J. Mol. Biol. (1992) 224, 487-499 Antibody Framework Residues Affecting the Conformation of the Hypervariable Loops Jefferson Foote and Greg Winter MRC Laboratory of Molecular Biology Hills Road, Cambridge CB2 Z&H, England (Received 17 September 1991; accepted 25 November 1991) Rodent monoclonal antibodies have been “humanized” or “reshaped” for therapy by transplanting the antigen-binding loops from their variable domains onto the /?-sheet framework regions of human antibodies. However, additional substitutions in the human framework regions are sometimes required for high a&nity antigen binding. Here we describe antigen binding by a reshaped antibody derived from the mouse anti-lysozyme antibody D1.3, and several variants in which point mutations had been introduced into framework positions to improve its afinity. The affinities were determined from the relaxation kinetics of reactant mixtures using quenching of fluorescence that occurs upon formation of the antibody-antigen complex. The dissociation constant of lysozyme ranged from 3.7 nM (for D1.3) to 260 nM. Measurement of antibody-antigen association kinetics using stopped-flow showed that D1.3 and most of the reshaped antibodies had bimolecular rate constants of 1.4 x lo6 s-l M-l, indicating that differences in equilibrium constant were predominantly due to different rates of dissociation of lysozyme from immune complexes. Mutations in a triad of heavy chain residues, 27, 29 and 71, contributed 99 kcal/mol in antigen binding free energy, and a Phe to Tyr substitution of light chain residue 71 contributed an additional 08 kcal/mol. The combined effect of all these mutations brought the affinity of the reshaped antibody to within a factor of 4 of D1.3. All of these substitutions were in the B-sheet framework closely underlying the complementarity- determining regions, and do not participate in a direct interaction with antigen. The informed selection of residues in such positions may prove essential for the success of loop transplants in antibodies. Variation of these sites may also have a role in shaping the diversity of structures found in the primary repertoire, and in afinity maturation. Keywords: humanized antibody; kinetics; site-directed mutagenesis; hypervariable loop; lysozyme 1. Introduction body, by transplanting these loops (Jones et al., 1986; Verhoeyen et al., 1988; Riechmann et al., The antibody is a Y-shaped molecule, in which 1988). The clinical success of one such “reshaped” the variable (V) domains forming the tips of the human anti-lymphocyte antibody, CAMPATH- arms bind to antigen and those forming the stem (Riechmann et al., 1988) in treating B-cell (C-domains) are responsible for triggering effector lymphoma (Hale et al., 1988) and vasculitis functions that eliminate antigen. X-ray diffraction (Mathieson et al., 1990)) has prompted the reshaping studies of crystalline antibody fragments reveal the of antibodies directed against the interleukin-2 V and C-domains as two layers of /?-pleated sheet, receptor (Queen et al., 1989), CD4 (Gorman et d., with loops connecting the ends of the b-strands 1991) respiratory syncytial virus (Tempest et al., (Poljak et al., 1973; Schiffer et al., 1973; Segal et al, 1991), herpes simplex virus (Co et al., 1991), human 1974). In the heavy and light chain V-domains the immunodeficiency virus (Maeda et al., 1991) and loops at one end of the sheet are hypervariable in epidermal growth factor receptor (Kettleborough et sequence (Wu & Kabat, 1970; Kabat & Wu, 1971), al., 1991). and form the antigen-binding site. The P-sheet However, reshaping requires that the rodent and provides a scaffold for mounting a diversity of human framework regions are structurally loops, and indeed the antigen binding site can be conserved, both in the orientation of the two transplanted from a rodent antibody to a human b-sheets of each domain and in the packing of the antibody, thereby “humanizing” the rodent anti- heavy and light chain V-domains together; that the 487 0022-2S36/92/060487-13 $03.00/O 0 1992 Academic Press Limited
`
