
`
`
`
`flexibility of the molecule, in particular in the region between the Fab and
`
`
`the Fc, may preclude for some time visualizing directly by X-ray diffraction
`
`
`
`
`
`an intact molecule with intact hinge at atomic resolution. However, the
`
`
`checks that can be made on this composite three-dimensional picture of the
`
`
`
`
`antibody molecule are reassuring. Thus, the Kol Fab in isolated form in the
`
`
`
`
`crystal is quite similar to the Fab of the whole molecule in its crystal form.
`
`Also, the Fc in Dob has, within the experimental error of the comparison,
`
`
`
`the same overall structure as does the Fc in the isolated Fc crystals.
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1058 Page 2 of 31
`
`

`
`ANTIBODY FUNCTION 89
`
`
`
`Fab Structure and the Antibody Combining Site
`
`McPC603 AND THE PHOSPHOCHOLINE BINDING SITE The structure
`
`of McPC603 Fab, a mouse myeloma IgA (K) with phosphocholine (PC)
`
`
`binding capability, has been determined at 3.l-A resolution (41, 42, 125,
`
`
`1 52) and is being refined to 2.7 A (Y. Satow, D. R. Davies, manuscript in
`
`
`
`preparation). The overall three-dimensional structure of the Fab is illus­
`
`
`
`
`trated in Figure 1, which demonstrates the strong lateral association be­
`
`
`tween domains of the light (L) and heavy (H) chains, together with the
`relatively weak longitudinal interactions along each chain. Figure 1 also
`
`
`shows the clustering of six of the seven hypervariable regions at the tip of
`
`the Fab, forming the complementarity-determining surface (89, 90). The
`
`
`
`
`
`variable domains have a very similar three-dimensional structure for both
`
`
`
`
`the Land H chains and across species (4, 42, 123). The constant domains
`
`
`
`CL and CHI are also very similar. Both the variable and the constant pairs
`
`
`
`of domains are related by approximately twofold (rotation about the V axis
`
`
`
`of 1 800 will superpose V L on V H) and the angle between the two axes has
`
`been referred to as the elbow bend of the Fab and has been observed to vary
`
`G 52C
`
`Figure 1 The a carbon backbone of McPC603 Fab. The heavy chain
`is represented by the
`
`
`
`thick line. The two variable domains are at the top and the constant domains are at the bottom
`residues (CDR) are shown as filed circles.
`
`of the figure. The complementaritydetermining
`Two residues in each CDR loop have been labeled.
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1058 Page 3 of 31
`
`

`
`90 DAVIES & METZGER
`
`between approximately 1 37° for Fab Newm (4), 1 35° for McPC603 (152),
`
`
`147° for Dob ( 161), and approximately 1700 for Kol ( 106).
`Figure 2 shows the combining site of McPC603 with PC bound. (Y.
`
`
`
`
`Satow, E. A. Padlan, G. H. Cohen, D. R. Davies, manuscript in prepara­
`
`
`
`
`tion). The choline is attached at the bottom of a pocket located principally
`
`
`
`between the hypervariable regions H3 and L3. The phosphate is on the
`
`
`
`surface and contacts residues from the heavy chain. It is apparent that PC
`
`
`
`is a small molecule and that the greater part of the hypervariable surface
`
`
`is not directly in contact with it. At the front of the pocket there are two
`
`
`
`
`
`hydrogen bond donors positioned within reasonable hydrogen bonding dis­
`
`
`
`tance of the phosphate oxygens; these are the hydroxyl group of Tyr 33H
`
`and the guanidinium group of Arg 52H (152). The residues lining the inside
`
`
`of the pocket are Tyr 94L on the right side, Asp 9 1L on the left, Leu 96L
`at the back, (125, 146), and the side chain ofTrp l00aH at the top left. In
`
`
`addition, the backbone ofresidues 92-94L form the lower rim of the front
`of the pocket.
`
`CONFORMATIONAL CHANGE One of the mechanisms proposed for
`
`
`
`
`
`
`
`effector function activation involves an allosteric change upon antigen bind­
`
`
`ing ( 1 10). Since crystals of immunoglobulins have large solvent channels
`
`
`
`
`
`and can bind to haptens soaked in through these channels, crystallographic
`
`
`
`investigation offers a direct way for observing conformational changes,
`when they occur.
