`Vol. 71, No. 11, pp. 4298-4302, November 1974
`
`The Three-Dimensional Structure of a Phosphorylcholine-Binding Mouse
`Immunoglobulin Fab and the Nature of the Antigen Binding Site
`(x-ray dift'raction/domain structure/hypervariable cavityfhapten binding/antibody diversity)
`
`DAVID M. SEGAL*t, EDUARDO A. PADLAN*, GERSON H. COHEN*, STUART RUDIKOFFt, MICHAEL
`POTTERt, AND DAVID R. DAVIES*
`•Laboratory of Molecular Biology, NIAMDD; t Immunology Branch, NCI; f Laboratory of Cell BioloKY, NCI; National Institutes of
`Health, Bethesda, Maryland 20014
`
`Commun~d by Ewin A. Kabat, August 1~, 1974
`
`The structure of the Fab of McPC 603, a
`ABSTRACT
`mouse myeloma protein with p~osphorylcholine binding
`activity, has been determined to 3.1-A resolution. The
`four domains are found to be structurally similar with a
`well-defined double-layer structure. A large cavity exists
`at one end of the fragment, the walls of which are formed
`exclusively of hypervariable residues. Phosphorylcholine
`binds in this cavity and forms specific interactions with
`several well-defined amino-acid side chains of the protein.
`The hapten is bound asymmetrically and interacts more
`with the heavy chain than with the light chain.
`
`We have been investigating the nature of antibody-antigen
`interactions by means of a crystallographic investigation of
`the Fab fragments of mouse myeloma proteins with known
`binding specificity (1, 2). The testing of myeloma proteins in
`mice for antigen-binding activity has revealed that many of
`them interact by precipitation, agglutination, or complement(cid:173)
`fixation with common natural antigens from the environment
`of the mouse (3, 4). Immunochemical analysis of inyeloma
`protein-antigen interactions has in many cases led to the
`identification of the chemical group (hapten) on the antigen
`molecule.
`Antibodies, induced to dextran, levan, and phosphoryl(cid:173)
`choline in mice, share cross-specific antigenic (idiotypic)
`determinants with myeloma proteins of corresponding specific·
`ity (5, 6; M. Potter and R. Lieberman, unpublished observa(cid:173)
`tions). These findings strongly suggest the close relationship of
`antigen-binding myeloma proteins in mice to natural or in(cid:173)
`duced antibodies. One of these myeloma proteins, from McPC
`603, precipitates with antigens from several pathogenic
`organisms, and has been demonstrated to bind specifically to
`phosphorylcholine (3). We have previously reported crystal(cid:173)
`lization of the Fab fragment of McPC 603 protein (1) and its
`structure at 4.5-A resolution together with the location of its
`phosphorylcholine-binding site (2). Here we present the results
`of a 3.1-A structure determination and the molecular details of
`phosphorylcholine binding. We compare these results with the
`crystallographic results on a human Fab fragment with
`hydroxyvitamin K1 binding properties (7, 8), and with two
`human Bence-Jones proteins (9, 10).
`MATERIALS AND METHODS
`The preparation and crystallization of McPC 603 pepsin
`Fab has been described (1). The crystals used were obtained
`
`Abbreviations: Cul and Vu; CL and VL signify the constant and
`variable domains of the heavy (H) and light (L) chains, re(cid:173)
`spectively, of the Fab. Hl, H2, and H3; Ll, L2, and L3 are the
`first, second, and third hypervariable regions of the H and L
`chains, respectively.
`
`from 42% saturated ammonium sulfate solutions, 0.1 M
`imidazole (pH 7.0). They belong to the space group P6a with
`a = b = 162.5, c = 60.8 A. Heavy atom derivatives were
`prepared by soaking the crystals in solutions containing
`TmCia, K2Pt(CNS)8,
`iodine, and combinations of these
`(E. A. Padlan, D. M. Segal, G. H. Cohen, T. Spande, and D.R.
`Davies, in preparation). Approximately 11,000 reflections
`(mean figure of merit 0.73) were used to calculate a 3.1-A
`electron density map. Details of the crystallographic analysis
`are being published elsewhere. A Kendrew skeletal model of
`the entire Fab has been constructed with the aid of the amino(cid:173)
`acid sequence, which is for the most part known. An optical
`comparator of the type proposed by Richards (11), but with
`the mirror parallel to the electron density sections, was used to
`facilitate the model building. Coordinates have been measured
`for the variable region of the molecule and improved by the
`Diamond model-building program (12). Hapten binding has
`been studied by difference-Fourier techniques using data from
`crystals of McPC 603 Fab soaked in saturating concentrations
`of phosphorylcholine as described (2)' but here extended to
`3.1-A resolution.