`Lassen - Exhibit 1033, p. 1
`
`

`

`488 J. Foote and b. Winter hypervariable loops make the majority of contacts with antigen; and that the loops are supported in a similar way by the underlying /?-sheet framework. Although these are likely to be true for some anti- bodies, the restitution of key contacts between loops and framework has proved necessary in others, and has been assisted by molecular modelling (Riech- mann et al., 1988; Tempest et al., 1991) and system- atic matching of rodent and human framework regions to minimize differences in primary sequences (Queen et al., 1989; Gorman et al., 1991; Maeda et al., 1991). As a model we have reshaped a human anti- body based on the hypervariable regions of mouse antibody D1.3 (Amit et al., 1986; Bhat et al., 1990) and the framework regions of human Vlc(1) family (Kabat et aE., 1987) and the myeloma protein NEW (Poljak et al., 1973). The parent antibody structures have all been solved crystallographically. The Bence-Jones protein REI is in the Vrc(1) family (Epp et al., 1974), and the D1.3 antibody was solved alone and in complex with the antigen lysozyme (Amit et al., 1986; Bhat et al., 1990). We achieved large enhancement of antigen affinity by substitu- tion of several framework residues of a group that may exert a determining influence on the conforma- tion of the CDRst. 2. Antibody Design The original distinction between hypervariable regions and framework residues (Wu & Kabat, 1970) had its basis in homologies between the primary sequences of immunoglobulins known at that time. However, as X-ray crystallographic structures became available, it became apparent that residues in the hypervariable regions formed the apical loops connecting the B-strands of the immunoglobulin fold, but could also extend part way along the P-strands themselves (Poljak et aE., 1973; Schiffer et al., 1973; Segal et al., 1974). Indeed structural analyses have invoked fewer, and in some cases different, residue positions as CDRs (for example, see Chothia & Lesk, 1987) In X-ray crys- tallographic structures of antigen-antibody complexes, all the CDRs do not necessarily make contact to antigen (Tulip et al., 1989) and in parti- cular the C-terminal portion of VH-CDR2 has never been shown to interact directly with antigen. We based our designs on Kabat (Kabat et al., 1987). The construction of the reshaped heavy chain has been described (Verhoeyen et aZ., 1988). The framework amino acid sequences chosen for the light chain CDR acceptor were designed de novo. They are a consensus of human IC subgroup I sequences (Kabat et al., 1987) and closely related (but not identical) to the sequence of the myeloma protein REI (Palm & Hilschmann, 1973). The framework sequences of the first reshaped antibody t Abbreviations used: CDR, complementarity- determining region; ELISA, enzyme-linked immunosorbent assay; PBS, phosphate-buffered saline; Keq, equilibrium dissociation constant; kb, IO3 bases or base-pairs. differ from DI.3 at 18 positions in the light chain and 30 in the heavy chain. The location of these residues is shown in the a-carbon traces of the variable domains of the D1.3 structure in Figure 1 (a). The differences are dispersed over the entire variable region of the antibody: all eight framework segments have multiple replacements. The most numerous, half of the total, occur in framework 3 (between CDR2 and CDRS) of both the light and heavy chain. As expected, few differ- ences are seen at the VH-Vk- framework interface, as this region is generally conserved. Two-thirds of the framework differences are located on the molecule’s surface and are unlikely to affect antigen binding. There are also differences in buried residues, particularly at the interfaces with the hypervariable regions. We considered retaining these buried framework residues from mouse D1.3 in the reshaped antibody (Padlan, 1991), but even buried residues can form the critical element of a T-cell epitope if presented as a denatured peptide by a class II MHC molecule (Allen et al., 1985). This could exacerbate any humoral response to the