J. Mol. Biol. (1986) 190, 593-604 Phosphocholine Binding Immunoglobulin Fab McPC603 An X-ray Diffraction Study at 2.7 A Yoshinori Satowf-, Gerson H. Cohen, Eduardo A. Padlan and David R. Davies Laboratory of Molecular Biology National Institute of Arthritis, Diabetes and Digestive and Kidney Diseases National Institutes of Health, Bethesda, MD 20892, U.X.A. (Received 27 December 1985) The crystal structure of the Fab of McPC603, a phosphocholine-binding mouse myeloma protein, has been refined at, 2.7 .& resolution by a combination of restrained least-squares refinement and molecular modeling. The overall structure remains as previously reported, with an elbow bend angle between the variable and constant modules of 133”. Some adjustments have been made in the structure of the loops as a result of the refinement. The hypervariable loops are all visible in the electron density map with the exception of three residues in the first hypervariable loop of the light chain. A sulfate ion occupies the site of binding of the phosphate moiety of phosphocholine. 1. Introduction Direct information about the three-dimensional structure of the antibody combining site is based on the crystal structures of three Fab species: Kol, a human myeloma protein with unknown specificity (Marquart et al., 1980); New, a human myeloma protein that binds to a vitamin Kl derivative (Amzel et al., 1974; Saul et al., 1978); and McPC603, a mouse plasmacytoma protein specific for phospho- choline (Segal et al., 1974). Although the structures of several monoclonal antibodies with known antigen binding specificities are being investigated (Mariuzza et al., 1983; Silverton et al., 1984; Colman et al., 1981; Gibson et al., 1985; Amit et al., 1985), refined, high-resolution structures from them are not available. The three-dimensional structure for the Fab of McPC603 (IgA,rc) protein that was previously reported (Segal et al., 1974) was an unrefined structure based on 3.1 .& diffractometer data and with only partial sequence information. This work has now been extended through‘the collection of a complete set of 2.7 d diffraction data using oscillation photography, augmented by the knowledge of the complete amino acid sequence and with the help of interactive molecular graphics procedures (Lipscomb et al., 1981; Diamond, 1981). The crystal structure has been refined using restrained least-squares procedures (Hendrickson & t Present address: Photon Factory, National Laboratory for High Energy Physics, Oho-machi, Tsukuba-gum, Ibaraki-ken, 305, Japan. Konnert, 1981). In this paper, we report the results of this investigation. 2. Materials and Methods (a) Crystal structure analysis and preliminary rejinement at 3.1 B resolution Crystals were prepared from concentrated solutions of ammonium sulfate as described (Rudikoff et al., 1972). After preparation, the crystals were transferred into stabilizing solutions consisting of 50% saturated ammonium sulfate (pH 7.0), 0.2 M-imidazole or 0.1 M- sodium cacodylate. Isomorphous heavy-atom derivatives were prepared using TmCl, (American Potash & Chemical Corp.) K,Pt(CNS)6 (K&K Laboratories, Inc.), and KI (Fisher Scientific Co.). The thulium and platinum derivatives were prepared by soaking crystals in solutions containing cacodylate and 30 m&r-TmCl, or 0.4 mM- K,Pt(CPU’S),, respectively, for 3 to 4 weeks. An iodine derivative was prepared by soaking crystals in 50 m;M-KI, 2.5 miv-Chloramine T (Eastman Kodak Co.) for 2 weeks, after which the iodinated crystals were washed with stabilizing solution to remove excess iodine. Double and triple derivatives were prepared using these heavy-atom compounds. The native and iodinated crystals were prepared using imidazole buffer; all other derivatives were prepared in cacodylate buffer. Most of the 3.1 d data were collected using a Picker FACS-I diffractometer (Segal et al., 1974); the TmCl, derivative data set was collected from precession photographs. Co-ordinates for the heavy-atom sites in the thulium and platinum derivatives were obtained from difference Patterson syntheses at 4.5 A resolution (Padlan et al., 1973). Those for the iodine derivative were obtained from a difference Fourier synthesis with phases computed using the first 2 derivatives. Alternating cycles of heavy- 002%2836/86/160190-12 v$O3.00/0 593 0 1986 Academic Press Inc. (London) Ltd.