`The amino-acid sequence of the variable portion of the
`heavy chain (Va) is completely known (Table 1). Approxi(cid:173)
`mately half the Cal region sequence has been determined
`(Rudikoff, unpublished), and a substantial part of the re(cid:173)
`mainder may be surmised from its expected homology with
`MOPC 315 (13). The VL sequence is mostly known and its
`determination is being actively pursued. The constant portion
`of the light chain (CL) is assumed to be identical with the
`sequence of the C region of other murine K light chains
`(14, 15). Light chain hypervariable regions were assigned
`according to Wu and Kabat (16). In the heavy chain these
`regions were identified by comparison with other heavy chain
`sequences, with deletions and insertions introduced to maxi(cid:173)
`mize homology (17).
`
`RESULTS
`The electron density map is clear enough to permit tracing the
`course of the polypeptide chain with confidence. While many
`of the smaller side groups are lost in the background noise,
`particularly those on the exterior of the molecule, the larger
`aromatic side chains are for the most part clearly visible, and
`serve as markers in fitting the amino-acid sequence to the
`map. Although the four disulfide bridges were located, they do
`not exhibit the highest peak density in each domain.
`
`Overall Structure. The overall appearance of the IgA(K) 603
`Fab is very similar to that previously observed in the human
`
`4298
`
`1 of 5
`
`Celltrion, Inc., Exhibit 1126
`
`
`
`Proc. Nat. Acad. Sci. USA 71 (1974)
`
`Structure of a Phosphorylcholine-Bindiiig Fab
`
`4299
`
`TABLE 1. Variable region sequence of McPC 603*
`
`I
`10
`20
`30
`40
`50
`.'>8a .58b
`v B: E-V-K-L-V-E-S-G-G-G-L-V-Q-P-G-G-S-L-R-W-C-A-T-S-G-F-T-F-S-B-F-Y-M-E-W-V-R-Q-P-P-G-K-R-L-E-W-I-A-A-S-R-B-K-G-B-K-Y ·T-T-
`60
`70
`80
`90
`100
`a b
`110
`Z-Y-S-A-S-V-K-G-R-F-I-V-S-R-B-T-S-Z-S-I-L-Y-L-Q-M-N-A-L-R-A-E-D-T-A-i-Y-Y-C-A-R-N·Y-Y-G-S-T-W-Y-F-D-V-W-G-A-G-T-T-V·T-V
`w
`~~m~m~m w
`m
`1
`10
`V Lt: D-I-V-M-T-Q-S-P-S-S-L-S-V-S-A-G-E-R-V-T-M-S-C-K-S-S-Z-S-L-L-B-S-G-B-Z-K-B-F-L-A-W-Y-Z-(Z )-K-P-G-Z-P-P-K-L-1-Y
`
`* Heavy chain numbering is that previously reported (17), with gaps and insertions introduced to maximize homologies with other
`heavy chain sequences. Light chain numbering is according to Wu and Kabat (16).
`t The first 35 residues of this light chain sequence have previously been reported (25 ).
`
`domain disulfides of V L and CL compared with a distance of
`about 30 A between VH and CHL Similar results have also
`been observed in the human Fab New (7) and in the human
`Meg Bence-Jones dimer (9). A possible explanation of this
`distortion of the symmetry of the McPC 603 fragment may lie
`in the more extensive interaction between the constant and
`variable domains of the heavy chain when compared with
`those of the light chain.
`The close approach of the two heavy domains (VH, CHI)
`appears to be facilitated by the presence at the interface of
`amino acids with small side chains. For instance, involved in
`this contact are the stretch Gly-Gly-Gly (residues 8-10) in SI
`and the stretch Gly-Ala-Gly (residues 109-111) in S9. In addi-
`
`IgG(>.) New Fab (7) and in the human Meg Bence-Jones
`dimer (9). The fragment has the overall dimensions 40 X 50 X
`75 A. It consists of two globular regions, each approximately
`40 X 40 X 50 A. Each globular region in turn is made up of
`two homologous domains; VH:VL in the variable region, and
`CHI: CL in the constant region. These pairs of domains are
`related to one another by an approximate 2-fold axis of sym(cid:173)
`metry. The pseudo 2-fold axes from the two regions are not
`colinear, but make an angle of approximately 135°.