native antibody, for example an anti-idiotype reac- tion, and we therefore used the entire human frame- work regions, whether the residues were buried or not. To probe the role of specific framework residues in supporting the CDR conformations, we introduced several additional framework substitutions into the reshaped antibody (VL: Phe71 to Tyr; VH Ser27 to Phe, Thr28 Phe29 Ser30 to Ser Leu Thr, and Lys71 to Val; see Fig. l(b)). Residue 71 of the light chain lies in a loop connecting b-strands and the tyrosine ring of D1.3 protrudes inward and is sandwiched between this loop and VL-CDRl loop. The phenolic oxygen appears t.o be important: it hydrogen-bonds to the amide nitrogen and carbonyl of Gly68 and the back- bone amide nitrogen of Asn31, thus forming a bridge between t,he two loops. The neighbouring residue Tyr32 is in intimate contact with lysozyme residue Gln121, previously identified as a central feature of the D1.3 recognition site (Amit et al., 1986). Residue Phe71 was therefore changed to Tyr. Residues 27 to 30 of the heavy chain are part of a structural loop which includes VH-CDRl (Kabat et al., 1987). Indeed they are included within a strue- ture-based hypervariable loop that comprises residues 26 to 32 (Chothia & Lesk, 1987) compared with VH-CDRl that comprises 31 to 35 (Kabat et al., 1987). These residues appear to make important interactions with CDRl and CDRS. For example, in. the D1.3 structure, Leu29 packs against Phe27 and Lys71 (see below), and Phe27 packs against residues 31 to 34. In turn Gly31 and Tyr32 of VH-CDRI make contact with Lysl6 of lysozyme. Residue Ser27 would be expected to create a cavity, and was therefore changed to Phe as in D1.3 and Riechman et aE. (1988). Residues Thr28, Phe 29, Ser30 were changed en bloc to Ser, Leu, Thr. Residue 71 of the heavy chain may fix the relative dispositions of CDRl and CDRS, according to
`
`Lassen - Exhibit 1033, p. 2
`
`

`

`Residues Affecting Hypervariable Loops 489 Fig white differs frame ;ure 1. FabD1.3 complex with lysozyme. The cc-carbon trace of the FabD1.3 (Fischmann et al., 1991) is marked (framework region) and red (CDRs), and the a-carbon trace of lysozyme in blue. Residues corresponding to mces between framework residues of mouse (D1.3) and reshaped antibody, (b) point mutations introduced into I :work of the reshaped antibody and (c) the Vernier zone, are highlighted in green. whether there is bulky side-chain (Lys or Arg), or a smaller side-chain (Val, Ala) present (Tramontano et al., 1990). We therefore changed Va171 to Lys as in D1.3. 3. Methods (a) Construction of reshaped light chain variable gene and Phe’ll + Tyr mutant A myeloma expression vector used to produce engin- eered immunoglobulin heavy chains (Neuberger, 1983; Neuberger et al., 1985) was adapted for light chain expression. This entailed synthesis and cloning the reshaped light chain V-gene and its introduction into the vector M13-HuVNP (Jones et al., 1986) to replace the heavy chain variable gene. A set of oligonucleotides was designed to encode the reshaped light chain V-gene with codon usage of mouse immunoglobulin sequences. However, the sequence encoding residues beyond number 96 of the mature pro- tein was taken directly from the human Jl segment (Hieter et al., 1982) including 30 nucleotides 3’ to the
`
`Lassen - Exhibit 1033, p. 3
`
`

`

`490 J. Foote and Q. Winter VL-HuLysll Hiildrn matamer B RNA slan pssl~ RNA start ~TATG,~*~~CCT~T~AAT~TA~AT~~TAAATATAGGTTT~TCTATACCACAA*~*GA*,~*ACAT~AGATCA~A~TT~T~~T~T*C*~TTA 20 60 so $0 03 ,w MGWSC,,LFLvAJ*r CTGAGCAC~,~AGGACCTCACCATGGGAT~~AGCTGTATCATCCTCTTCTTGGTAGCAACAGCTACAG~~AA~~G~CTCACAGTAGCAGGCTTGAGGTCTG 120 140 150 lM r 110 180 1P xc 0MT0SPSSLS.4 , ,‘fr T T T ‘ A 010 V P : R F S 0 S 0 S G T D Y T F r I S S L 0 GATCTACTACACCACCACCCTGGCTGACGGTG~GCCAAGCAGATTCAGCGGTAGCGGTA~CGGTACCG~,~TACACCTT~~CCATCAGC~~CCTCCAGC~~ 170 424 4% 4 CD;3 w E 0 ATYYcI&+HFWSTPRT~ G’:G F T K Y E I I( ‘a GAGGACA:~,~CCACCTACTACTGCCA~C~~TTCTGGA~CACCCCAA~~~~~TTCGGCCAAG~GACCAA~~TG~AAATCAA~C~TGA~?