`When PC binds to McPC603 in the crystal, no significant conformational
`
`
`
`
`
`change occurs in the protein. There is a small movement ofTrp l 04a away
`
`Figure 2 Stereo drawing of the combining
`site of McPC603 with phosphocholine bound. The
`
`
`
`
`lower residues (91 96 and F32) are from the light chain. The remaining residues belong to
`the three complementarity determining
`
`regions of the heavy chain. The phosphocholine has
`group in front with the choline moiety buried in a pocket.
`the phosphate
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1058 Page 4 of 31
`
`

`
`ANTIBODY FUNCTION 9 1
`
`from the pocket, but no other change of any significance. However, there
`
`
`
`
`
`
`
`are several reasons why it cannot be concluded from this observation that
`
`
`
`
`antigen-antibody interaction results in no conformation change:
`
`
`
`
`1. The crystals are grown in a concentrated ammonium sulfate solution,
`
`
`
`
`and it has been observed that in the absence of PC there is a peak at the
`
`
`
`
`phosphate binding site interpreted to be a sulfate ion (126). A conforma­
`
`
`
`
`tional change might have been triggered by the presence of this sulfate ion,
`
`
`
`so that no additional change would be observed upon PC binding. However,
`
`
`in this respect it should be noted that in Fab Newm, (148) no conforma­
`
`
`tional change occurs upon binding of a neutral vitamin Ki derivative.
`
`
`
`
`
`
`2. PC is small and the association constant ('"105 M i) with McPC603
`
`
`
`might be insufficient to trigger a conformational change that could be
`
`
`
`induced by a larger, more tightly binding antigen. The same consideration
`
`
`applies to Fab Newm.
`
`
`
`known binding specif
`
`
`
`
`might not occur with fragments (81).
`Thus, although there is no support from X-ray diffraction for a conforma­
`
`
`
`
`
`
`tional change associated with antigen binding, such a change cannot be
`
`rigorously excluded.
`
`
`
`3. The only two structures at atomic resolution of immunoglobulins with
`
`rule out the possibility that changes that occur with the intact molecule
`
`McPC603; THE CONTACTING RESIDUES AND THE EFFECT OF
`
`
`CHANGES IN THE COMBINING SITE The PC molecule is in direct con­
`
`
`
`
`tact with only a limited number of residues. They include side chains from
`
`
`
`all three heavy chain hypervariable regions and from one (L3) light chain
`
`
`
`
`
`region. The next most distant range of contacts contain many residues that
`
`play a role in positioning the directly contacting residues and changes in
`
`
`
`these might be expected to influence PC binding. An example of such a side
`
`chain is Glu 35H, a residue in the interface between VH and VL that makes
`
`a hydrogen bond with the hydroxyl of Tyr 94L, which is in tum a major
`
`
`
`
`
`
`contacting residue with hapten. A mutant ofSI07, a PC-binding myeloma
`
`
`
`
`
`protein, has been observed that has lost the ability to bind PC and also that
`
`
`fails to agglutinate PC-SRBC (144). Amino acid analysis showed that the
`
`
`
`
`
`mutation results in substitution of an alanine for glutamic acid in position
`
`
`
`
`35H. Although a change of this magnitude is likely to produce a significant
`
`
`
`
`rearrangement of side chains in its vicinity simply because of the difference
`
`
`in volume of the two side chains, the loss of contact with Tyr 94L reduces
`
`
`
`
`
`with hapten. contact on a residue in direct an important constraint
`
`
`Another mutant observed by Cook et al (38) is more puzzling. The
`
`mutant still bound PC, but it bound less well than S 107 to PC coupled to
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1058 Page 5 of 31
`
`

`
`92 DAVIES & METZGER
`
`different carriers, and showed a decrease in affinity for a variety of PC­
`
`
`
`
`
`
`
`carrier conjugates. The only amino acid change observed was that of Asp
`
`
`
`
`� Ala in the fifth position of the heavy chain J region. Changes in the
`
`
`
`carboxy terminal half of V L were not entirely ruled out, but they did not
`
`
`
`
`
`
`appear in a tryptic peptide analysis. The strange aspect of this mutation site
`
`
`
`is that it is spatially well removed from the PC pocket so that it might not
`
`
`
`
`
`have been expected to be involved in antigen binding. Another curious
`
`
`
`
`feature is that diverse carriers coupled to PC were all affected, which implies
`
`
`
`
`
`they all have some contact with this residue, or with a region influenced
`it.