`
`Domain Structure. In both V and C regions the basic domain
`structure consists of straight segments of polypeptide chain
`arranged in two layers and forming a sandwich-like structure.
`The chains in each layer have the frequently observed left(cid:173)
`handed twist.
`Fig. 1 is a schematic representation of the folding for the
`variable and constant domains. Adjacent segments within
`each layer are antiparallel to each other and frequently assume
`a 13-pleated sheet configuration. The interior of the sandwich
`contains principally hydrophobic residues. The intradomain
`disulfide and the invariant tryptophan are in close proximity
`and are located approximately in the middle of the domain.
`The extended segments are linked by bends with varying
`degrees of sharpness. The electron density in many of the
`tight bends can be fitted with 13-bend configurations (I8).
`We have adopted the numbering scheme shown in Fig. I,
`where we have denoted the stretches as SI, S2i S3, ... 89 and
`bends as B12, B23, B34, ... B89 with the bend numbers repre(cid:173)
`senting the stretches connected. Thus, we can think of the
`variable domains as consisting of nine stretches with eight
`bends and the constant domains as having seven stretches
`with six bends. Segments S4 and 85 and bend B45 are missing
`in the constant domains; this additional loop in the variable
`domains contains the second hypervariable regions of the light
`and heavy chains. Segments SI, S2, S6, and S7 form one layer
`of the sandwich and segments S3, 88, and S9 form the other.
`In the variable domains the additional loop occurs at the edge
`of the sandwich and is located approximately in the three(cid:173)
`segment layer. The extended segments in the constant do(cid:173)
`.mains, are on the average, slightly longer than the homologous
`segments in the variable domains.
`
`Quaternary Structure. It was clear even at low resolution
`that the two domains of each chain were oriented approxi(cid:173)
`mately at right angles to each other. In addition, the four
`domains of the fragment can be regarded as being located on
`the vertices of an elongated but distorted tetrahedron (2). The
`long axes of the domains of the light chain make an angle of
`approximately 100° with each other, while the corresponding
`angle for the heavy chain is approximately 80°. This results in
`a center to center distance of about 40 A between the intra-
`
`- - - - - - - - - - - - •
`
`1f3
`
`Bl2
`
`·-------------127
`----------183
`-------161 836
`
`878
`
`----~-----151
`-----~--------190
`-------------211
`(b)
`(a) A schematic representation of the heavy chain
`Fro. 1.
`variable domain. The relatively extended segments are denoted
`by S and are numbered sequentially from the NH, terminus.
`The bends B are numbered according to the segments they
`connect. The hypervariable regions (31-35, 49-66, 99-104b) are
`represented by the perpendicular bars. The segments drawn
`with thick lines belong to one layer of the sandwich structure,
`and the thin segments belong to the other layer. (b) A com(cid:173)
`parable schematic drawing of the CL domain. Note the absence
`of the loop 84, B45, 85, which is only found in the variable
`domains.
`
`2 of 5
`
`Celltrion, Inc., Exhibit 1126
`
`
`
`4300
`
`IDimunology: Segal et al.
`
`Proc. Nat; Acad. Sci. USA 71 (1974)
`
`:i;:10. 2. Stereo dr~wing o~ the a-carbon skeleton of the variable ~egion. Open lines represent the light chain and solid lines the heavy
`chain. The hypervariable residues are represented by the large open circles. The star marks the location of the large hypervariahle cavity.
`
`tion, this heavy chain interdomain interaction involves no
`charged amino-acid side chains.
`It is apparent that the dominant force maintaining the
`structural integrity of the McPC 603 Fab is due to the strong
`lateral interactions between homologous domains <VrVL;
`CHI-Ci.) rather than the much weaker longitudinal interac(cid:173)
`tions between domains of the same chain. The interactions
`between the two constant domains occurs principally between
`the layers containing segments 81, 82, 87, and 86. The in(cid:173)
`teraction between the variable domains is quite different, and
`involves the segments 83, 88, and 89, as well as a major por(cid:173)
`tion of the extra loops 84-85. Additional intervariable domain
`contacts occur between parts of the hypervariable loops, es(cid:173)
`pecially in the regions B23 of the L chain and B89 of the H
`chain.