~~AATT~AAACT 520 w Lw 5M 630 BornHI TTGCTTCCTCAGTTi m 6eJ VH-HIJL~s~ 1 Hindm o*amw _octamer RNA start ~TAT~A*T~CCTGCTCATGAATATGCAAATCCTCTGAA~CTACATGGT~~AAT*T*GGT,~~GT~TATAC~*~*AACAGA~AAACATGA~~ II)* RNA%i M 30 49 TCACAGTT~~~CTCTACAGTTACTGAGCA~~CA~GACCTCACCA~~G~A~~A~CT~TA~CA~CC~CT~~,~~~~~A~~AA~A~~TA~A~~TAA~G~~CTCA 120 14a 160 la, , 180 200 VHSi)VOI CAGTAGCA~~CTTQAGGTCTOGACATATA~ATGGGTOACAATGACATC~~CTTTGCCTTTCTCTCCAC~,~~TGTCCAC~~CCAGGTCC~~CTGC~GC:OA 220 240 iao Jo0 m SL) CDR 1 e S f L GC~~TC’~A~,~TCT~G:BA~AC~TA~CC~~~~CC~~A~CC~~A~CT~CA~C~~GT~TG~CT~CA~CC~GA~ 320 340 350 3M) 310 t-w 400 70 AC~TG~AC~,~G”GTC:T~“A~T~~A~T~~ A~AG~GAEAA”;‘GC:GA:o~~CAEC 4F.v 430 440 den 5co Bo “YFFF 7 zbR3 ,: A~CA:,A~~,~~GT:CA~CC:GA”GAC:CA~~A~CGTGAC~~~CGdCBD*CA~C~~~G~CT~~T:TT~TG~~,~~A AOAGAG T&G ,F Ho 183 590 MO 0 ‘I;’ GTC~AG~C~,~CC:CG~CACAG~CT~CTC~~~T~A~TCCTTACAACCTC~~TCTTCTATTCAGCTTAAA~,~~ATT~TACTGCAT~TGTT~~~GGG~AAAT~ 520 64 660 Em 700 TGTGTATCTGAATTTCAGGTCAT~AAGGG*CACTA~~GACACCTTG~GA~TCAGAAA~G~TCATTGGGAGCC~TGGCTGAT~CA~ACA~ACATCCTCA~CTCC 710 120 8%+1 740 750 m no 780 190 800 CAGACTTC;,;GGCCAGAGATTTATAG?&&W 820 Figure 2. Nucleotide and encoded protein sequences of reshape4 heavy and light, chain variable domains. The DNA sequence encoding the highest-affinity reshaped antibody (HuLysll) is shown. The location of the mutations, transcriptional control motifs and protein translations appear above the corresponding nucleotide sequence, with nucleotide numbering underneath. Pe’ptide numbering begins from the N-terminal residue of the mature protein. CDR sequences are boxed. splice junction. Eighteen oligonucleotides were used to encode both strands of the 370 base-pair construct, and were assembled and cloned in three separate blocks, deli- mited by P&I-KpnI, KpnI-Kpnl, and KpnI-EcoRI sites. 50 pmol of each oligonucleotide was phosphorylated for 30 min at 37°C in a 20 ~1 reaction mixture with 1 mM-ATP, 5 units polynucleotide kinase, 5 mM- dithiothreitol, 50 mM-Tris, 10 mm-MgCl, (pH 8). Portions (4 ~1) of each phosphorylation mixture were annealed by heating together at 80°C for 5 min, 67°C for 30 min, and cooling gradually (30 min) to room temperature. 200 nl of annealing mix was ligated into 20 ng M13-mp19 vector (Yanisch-Perron et al., 1985) in a volume of 10 ~1 (com- position as above except for 120 units of T4 DNA ligase and a pH of 7.5). Recombinant clones were sequenced and the 3 blocks excised, gel-purified and assembled by liga- tion into M13-mp19 as a P&I-EcoRI fragment. The synthetic gene was introduced as a PstI-BarnHI fragment into the vector M13-pHVNP and joined in- frame to the signal sequence with a mutagenic oligo- nucleotide (5’-TCA TCT GGA TGT CGG AGT GGA CAC CT-3’) (Zoller & Smith, 1982). Further derivatives were made: about 1.2 kb of sequence 5’ to the immunoglobulin oetamer transcriptional control element (Falkner & Zachau, 1984) was deleted using the oligonucleotide B’-TAGATTCAGAGGATTTGCATATTCATAAGC TTG GGC TAA TCA T-3’, and a point mutation, Phe7B to Tyr, was introduced using the oligonucleotide 5’-GGT GAA GGT GTA GTC GGT ACC-3’. The annotated sequence of the mutated gene is given in Fig. 2. (b) Construction of mutants of reshaped heavy chain variable gene Mutants of the reshaped heavy chain variable gene (Verhoeyen et al., 1988) were constructed by oligonucleo- tide-directed mutagenesis. As above, 1.2 kb of sequence 5’ to the octamer motif was deleted, but here it also resulted in spurious duplication of the motif. Residues 27 to 30, initially Ser-Thr-Phe-Ser, were changed to Phe-Ser-Leu-Thr with the mutagenic oligonucleotide 5’-TAC ACC ATA GCC GOT TAA GCT GAA GCC AGA CAC GGT-3’, Ser27 to Phe with the oligonucleotide 5’-GCT GAA GGT GAA GCC AGA CAC G-3’, and Val7l to Lys with the oligonucleotide 5’-TGC TGG TGT CCT TCA GCA TTG TC-3’. The annotated sequence of the mutated gene is shown in Fig. 2. (c) Eukaryotic expression vectors Fragments carrying heavy and light chain variabie genes with signal sequences and promoters (as above)
`
`Lassen - Exhibit 1033, p. 4
`
`

`