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1085 Page 1 of 12
`
`

`
`594
`
`Y. Satow eL all
`
`atom refinement and phase calculation (Dickerson et al.;
`1961) were then computed using local versions of the
`programs of Busing et aE. (1962) and Matthews (1966).
`Table 1 shows the refined heavy-atom para,meters. A
`“best” Fourier synthesis
`(Blow & Crick, 1959) was
`computed and a Kendrew skeletal model was fitted to the
`electron density using an optical comparator
`(Richards,
`improved using the computer
`1968). The model was
`graphics system GRIP at the Deparbment of Computer
`Science, University of Pu’orth Carolina (Tsernoglou et al..
`1977) and was then subjected to restrained least-squares
`refinement
`(Konnert, 1976; Hendrickson & Konnert,
`1981) on the TI-ASC
`computer at the U.S. Naval
`Research Laboratory.
`The
`initial
`value
`of
`the
`conventional crystallographic R-factor was 0.41. This was
`reduced to 0.30 after 5 cycles with a single overall
`temperature
`factor and
`tight
`structural
`restraints.
`Further refinement using individual temperature factors
`for the atoms reduced the residual to 0.27. The r.m.s.t
`total shift from the original position was 0.76 8. At this
`point,
`the fit of the model to a 2F,- F, map was
`using BILDER
`examined
`(Diamond,
`1981),
`as
`implemented on a PDP 11/70 under the RSX-11M
`operating system (G. H. Cohen, unpublished results). In
`the second stage of refinement, the protein geometry was
`initially
`less restrained and subsequently
`tightened,
`yielding a final residual of 0.24. The r.m.s. total shift from
`the atomic positions at the start of the second stage of
`refinement was 0.42 8. A sulfate
`ion, t’hat had been
`located in the hapten binding cavity (Padlan et al., 1973;
`included
`throughout
`the
`Segal et al., 1974), was
`refinement.
`The regions corresponding to residues 101 to 108 in the
`heavy chain and residues 31 to 35 in the light chain were
`not clearly defined
`in the original electron density
`function based on the heavy-atom phases. These regions
`remained poorly defined after the preliminary refinement.
`
`(b) Crystal structure analysis and re$nement
`at 2.7 d resolution
`
`A new set of crystals was prepared for the higher
`resolution, 2.7 ip, phase of this study. The imidazole
`buffer employed during crystallization was replaced by
`cacodylate when
`the
`crystals were
`transferred
`to
`stabilizing
`solutions.
`Intensity data
`to 2.7 a were
`collected by rotation photography
`(Arndt & Wonacott;
`1977) with Kodak 1Vo-Screen Medical X-ray films (3 in a
`pack) using Ni-filtered CuKcc radiation from an Elliott
`GX-6 rotating anode X-ray generator operated at 40 kV
`and 40 mA. A Franks double bent mirror system
`(Harrison, 1968) purchased from Brandeis University was
`used to focus the X-ray beam. Data were recorded on an
`Enraf-Nonius Arndt-Wonacott
`rotation camera with a
`nominal
`87 mm
`crystal-to-film
`distance. The data
`consisting of 88 film packs were collected from 26 crystals
`by oscillation about the 2 crystal axes, b and c; with an
`
`t Abbreviations used: r.m.s., root-mean-square; F,,
`observed structure factor amplitude; F,, calculated
`structure factor amplitude; w, weight,; CDR.
`complementarity determining region (Kabat et al.: 1983);
`VL, light chain variable domain; VH, heavy chain
`variable domain; CL, light chain constant domain; CHl,
`first constant domain of heavy chain; m.i.r., multiple
`isomorphous replacement; Ll; L2 and L3, lst, 2nd and
`3rd CDR of the light chain; Hl, H2 and H3, lst, 2nd
`and 3rd CDR of the heavy chain.
