1 of 5
`NATURE VOL. 332 24 MARCH 1988
`~
`——ARTICLES
`
`OnCusp
`Ex. 1035
`
`- ee 323
`
`different paths, and by ensuring that an azimuthally uniform
`coverage of stations is used in the averaging calculation. To
`compensatefor other factors, such as focal depth, fault geometry
`and corner frequency would require such a detailed knowledge
`of the earthquake source that the M, measurementitself would
`be redundant.
`Theresults of this analysis can be summarizedin five points.
`(1) A global average moment-magnitude relationship M, has
`been defined which can be used to predict M, over a wide range
`of magnitudes and scalar moments.
`(2) The variance of surface wave measurements for an event
`of a particular scalar moment is ~0.2 magnitude units.
`
`(3) Large regional biases in M, exist.
`(4) Differences in source scaling may explain some of the
`differences. Specifically, observations show that the transition
`from a slope of unity to a smaller value occurs at large moments
`for continental events than for ridge and fracture zone events,
`suggesting systematic differences in stress drop.
`(5) Other systematic factors affecting the calculation of M,
`also appear to contribute to the observed regionalbias.
`Wethank Professor J. H. Woodhouse for reading and correct-
`ing the manuscript and Professor H. Kanamori for constructive
`criticism throughout our work on this subject. This work was
`supported by the NSF.
`
`Received 20 October 1987; accepted 4 February 1988.
`
`12. Dziewonski, A. M., Ekstrom, G., Woodhouse, J. H. & Zwart, G. Phys. Earth planet. Inter.
`{in the press).
`1. Richter, C. F. Bull, seism. Soc. Am, 25, 1-32 (1935).
`13. Kanamori, H. J. geophys. Res. 82, 2981-2987 (1977).
`2. Wanek, J. et al. Izv. akad. Nauk. USSR, Ser. Geophys. 2, 153-158 (1962).
`14. Richter, C. F. Elementary Seismology (W. H. Freeman, San Fransisco, 1958).
`15. Lienkamper, J. J. Bull. seism. Soc. Am. 74, 2357-2378 (1984).
`3. Aki, K. Bull Earthqu. Res. Inst. Tokyo Univ, 44, 23-88 (1966).
`16. Kanamori, H. Anderson, D. L. Bull. seism. Soc. Am. 65, 1073-1095 (1975).
`4. Agnew, D., Berger, J., Buland, R., Farrell, W. & Gilbert, F. Eos 57, 280-288 (1976).
`5. Peterson, J., Butler, H. M., Holcomb, L. G. & Hutt, C. R. Bull seism. Soc. Am. 66, 2049-2068
`17. Ekstrom, G. & Dziewonski, A. M. Bull. seism. Soc. Am. 75, 23-39 (1985).
`(1976).
`18. Sipkin, S. A. Bull. seism. Soc. Am. 76, 1515-1541 (1986).
`19. Harkrider, D. G. Bull. seism. Sac. Am. 54, 627-679 (1964).
`Kanamori, H. & Given, J. W. Phys. Earth planet. Inter. 27, 8-31 (1981).
`. Dziewonski, A. M., Chou, T. A. & Woodhouse, J. H. 4. geophys. Res. 86, 2825-2852 (1981).
`20. Gutenberg, B. & Richter, C. F. Gerlands Beitr. z. Geophysik 47, 73-131 (1936).
`21, Gutenberg, B. Bull. seism. Soc. Am. 35, 3-12 (1945).
`. Woodhouse, J. H. & Dziewonski, A. M. J. geophys. Res, 88, 3247-3271 (1983).
`22. Von Seggern, D. Bull. seism. Soc. Am. 60, 503-516 (1970).
`Woodhouse, J. H. & Dziewonski, A. M. J. geophys. Res. 89, 5953-5986 (1984).
`. Dziewonski, A. M., Franzen, J. E. & Woodhouse, J. H. Phys. Earth planet. Inter, 34, 209-219
`23. Nuttli, O. Tectonophysics 118, 161-174 (1985).
`(1984).
`24, Kanamori, H. & Allen, C. R. in Maurice Ewing Series Vol. 6, Earthquake Source Mechanics
`11. Dziewonski, A. M., Ekstrom, G., Franzen, J. E. & Woodhouse, J. H. Phys. Earth planet.
`(American Geophysical Union, Washington, DC, 1986).
`Inter. 45, 11-36 (1987).