`Residues Aflecting Hypervariable Loops 491 Figure 3. Vectors for expression of reshaped heavy (aLys37) and light (aLys27) chain variable regions as human yl and IC chains in myeloma cells. Immunoglobulin exons are indicated by wide boxes, the coding regions of Z?-lactamase, aminoglycoside phosphotransferase (hygro- mycin resistance) and bacterial xanthine-guanine phos- phoribosyl transferase by narrow boxes, and splicing between immunoglobulin exons by broken lines. were now introduced into expression vectors. A vector pSV-V,,He (Neuberger et al., 1985) based on the vector pSV2gpt (Mulligan & Berg, 1980) was further modified to facilitate rapid cloning of the variable genes and their co- expression as heavy and light chains in cultured myeloma cells. A human K constant gene was built into a vector utilizing hygromycin resistance as a selectable marker, and a human heavy chain yl gene built into a vector utilizing mycophenolic acid resistance (Fig. 3). Light and heavy chain constructs were introduced into myeloma cells by electroporation (Potter et al., 1984), and stably transformed cells selected on the basis of drug resistance and cloned. Generally we co-transformed cells with both light and heavy chain constructs simul- taneously, but we also tried transforming first with light chain, cloning the intermediate light chain producing transfeotoma, then transforming with the heavy chain construct. The sequential method gave higher transfec- tion frequencies, and was a convenient strategy in creating a family of antibodies with identical light chains, but similar yields of antibody were purified from clones obtained either through sequential or cotransformation. The level of antibody production by individual clones of cells varied widely: the better clones were identified by ELISA and picked for large scale growth. The assembly of the vector for light chain expression involved a ligation of 3 fragments: a 2 kb HindIII- BamHI fragment containing the synthetic variable domain described above; a HindIII-Hind111 fragment containing the immunoglobulin heavy chain locus 5’ enhancer, /?-lactamase gene, and SV40 origin of replica- tion and promoters (M. S. Neuberger and L. Riechmann, unpublished results); a HindIII-BamHI fragment containing the coding sequence of a hygromycin-specific aminoglycoside phosphotransferase, SV40 t-antigen splice site and poly(A) sequences, excised from the plasmid pSV2*hyg (A. Smith, D. Strehlow and A. Miyajima, unpublished results). Next, the vector was partially digested with HindIII, and the Hind111 site at the end of the hygromycin fragment removed by fill-in extension with DNA polymerase I (Klenow fragment) and religa- tion. The human Clc constant region was then introduced as a BarnHI fragment at the unique BamHI site of the vector. (A genomic fragment containing the human J and CK gene segments (Hieter et al., 1982) was first cloned as a 10 kb BamHI fragment in pUC7 (Vieira & Messing, 1982) the J-segments excised as a 5 kb Hind111 fragment and the vector religated. The 5 kb BumHI fragment containing the CK coding sequence and the TC enhancer, was used in the construction.) The BamHI site at the 3’ end of the Clc constant region, and a Hind111 site internal to the IC fragment were then eliminated by fill-in and religation of partial digests. Plasmids were purified from 1 1 of bacterial culture by alkaline detergent lysis and CsCl/ethidium bromide equili- brium density gradient centrifugation (Ish-Horowitz & Burke, 1981). Heavy chain constructs were linearized with PvuI, a step which increased transfection frequency by more than lo-fold. Light chain constructs do not have a convenient restriction site for linearization, so the circular form was used. The cell line NSO (Galfre & Milstein, 1981), a myeloma which produces neither endo- genous immunoglobulin chain due to abolition of heavy chain transcription and a defect in the IC transcript (Carroll et al., 1988) was grown in Dulbecco’s modified Eagle’s medium supplemented with 10% (v/v) fetal calf serum, 110 mg sodium pyruvate/l, 100 mg streptomycin/l, and 100,000 units penicillin/l. Cells were harvested at a density of lo6 cells/ml, and held on ice. 10’ cells were suspended in @l ml