`
`Heavy-aiom parameters
`
`Co-ordinates
`
`Heavy-atom
`compound
`
`K,Pt(CNS),
`TmC13
`Iodine-l
`iodine-2
`Iodine-3
`Iodine-4
`
`.F
`
`Y
`
`2
`
`occupancyt
`
`0.3807 0.3936 0.5422
`0.3427
`0.6617
`0.2523
`0.2336
`0.7083
`0.5569
`0.2409
`0.6794
`0.5695
`0.2328
`0.6328
`0.6404
`0.1241
`0.7017
`0.4838
`
`0.0218
`0.0245
`0.0119
`0.0184
`0.0125
`0.0044
`
`Thermal
`facior
`(AZ)
`
`19,6
`34.8
`28.3
`7.5
`6.7
`6.8
`
`is on an arbitrary scale in which I-he
`“r The site occupancy
`average stru&ure amplitude of the native protein is 14.3.
`
`oscillation range of 1.25” and an overlap of 0.25”. The
`time for each exposure was 14 to 21 h.
`The
`films were scanned at 100 pm steps on an
`Optronics P-1000 film scanner. The initial film processing.
`which
`included preliminary refinement of the crystal
`orientations and successive evaluation of the integrated
`intensities, was made on a, PDP 1 l/70 computer using the
`rotation program package writ,ten by G. Cornick & 11. A.
`Navia (unpublished results). Intensities from each pack
`were further processed through intra-film-pack scaling.
`post, refinement. scaling and averaging using programs
`for
`these
`purposes
`specially written
`(Y. Satow.
`unpublished results). Intensities
`in a single pack were
`scaled by refining non-linear response correction factors
`(Matthews et al., 1972) and absorption factors for the film
`base and emulsion, then were corrected for Lorentz and
`polarization
`factors. The measured int’ensities from the
`films were merged; reduced to a unique set and processed
`by a scaling and averaging program, which refines
`relative
`scale and exponential
`fall-off
`factors
`for
`individual
`films. This program
`follows
`closely
`the
`formalism of the established algorithms of Hamilton et al.
`(1965) and Rossmann et al. (1979). Post refinement of the
`crystal orientations, lattice constants and rocking curve
`parameters were done as proposed by Winkler et a,E.
`(1979) and Rossmann et al.
`(1979). The total number
`of 147,706
`intensities,
`including partially
`recorded
`reflections, were finally scaled and symmetry averaged.
`yielding 24,235 unique reflections for the resolution range
`of 10 f% to 2.7 d. The agreement fa,ctor:
`
`h j h
`
`where the intensity Ihj for reflection h was measured ATh
`times, was 0.077 or 0.054
`for the structure
`fa,ctor
`amplitudes Fhj. Within a 3 A sphere, 95% of the data
`were retained; wit,hin the complete 2.7 A sphere, 93% were
`retained. The earlier diffractometer data were not
`included in this final set of 2.7 A film data.
`The initial model for the 2.7 a work was taken from
`the 3.1 ,!% results. It was subjected to a number of cycles
`of least-squares refinement
`(Hendrickson & Konnert.
`1981) with periodic examinat,ion and rebuilding of the
`model using BILDER
`on a VAX 111780 (R. Ladner,
`unpublished results). The sulfate ion located in the 3.1 a
`analysis was not included in the early stages of least-
`squares refinement. With an overall temperature factor,
`the initial value of the R-factor was O”41. After 5 cycles of
`positiona,
`parameter
`refinement with
`an
`overall
`temperature
`factor;
`the value was
`reduced to 0.33.