`25. Zhuo, T. & Kanamori, H. Bull seism. Soc. Am. 77, 514-529 (1987).
`
`
`Sewers
`
`Reshaping human antibodies for therapy
`
`Lutz Riechmann’, Michael Clark’, Herman Waldmann’ & Greg Winter*
`
`MRCLaboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, UK
`* Department of Pathology, University of Cambridge, Tennis Court Road, Cambridge CB2 1QP, UK
`
`A human IgGI antibody has been reshaped for serotherapy in humans by introducing the six hypervariable regions from
`the heavy- and light-chain variable domains of a rat antibody directed against human lymphocytes. The reshaped human
`antibody is as effective as the rat antibody in complement andis moreeffective in cell-mediated lysis of human lymphocytes.
`
`IN 1890 it was shownthat resistance to diphtheria toxin could
`be transferred from one animal to another by the transfer of
`serum. It was concluded that the immune serum contained an
`anti-toxin, later called an antibody’. For many years animal
`antisera were used in the treatment of microbial infections and
`for the neutralization of toxins in man”. More recently rodent
`monoclonal antibodies (mAbs)* have been used as ‘magic bul-
`lets to kill and to image tumours™*®. The foreign immuno-
`globulin, however, can elicit an anti-globulin response which
`may interefere with therapy’ or cause allergic or immune com-
`plex hypersensitivity*. Thus ideally human antibodies would be
`used. Human immunoglobulins are widely used as both prophy-
`lactic and microbicidal agents®, but it would be far better to
`have available human mAbsof the desired specificity. [t has
`proven difficult, however, to make such mAbs by the conven-
`tional route of immortalization of human antibody-producing
`cells’.
`There is an alternative approach. Antibody genes have been
`transfected into lymphoid cells, and the encoded antibodies
`expressed and secreted; by shuffling genomic exons, simple
`chimaeric antibodies with mouse variable regions and human
`constant regions have been made'*"'’. Such chimaeric antibodies
`
`
`t Address from April 1988: Department of Molecular Biology, The
`Research Institute of Scripps Clinic, North Torrey Pines Road, La Jolla,
`California 02937, USA
`+ To whom correspondence should be addressed.
`
`have at least two advantages over mouse antibodies. First, the
`effector functions can be selected or tailored as desired. For
`example, of the human IgGisotypes, IgG1 and IgG3 appear to
`be the mosteffective for complement andcell-mediated lysis'*"",
`and therefore for killing tumour cells. Second, the use of human
`rather than mouse isotypes should minimize the anti-globulin
`responses during therapy'®'’ by avoiding anti-isotypic anti-
`bodies. The extent to which anti-idiotypic responses to rodent
`antibodies in therapy are dictated by foreign components of the
`variable versus the constant region is not known, but the use of
`human isotypes should reduce the anti-idiotypic response. For
`example, when mice were madetolerant to rat immunoglobulin
`constant-region determinants,
`administration of
`rat
`anti-
`lymphocyte antibodies did evoke anti-idiotypic responses, but
`these were delayed and weaker than in animals that had not
`been madetolerant’’. Nevertheless,it is likely that a chimaeric
`antibody would provoke a greater immune response than a
`human mAb.
`We have attempted to build rodent antigen binding sites
`directly into human antibodies by transplanting only the antigen
`binding site, rather than the entire variable domain, from a
`rodent antibody. The antigen bindingsite is essentially encoded
`by the hypervariable loops at one end of the B-sheet framework.
`The hypervariable regions of the heavy chain of mouse anti-
`bodies against a hapten’® or a protein antigen*’ were previously
`transplanted into a human heavychain, and, in association with
`the mouse light chain, the antigen binding site was retained.