PBS or medium, placed in a sterile plastic 64 cm x 1 cm cuvette and mixed with 10 pg of each DNA construct. After several minutes, three 2 kV pulses from an Apelex (Bagneux, France) cell porator were applied, 1 s apart, then the cells were returned to ice. The electroporation mixture was washed into a flask with 25 ml medium, and grown overnight. An equal volume of selective medium was added, and the suspended cells distributed over a 24-well plate. For selection of hygro- mycin resistance alone, cells were exposed to a drug concentration of @4 g/l. For co-selection of hygromycin and mycophenolic acid resistance, culture medium contained 0.2 g hygromycin/l, 98 mg mycophenolic acid/l, and 025 g xanthine/l. As stock solutions of xanthine were made in 91 M-NaOH, an equivalent of HCl was also added to preserve the pH of the growth medium. Construction of a reshaped anti-lysozyme antibody An antigen-based ELISA procedure was devised to with human y2 constant region has been described facilitate screening of transformants. Microtitre plates (Verhoeyen et aZ., 1988); our interest shifted to the yl (Dynatech) were coated overnight with 63 g lysozyme/l isotype, as it is more potent in activation of the comple- in 50 mM-NaHCG3 (pH 96). The plates were washed with ment cascade and of antibody-dependent cell-mediated ,PBS, then blocked with 1% (w/v) bovine serum albumin cytotoxicity (Briiggeman et al., 1987; Riechmann et al., in PBS for 5 min. 91 ml portions of culture supernatant 1988). A clone of the human yl constant region exons were allowed to react for 2 h, and after washing, adsorbed (Takahashi et al., 1982), was provided as a 2 kb BgZII antibody was quantified with peroxidase-conjugated fragment in the Ml3 phage tg131 (Kieny et al., 1983) by rabbit anti-human IgG (Dakopatts, Denmark) diluted to M. Brtiggemann. After Hind111 digestion, fill-in and reli- gation to destroy the internal Hind111 site, the BgZII fragment was introduced into the BarnHI-digested pSVgpt-HuVuLYS-HulgG2 vector (Verhoeyen et al., 1988) to replace the y2 constant region exons. Although the ligation destroys the BumHI sites of the vector back- bone, an internal BamHI site (present in the polylinker of the M13-tg 131 vector) remains at the V-proximal end of the insert. (d) Construction of transfectoma lines
`
`Lassen - Exhibit 1033, p. 5
`
`

`

`492 J. Foote and 6. Winter 2 pg/ml in 1% bovine serum albumin/PBS. After 1 h, plates were washed and substrates for the colour reaction added: 4 mmH,O,, 1 m&i-diammonium 2, 2’-azino bis (3-ethylbenzthiazoline sulphonic acid), in @l M-citrate buffer (pH 45). For screening of light chain producing transfectomas, a similar procedure was used. Plates were coated with a 1: 500 dilution of ascites fluid of the anti- human kappa hybridoma NH3/41 (Downie et al., 1983), supplied by J. Jarvis. Peroxidase-conjugated rabbit anti- human kappa was used for quantitation. equilibrated in fresh dialysis buffer, and eluted with 0.1 M-citrate (pH 6). Prior to storage, antibody was checked on an SDS/polyacrylamide gel (Laemmli, 1971) to verify purity, and an ultraviolet spectrum taken to ascertain concentra- tion. Yields of purified antibody were approximately 15 mg/l of culture in the case of the hybridoma, and typically 2 to 3 mg/l for the transfectoma lines. Antibody solutions were filter-sterilized and stored at 4°C under N, in acid-washed septum-top vials. Once visible colonies had appeared in the 24-well plates, after 2 weeks incubation, antibody production was assayed by ELISA. Cells from wells giving the strongest signals were cloned by limiting dilution in selective medium and regrowth on 96-well plates. The ELISA screening was repeated, usually on 20 or more clones, and the cell lines giving the strongest signals retained for permanent storage and large-scale growth for antibody production. (f) PhysieaE methods (e) Antibody production Antibodies were purified to homogeneity from several litres of culture medium by affinity chromatography on both lysozyme-Sepharose and Protein A-Sepharose. Sepharose CL-4B (Pharmacia), chilled in 