`Individual temperature factors for the atoms were used in
`t’he succeeding cycles. The program restrains the values of
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1085 Page 2 of 12
`
`Table B
`

`
`Phosphocholine Binding Immunoglobulin Fab McPC603 595 the B-factors so that each is influenced by the B-factors of the atoms to which it is bound as well as the atoms 1 removed along a chain. In the final stages of least-squares refinement, particular care was taken to ensure that the stereochemistry was kept reasonable and that w(lF,I - lF,[)’ was approximately constant over the range of the data used (80 A through 2.7 A). The main-chain stereochemistry was constantly monitored by the program GEOM (G. H. Cohen, unpublished results) and points of significant departure from expected stereochemistry were examined and corrected via interactive computer graphics. In addition to a sulfate ion, 138 water molecules of variable occupancy were identified from examination of AF and W-F, maps and refined with the protein molecule. As sequence data became available, the “working” sequence was updated appropriately. The VL and VH sequences are listed (sequences 12, p. 45 and 1, p. 128, respectively) by Kabat et al. (1983), quoted from Rudikoff et aZ. (1981) and Rudikoff & Potter (1974), with the addition of the tetrapeptide Leu-Glu-Ile-Lys, which occurs at the end of VL (Rudikoff, unpublished results). The CHl sequence given by Auffray et al. (1981), obtained by translation of the nucleotide sequence of cDNA complementary to alpha-chain mRNA from 5558 tumor cells, (sequence 60, p. 175, Kabat et al., 1983) was used. This sequence differs in 4 places from that obtained by Tucker et aZ. (1981) from BALB/c genomic DNA, while they both differ in 2 of these positions from the sequence reported by Robinson $ Appella (1980) for MOPC511, as quoted by Kabat et al. (1983). Examination of the final electron density map at these positions did not permit a clear distinction to be made between these sequences. The sequence of MOPC21 (sequence 23, p. 167, Kabat et al.: 1983 from Svasti & Milstein, 1972) was used for the CL domain of McPC603. Throughout this paper, the amino acid numbering is serial, starting from number 1 with the 1st residue of each chain of the molecule. The correspondence between our numbering scheme and that of Kabat et al. (1983) is presented in Table 2. As noted by Segal et al. (1974), the molecule possesses 2 approximate local dyad symmetry axes, which relate the pair of variable domains and the pair of constant domains to each other. We examined the relationship and the Table 2 Correspondence between the numbering scheme used here and that of Kabat et al. (1983) Light chain Heavy chain This work Kabat et al. This work Kabat et al. l-27 l-27 28-33 27a-27f 34-220 28-214 1-52 1-52 53-55 52a-52c 56-85 53-82 86-88 82a-82c 89-106 833100 107-109 lOOa-100c 110-139 101-130 140-142 133-135 143-163 137-157 164-171 162-169 1722184 171-183 185-200 185-200 201-205 202-206 206-216 208-218 217-222 220-225 similarity of these pairs of domains using the program ALIGN (G. H. Cohen, unpublished results), which iteratively rotates one set of atoms to another set to optimize their fit while preserving the order of the linear sequences of the 2 sets. The program uses the algorithm of Needleman & Wunsch (1970) to identify the structural homology while accounting for insertions and deletions. Interdomain and intermolecular contacts were calculated with the aid of the program CONTAX (E. A. Padlan, unpublished results). Two atoms are defined to be in contact if their co-ordinates lie within the sum of their van der Waals radii plus 1.0 A. Intramolecular main-chain hydrogen bonds were calculated by the program EREF (M. Levitt, personal communicati~on). 3. Results and Discussion Table 3 shows an estimate of the quality of the stereochemical parameters for the final model. It is expressed in terms of the r.m.s. deviations of the various classes of parameters from accepted values (Sielecki et al., 1979). The ~,IJJ plot for the main chain is shown in Figure 1. There are a few residues that have “forbidden” C#J,I+!J values. The quality of the map in these regions together with the low Table 3 Summary of stereochemical criteria (Hendrickson & Konnert, 1981; Sielecki et al., 1979; Cohen et al., 1981) Final model Target (r Interatomic distances (A) 1-2 l-3 l-4 Planarity (A) Chiral volumes (A3) Non-bonded contacts (A) l-4 Other Angles (deg.) a> xs (AM Xl> ‘> x4 Temperature factors (A’) Main chain l-2 Main chain l-3 Side chain l-2 Side chain l-3 0.225 136 0.020 0615 0.040 0620 0.037 0.025 0.027 0920 0.257 0.150 0.24 O.!jO 0.34 O.!jO 27.0 15.0 5.0 5.0 0.5 1.0 0.4 0.7 t 05 0.#7 0..5 0.7 The standard groups dictionary used is specified in T.able 2 of Sielecki et al. (1979). In this Table, F0 refers to the observed structure factor, F, is the calculated structure factor and AF is the quantity llF,l-lF,ll. The target 0 represents the inverse square-root of the weights used for the parameters listed. The values given are the r.m.s. deviations from the respective ideal values. t The weight chosen for the structure factor refinement, the “target u” of AF, was modeled by the function w = (l/d)’ with: d = 40--5OO(sin(tl)/,l-l/6). By this means, w(AF) was approximately constant over all of the data (W representing the weight for a given reflection). Other details of this Table are explained fully by Cohen et al. 1(1981).