`
`1 of 5
`
`OnCusp
`Ex. 1035
`
`

`

`
`
`110 113,|Spi |[RPOR TIF GTGTkKLELKRtce
`
`Spl i
`
`OnCusp
`2 0f 5
`Ex. 1035
`
`324
`; ee :
`ARTICLES—
`_
`NATURE VOL. 332 24 MARCH 1988
`a
`b
`Hind 11]
`Hind VI]
`——
`—
`—
`St... Mee ATGCAARTCCTCTGAATCTACATGGTAARATATAGGTTTGTCTATACC
`Sooo. ATGCARATCCTCTGARTCTACATGGTARATATAGGTTTGTCTATACC
`
`g@—>RNA starts g@—>RNA startsBA» RNA starts m—— RNA starts
`
`
`ACARACAGARAAACRTGAGATCACAGTTCTCTCTACAGTTACTGAGCACACAGGACCTCA +60
`ACAAACAGARARACATGAGATCACAGTTCTCTCTACAGTTACTGAGCACRCAGGACCTCA +60
`signa
`Splice
`ATGA
`MGW S Cl
`itt*f?rFLUA TAT
`CCATGGGATGGAGCTGTATCATCCTCTTCTTGGTAGCAACAGCTACAGGTAAGGGGCTCA +120
`ATGAAGTTGTGGCTGAACTGGATTTTCCTTTTARCACTTTTAAAT
`
`MoK EL WoL N Wo}
`F
`bt
`bk TEL N
`
`Splice
`signal
`"GoW S
`¢ tI LF LU AT AT
`CCATGGGATGGAGCTGTATCATCCTCTTCTTOGIAGCAACAGCTACAGGTARGGGGCTCA +120
`TGGCTGCACTTCAACTCTTAGGGGTAGCTGCTAGCTCTGGCTCCCAG
`
`CAGTAGCAGGCTTGAGGTCTGGACATATATATGGGTGACAATGACATCCACTITGCCTTT +180
`(h#
`A
`A LQLlLGUAAS § GS Q)
`splice}
`__
`oligos Ill, IV, Vil
`signal
`4
`5
`10
`CAGTAGCAGGCTTGAGGTCTGGACATATATATGGGTGACAATGACATCCACTTTGCCTTT +180
`GVAHSavatlaqaeEescopPpGateueR
`Soli
`CTCTCCACAGGTGTCCACTCCCAGGTCCAACTGCAGGAGAGCGGTCCAGGTCTTGTGAGA +240
`10
`5
`1
`signal
`pitcel
`GGTATCCAGTGTGAGGTGARACTGTTGGARTCTGGAGGAGGCTTGGTACAG
`GUA SDI!aqantasPSSLSA
`Gii@q Orteuvuk
`tl LESGEGLY O
`CTCTCCACAGGTGTCCACTCCGACATCCAGATGACCCAGAGCCCARGCAGCCTGAGCGCC +240
`
`
`
`
`90 9s oligo XI con9° %5 GCCATGAGATGTGACATCAAGATGACCCAGTCTCCCTCATTCCTGTCTGCA@MR ODI KH T $s PS FLSA
`
`PsSQTL$S LTCTUS GS TF S*D FV
`OS
`igo KV
`CCTAGCCAGACCCTGAGCCTGACCTGCACCGTGTCTGGCAGCACCTTCAGCGATTTCTAC +300
`15
`20
`25
`30
`CDR 1
`CCGGGGGGTTCTATGAGACTCTCCTGTGCAGGTTCTGGATTCACCTTCACTGATTTCTAC
`sUGDRUTITE
`PGGSMHRLS € AGS GF TF T
`AGCGTGGGTGACAGAGTGACCATCACCTGTAARGCAAGTCAGAATATTGACAAATACTTA +300
`oligo IX
`TCTGTGGGAGACAGAGTCACTCTCRACTGCARAGCAAGTCAGAATATTGACAAATACTTA
`
`
`
`
`35 sU GORY TLN ClkASQN1OKvt)40 45 5Q__52_a
`[MN] W YVR QPPGRGL ERI GfF TROD]oligo XV
`ATGAACTGGGTGAGACAGCCACCTGGACGAGGTCTTGAGTGGATTGGATTTATTAGAGAC +360
`35
`40
`45
`50
`CDR 2
`ATGAACTGGATCCGCCAGCCTGCAGGGAAGGCACCTGAGTGGCTGGGTTTTATTAGAGAC
`If]u ¥ aaqKkPGKAPKLLIY
`
`[HMN]JW 1 R QP AGK AP ERL G AACTGGTACCAGCAGAAGCCAGGTARGGCTCCAARGCTGCTGATCTACAATACAAACAAT +360
`oligo XI
`AACTGGTATCAGCAAARGCTTGGAGAATCTCCCARACTCCTGATATATAATACAARCART
`b_NYRUTH L¢ 53 55 CDR 2 60 65 70 [N]}¥ ¥ aaQqkLGESPKLEL I ¥ [NTN
`
`
`
`
`
`
`
`
`
`
`AAAGCTARAGGTTACACAACAGAGTACAATCCATCTGTGAAGGGGAGAGTGACAATGCTG +420
`55
`60
`65
`70
`ARAGCTARAGGTTACACARCAGAGTACAATCCATCIGTGAAGGGGCGGTTCACCATCTCC
`GUPSRFES$6SG6GSGTDFTE
`R FT 1
`S$
`TTGCARACGGGTGTGCCARGCAGATTCAGCGGTAGCGGTAGCGGTACCGACTTCACCTTC +420
`TIGCARACGGGCATCCCATCAAGGTTCAGTGGCAGTGGATCTGGTACTGATTTCACACTC
`LoT]G i PSRFSGSGSEEGETDEFETL
`85
`c 83
`b
`a
`82
`80
`75
`oligo XVI
`vDTS KNQFSLALSSVUTAAOD T
`75
`80
`85
`90.
