Proc. Natl. Acad. Sci. USA
`Vol. 89, pp. 4285-4289, May 1992
`Immunology
`
`Humanization of an anti-p18
`cancer therapy
`(antibody engineering/site-directed mutagenesis/c—erbB-Z/neu)
`
`5HER2 antibody for human
`
`PAUL CARTER*, LEN PRESTA*, CORNELIA M. GORMANT, JOHN B. B. RIDGWAYT, DENNIS HENNERT,
`WA] LEE T. WONoi, ANN M. ROWLANDi, CLAIRE Korrsi, MONIQUE E. CARVERi,
`AND H. MICHAEL SHEPARD§
`
`Departments of ‘Protein Engineering, TCell Genetics, *Medicinal and Analytical Chemistry, and 5Cell Biology, Genentech Inc., 460 Point San Bruno
`Boulevard, South San Francisco, CA 94080
`
`Communicated by Hilary Koprowski, January 16, 1992 (received for review February 15, I991)
`
`The murine monoclonal antibody mumAb4D5,
`ABSTRACT
`directedagaimthumnepidermalgrowthfactorreceptorz
`(pl85mz), specifically inhibits proliferation of human tumor
`cells overexpressing pissflm. However,
`the efficacy of
`mumAMDShhumancancertherapyislikelytobelimitedbya
`human anti-mouseantibodyresponseandlackofefl'ectorfunc-
`tious. A “humanized” antibody, humAb4D5-l, containing only
`theautigenbindingloopsfrommumAbttDSandhumanvariable
`region framework residues plus IgGl constant domains was
`constructed. Light- and heavy-chain variable regions were simul-
`taneously humanized in one step by “gene conversion mutagen-
`em” using 3ll-mer and 361-mer pmembled oligonudeotides,
`respectively. The humAMDS-l variant does not block the pro-
`liferation of human brew carcinoma SK-BR-3 cells, which
`overexpress p185“, deflate tight antigen binding (K., = 25
`nM).0neofsevenadditionalhumanizedvaflantsdedgnedby
`molecular modeling (humAb4DS-8) binds the [318sz antigen
`ISO-fold and 3-fold more tightly than humAMDS-l and
`mumAMDS, respectively. In addition, humAMDS-s has potency
`comparable to the murine antibody in blocking SK-BR-3 cell
`proliferation. Furthermore, humAMDS-s is much more efficient
`in supporting antibody-dependent cellular cytotoxicity against
`SK-BR-3 celk than mumAMDS, but it does not clficiently kl]
`WI-38 cells, which express p185“Em at lower levels.
`
`The protooneogene HER2 encodes a protein tyrosine kinase
`(p185HER2)
`that
`is homologous to the human epidermal
`growth factor receptor (1—3). Amplification and/or overex-
`pression of HER2 is associated with multiple human malig-
`nancies and appears to be integrally involved in progression
`of 25-30% of human breast and ovarian cancers (4, 5).
`Furthermore, the extent of amplification is inversely corre-
`lated with the observed median patient survival time (5). The
`murine monoclonal antibody mumAb4D5 (6), directed
`against the extracellular domain (ECD) of pISSHERZ, specif~
`ically inhibits the grthh of tumor cell lines overexpressing
`p185HER2 in monolayer culture or in soft agar (7, 8).
`mumAb4D5 also has the potential of enhancing tumor cell
`sensitivity to tumor necrosis factor (7, 9). Thus, mumAb4D5
`has potential for clinical intervention in carcinomas involving
`the overexpression of p18SHER2.
`A major limitation in the clinical use of rodent mAbs is an
`anti-globulin response during therapy (10, 11). A partial
`solution to this problem is to construct chimeric antibodies by
`coupling the rodent antigen-binding variable (V) domains to
`human constant (C) domains (12—14). The isotype of the
`human C domains may be varied to tailor the chimeric
`antibody for participation in antibody-dependent cellular
`
`The publication costs of this article were defrayed in part by page charge
`payment. This article must therefore be hereby marked “advertisement“
`in accordance with 18 U.S.C. §1734 solely to indicate this fact.
`
`cytotoxicity (ADCC) and complement-dependent cytotoxic-
`ity (CDC) (15). Such chimeric antibody molecules are still
`~30% rodent in sequence and are capable of eliciting a
`significant anti-globulin response.