`
`SEQUENCE COMPARISON OF PC-BINDING ANTIBODIES Sequences
`
`
`
`
`and binding data are available for a variety of PC-binding myeloma proteins
`
`
`
`
`and monoclonal antibodies. They have recently been reviewed by Rudikoff
`
`
`
`(145) and are only discussed here in relation to the three-dimensional
`
`structure of McPC603.
`
`The heavy chain The sequences of 19 heavy chains of BALB/c immuno­
`
`
`
`
`
`
`globulins that bind phosphocholine have been analyzed (74, 145). Ten of
`
`
`
`
`these are identical and employ the T15 sequence. The remaining nine differ
`
`
`
`by one to eight residues from the T15 sequence. Gene isolation and analysis
`
`
`
`have revealed that all 19 of the V H regions must have arisen from the single
`
`
`
`germline T15 VH gene segment (39). M 167 is the most divergent protein
`
`with eight VH substitutions. In both M167 and HPCG1 3 the same change
`
`
`
`
`(Thr at position 40) occurs, but all the other substitutions are unique,
`
`occurring in only one protein.
`The PC-contacting residues Tyr33 and Arg52, together with Glu35, are
`
`
`
`
`
`
`
`present in all of these sequences. Similarily, all of the BALB/c sequences
`
`
`
`with the single exception of M 1 67 contain a tryptophan
`at lOOb. There is
`
`
`
`
`
`considerable variation in the D region for these proteins, accompanied by
`
`
`
`only relatively small changes in PC affinity, consistent with the fact that,
`
`
`with the exception of TrplOOb, CDR3 of the heavy chain does not playa
`
`major role in defning the PC pocket.
`
`chains The light chains of PC-binding antibodies can be repre­
`
`
`
`The light
`
`
`sented by the three BALB/c myeloma proteins, T15, M603, and M167 (34,
`
`
`
`
`
`35). These light chains differ considerably in sequence, despite the similarity
`
`
`
`
`of their corresponding heavy chain sequences. However, in the PC-binding
`
`
`
`region, in particular with the contacting residues Tyr94, Pr095, and Leu96,
`
`
`their sequences are the same. They also all employ the same h sequence
`(J5).
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1058 Page 6 of 31
`
`

`
`93
`ANTIBODY FUNCTION
`
`The invariant residues in these light and heavy chains provide strong
`
`
`
`
`
`
`
`
`support for the suggestion that these PC-binding antibodies all have the
`
`
`
`same overall combining site for PC (125), with differences in binding speci­
`
`
`
`ficity being contributed by the amino acid substitutions.
`
`Fab Newm The structure of the Fab of Newm, a human IgG (A) myeloma
`
`
`protein, has been refined to a nominal 2-1 resolution (4, 148). The elbow
`
`
`bend is 137° and the L chain has a seven-residue deletion that includes
`
`residues 55 and 56 of CDR2, and 56 to 62 of the framework region FRIU.
`
`
`
`The combining site is formed by the association of the remaining five
`
`
`hypervariable regions. The principal feature of the site is a shallow groove
`15 X 6 X 6 A deep, bordered by residues
`
`from the Hand L chains. Fab
`
`
`
`Newm binds several haptens at this site with affinity constants ranging from
`
`
`
`
`103 to lOS M I. A derivative of vitamin KI binds with the higher affinity and
`
`
`
`a crystallographic investigation has demonstrated that the menadione ring
`
`
`
`system binds in the shallow groove with the phytyl chain draped along the
`
`
`
`surface, making contact with a number of residues from the light and heavy
`
`
`
`chains (5). The number of contacts provided by the phytyl chain can
`
`
`
`
`
`probably account for the difference in binding between menadione and the
`
`
`vitamin KI derivative (103 vs 1.7 X 105 M l). As noted above, no conforma­
`
`
`tional change is observed upon the hapten binding.
`
`KoL The human IgG A cryoglobulin Kol and its Fab both crystallize and
`
`
`
`
`both structures have been solved, the intact molecule at 3.5 1 and the Fab,
`
`
`at 1 .9 1 resolution
`
`for which the combining site specificity is unknown,
`
`
`(106). The crystals of the intact molecule display an unusual form of disor­
`
`
`
`
`der, described above, that prevents visualization of the Fc.