`CDR 3
`GTAGACACCAGCAAGAACCAGTTCAGCCTGAGACTCAGCAGCGTGACAGCCGCCGACACC +480
`rT 1S S$LQPEDtINATYVYC
`AGAGATAATACCCARARCATGCTCTATCTTCARATGARCACCCTAAGAGCTGAGGACACT
`ACCATCAGCAGCCTCCAGCCAGAGGACATCGCCACCTACTACTGCTTGCAGCATATARGT +480
`RONTQNANHM*LY¥LQmMNTLRAE OT
`ACCATCAGCAGCCTGCAGCCTGAAGATGTTGCCACATATTTCTGCTIGCAGCATATAAGT
`oligo XIl
` 95 CDR 3 100_a 6101 105Vv 90 TISSLQPEDVATY FCcft QHtTs]A ¥Y ¥ C A RLE GH TA AP FD YIN GQ
`
`
`
`
`
`
`
`
`
`
`
`GCGGTCTATTATTGTGCAAGAGAGGGCCACACTOCTGCTCCTTTTGATTACTGGGGTCAA +540
`95
`100
`105
`108
`GCCACTTACTACTGTGCARGAGAGGGCCACACTGCTGCTCCTTTIGATTACTGGGGCCAA
`FGQGTKUEIKR
`A TY Y CARR
`WG Q
`AGGCCGCGCACGTTCGGCCRAGGGACCARGGTGGARATCARACGTGAGTAGAATTTARAC +540
`oligos V, VI, Vil
`AGGCCGCGCACGTTTGGARCTGGGACCARGCTGGAGCTGARACGG
`BamH|
`6S LUTUSs
`BamH|
`GGCAGCCTCGTCACAGTCTCCTCAGGT.0... 3° +600
`TTTGCTTCCTCAGTTGGATCC-3
`GGAGTCATGGTCACAGTCTCCTCA
`GUM UTuUs 5s
`
`
`
`
`
`
`
`Oligonucleotides: I: 5'-GGC CAG TGG ATA GAC-3", Ill: 5'-CAG TTT CAT CTA
`GAA CTS GAT A-3', IV: 5'-GCA GTT GGG TCT AGA AGT GGA CAC C-3',
`V:5'-TCA GCT GAG TCG ACT GTG AC-3', VI: 5'-TCA CCT GAG TCG ACT GTG
`AC-3', Vil: 5'-AGT TTC ACC TCG GAG TGG ACA CCT-3', VIII: 5°-TCA CCT GAG
`GAG ACT GTG AC-3'; IX: S'-GGC TGG CGA ATC CAG TT-3', X:5'-CTG TCT CAC
`CCA GTT CAT GTA GAA ATC GCT GAA GGT GCT-3', XI: 5'-CAT TGT CAC TCT
`CCC CTT CAC AGA TGG ATT GTACTC TGT TGT GTA ACC TTT AGC TTT GTC
`TCT AAT AAA TCC AAT CCA CTC~3', XII: S'-GCC TTG ACC CCA GTA ATC AAA
`AGG AGC AGC AGT GTG GCC CTC TCT TGC ACA ATA-3', XII: S'~AGA AAT
`CGG/C TGA AGG TGA AGC CAG ACAC-3'.
`
`Oligonucleotides : I}: S'-TGC AGC ATC AGC C-3', XIV: 5'-CTG CTG GTACCA
`GTT TAA GTA TTT GTC AAT ATT CTG ACT TGC TTT ACA GGTGAT GGT-3',
`XV: 5'-GCT TGG CAC ACC CGT TTG CAA ATT GTT TGT ATT GTA GAT CAG
`CAG-3', XVI: 5'-CCC TTG GCC GAA CGT GCG CGG CCT ACT TAT ATG CTG CAA
`GCA GTA GTA GGT-3'.