`The constant region is more compact than the variable
`region, probably as a result of having fewer bulky residues in
`the interface between CL and CHL The interacting layers in
`this region are about 10 A apart; this results in a distance
`between the two S-S bridges of 18 A. In the absence of com(cid:173)
`plete Sequence iruormation for CHl, the interactions across the
`C region interlace cannot be deterinined with confidence.
`In the variable region the interacting segments are 12-15 A
`apart, and this greater separation is. reflected in the distance
`between the intervariable domain disulfides of approximately
`25 A. In spite of our lack of complete light chain sequence,
`many of the interdomain contact residues can be identified.
`They span the complete range of variability. Some, for exam(cid:173)
`ple, Leu 45 and Trp 47, have been found in all heavy chain
`sequences, and Gln 39 has been found in most (17, 19).
`Similarly in the light chain, Tyr 36, which makes contact
`across the interface, is the residue most frequently found in
`this position in kappa chains (19). In addition, many of the
`interacting residues are found in the hypervanable regions.
`Indeed; the large loop produced by the six-residue insertion in
`Ll is in intimate contact with most of the residues in H3.
`Moreover, some residues in L3 interact with residues in both
`hypervariable and nonhypervariable portions of the V H
`domain.
`
`The Antigen-Binding Site. The 4.5-A investigation revealed
`the location of the hapten-binding site (2). The appearance of
`two peaks on the difference electron density map led to the
`
`suggestion that, in the native crystals, the phosphate binding
`site was occupied by a sulfate ion. The present native electron
`density map at 3.1-A resolution shows a large peak, tetra(cid:173)
`hedral in appearance, and making contact with side chains, in
`the phosphate position. The choline site is revealed by a 3.1-A
`dltrerence map between native and phosphorylcholine-soaked
`crystals.
`The site of hapten binding is in a large wedge-shaped
`cavity, which exists at the ainino-terminal end of the molecule
`(Fig. 2). The cavity is approximately 12 -A deep, 15-A wide at
`the mouth, and 20-A long, the walls of which are lined exclu(cid:173)
`sively with hypervariable residues (16). Only five of the
`hypervariable regions contribute to the formation of the
`cavity; these are the LI, L3, Hl, H2, and H3. The portion
`homologous to the fourth hypervariable region (residues 82-
`89), reported in human heavy chains (20), occurs predoini(cid:173)
`nantly in the bend B78 at the COOH-terininal end of the
`variable region and is quite removed from the cavity. L2
`likewise does not form a part of the walls of the cavity, being
`screened from it by the large loop containing LI. The inser(cid:173)
`tions in Ll, H2, and H3 cause these hypervaria.ble loops to
`extend farther out, thus ll\creasmg the cavity depth.
`The phosphorylcholine occupies only a small.part of the
`cavity and is bound asymmetrically (Fig. 3), being closer to
`the H chain than to the L chain. The choline moiety is bound
`in the interior of the cavity with the phosphate group more
`towards the exterior. The phosphate group appears to be
`exclusively bound by the H chain; in particular, specific
`interactio.QS are formed between the phosphate and Tyr a3(Ii)
`and Arg 52(H). The hydroxyl group of Tyr 33(H) is appar(cid:173)
`ently hydrogen bonded to one of the oxygens of the phos(cid:173)
`phate, as is one of the amino groups of the arginine side chain.
`Moreover, the close proximity of the positively charged
`guanidinium group of Arg 52(H) to the negatively charged
`phosphate should produce a large favorable electrostatic
`interaction. There is also in the immediate vicinity of the
`phosphate another positive group, that of Lys 54(H), also of
`the heavy chain which could help neutralize the. negative
`charge of this portion of the hapten. The choline group
`appears to interact with both the L and H chains. The acidic
`side chain of Glu 35(H) is about 5 A away from the positively
`charged nitrogen of the choline. In addition, there are van der
`Waals interactions between the choline and main chain atoms
`
`3 of 5
`
`Celltrion, Inc., Exhibit 1126
`
`
`
`Proc. Nat. Acad. Sci. USA 71 (1974)
`
`Structure of a Phosphorylcholine-Binding Fab
`
`4301
`
`There are two consequences of these insertions in McPC
`603. The first is to provide an extensive hypervariable surface
`with a large cavity. The second consequence of these inser(cid:173)
`tions is to make the cavity quite asymmetric, making the
`antigen-binding site of McPC 603 quite different from the
`symmetric hypervariable cavities of the two human Bence(cid:173)
`Jones dimers (9, 10). By contrast, the site of hydroxy vitamin
`K1 binding in Fab New is relatively flat, with a shallow groove
`between the two domains (8).