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1085 Page 3 of 12
`
`

`
`596 Y. Satow et al. 0 Q Q x Y :: (II Q i ,x m I!fQX c) ‘mo.00 -135.00 X -90.00 -h. 00 p’ i5 HI xx x G 0 0 Q x x Q x X X E IO u5.00 90.00 135.00 I Figure 1. A plot of the dihedral angles at’ each alpha-carbon atom for the refined co-ordinates. Asterisk, Pro; cross, Gly; circle, any other residue. resolution thwarted the assignment’ of more satisfactory geometry. The estimate of error in this model is 0.3 A as determined by the method of Luzzati (1952). The final crystallographic R-factor is 0.225 for 23,737 reflections in the range of 8.0 A to 2.7 A. The diffraction data, co-ordinate data, temperature factors and solvent occupancies for the Fab of McPC603 have been deposited in the Protein Data Bank at Brookhaven National Laboratory (Bernstein et al., 1977). (a) General structure of the Fab Several changes were observed in the molecule as a consequence of this refinement, although the overall structure remained similar to that described earlier (Segal et al., 1974). Most of these changes are to be found in the details of the loops, in particular Ll and H3. The region of Ll is poorly defined in the electron density maps, even with the higher- resolution data. We have now rebuilt the neighboring H3, whose density has become less ambiguous and, concurrently, have altered the conformation of Ll. In Figure 2! Ll may be noted to protrude from the general surface of the molecule. The direction of this protrusion is reasonably correct but the electron density is too weak to define the positions of the three outermost amino acid residues. Three other residues have tentative placements due to problems in maintaining proper stereo- chemistry while fitting the electron density: Gln162 and Asn163 (residues 156 and 157 in the numbering used by Kabat et al., 1983) of CL and Glu202 (residue 203, Kabat et al., 1983) of CHl. In these cases, we permitted the stereochemical constraints to dictate the final placement. Also) several disordered side-chains were observed within the entire molecule. In all t,hese cases: the configuration corresponding to the stronger density was chosen for the model. The 442 residues of &PC603 Fab include 25 proline residues. Of these, five are cis-proline: residues 8 and 101 in VL, 147 in CL, and 143 and I55 in CHI. The assignment of the configuration of all five cis-proline residues was unambiguous. The structurally homologous Pro& and Pro95 of Rei (Huber & Steigemann, 1974), and Pro147 (CL) and Pro155 (CHl) of Kol (Marquart et al., 1980), also have been found in the cis conformation. Pro143 (CHl ) of McPC603 has no counterpart in previously reported antibody structures. (b) Crystal pa&@ The McPC603 Fab crystallizes in the space group P6, (Rudikoff et al., 1972), with a = 162.53 8: c = 60.72 A. The molecules are situated in clusters
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1085 Page 4 of 12
`
`

`
`Phosphocholine Binding Immunoglobulin Fab McPC603 597 Figure 2. A stereo drawing of the alpha-carbon skeleton of McPC603. Continuous lines denote the heavy chain. The filled circles show the complementarity-determining residues (Kabat et al.; 1983). of three at each of the crystallographic 3-fold axes of the unit cell (B’ig. 3). Each cluster is related to neighboring clusters via the 2-fold screw situated midway between the 3-fold axes (Brigs 3 and 4). The clusters of three are maintained principally by a set of hydrogen-bond and van der Waals contacts between neighboring molecules. Residues 14 to 20 and 71 of VL interact with residues 1, 26, 2’7, 100, 102, 104, 105 108 and 110 of the VH Iof the neighboring molecule while residues 18, 66, 67, 69, Figure 3. A projection of the alpha-carbon skeletons of 4 unit cells and the ab plane, illustrating the large channels that run parallel to the c axis through the crystal. The heavy chains have been drawn bolder in the 3 molecules that are clustered about a crystallographic 3-fold axis of the lower left-hand cell. The 3-fold axes are indicated by the symbol A: Z-fold screw axes are located midway between each adjacent pair of 3-fold axes; a 6, axis is located at each corner of each cell.