`
`Fig. 1 Heavy-chain (a) and light-chain (b) sequences of the variable domains of reshaped (upper line) or rat YTH 34.5HL (lowerline)
`antibodies. The reshaped heavy-chain variable domain HuVHCAMP wasbased on the HuVHNP gene!”!”, with the framework regions of
`human NEW (see note) alternating with the hypervariable regions of rat YTH 34.5HL. The reshaped light-chain variable domain HuVLCAMP
`is a similar construct, except with the framework regions of the human myelomaprotein REI, with the C-terminal and the 3’ non-coding
`sequence taken from a human J,-region sequence**. The sequencesof oligonucleotide primers are given and their locations on the genes are
`marked.
`Methods. Messenger mRNAwaspurified*’ from the hybridoma clone YTH 34.5HL (y2a, «”). First strand cDNAwassynthesized by priming
`with oligonucleotides complementary to the 5’ end of the CH1 (oligonucleotide I) and the Cx exons (oligonucleotide II), and then cloned
`and sequencedas described previously***?. Two restriction sites (Xbal and SalI) were introduced at each end of the rat heavy-chain variable
`region RaVHCAMP cDNAclone in M13 using mutagenic oligonucleotides III and V respectively, and the XbaI-Sall fragment was excised.
`The corresponding sites were introduced into the M13-HuVHNP geneusing oligonucleotides IV and VI, and the region between thesites
`was then exchanged. The sequence at the junctions was corrected with oligonucleotides VII and VIII, and an internal BamHI site removed
`using the oligonucleotide IX, to create the M13-RaVHCAMP gene. The encoded sequence of the mature domainis thus identical to that of
`YTH 34.5HL. The reshaped heavy-chain variable domain (HuVHCAMP) was constructed in an M13 vector by priming with three long
`oligonucleotides simultaneously on the single strand containing the M13-HuVHNP gene’*'®. Each oligonucleotide (X, XI and XII) was
`designed to replace each of the hypervariable regions with the corresponding region from the heavy chain of the YTH 34.5HL antibody.
`Colony blots were probedinitially with the oligonucleotide X and hybridization positives were sequenced: the overall yield of the triple mutant
`was 5%. The (Ser27—> Phe) and (Ser27 > Phe, Ser30> Thr) mutants of M13mp8-HuVHCAMP were made with the mixed oligonucleotide
`XII. The reshaped light-chain variable domain (HuVLCAMP) was constructed in M13 from a gene with framework regions based on human
`REI (J. Foote, unpublished data). As above, three long oligonucleotides (XIV, XV and XVI) were used to introduce the hypervariable regions
`of the YTH 34.5HL light chain.
`Note: There are discrepancies involving the first framework region andthe first hypervariable loop of the NEW heavy chain between the
`published sequence”’ used here and the sequence deposited in the Brookhaven data base (in parentheses): Ser27 (>Thr), Thr28 (Ser) and
`Ser30 (> Asp). Neither version is definitive (R. J. Poljak, personal communication) and the discrepancies do not affect our interpretations.
`
`2 of 5
`
`OnCusp
`Ex. 1035
`
`

`

`OnCusp
`3 of 5
`Ex. 1035
`
`
`NATURE VOL. 332.24 MARCH 1988
`ARTICLES
`-
`oo
`(325
`
`HuVHCAMP-RalgG2b
`
`1
`oe
`
`Y aavilcaneauger 24
`
`.
`
`3
`
`rat
`
`
`
`
`
`
`
`HuVHCAMP-HulgG|
`HuVE-CAMP-Hulgs
`
`
`
`
`
`
`
`
`
`human
`
`
`Table 1 Reshaping the heavy-chain variable domain
`
`Concentration of antibody
`in pg mi at
`50%
`50%
`antigen
`complement
`binding
`lysis
`Heavy chain variable domain
`2.1
`0.7
`RaVHCAMP
`*
`27.3
`HuVHCAMP
`18
`163
`HuVHCAMP(Ser27- Phe)
`
` HuVHCAMP (Ser 27> Phe, Ser 30~ Thr) 2.0 17.6
`
`
`
`Antibodies with the heavy-chain variable domainslisted above, rat
`IgG2b constant domains andrat light chains were collected from super-
`natants of cells at stationary phase and concentrated by precipitation
`with ammonium sulphate, followed by ion exchange chromatography
`ona Pharmacia MonoQ column. Theyields of antibody were measured
`by an enzyme-linked immunosorbent assay (ELISA) directed against
`the rat IgG2b isotype, and each wasadjusted to the same concentration®>.