`
`
`The Fab crystallizes well and has provided a detailed high resolution
`
`structure. The L chain conformation is quite similar to that of Fab Newm,
`
`
`
`except for the presence of the seven additional residues around CDR2.
`
`
`
`
`However, in the H chain, CDR3 is eight residues longer than Newm (and
`
`
`six longer than McPC603), and these additional residues fold into the
`
`combining site and f
`in Newm. This combining site is rich in aromatic side chains, being filled
`
`
`
`
`largely by Trp (47H, 52H, 90L and 108H), Tyr (35H, 35L and 97L), and
`His (59H).
`In both the crystals of the intact molecule and the Fab, the same contact
`
`
`
`
`is made between the hypervariable surface and the hinge segment, Cys221-
`
`Cys230, the light chain C terminus G 1 u212-Ser214, and the residues 1 3 3-
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1058 Page 7 of 31
`
`

`
`94 DAVIES & METZGER
`
`138 and 196-199 of CH1 of a neighboring molecule. This contact involves
`
`
`considerable surface area and it has been estimated that 1314A2 ofthe Kol
`
`
`
`hypervariable surface is excluded from solvent. The contact is tight and
`
`
`
`
`closely packed and involves hydrophobic interactions as well as salt links
`
`
`and hydrogen bonds. Marquart et al (105) suggest that this interaction
`
`
`
`
`could be a prototype antibody-antigen interaction and might be responsible
`for the cryo properties of this molecule.
`
`ANTIBODY COMBINING SITES Data on antibody combining sites have
`
`
`
`been comprehensively reviewed by Givol (76). Here, we only highlight a few
`
`structural topics.
`What common features, if any, will the three-dimensional structures of
`
`
`
`
`the combining sites share? Givol (76) notes that the sizes of these sites are
`
`
`
`
`
`comparable to those of some enzymes. Lysozyme, for example, has a groove
`
`
`
`
`that will accommodate a hexasaccharide, and this is believed to represent
`
`
`about the upper limit in size for this kind of antigen (88, 102). The binding
`
`
`
`of antidextrans can be divided into two classes: end-binders like W3129,
`
`
`
`
`which bind only the nonreducing end of the dextran; and middle-binders
`
`
`like QUPC52, which bind in the middle of the chain to runs of six glucose
`
`
`
`units. It has been suggested that the former type of antibody might have
`
`
`
`
`a pocketlike site, whereas the latter might be more likely to have a lengthy
`groove (33, 89).
`Since three-dimensional structures are known for but two Fabs with
`
`
`
`
`
`known binding specificities, only a limited picture can be obtained from
`
`
`
`direct observation. It is nevertheless suggestive that one of these structures,
`
`
`McPC603, has a pocket for binding PC, whereas the other, Newm, has a
`
`
`shallow groove where the menadione binding site is located. Again, if we
`
`
`
`were to argue from analogy with enzymes we might expect a pocket or
`
`
`groove in most antibodies to provide specificity. However, specificity can
`
`
`
`
`also be produced by complementarity between two interacting surfaces
`
`
`
`
`without necessarily invoking grooves or pockets, and significant contribu­
`
`
`tions to the free energy of interaction can come from exclusion of hydro­
`
`
`phobic groups from contact with water. This kind of specificity, like the
`
`
`
`t protein, can be quite precise interactions between subunits in a multisubuni
`
`
`
`and can be destroyed by single amino acid changes. For instance, not only
`
`
`
`
`is it necessary to maintain complementarity of two interacting surfaces, but
`
`
`
`there also needs to be a suitable juxtaposition of oppositely charged groups
`
`forming salt bridges as well as appropriate disposition of hydrogen bond
`
`
`
`
`donors and receptors. Accordingly, although antibodies specific for small
`
`
`
`antigenic determinants will probably have a groove or pocket, other anti­
`
`
`
`bodies specific for an array of amino acids such as epitopes on the surface
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1058 Page 8 of 31
`
`

`
`ANTIBODY FUNCTION 95
`
`of a protein need not necessarily have these features. The Kol protein
`
`
`
`
`
`
`
`
`
`combining site, although similar in some respects to Newm, has its groove
`
`partially f
`this site interacts strongly with the C-terminal portion of another Fab in
`
`
`
`
`
`
`
`
`what could be a prototype antibody/protein antigen complex.