1 M-Na,CO,, was activated for 2 min with 200 mg CNBr per ml resin, added as 1: 1 solution in CH,CN. The buffer was changed to @1 M-NaHCO,, 05 iv-NaCl (pH 8-5), and 5 mg lyso- zyme per ml. Sepharose was added, as a 1% solution in 10 mM-acetate, 40 mM-NaCl (pH 4%). The slurry was stirred intermittently for 4 h and then residual imidocar- bonate groups hydrolysed by overnight treatment with @l M-NaHCO, (pH 1@9). To minimize the effect of leakage of the immobilized lysozyme during ehromato- graphy, the columns were washed with high-salt and elution buffers immediately before use. The lysozyme- Sepharose columns were used once and discarded. Protein A-Sepharose was washed with @l M-Citric acid before use; resin was in this case re-used. The conditions chosen for all physical measurements were a temperature of 20°C and a standard buffer of 25 mmNaH,PO,, 125 mM-NaCl, @2 mM-EDTA (pH 7-O). Ultraviolet extinction coefficients (+sO ,,) were deter- mined for the mouse and reshaped antibodies to facilitate routine, accurate determination of concentration. Quantitative amino acid analysis was performed on samples of known optical density, with norleucine added prior to hydrolysis as an internal standard. Correlation of the observed amino acid composition with that deduced from the encoded protein sequence gave c2s,, nm values of 22 2 l@ (M-l em-‘) for mouse and %l x IO5 (~2 isotype) or 1.9 x lo5 (71 isotype) for reshaped antibodies. Antibody preparations were quantified at the time of storage. Portions were withdrawn as needed and used directly, assuming that the concentration at the time of storage was unchanged. Hen egg lysozyme and bovine ubiquitin were from Sigma. These were dissolved in PBS and dialysed against the same buffer. Concentrations were determined spectrophotometrically, using c2s0 nm of 37,600 for lysozyme (Imoto et al., 19’72a) and @16 @g/ml)-’ em-’ for ubiquitin (Ciechanover et al., 1980). (g) Fluorescence spectroscopy Transfectoma cultures were grown from cloned stocks in flasks to 100 to 200 ml, in medium with both myeo- phenolic acid and hygromycin present, and 10% fetal calf serum. Each culture was then transferred to a roller bottle, and grown to 2 1 with 5% serum with mycophe- nolic acid. Cells were grown until the culture medium was exhausted, leading to the demise of most cells. Cultures were cleared by eentrifugation at 11,000 g for 15 min. Solid (NH&SO, was added slowly at a ratio of @313 kg/l of supernatant, and the mixture stirred for 1 h at 4°C. Precipitate was harvested by a 45 min centrifugation dissolved in 20 ml phosphate buffered saline (125 mM-Nacl, 84 mmNaH,PO,, 166 mM-Na,HPO,) (PBS) and dialysed against the same. Emission spectra revealed a large quench of native Trp fluorescence upon antibody-lysozyme complex formation; fluorescence measurements at fixed wavelengths were then used to determine the stoichiometry and approxi- mate equilibrium constant. A Perkin-Elmer LSdB spectrofluorimeter interfaced to a Macintosh computer was used for all measurements. Of significance was the 50 Hz stroboscopic Xe light source, which gives a l50 W pulse, but has a root-mean-square intensity.of 8 W, reducing photobleaching of samples to negligible levels. There was virtually no signal drift over the course of an experiment. Temperature control of the cuvette block was maintained by a thermostatted circu- lating water bath. The dialysed protein was centrifuged to remove any insoluble material, and the reshaped antibodies applied to a 5 ml column of lysozyme-Sepharose, previously equili- brated in PBS. The column was washed extensively with high-salt buffer, @I M-Tris, 05 M-Nacl (pH 85), and anti- body eluted with 50 mr+-diethylamine. 