`To measuring binding to antigen, partially purified CAMPATH-1 anti-
`gen wascoated onto microtitre wells and bound antibody was detected
`via a biotin-labelled anti-rat IgG2b mAb*, developed with a strep-
`tavidin-peroxidase conjugate (Amersham). Complementlysis of human
`lymphocytes was with human serum as the complement source”!. For
`both binding and complement assays, antibody titres were determined
`by fitting the data to a sigmoid curve by at least squares iterative
`procedure”!.
`* Complementlysis with the HuVHCAMP variable domain was too
`weak for the estimation oflytic titre.
`
`Fig. 2 Strategy for reshaping a human antibody for therapy.
`Sequences of rat origin are marked in black, and those of human
`origin in white. The recombinant heavy and light chains are also
`marked using a systematic nomenclature. See text for description
`of stages 1, 2 and 3. The genes encoding the variable domains were
`excised from the M13 vectors as HindIII- BamHI fragments, and
`recloned into pSV2gpt?® (heavy chains) or pSV2neo*° (light
`chains), expression vectors containing the immunoglobulin en-
`hancer!?. The human yi (ref. 40), y2 (ref. 41), y3 (ref. 42), v4
`(ref. 41) and « (ref. 36) and the rat y2b (ref. 43) constant domains
`were introduced as BamHI fragments. The following plasmids
`were constructed and transfected into lymphoid cell
`lines by
`electroporation™. In stage 1, the pSVgpt plasmids HuVHCAMP-
`RalgG2B, HuVHCAMP(Ser- Phe)-RalgG2B, HuVHCAMP-
`(Ser27 > Phe, Ser30~> Thr)-RalgG2B were introduced into the
`heavy chain loss variant of YTH 34.5HL.In stage 2, the pSVgpt
`
`plasmids©RaVHCAMP-RalgG2B, RaVHCAMP-HulgGi,
`CAMPATH-1 antigen andthe selection of humaneffector func-
`RaVHCAMP-HulgG2, RaVHCAMP-HulgG3, RaVHCAMP-
`tions to match the lytic potential of the rat IgG2b isotype.
`HulgG4 weretransfected as above.In stage 3, the pSV-gpt plasmid
`Hu(Ser27 > Phe, Ser30> Thr)VHCAMP-HulgG1 was co-trans-
`fected with the pSV-neo plasmid HuVLCAMP-HulgkK intothe rat
`myelomacell line YO (Y B2/3.0 Ag 20 (ref. 31). In each of the
`three stages, clones resistant to mycophenolic acid were selected
`and screened for antibody production by ELISA assays. Clones
`secreting antibody were subcloned bylimiting dilution (for YO) or
`the soft agar method(for theloss variant) and assayed again before
`1 litre growth in roller bottles.
`
`Strategy
`The amino-acid sequencesof the heavy- and light-chain variable
`domains of the rat IgG2a CAMPATH-1 antibody YTH 34.5HL
`were determined from the cloned complementary DNA(Fig. 1),
`and the hypervariable regions were identified according to
`Kabat”’. In the heavy-chain variable domain there is an unusual
`feature in the framework region. In most known heavy-chain
`sequences Pro4] and Leu45 are highly conserved: Pro41 helps
`turn a loop distant from the antigen binding site and Leu45is
`in the 8 bulge which forms part of the conserved packing
`between heavy- and light-chain variable domains**. In YTH
`34.5HL these residues are replaced by Ala41 and Pro45 and
`presumably this could have some effect on the packing of the
`heavy- and light-chain variable domains. Working at the level
`of the gene and using three large mutagenic oligonucleotides
`for each variable domain, the rat hypervariable regions were
`mounted in a single step on the human heavy- or light-chain
`framework regions taken from the crystallographically solved
`proteins NEW”’ and REI”® respectively (Fig. 1). The REI light
`chain was used because there is a deletion at the beginning of
`the third framework region in NEW. The reshaped human
`heavy- and light-chain variable domains were then assembled
`with constant domains in three stage (Fig. 2). This permits a
`step-wise check on the reshaping of the heavy-chain variable
`domain (stage 1), the selection of the humanisotype (stage 2),
`and the reshaping of the light-chain variable domain and the
`assembly of human antibody (stage 3). The plasmid construc-
`tions were genomic, with the sequences encoding variable
`domains cloned as HindII]- BamHI fragments and those encod-
`ing the constant domains as BamHI- BamHI fragmentsin either
`pSVegpt (heavy chain)? or pSVneo (light chain)*° vectors. The
`heavy-chain enhancer sequence was included on the S’ side of
`the variable domain, and expression of both light and heavy
`chains was driven from the heavy-chain promoterand the heavy-
`chain signal sequence.