`
`
`
`The availability of monoclonal antibodies to specif
`
`
`
`
`hybridoma technology should provide a significant increase in the diversity
`
`
`
`
`of antibodies studies by X-ray diffraction. In particular, they offer the
`
`
`
`opportunity to study interactions with ligands that are larger than simple
`
`
`
`
`
`haptens and that could completely fll the combining site. One example is
`
`
`
`
`been crystallized an anti-influenza virus neuraminidase that has already
`
`
`
`
`(37). More of these anti-protein antibodies need to be studied, both alone
`
`
`
`
`and complexed with antigen, to obtain structural comparisons with other
`
`
`
`
`studies (12, 57, 164). Other monoclonal antibodies to polysaccharides such
`
`
`
`as 0' 1 � 6 linked dextran should clarify the mechanism of binding for linear
`
`
`
`
`antigens (90) and the effects of amino acid changes on specificity.
`
`MODEL BUILDING STUDIES OF Fv Since in the last 10 years high reso­
`
`
`
`lution structures have been determined for only three different Fab's, and
`
`
`
`only two (M603 and Newm) have known binding specificities, it appears
`
`
`
`that X-ray crystallography can only provide a small fraction of three­
`
`
`
`
`
`
`dimensional structures for interesting antibody combining sites. There is
`
`
`
`also the difficulty that only some Fab's can be induced to crystallize. An
`
`
`
`
`alternative approach to direct X-ray analysis is to utilize the knowledge
`
`
`
`available from crystallography together with known amino acid sequences
`
`
`
`to construct models of the Fv's of interesting antibodies. The general prob­
`
`
`
`
`
`lem of protein folding, i.e. how a polypeptide chain several hundred amino
`
`acid residues long folds into its fnal globular form, is being extensively
`
`
`
`
`investigated. However, it is most unlikely that it will soon be possible to
`
`predict correctly the f
`ertheless, the forces that contribute to the stability of a protein continue to
`
`
`
`
`
`
`
`
`be stu<;lied and are being refined. There is also an increased appreciation of
`
`
`
`
`the dynamic aspects of protein structure.
`The problem of predicting antibody combining site structures is a special
`
`
`
`
`
`
`
`
`
`case with some features that make it an attractive candidate for investiga­
`
`
`
`tion by molecular modeling. The similarity of variable domain structures
`
`
`
`
`is quite remarkable (4, 120-122), which indicates a strong conservation of
`
`
`
`the three-dimensional structure of the framework part of the variable do­
`
`
`
`mains. Also, the most variable parts of the V domains occur largely at one
`
`
`
`
`
`end of the domain in the hypervariable loops. As a result, a comparison of
`
`
`
`with that of V domains of known structure could, in optimal
`the sequence
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1058 Page 9 of 31
`
`

`
`96 DAVIES & METZGER
`
`cases, directly lead to a preliminary model that could then be refned by a
`
`
`
`
`
`
`
`
`suitable energy minimization technique. The problem becomes more com­
`
`
`
`plicated when there are amino acid insertions and deletions that change the
`
`
`
`length of the hypervariable loops. The new loops could be copied wherever
`
`
`
`
`possible from similar size loops in proteins whose structures have been
`The V L : V L dimers Rei and Au (60) provide
`determined.
`
`an example of very
`
`
`
`
`similar three-dimensional structures associated with loops of the same
`
`
`
`
`length, although there are 18 amino acid differences in each domain. A
`
`
`different point of view would be inferred from the L· chain dimer of Mcg
`
`
`
`(58) where corresponding hypervariable loops do not preserve the local
`of the V domains
`
`twofold symmetry
`
`
`but have different conformations as a
`
`
`
`
`result of interaction with neighboring molecules in the crystal.