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1085 Page 5 of 12
`
`

`
`598 Y. Satow et al. Figure 4. The projected structure viewed perpendicular to the ac plane, illustrating the effect of the dyad screw axis. The plane of projection is along the long diagonal of the unit cell (see Fig. 2). It contains the two 3-fold axes and the Z-fold screw midway between them. For clarity, only the 4 molecules closest to the plane have been drawn. The heavy line corresponds to the heavy chain of the molecule. 73 and 82 of the same VL meet the VL of the neighboring molecule at residues 35, 55 and 61 to 63. This pattern repeats three times around the axis. Other intermolecular contacts are found between VW of one molecule and CL and CHl of a second molecule related to the first by the 2-fold screw axis. Several additional contacts are found between VL and CHl of a neighbor via the x unit cell translation. It has been noted (Padlan et al., 1973) that roughly 70% of the cell volume is occupied by solvent, thus permitting the diffusion of hapten to the molecule in the crystal. The long axis of the molecule makes an angle of approximately 50” with the xy plane of the unit cell (Fig. 4). (c) Intramolecular pseudosymmetry Analysis of the approximate 2-fold axes relating pairs of domains yielded the following results: for the VLjVH relationship, we find a 173” rotation with a r.m.s. deviation of 1.3 A (3.5 A for poorest agreement) between 385 matched pairs of main- chain atoms out of 456 atoms. A number of the residues of the hypervariable loops align quite well, particularly at t,he beginning and end of each loop. The departure from a more exact 2-fold symmetry is probably due to the fact that the interface residues can be described as fitting a cylinder containing four strands from the light chain and five strands from the heavy chain (NTovotny et al., 1983). This asymmetry of the VHjVL interaction distinguishes it from most of the VLjVL inter- actions observed in Bence-Jones dimers which, in many cases, display exact 2-fold symmetry. The dyad symmetry of the CLjCHl pair is not as close. The best alignment for the constant modules uses 358 of the possible 400 pairs of main-chain atoms to achieve a r.m.s. deviation of 2.0 A (5.1 A for poorest agreement) between the pairs of matched atoms. The best fit between CL and CHI involves a rotation of 169” together with a translation of 2.6 A along the rotation axis. A comparison of these two axes of rotation yields an elbow bend of 133”, essentially unchanged from the earlier report, (Segal et al., 1974). (d) Bntramotecular contacts As noted elsewhere (reviewed by Davies & Metzger, 1983; Amzel R: Poljak, I979), VLjVH contacts involve hypervariable residues as well as framework residues. The boxed regions of Table 4 indicate hypervariable t,o hypervariable residue interactions, while other regions of the Table involve framework residues. Nearly half (46 in 105) of the interactions between VL and VH involve only hypervariable residues, with most of t,hese being located at the upper end of the interface (Fig. 2) in the vicinity of the combining site a,nd with the framework interactions at the other end. A number of good hydrogen-bonded contacts occur: notably two between the highly conserved residues Gln44L and Gln39H, between Gln39H and Tyr93L, between the CDR residues Asp97L and AsnlOlH, and between the hydroxyl group of TyrlOOL and Glu35H. The latter interact’ion appears to be important in maintaining the integrity of the phosphocholine binding pocket (Rudikoff et al., 1981) as it has been observed (Rudikoff et al.? 1982) that a mutation of Glu35H to Ala results in a loss of phosphocholine binding ability. Other interdomain, intramolecular contacts are presented in Ta’bles 5 to 7. There are no interactions
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1085 Page 6 of 12
`
`

`