`The hapten binding in the McPC 603 is predominantly with
`the heavy chain. This observation is consistent with affiruty
`labeling studies (21), which demonstrated an 8: 1 preference
`for H chaill labeling. In contrast, the methylnaphthoquinone
`ring of hydroxyvitamin K1 intera.cts principally with Ll, L21
`and H3 of the Fab New with only the end of the phytyl chain
`interacting with Hl and H2. In addition to this difference in
`the location of the binding site, there are also differences in the
`nature of the protein-ligand interactions. At pH 7, phos(cid:173)
`phorylcholine can be expected to have two negative charges on
`the phosphate and one positive charge on the quaternary
`nitrogen of the choline. It is not surprising, therefore, that
`strong interactions occur between this hapten and charged
`groups such as Arg 52(H), Lys 54(H), and Glu 35(H). The
`electronegative oxygens of the phosphate groups form strong
`hydrogen bonds with the phenolic hydroxyl of Tyr 33(H) and
`with an amino group of Arg 52(H). Hydroxyvitamin K1 on
`the other hand, cannot be expected to participate in electro(cid:173)
`static interactions with charged groups, and it is not entirely
`surprising that the aromatic methylnaphthoquinone moiety
`forms a close contact with the ring of Tyr 90(L) (8). It should
`be noted that despite the disparity in size between hydroxy(cid:173)
`vitamin K1 and phosphorylcholine, they both bind to their
`respective proteins with roughly equal affinity (8, 1).
`These studies demonstrate the involvement of many amino
`acids from the hypervariable loops of the two chains in
`hapten binding, in spectacular confirmation of the hypothesis
`of Wu and Kabat (16). In the case of McPC 603, in particular,
`the binding studies illustrate the stringent requirements for
`complementarity for the hapten-protein interactions. Changes
`in specificity might be produced in at least three different
`ways. First, there are simple substitutions of amino acids; in
`the case of McPC 603, for example, a substitution of glutamic
`acid for Arg 52(H) would probably not affect the overall
`topology of the binding surface, but should dramatically
`change its binding properties. Second, there are insertions and
`deletions within the hypervariable regions. They can pro(cid:173)
`foundly alter the folding of the hypervariable loops, thus
`producing an effect that is greater than that due to merely
`substituting additional amino-acid side chains into the cavity.
`Insertions and deletions can also alter the shape and extent of
`the hypervariable cavity. Finally, there is the possibility that
`either by amino-acid substitutions or by insertions or deletions
`one may alter the interactions between the two variable
`domains, resulting in a change in their relative positions or
`orientations. Thus, .by altering the relative positions of the
`light and heavy chain hypervariable loops, significant changes
`in specificity can be produced. While present data do not per(cid:173)
`mit a test of this possibility to be made, further comparative
`studies [e.g., with J539, a mouse lgA(K) Fab with galactan
`specificity] should reveal its importance as a mechanism of
`antibody variability.
`Phosphorylcholine, being a small molecule, interacts with
`relatively few amino acids at the binding site. It is interesting
`
`FIG. 3. Stereo drawing of the hypervariable cavity with phos(cid:173)
`phorylcholine bound. The six hypervariable loops are shown to(cid:173)
`gether with the side chains of Tyr 33(H) and Arg 52(H). Note
`the position of L2 which is screened from interaction with the
`hapten by the extended Ll loop. The orientation is approximately
`the same as in Fig. 4 of Amzel tt al. (8).
`
`of residues 102-103 of the H chain and residues 91-94 of the L
`chain. The entire hapten is in close van der Waals contact
`with the ring atoms of Tyr 33(H).