`
`Heavy-chain variable domain
`In stage
`1,
`the
`reshaped heavy-chain variable domain
`(HuVHCAMP)was attached to constant domains of the rat
`
`Since, to a first approximation, the sequences of hypervariable
`regions do not contain characteristic rodent or human motifs,
`such ‘reshaped’ antibodies should be indistinguishable in
`sequence from humanantibodies.
`There are mAbs to manycell-type-specific differentiation anti-
`gens, but only a few have therapeutic potential. Of particular
`interest is a group of rat mAbs directed against an antigen, the
`‘CAMPATH-1’ antigen, whichis strongly expressed onvirtually
`all human lymphocytes and monocytes, but is absent from other
`blood cells
`including the haemopoietic stem cells”°. The
`CAMPATH-1 series contains rat mAbof IgM, IgG2a and IgG2c
`isotypes”', and more recently IgG1 and IgG2b isotypes which
`were isolated as class-switch variants from the IgG2a-secreting
`cell line YTH 34.5HL”. All of these antibodies, except the rat
`IgG2c isotype, are able to lyse human lymphocytes efficiently
`with human complement. Also the IgG2b antibody YTH
`34.5HL-G2b, butnotthe otherisotypes, is effective in antibody-
`dependent cell-mediated cytotoxicity (ADCC) with human
`effector cells’?. These rat mAbs have important applicationsin
`problems of immunosuppression: for example control of graft-
`versus-host disease in bone-marrow transplantation”,
`the
`management of organ rejection’; the prevention of marrow
`rejection; and the treatment of various lymphoid malignancies
`(ref. 24 and M. J. Dyer, Hale, G., Hayhoe, F. G. J. and
`Waldmann,H., unpublished observations). The [gG2b antibody
`YTH 34.5HL-G2b seems to be the most effective at depleting
`lymphocytes in vivo but the use of all of these antibodies is
`limited by the anti-globulin response which can occur within
`two weeksof the initiation of treatment**. Here we describe the
`reshaping of humanheavyandlight chains towards binding the
`
`3 of 5
`
`OnCusp
`Ex. 1035
`
`

`

`4 of 5
`we
`——-———
`
`
`
`_|-
`ARTICLES-—
`
`OnCusp
`Ex. 1035
`
`NATURE VOL. 332 24 MARCH 1988
`—
`-
`-
`
`Fig. 3. Loop Phe27 to Tyr35 in the heavy-chain variable
`domain of the human myeloma protein KOL, which has
`been solved crystallographically**. The backbone of the
`hypervariable region according to Kabat”is highlighted,
`and a 200%van der Waal surface is thrown around Phe 27
`to show the interactions with Tyr32 and Met 34 of the
`Kabat hypervariable region. In the rat YTH 34.5HL heavy
`chain, these three side chains are conserved in character,
`but in HoVHCAMP,Phe27 is replaced by Ser.
`
`
`
`We therefore selected the human IgG1 isotype for the
`reshaped antibody. Other recent work also favours the use of
`IgG1 isotype for therapeutic application. When the effector
`functions of human isotypes were compared using a set of
`chimaeric antibodies with an anti-hapten variable domain, the
`IgG1 isotype appeared superior to the [gG3 in both complement
`and cell-mediated lysis’®. Also, of two mouse chimaeric anti-
`bodies with human IgG1 or IgG3 isotypes directed against cell
`surface antigens as tumourcell markers, only the IgG1 isotype
`mediated complementlysis'*""*.
`
`isotype [gG2b and transfected into a heavy-chain loss variant
`of the YTH 34.5 hybridoma. Thisvariantcarries twolight chains,
`one derived from the Y3 fusion partner*’. The cloned rat heavy-
`chain variable domain (RaVHCAMP) was also expressed as
`above, and the antibodies were purified and quantified (Table
`1). The HoVHCAMP and RaVHCAMP antibodies, each of the
`rat IgG2b isotype, were compared to the CAMPATH-1 antigen
`in a direct binding assay and in complement lysis of human
`lymphocytes (Table 1). Comparedwith the original rat antibody,
`or the engineered equivalent, the antibody with the reshaped
`heavy-chain domain bound poorly to the CAMPATH-1 antigen
`and was weakly lytic. This suggested an error in the design of
`the reshaped domain.