`Winter and coworkers (16—18) pioneered the “humaniza-
`tion” of antibody V domains by transplanting the comple-
`mentarity determining regions (CDRs), which are the hyper-
`variable loops involved in antigen binding, from rodent
`antibodies into human V domains. The validity of this ap-
`proach is supported by the clinical efficacy of a humanized
`antibody specific for the CAMPATH—l antigen with two
`non-Hodgkin lymphoma patients, one of whom had previ-
`ously developed an anti—globulin response to the parental rat
`antibody (17, 19). In some cases, transplanting hypervariable
`loops from rodent antibodies into human frameworks is
`sufficient to transfer high antigen binding affinity (16, 18),
`whereas in other cases it has been necessary to also replace
`one (17) or several (20) framework region (FR) residues. For
`a given antibody, a small number of FR residues are antici—
`pated to be important for antigen binding. First, there are a
`few FR residues that directly contact antigen in crystal
`structures of antibody—antigen complexes (21). Second, a
`number of FR residues have been proposed (22—24) as
`critically affecting the conformation of particular CDRs and
`thus their contribution to antigen binding.
`Here we report the rapid and simultaneous humanization of
`heavy-chain (VH) and light-chain (VI) V region genes of
`mumAb4D5 by using a “gene conversion mutagenesis" strat-
`egy (43). Eight humanized variants (humAb4D5) were con-
`structed to probe the importance of several FR residues
`identified by our molecular modeling or previously by others
`(22—24). Efficient transient expression of humanized variants
`in nonmyeloma cells allowed us to rapidly investigate the
`relationship between binding affinity for p185HER2 ECD and
`antiproliferative activity against p185HER2 overexpressing car-
`cinoma cells.
`
`MATERIALS AND METHODS
`
`Cloning of V Region Genes. The mumAb4D5 VH and VL
`genes were isolated by PCR amplification of mRNA from the
`corresponding hybridoma (6) as described (25). N-terminal
`sequencing of mumAb4D5 VL and VH was used to design the
`sense—strand PCR primers, whereas the anti-sense PCR prim-
`ers were based on consensus sequences of murine FR resi-
`
`Abbreviations: mumAb4D5 and humAb4D5, mun‘ne and humanized
`versions of the monoclonal antibody 4D5, respectively; ECD, ex-
`tracellular domain; ADCC, antibody-dependent cellular cytotoxic-
`ity; CDC, complement-dependent cytotoxicity; CDR, complemen-
`tarity—determining region; FR, framework region; VH and VL, vari-
`able heavy and light domains, respectively; C region, constant
`region; V region, variable region.
`
`4285
`
`(cid:43)(cid:82)(cid:86)(cid:83)(cid:76)(cid:85)(cid:68)(cid:3)(cid:89)(cid:17)(cid:3)(cid:42)(cid:72)(cid:81)(cid:72)(cid:81)(cid:87)(cid:72)(cid:70)(cid:75)(cid:3)
`Hospira v. Geuentech
`(cid:44)(cid:51)(cid:53)(cid:21)(cid:19)(cid:20)(cid:26)(cid:16)(cid:19)(cid:19)(cid:27)(cid:19)(cid:24)(cid:3)
`IPR2017-00805
`Geueutech Exhibit 2057
`(cid:42)(cid:72)(cid:81)(cid:72)(cid:81)(cid:87)(cid:72)(cid:70)(cid:75)(cid:3)(cid:40)(cid:91)(cid:75)(cid:76)(cid:69)(cid:76)(cid:87)(cid:3)(cid:21)(cid:19)(cid:24)(cid:26)
`
`

`

`4286
`
`Immunology: Carter et al.
`
`Proc. Natl. Acad. Sci. USA 89 (1992)
`
`muMAb4D5 VL
`40
`30
`20
`10
`I T C
`D
`H
`D
`I V M T Q S
`P
`K A S Q D V N T A V A W Y Q Q K
`R V S
`K F H S T S V G
`GCTGATATCGTGATGACCCAGTCCCACAAGTTCND3DJHKI3H7RTFiIX1HUDMIXflTmGCATCAEIflYIIAGGCtAGTCAGGA1GTGAAHICTQfflTTNEJTRIFTATCAACAGAAACCA
`no u. n -
`GCTGATATCCAGATGACCCAGTIX1XJ3AGC1YIKID3flXXIITFCF3F3GGCGATAIKXfD2ACCATCAKI7H33I3F33CAG1C1GGA1V1D3AAflTKID3IDSTAGCKHVITTAJCAACAGAAACCA
`D
`I Q M T Q S
`P
`S
`S
`L
`S A S V G
`D R V T I T C
`R A S 9 D V N T A V A W Y Q Q K
`P
`vL——CDR1
`huMAMDS-S VL
`
`50
`Y S A S
`
`F R Y T G V
`
`P
`
`60
`D R
`
`d
`I
`F T G N R
`
`0
`S
`
`70
`G T D
`
`P T F T I
`
`Y
`
`F
`S A S
`vL—cuu
`
`L
`
`E S G V P
`
`S
`
`R
`
`F
`
`S
`
`G S
`
`R
`
`S G T D
`
`F T L T I
`
`80
`S V 0 A E
`
`S
`
`L Q
`
`P
`
`E
`
`S
`
`S
`
`G H S
`
`F K L L
`
`G K A P
`
`K
`
`L L
`
`I
`
`I
`
`A
`
`B
`
`FIG. 1. Nucleotide and amino
`acid sequences of mumAb4D5 and
`humAMDS-S V; (A) and VH (B)
`(numbered according to ref. 26).