`
`DISCUSSION
`The close similarity between this mouse 603 IgA(K) Fab and
`the human New IgG(X) Fab (7) in both their tertiary and
`quaternary structures, coupled with the similar structures
`observed for human Bence-Jones dimers (9, 10), strongly
`implies the invariance of many structural features in the
`architecture of antibodies. However, the considerable differ(cid:173)
`ences in the overall shai>e, size, and general chemical nature of
`the antigen-binding sites provides a plausible and sufficient
`structural explanation for antibody diversity.
`A comparison of the two variable domains of McPC 603
`shows them also to be very similar except in the hypervariable
`regions. When one domain is fitted to the other by a least
`squares procedure, the root-mean-square distance between the
`alpha carbon atoms of 74 homologous residues is 1.9 A.
`It is interesting that a similar comparison of VL (603) with
`VL(REI) (10) shows that they resemble each other. even more
`closely, with a root-mean-square distance of 1.4 A between 94
`homologous residues in the nonhypervariable regions. This is
`very close agreement in view of the fact that the 603 coordi(cid:173)
`nates have not yet been refined. A detailed comparison is not
`yet possible with the human Fab (7) and the Bence-Jones
`dimer (9), and will have to await further data.
`Such comparisons are meaningless when applied to the
`hypervariable regiOns, because of the numerous insertions and
`deletions that occur in these portions of the molecule. It is
`nevertheless clear that the folding of these regions varies
`greatly from one molecule to the next. Sequence comparisons
`of the light and heavy chains of 603 with other iminuno(cid:173)
`globulins reveal a six-residue insertion in LL This results in a
`large convoluted loop which provides a larger surface for
`interaction with antigen while at the same time masking the
`potential involvement of L2 in binding with hapten.
`In the heavy chain, by contrast, Hl is of normal length,
`while H2 has a two-residue insertion. These two residues are in
`addition to the five by which the heavy chain is usually
`greater than the light in this region (19). The effect of these
`extra residues is to make H2 extend much farther out, provid(cid:173)
`ing a much bigger surface for interaction with antigen. The
`H3 region also has a two-residue insertion.
`
`4 of 5
`
`Celltrion, Inc., Exhibit 1126
`
`
`
`4302
`
`Immunology: Segal et al.
`
`Proc. Nat. Acad. Sci. USA 71 (1974)
`
`to speculate what might be the function of the remainder of
`hypervariable surface of the molecule.
`The most likely explanation is that the true antigenic deter(cid:173)
`minant is much larger than phosphorylcholine and would
`probably fill most of the site.
`Another possible use of the remaining portion of the McPC
`603 hypervariable surface would be to bind a completely
`unrelated, and as yet undefined, antigen. In this regard, either
`phosphorylcholine or any other hypothetical unrelated sub(cid:173)
`stance would be considered an antigen for McPC 603 if it
`elicited antibody production. Indeed, antibody populations
`typically contain members capable of crossreacting and being
`elicited by unrelated antigens (22). Of course, the remainder of
`the potential binding site may not recognize any foreign
`substance and merely be, in this case, vestigial.
`The location of the phosphorylcholine binding site, sur(cid:173)
`rounded as it is by the high walls containing hypervariable
`residues, provides a simple structural basis for the existence of
`two classes of hypervariable region related idiotypic deter(cid:173)
`minants: those which act independently of hapten binding
`(e.g., L2 in 603) and those which are correlated with hapten
`binding. Other non-hypervariable areas in the variable domain
`could also provide sources of idiotypic determinants.
`The results of the 4.5-A study on phosphorylcholine binding
`to McPC 603 (2), supported by the hydroxyvitamin K1
`binding studies on Fab New (8), show very clearly that bind(cid:173)
`ing is unaccompanied by a major conformational change in
`the Fab fragment in the crystal. This result implies that hap(cid:173)
`ten binding in solution will leave the overall configuration of
`the fragment unchanged in agreement with low-angle x-ray
`scattering observations (23). It thus seems likely that the
`structural basis of antibody effector reactions, such as comple(cid:173)
`ment fixation or B cell activation, must occur through a
`mechanism not involving a major rearrangement of the Fab
`conformation.
`It is interesting to note that although the tertiary structures
`of the variable and constant domains are very similar, their
`interaction in pairs is very different. The interface between
`variable domains involves one surface of the domain bilayer
`sandwich, while the constant domain interface involves the
`other side. The relatively strong homology between CL, CHl,
`CH2, and CH3 domains (24) leads us to conclude that the
`quaternary structure of the CH2 and CH3 regions of the Fe
`part of the molecule will closely resemble the structure ob(cid:173)
`served here for the constant region of the Fab.