`
`
`
`
`The earliest modeling studies involved the 2,4-dinitrophenol-binding
`
`
`mouse myeloma protein MOPC315. The combining site of MOPC315 has
`
`
`
`been discussed (135) relative to the structure of Newm. Padlan et al (124)
`
`
`
`constructed a molecular model for MOPC315 that utilizes the framework
`from other V regions
`
`
`stru9ture ofM603 and the hypervariable loops derived
`
`
`
`
`that have CDR's of similar length whose structure had been determined by
`
`
`
`
`
`X-ray diffraction. An extensive nuclear magnetic resonance investigation of
`
`
`MOPC315 (54,55, 186) and its interaction with ligands has led to a refined
`
`
`
`model that is similar to the original model, but has a different orientation
`
`
`
`for the DNP binding site (89). The crystal structure of the Fv ofMOPC315
`
`
`
`is being investigated and should ultimately provide a basis for evaluating
`Davies & Padlan (40) constructed
`these models (6). Subsequently,
`a model
`
`
`
`
`poly­for the homogeneous rabbit antibody (BS5) to type III pneumococcal
`
`
`saccharide. Potter et al (136) presented a model for the inulin-binding
`& Wu (167) have con­
`
`
`
`myeloma protein EPC109. More recently, Stanford
`
`
`
`structed a backbone model for MOPC325, and Feldmann et al (62) have
`
`
`
`
`proposed models for J539 and included a proposal for the binding of hex­
`at 4.5 A
`
`
`
`asaccharide. The crystal structure for J539 has been determined
`
`
`
`and the atomic resolution structure is under investigation so that it should
`
`soon be possible to test this model.
`None of the models described above has been subjected to any form of
`
`
`
`
`
`
`energy minimization. They may give an approximate general, low esolu­
`
`
`
`
`tion picture of the combining sites particularly where they illustrate some
`
`
`
`
`striking insertion or deletion, as in EPC109. However, they are unlikely to
`
`
`
`
`
`
`be accurate to better than several angstroms for the backbone atoms, and
`
`
`
`
`could be quite incorrect in positioning the amino acid side chains. Until, for
`
`
`
`at least a few cases, they are compared with the results of X-ray diffraction
`
`
`these models should be regarded as being quite hypothetical and should be
`
`
`
`
`treated with caution. In the case of MOPC315, the structure investigation
`
`
`by Padlan et al (124) was consistent with the known chemical data from
`
`'r
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1058 Page 10 of 31
`
`

`
`ANTIBODY FUNCTION 97
`
`affinity labeling and did lead to the discovery of an error in the sequence
`
`
`
`
`determination, but these correlations derive from coarse rather than fne
`
`
`
`
`detail in the model. Since single amino acid changes can produce large
`
`
`
`
`effects on structure, it is perhaps optimistic at this stage to expect to define
`
`
`
`
`
`an antibody site with reasonable precision. Certainly, some powerful form
`
`
`of energy minimization will be necessary to ensure that the models pro­
`
`
`
`duced do at least satisfy the basic requirement of stereo-chemistry. How­
`
`
`
`ever, satisfactory prediction also requires a larger library of known
`
`
`
`
`structures of antibodies to a greater variety of antigens.
`
`THE H: L ASSOCIATION For the combinatorial mechanism for generat­
`
`
`
`
`
`
`ing antibody diversity to be reasonably effective, most light chains should
`
`
`
`
`have the ability to combine with most heavy chains. This requirement has
`
`been examined both in vitro and in vivo.
`kinetics (7, 13, The '}'-L interaction has been shown to obey second-order
`
`
`
`22, 69) and has a high affinity with Ka> 1010 M-i (13). When the competi­
`
`
`
`tive association of autologous and heterologous pairs of chains (Le. pairs
`
`
`
`derived or not from the same myeloma) was examined, it was discovered
`
`
`
`that there was a preferred association with the autologous chain (46, 78,
`
`
`associa­104, 170). When heterologous VI( was added to Fd', no significant
`
`tion took place except in the presence of CI(. This was not true for the
`
`
`
`autologous VI(, which did not require the presence of CI( (96). Isolated
`
`
`
`VI( recombined with the autologous V H, but heterologous pairs did not
`of UV difference
`
`associate by the criterion
`
`spectra (80). It is clear that the
`
`
`
`association of the two complete chains must be significantly aided by the
`
`
`
`
`presence of the constr.nt domain of the light chain. Nevertheless, the prefer­
`
`
`
`
`ential association for the autologous pair observed in these in vitro experi­
`
`ments does imply some form of prior selection. Whether or not any
`
`
`
`
`additional selection is needed other than that for antigen binding per se is
`
`
`
`not clear. Clearly, if the VH and VL do not associate at all or associate very
`
`
`
`
`weakly, then their effectiveness in forming a combining site will be greatly
`reduced.