`The CDR residues according to a
`sequence definition (26) and a
`structural definition (22) are un-
`derlined and overlined, respec-
`tively. The 5’ and 3’ ends of the
`oligonucleotides used for gene
`conversion mutagenesis are
`shown by arrows and mismatches
`30
`20
`10
`muMAMDS VH
`QVQLQQSGPELVKPGASLKLSCTASGFNIKDTYIHWVK betweengenesareshownbyas-
`on
`,.
`@JEHKXXflCN1H1CAGCFI1GCN3K3flIIXXflGAGCDIflGANIKJGGGI37N3KHIJAGTD1flXflGTACAGCTDTKKXHHCAACNPTAAAGACNJH“HRTACACEIXHGAAA
`terisks. The asparagine-linked gly-
`GaETACXi3D3AlXFFF:AGIflVXFRIflKIDIFIIJ3SRIiJ3HITn33UI32ALKIIII3ICAEflVXI1PTn3DI3flFDI3VSTTD3FI33rF2AACAH1AAAGACAKIHTJITTCACHTXIFHIJEF
`cosylation site (#) in mumAb4D5
`E V Q L V B
`S G G G L V Q P G G S
`L
`R L
`S
`C A A S G F N I
`K D T Y
`I
`H W V R
`vH-cuu
`VL is used in some mumAb4D5
`molecules derived from the corre-
`sponding hybridoma. However,
`mumAb4D5 variants, which are
`glycosylated or aglycosylated in
`VL, are indistinguishable in their
`binding affinity for the p18SHER2
`ECD and in their antiproliferative
`activity with SK-BR-3 cells (C.K. ,
`M. Spellman, and B. Hutchins,
`unpublished data).
`
`100
`90
`K
`I
`P T F G G G T K V E
`P
`D L A V Y Y C Q Q H Y T T
`GAKIIn332AGT11A1TACflTFD2AGCAACAT1RTIEfl2KflEIflLXX3KIFFDI1EU3331STACCAAGGTGGAGATCAAA
`GACTFJXIAIH1A1TRCTGTCAGCAACNTTATACTAEflKXfltXImCGT1CGGACAGGGTACCAAGGTGGAGATCAAA
`D F A T Y Y C Q 9 H Y T T P
`P T F
`G Q G T
`K V E
`I
`K
`VL-CDRB
`
`huMAMDS-S VH
`60
`40
`a
`50
`7°
`S N T A
`P K F O D K A T I T A D T S
`Y D
`Y
`Q R
`P
`E Q G L
`P T N G Y T R
`E W I G R I
`1V3AKXII3AAK:TTISIAKIEACAAEXIIIAKTPA
`TAACAGCAGACACATCKHEIZACACAGCC
`CN3I1XHGGAAJGGNPRIHAGGATTTAflJflTCGANDIHHEflACHAGATA
`a .*. «'1 e
`a tt‘t'
`non
`. ttt
`CAGGCKXXXII1PAAIXIIIIF33AAflKII§Pn32AAGGAT1111IIflWKl3AA1VIFTTAHACTAGATTflTXX33ATAIKIFF3AA(XII333TTF2ACTATAAKIJJCAGACACATCCAAAAACACAGCC
`T N G Y T R
`Q A P
`G
`K G L
`E W V A R
`I
`Y P
`Y A D S V K G R
`F T I
`S A D T S
`K N T A
`vH—cmz
`
`110
`100 a
`90
`c
`b
`a
`80
`S
`R W G G D G F
`Y A H D Y W G Q G A S V T V S
`Y C S
`E D T A V Y
`L T S
`R
`Y L O V S
`tit.