`
`We thank Dr. Matthew D. Scharff for preserving the McPC
`603 cell line, and Dr. Enid Silverton for allowing us to quote re(cid:173)
`sults prior to publication.
`1. Rudikoff, S., Potter, M., Segal, D. M., Padlan, E. A. &
`Davies, D. R. (1972) Proc. Nat. Acad. Sci. USA 69, 3689-
`3692.
`2. Padlan, E. A., Segal, D. M., Spande, T., Davies, D. R.,
`Rudikoff, S. & Potter, M. (1973) Nature New Biol. 245,
`165-167.
`3. Potter, M. (1972) Physiol. Rev. 52, 631-719.
`4. Potter, M. (1971) Ann. N.Y. Acad. Sci. 190, 306-321.
`5. Blomberg, B., Carlson, D. & Weigert, M. (1972) in 3rd
`(Karger,
`Int. Convocation
`Immunology, Buffalo, N.Y.
`Basel, 1973), pp. 285-293.
`6. Lieberman, R., Potter, M., Mushinski, E. B., Humphrey,
`W., Jr. & Rudikoff, S. (1974) J. Exp. Med. 139, 983-1001.
`7. Poljak, R. J., Amzel, L. M., Avey, H. P., Chen, B. L.,
`Phizackerly, R. P. & Saul, F. (1973) Proc. Nat. Acad. Sci.
`USA 70, 3305-3310.
`8. Amzel, L. M., Poljak, R. J., Saul, F., Varga, J. M. &
`Richards, F. F. (1974) Proc. Nat. Acad. Sci. USA 71, 1427-
`1430.
`9. Schiffer, M., Girling, R. L., Ely, K. R. & Edmundson, A. B.
`(1973) Biochemistry 12, 4620-4631.
`10. Epp, 0., Colman, P., Fehlhammer, H., Bode, W., Schiffer,
`M., Huber, R. & Palm, W. (1974) Eur. J. Biochem. 45, 513-
`524.
`11. Richards, F. M. (1968) J. Mol. Biol. 37, 225-230.
`12. Diamond, R. (1966) Acta Crystallogr. 21, 253-266.
`13. Francis, S. H., Leslie, G. Q., Hood, L. & Eisen, H. N. (1974)
`Proc. Nat. Acad. Sci. USA 71, 112&-1127.
`14. Svasti, J. & Milstein, C. (1972) Biochem. J. 128, 427-444.
`15. Gray, W. R., Dreyer, W. J. & Hood, L. (1967) Science 155,
`465-467.
`16. Wu, T. T. & Kabat, E. A. (1970) J. Exp. Med. 132, 211-250.
`17. Rudikoff, S. & Potter, M. (1974) Biochemistry 13, 4033-4038.
`18. Venkatachalam, C. M. (1968) Biopolymers 6, 1425-1436.
`19. Dayhoff, M. 0., Ed. (1972) Atlas of Protein Sequence and
`(National Biomedical Research Foundation,
`Structure
`Washington, D.C.).
`20. Capra, J. D. & Kehoe, J. M. (1974) Proc. Nat. Acad. Sci.
`USA 71, 845-848.
`21. Cheesebro, B., Hadler, N. & Metzger, H. (1972) ih 3rd Int.
`Convocation Immunology, Buffalo, N. Y. (Karger, Basel,
`1973), pp. 20.5-217.
`22. Rosenstein, R. W., Musson, R. A., Armstrong, M. Y. K.,
`Konigsberg, W. H. & Richards, F. F. (1972) Proc. Nat.
`Acad. Sci. USA 69, 877-881.
`23. Pilz, I., Kratky, 0. & Karush, F. (1974) Eur. J. Biochem.
`41, 91-96.
`24. Edelman, G. M., Cunningham, B. A., Gall, W. E., Gott(cid:173)
`lieb, P. D., Rutishauser, U. & Waxdal, M. J. (1969) Proc.
`Nat. Acad. Sci. USA 63, 78-85.
`25. Barstad, P., Rudikoff, S., Potter, M., Cohn, M., Konigsberg,
`W. & Hood, L. (1974) Science 183, 962-964.
`
`5 of 5
`
`Celltrion, Inc., Exhibit 1126
`
`