`There are several assumptions underlying the transfer of
`hypervariable loops from one antibody to another”, in particular
`the assumption that the antigen binds mainly to the hypervari-
`able regions. These are defined as regions of sequence” or
`structural*? hypervariability,
`the locations of hypervariable
`regions being similar by both criteria except for the first hyper-
`variable loop of the heavy chain. By sequencethefirst hyper-
`variable loop extends from residues 31-35 (ref. 25) whereas by
`structure it extends from residues 26-32 (ref. 32). Residues 29
`and 30 form part of the surface loop, and residue 27, which is
`phenylalanine or tyrosine in most sequences, including YTH
`35.5HL, helps pack against residues 32 and 34 (Fig. 3). Unlike
`most human heavy chains, in NEW (see note in Fig. 1) the
`phenylalanine is replaced by serine, which would be unable to
`pack in the same way. To restore the packing of the loop, we
`made both a Ser 27> Phe mutation, and a Ser 27> Phe, Ser
`30 Thr double mutation in HuVHCAMP.These two mutants
`showed a significant increase in binding to CAMPATH-1 antigen
`and lysed human lymphocytes with human complement (Table
`1). Thus the affinity of the reshaped antibody could be restored
`by a single Ser 27 > Phe mutation, possibly as a consequence of
`an altered packing between the hypervariable regions and the
`framework. This
`suggests
`that alterations
`in the ‘Kabat’
`framework region can enhance theaffinity of the antibody and
`extends previous work in which an engineered changein the
`hypervariable region yielded an antibody with increased
`affinity®’.
`Heavy-chain constant domains
`In stage 2 (Fig. 2), the rat heavy-chain variable domain was
`attached to constant domains of the humanisotypes IgG1, 2, 3
`and 4, and transfected into the heavy-chain loss variant of the
`YTH34.5 hybridoma. In complementlysis (Fig. 4a), the human
`IgG1 isotype proved similar to the YTH 34.5HL-G2b, with the
`human IgG3 isotype being less effective. The human IgG2
`isotype was only weakly lytic and the IgG4 isotype was non-lytic.
`In ADCC (Fig. 4b) the human IgG1 was more lytic than the
`YTH 34.5HL-G2b antibody. The decrease in lysis at higher
`concentrations of the rat IgG2b and the human IgG1 antibody
`is due to an excess of antibody, which causesthe lysis of effector
`cells. The human IgG3 antibody was weakly lytic, and the IgG2
`and IgG4 isotypes were non-lytic.
`
`lysis 04
`lysis .0001
`
`A
`
`1
`
`10
`
`100
`
`1,000
`
`g P
`
`ercent
`
`Percent
`
`004
`
`014
`
`A
`
`1
`
`10
`
`100
`
`Antibody concentration, jg mi!
`
`a, Complement lysis and b, ADCC for antibodies with rat
`Fig. 4
`
`light-chain and rat heavy-chain variable domain attached to human
`
`
`
`IgG1 (Q), IgG2 (©), IgG3 (Ml), or IgG4 (V) isotypes. Lysis with
`the YTH 34.5HL antibody (@) is also shown.
`Methods. Antibody was collected from cells in stationary phase,
`concentrated by precipitation with ammonium sulphate and desal-
`ted into phosphate buffered saline (PBS). Antibodies bound to the
`CAMPATH-1 antigen-coated on microtitre plates, were assayed
`in ELISA directed against the rat « light chain*®, and each adjusted
`to the same concentration. The antibodies were assayed in comple-
`mentlysis (Table 1) and ADCC with activated human peripheral
`blood mononuclear cells**“*, Briefly, 5x 10* human peripheral
`blood cells were labelled with *'Cr and incubated for 30 min at
`room temperature with different concentrations of antibody. Excess
`antibody was removed anda 20-fold excess of activated cells added
`as effectors. After 4h at 37°C cell death was estimated by *'Cr
`release.
`
`4 of 5
`
`OnCusp
`Ex. 1035
`
`

`

`View publication stats
`
`5 of 5
`
`OnCusp
`Ex. 1035
`
`