`The source of the variability in the V H : V L association comes from the
`
`
`
`
`
`
`
`involvement of hypervariable residues in the V H : V L interface, as noted
`
`
`previously (41, 42, 148). Figure 3 shows the residues that make contact
`
`
`across the interface between VH and VL in McPC603 (Y. Satow, D.R.
`
`
`
`
`
`
`Davies, manuscript in preparation). There are many highly conserved resi­
`
`
`
`dues that interact, particularly those in FRII, such as Glu39. However, at
`
`
`
`one end of the interface a variety of interactions can be seen to occur
`
`
`
`
`hypervariable and framework residues and between pairs of hyper­
`between
`
`
`
`
`variable residues. These must contribute to the forces governing the exact
`
`
`positioning of VL relative to VH. That this is not always the same even in
`
`
`
`related pairs has been demonstrated by Marquart et al (106) for the proteins
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1058 Page 11 of 31
`
`

`
`98 DAVIES & METZGER
`
`Kol (IgG A) and Newm (IgGI A) where small differences ('"'-'9°) occur in
`
`
`
`the position of V H relative to V L that are outside the limit of experimental
`error.
`
`
`
`In vivo results from cell fusion experiments demonstrated the existence
`of most of the possible "artif
`most, if not all, H and L chain can randomly recombine to produce new
`
`
`
`
`
`
`
`immunoglobulins that do not interfere with cell viability (114). This appar­
`
`
`
`
`ent contradiction raises the possibility that the preferential association is an
`
`in vitro artifact, but this would appear to be ruled out by the numerous very
`
`
`
`
`carefully controlled experiments. Alternatively, the difference in the affini­
`
`
`
`
`ties between autologous and heterologous pairs of chains may be so small
`
`
`as to be unobservable under cellular conditions.
`
`When a hybrid H/L recombinant is formed, then there is the additional
`
`
`question of how its binding properties will relate to those of the parent
`
`
`molecules. This would appear to depend on how similar the parent chains
`
`
`
`are, and Sher et al (157) observed that with reconsituted H/L hybrids from
`
`H
`
`035
`
`039
`
`R44
`
`L45
`
`W47
`
`Y91
`
`S99
`
`T100
`
`W100a
`
`Y100b
`
`F100c
`
`W103
`
`A105
`
`L
`
`30K
`
`32F
`
`36Y
`
`380
`
`43P
`
`44P
`
`46L
`
`49Y
`
`87Y
`
`890
`
`910
`
`94Y
`
`95P
`
`96L
`
`98F
`
`99G
`
`100A
`Figure 3 The residues that fonn the interface
`between V H and V L in McPC603. The vertical
`bars indicate
`
`the compiementarity deteng residues. The connecting
`
`lines join those resi­
`of 4 A of one another.
`dues that have atoms within a distance
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1058 Page 12 of 31
`
`

`
`ANTIBODY FUNCTION 99
`
`three anti-PC myeloma proteins, those with common idiotypes retained
`
`
`
`
`
`
`
`specificity upon recombination, whereas those without common idiotypes
`
`
`did not. Similarily, Manjula et al (103) showed that all the recombinants
`
`
`
`of anti-galactan myeloma proteins they examined showed anti-galactan
`
`
`
`
`
`activity comparable to the parents. However, as reviewed by Rudikoff(145),
`
`
`
`
`these anti-galactan heavy and light chains are so similar to one another that
`
`
`
`
`it is not surprising they exhibit the same binding properties. Recently,
`
`
`
`binding of all the possible Kranz & Voss (97) examined the fluorescein
`
`
`
`
`recombinants from six monoclonal antibodies to fluorescein and found that
`
`
`
`
`no hybrid has anti-fluorescein activity; the sequence and idiotype of these
`
`antibodies was not analyzed.
`
`The Fe
`The human Fc has been crystallized and the structure has been deter­
`
`
`
`
`
`mined at resolution, Figure 4 (43, 44). The structure differs from that
`
`
`
`of a Fab in that there is no protein/protein contact between the two
`
`
`
`
`Instead, the complex carbohydrate attached to Asn297 occu­
`CH2 domains.
`
`
`
`pies the ihterface region between the two domains with weak interactions
`
`
`
`
`
`between the two carbohydrate chains. The two CH3 domains associate in
`
`
`a manner siinilar to the CHI: CL domains of the Fab. The two CH2 domains
`
`
`
`would seem to be positioned