`at. t a
`TNI3GCNRTn3MXXIXXHGACNFTDHMIaCACTEKIHTflRTTATF1rKflMGNFfl3XflGGGGNUiXflTtHATGCTNR3E“HECTEIXEVEAGGMIIICGGTCNJITRJVIHEB
`1ACIflVI2AGATGAACAGCCflYKI3NI3F3AGGACACflYIIJFFIPA1Tmfl1VTFN2IAGA1YKIXBU1133Afl1XIIFN3TATVI3TNn33AIXYDSDI3I3D2AAGGAALXXH‘XTDSNJJSFZDJSFJS
`V W G O G T L V T V S
`S
`Y L Q N N S
`L
`R A E D T A V Y
`Y C S
`R W G G D G F Y A M D
`Vg<w
`
`dues (25, 26) incorporating restriction sites for directional
`cloning shown by underlining and listed after the sequences:
`VL sense, 5'~TCCGATATCCAGCTGACCCAGTCTCCA-3’
`EcoRV; VL antisense, 5’-GTTTGATCTCCAGCTTG_G-
`TACCHSCDCCGAA-3’ Asp718; VH sense, 5’-AGGTSM-
`ARCTGCAGSAGTCWGG-3’ Pst 1; Va antisense, 5 ’-
`TGAGGAGACGGTGACCGTGGTCCC'l‘TGGCCCCAG-3’
`BstEII; where H is A, C, or T; S is C or G; D is A, G, or T;
`MisAorC; Ris AorG;WisAorT. The PCR products were
`cloned into pUC119 (27) and five clones for each V domain
`were sequenced by the dideoxynucleotide chain-termination
`method (28).
`Molecular Modeling. Models of mumAb4D5 VH and VL
`domains were constructed by using seven Fab crystal struc-
`tures from the Brookhaven Protein Data Bank (entries 2FB4,
`2RHE, 2MCP, 3FAB, 1FBJ, 2HFL, and 1REI) (29). VH and
`VL of each structure were superimposed on 2FB4 by using
`main-chain atom coordinates (INSIGHT program, Biosym
`Technologies, San Diego). The distances from each 2FB4 Caz
`to the analogous Ca1n each of the superimposed structures
`was calculated. For residues with all Ca—Ca distances<_1A,
`the average coordinates for individual N, Ca, C, O, and C3
`atoms were calculated and then corrected for resultant de-
`viations from standard bond geometry by 50 cycles of energy
`minimization (DISCOVER program, Biosym Technologies) us-
`ing the AMBER forcefield (30) and fixed Ca atoms. Side chains
`of FR residues were then incorporated, followed by inclusion
`of five of the six CDR loops (except VH—CDR3) using
`tabulations of CDR conformations (23) as a guide. Side-chain
`conformations were chosen on the basis of Fab crystal
`structures, rotamer libraries (31), and packing consider-
`ations. Three possible conformations of VH—CDR3 were
`taken from a search of similar sized loops in the Brookhaven
`Protein Data Bank or were modeled by using packing and
`solvent exposure considerations. Models were then sub-
`jected to 5000 cycles of energy minimization.
`A model of the humAb4D5 was generated by using consen-
`sus sequences derived from the most abundant human sub-
`classes—namely, VL K subgroup I and VH subgroup III (26).
`The six CDRs were transferred from the mumAb4D5 model
`onto a human Fab model. All humAb4D5 variants contain
`
`human replacements of mumAb4D5 residues at three positions
`within CDRs as defined by sequence variability (26) but not as
`defined by structural variability (22): VL—CDRI K24R, VL—
`CDR2 R54L and VL—CDRZ T56S.ll Differences between
`mumAb4D5 and the human consensus FR residues (Fig. 1)
`were individually modeled to investigate their possible influ-
`ence on CDR conformation and/or binding to p185HERZ ECD.
`Construction of Chimeric Genes. Genes encoding the chi-
`meric mAb4D5 light and heavy chains were separately as-
`sembled in previously described phagemid vectors contain-
`ing the human cytomegalovirus enhancer and promoter, a 5'
`intron, and simian virus 40 polyadenylylation signal (32).
`Briefly, gene segments encoding mumAb4D5 VL (Fig. 1A)
`and REI human K1 light-chain CL (33) were precisely joined
`as were genes for mumAb4D5 V" (Fig. IB) and human IgGl
`C region (34) by subcloning (35) and site-directed mutagen—
`esis as described (36). The IgGl isotype was chosen, as it is
`the preferred human isotype for supporting ADCC and CDC
`by using matched sets of chimeric (15) or humanized anti-
`bodies (17). The PCR-generated VL and VH fragments (Fig.
`1) were subsequently mutagenized so that they faithfully
`represent the sequence of mumAb4D5 determined at the
`protein level: VH, QlE; VL, V104L and T109A. The human
`IgGl C regions are identical to those reported (37) except for
`the mutations E359D and M361L (Eu numbering; ref. 26),
`which we installed to convert the antibody from the naturally
`rare A allotype to the much more common non-A allotype
`(26). This was an attempt to reduce the risk of anti-allotype
`antibodies interfering with therapy.