`Phosphocholine Binding Immunoglobulin Fab McPC603 599 Table 4 Contacts between residues of VL and VH Y33 E35 Q39 R44 L45 W47 E i::1 : : : : Y42 Q44 4 P49 . P50 1 : L52 Y55 G56 rl : 2 : : : 4 F104 1 3 G105 1 A106 . 6 A50 E61 Y97 NlOl Y103 S105 T106 W107 Y108 F109 W112 G113 All4 2 8 : : : 10 2 2 Residues listed in the left column are from VL; the residues listed above the columns are from VH. The numbers in the Table correspond to the number of pairs of atoms from 2 residues that are within potential van der Waals contact distance, as calculated by the program CONTAX (see Materials and Methods, section (b)). A dot indicates no contacts. The boxed regions delineate possible hypervariable/hypervariable interactions. involving VL with CHl or VH with CL. Tables 6 and 7 show that the number of interdomain intrachain contacts, i.e. VL with CL or VH with CHl, is small. (e) Domain structure The structures of the four domains of McPC603 are illustrated by Figure 5, which shows also the location of the hydrogen bonds between the main- chain amide groups. The general tertiary structure of the domains and the distribution of the hydrogen bonds are very like those of other antibody classes and subgroups. Thus, the a heavy-chain domains of McPC603 Fab resemble those of the y chain domains of the human proteins New and Kol, and the IC chain domains of McPC603 resemble the ;1 chain domains of New, Kol and Meg, as well as the IC chains of various hu.man VL dimers (Padlan & Davies, 1975; Padlan, 1977). A significant difference from the y chain structure occurs in CHl, where there is an additional disulfide bond formed between residues 198 and 222 (residues 198 anld 225, in Kabat et aE., 1983). The Cys222 residue should therefore be regarded as forming the end of CHl Table 5 Contacts between residues of CHl and CL Y131 P132 L133 T134 L135 P136 1145 V173 F175 P176 Al78 S188 Q196 1123 2 F124 i 4 5. 5 : : : : . . S127 i El29 3. Q130 14 s133 1 v139 2 F141 2 3 : : 1 2 N143 i : 4 L166 5168 i i 1 W169 T170 i 2 S180 1 : : Ml81 3 S182 5 i : R217 i : : . Residues listed in the left columns are from CHI; the residues listed above the columns are from CL. The numbers are as in Table 4.
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1085 Page 7 of 12
`
`

`
`600 Y. Satow et al. Table 5 Table 7 Contacts between residues on VL and CL Contacts between residues qf VJU and Cfib Y146 D171 Q172 Sl74 K175 D176 S177 Y179 El23 5124 Al25 F154 P155 G157 T158 IX11 S212 P46 2 A86 ; i : : : L89 K109 ; 4 El11 3 : : : : i 1112 2 : 5 1 K113 1 R114 6 s j : G9 2 Cl0 i Lll ; 2 ; 2 ; : : T117 1 3 T119 3 s121 4 The residues listed in the left column are from VL; those listed at the column heads are from CL. The numbers are as in Table 4. The residues listed in the left column are from VH; those listed at the column heads are from CHl. The numbers are aa in Table 4. Figure 5. Schematic drawings of the (a) VH, (b) VL; (c) CHl and (d) CL domains illustrating the hydrogen bonds between the main-chain amide residues. The circles marking hypervariable residues in VH and VL are drawn with a double line.
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1085 Page 8 of 12
`
`

`
`Phosphocholine Binding Immunoglobulin Fab McPC603 601 6-O 8:0 6.0 EVKLVESG 0 IVMTOSP ;FTFsD OSL LN! SRNKGNKYTTEYS AS”KG”Fl”6ROTSOS~LYLoMN, _ XNOKNFL~WYOMcP6OP~Ktt IYGA ~TFJESGVP ~II~T~~GSG TDFTLIISS LVOPGGSLRCSCATS( SVSAfEAV~MCKSS 10 20 30 40 50 60 70 80 (a) 8-O 7.0 6.0 5.0 4.0 3uo 2.0 l,O - -----___ - - I ESARNPTLYPLTLPPALS SDPVI IGCL~HDYFPSGTM~VTWG KSGKOI TTVNFPPALASGG RVTMSN$TLPAVECP EGESVKCSVOIJIS NP VOELLI~NC ADAAFVSI FPPSSE~LTSGGASVVCFLNNFYPK LILNVKWKI DGS~AO NGVLNSW~DODSKDSTVSMSSTLTLTK~EYERH NSYTCEATN KTSTSPIVKSFNRNEC 120 130 140 150 160 170 180 190 200 210 220 (b) Figure 6. The separation (in d) of corresponding alpha-carbon atoms of the superimposed domains of (a) VH and VL, and (b) CHl and CL. The upper sequence is that of the heavy chain and the lower sequence is of the light chain. Every 10th letter in each sequence is underlined. The numbers below the sequences refer to the residue number in th(e light chain. Where the structural alignment has matched identical residue types, the corresponding letters are shown in bold face. Caps occur when there are no corresponding residues for comparison. The horizontal broken lines indicate the r.m.s. separation.