1 M-Tris . HCl (pH 7.4) was present in fraction collector tubes to neutra- lize the eluate. The eluate was then applied directly to a 5 to 10 ml Protein A-Sepharose column equilibrated in PBS. The columns were washed, eluted with 4 column vols of a gradient, @l mNa,-citrate to @I M-Citrk acid, and the antibody dialysed against PBS, O-2 mm-EDTA. For emission spectra, an excitation wavelength of 280 nm was used, with a bandwidth of 10 nm; the emis- sion bandwidth was 5 nm. Samples were scanned from 300 to 420 nm at 50 nm/min, with an 8 s time constant. In a typical set of measurements, a buffer sample was scanned to obtain a baseline. Antibody was added to 100 nM for the next scan, following which lysozyme was added in a negligible volume to 200 nM. A 3rd scan was taken after a 2 min wait for complex formation to occnr. Similar measurements were made on a separate sample of buffer, then 2OOnmlysozyme alone. The spectra of pro- tein samples were then corrected by numerical subtrac- tion of the buffer blank. For purification of the D1.3 antibody (mouse yl In a typical titration, 15 samples of a concentrated isotype) on Protein A-Sepharose, the sample was dialysed lysozyme solution were added to a 3 ml sample of 10 to against either 3 M-Nacl, @I En-glycine (pH &9), or against @l M-Na,HPO, (pH S), before application to the column 200 m-antibody, culminating in a molar ratio of 4 lyso- zyme/antibody (i.e. 2 lysozyme added per combining site).
`
`Lassen - Exhibit 1033, p. 6
`
`

`

`Residues Affecting Hypervariable Loops 493 For higher antibody concentrations, a minimum interval of 2 min was maintained between adding lysozyme and reading the fluorescence. For lower concentrations, con- sideration was made of the length of time necessary for approach to equilibrium. An approximate relaxation time was estimated from any available kinetic and equilibrium data, and an interval several times that used, often in excess of 5 min. Titrations were most frequently done at an excitation wavelength of 280 nm and an emission wavelength of 400 nm. The excitation bandwidth was 10 nm. Emission bandwidths of 5, 10 or 20 nm were used, the choice for a given experiment depending on the magnitude of sample fluorescence. Each observation was obtained by averaging the fluorescence signal over a span of 16 s. A source of systematic error in titrations was a progres- sive loss of fluorescence perhaps by adsorption to glass. This was especially compromising at high dilutions of antibody. For example, the fluorescence of a 10 nM solu- tion of D1.3 antibody decreased 25% during a mock titration with samples of buffer. Use of ubiquitin as a carrier almost completely overcame this artifact. Ubiquitin has no tryptophan residues and a single tyro- sine (Schlesinger et aZ., 1975), hence is very weakly fluores- cent. A concentration of 30 pg/ml gave a stable background, which was a fraction of the total fluorescence of the antibody and antigen. Discounting background, the fluorescence, F, of a mixture of antibody and lysozyme at equilibrium is described by the equation: F = fAbrA +fLySL -j,:{ rA + L + K) -,/(rA+L+K)‘-4rAL), (1) where A and L are the total (complexed and uneom- plexed) antibody and lysozyme concentrations; r is the average number of combining sites per antibody molecule; fAb and X,, are the molar fluorescence of free antibody (on a per-site basis) and lysozyme; fd is the molar fluorescence change occurring on complexation of lysozyme, a positive value indicating a net quench; K is the equilibrium constant for complex dissociation. A computer program for analysis of experimental data according to eqn (1) was written in Pascal and implemented on the Apple Macintosh computer. A and L were treated as indepen- dent variables, F as the dependent variable, and fAb, fLysr fA, r and K were adjustable parameters optimized by the operation of the Marquardt least-squares algorithm on a first-order Taylor series expansion of eqn (1). (h) Rapid kinetics Measurement of the rate of antibody-antigen associa- tion was determined from a series of stopped-flow measurements under a pseudo-first-order regime. A commercial instrument of standard design (Hi-Tech Scientific, Salisbury, U.K.) was used for this work, in fluorescence mode. Light from a 75 W Xe lamp passed through a monochromator set at 280 nm, and thence via a light guide to the sample. Light emitted from the sample passed through a 320 or 360 nm cutoff (lower limit) filter before impinging