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1085 Page 9 of 12
`
`

`
`602 Y. Xatow et al. rather than the beginning of the hinge, consistent with the gene structure observed by Tucker et al. (1981). The structures of the light and heavy chain domains of MePC603 are compared in Figure 6(a) and (b). The sequence identity derived from the structural alignment is 25% for the variable pair and 22% for the constant pair, excluding gaps. In the variable domains, the framework residues superimpose rather well, except for the region from residues 60 to 74. Large differences are observed in the CDRs, but, these are frequently the result of differences in the lengths of the hypervariable loops. In the constant domains, the differences between the light and heavy chains are comparable to those found between the variable domains. (f) Bound carbohydrate Robinson & Appella (1979) noted the presence of carbohydrate attached to Asn155 (CHl) during their sequence analysis of the heavy chain of MOPC 47A. In electron density maps of McPC603 there is suggestive density in the vicinity of the homologous Asnl60 that indicates the possibility of carbo- hydrate here as well. This density is, however, insufficiently defined to permit fitting of any carbohydrate chain. It appears, therefore, that carbohydrate does not occupy a fixed position on the surface of the molecule. The lack of any nearby contacts from neighboring domains could also contribute to the delocalization of the carbohydrate moiety. An analogous poorly defined trace of density is found in the electron density map of the Fab of 5539 (Suh et aZ., unpublished results). (g) The sulfate ion in the combining site Figure 7 illustrates the combining site of McPC603 as seen in these crystals. In addition to several water molecules, the m.i.r. density map contains a large peak that has been assigned to a sulfate ion (Padlan et al., 1973). This occupies the same location as the phosphate group of phospho- choline when the latter binds to the site, and its presence is presumably due to the high concentra- tion of ammonium sulfate (about 2 M) in the crystal. This sulfate ion is in contact with a number of atoms that interact with it through salt bridges and hydrogen bonds. These include the int,eractions proposed for the binding of the phosphate of phosphocholine; namely, Tyr33 and Arg52 of the heavy chain (Segal et al., 1974). In addition, there is a water molecule within hydrogen-bonding distance of another of the sulfate oxygen atoms. This water is part of a network of hydrogen bonds involving another water molecule, TyrlOOL, Glu35H, AsnlOlH, Asp97L and Trpl07H as illustrated in Figure 7. There are, therefore, a total of four hydrogen bonds to three of the sulfate oxygen atoms which, combined with the charge attraction to Arg52, could provide sufficient attractive force to keep the sulfate in this position. The binding Figure 7. Part of the combining site residues of McPC603, showing the key residues that, surround the phosphocholine binding site. The sulfate ion that occupies the place of the phosphate group of phosphocholine is indicated by thick lines, and the 2 larger isolated circles are water molecules that partly fill the choline binding site. Residues F38 and D97 to Llo2 are from the light chain, the remaining residues being from the heavy chain. Potential hydrogen bonds are illustrated with thin cont,inuous lines. constant of the sulfate cannot be very large, because it is displaced in the crystal by phospho- choline, which itself does not bind very strongly. The association constant for phosphocholine in the crystal is 4 X 104 mol-1 ver8u8 2 X lo* mol-’ in an aqueous environment without ammonium sulfate (Rudikoff et al., 1972). Assuming the sulfate eoncent’ration in the crystal to be 2 M, and that the difference in the binding constants for phospho- choline is due only to the presence of the sulfate, these numbers lead to a value of