`Construction of Humanized Gem. Genes encoding chi-
`meric mAb4D5 light-chain and heavy-chain Fd fragment (VH
`and CHI domains) were subcloned together into pUC119 (27)
`to create pAKl and were simultaneously humanized in a
`single step (43). Briefly, sets of six contiguous oligonucleo—
`tides were designed to humanize VH and VL (Fig. 1). These
`oligonucleotides are 28—83 nucleotides long, contain 0—19
`mismatches to the murine antibody template, and are con-
`
`1Variants are denoted by the amino acid residue and number
`followed by the replacement amino acid.
`
`

`

`Immunology: Carter et al.
`
`Proc. Natl. Acad. Sci. USA 89 (1992)
`
`4287
`
`strained to have 8 or 9 perfectly matched residues at each end
`to promote efficient annealing and ligation of adjacent oligo-
`nucleotides. The sets of VH and VL humanization oligonu-
`cleotides (5 pmol each) were phosphorylated with either ATP
`or [7-32P]ATP (36) and separately annealed with 3.7 pmol of
`pAKl template in 40 pl of 10 mM Tris-HCl (pH 8.0) and 10
`mM MgCl2 by cooling from 100°C to =20°C over ~20 min.
`The annealed oligonucleotides were joined by incubation
`with T4 DNA ligase (12 units; New England Biolabs) in the
`presence of 2 p.1 of 5 mM ATP and 2 pl of 0.1 M dithiothreitol
`for 10 min at 14°C. After electrophoresis on a 6% acrylamide
`sequencing gel, the assembled oligonucleotides were located
`by autoradiography and recovered by electroelution. The
`assembled oligonucleotides (~03 pmol each) were simulta-
`neously annealed to 0.15 pmol of single-stranded deoxyuri-
`dine-containing pAKl prepared as described (38) in 10 pl of
`40 mM Tris-HCl (pH 7.5) and 16 mM Mng as described
`above. Heteroduplex DNA was constructed by extending the
`primers with T7 DNA polymerase and transformed into
`Escherichia coli BMH 71-18 mutL as described (36). The
`resultant phagemid DNA pool was enriched first for human
`VL by restriction purification using Xho I and then for human
`VH by restriction selection using Stu I as described (36, 39).
`Resultant clones containing both human VL and human VH
`genes were identified by nucleotide sequencing (28) and
`designated pAK2. Additional humanized variants were gen-
`erated by site-directed mutagenesis (36). The mumAb4D5 VL
`and V" gene segments in the transient expression vectors
`described above were then precisely replaced with their
`humanized versions.
`Expression and Purification of mAb4D5 Variants. Appro-
`priate mAb4D5 light- and heavy-chain cDNA expression
`vectors were cotransfected into adenovirus-transformed hu-
`man embryonic kidney cell line 293 by a high-efficiency
`procedure (32). Media were harvested daily for up to 5 days
`and the cells were refed with serum-free medium. Antibodies
`were recovered from the media and affinity purified on
`protein A-Sepharose CL-4B (Pharmacia) as described by the
`manufacturer. The eluted antibody was buffer-exchanged
`into phosphate-buffered saline by 025 gel filtration, concen-
`trated by ultrafiltration (Amicon), sterile-filtered, and stored
`at 4°C. The concentration of antibody was determined by
`both total IgG and antigen binding ELISAs. The standard
`used was humAb4D5-5, whose concentration had been de-
`termined by amino acid composition analysis.
`Cell Proliferation Assay. The effect of mAb4D5 variants on
`proliferation of the human mammary adenocarcinoma cell
`line SK-BR-3 was investigated as described (6) by using
`saturating mAb4D5 concentrations.
`
`Affinity Measurements. mAb4D5 variant antibodies and
`p185HERz ECD were prepared as described (40) and incubated
`in solution until equilibrium was found to be reached. The
`concentration of free antibody was then determined by
`ELISA using immobilized p185HER2 ECD and was used to
`calculate affinity (Kd) as described (41). The solution-phase
`equilibrium between p185HERZ ECD and mAb4D5 variants
`was found not to be grossly perturbed during the immobilized
`ECD ELISA measurement of free antibody.
`
`RESULTS
`
`Humanization of mumAb4D5. The mumAb4D5 VL and VH
`gene segments were first cloned by PCR and sequenced (Fig. 1).
`The V region genes were then simultaneously humanized by
`gene conversion mutagenesis using preassembled oligonucleo-
`tides. A 311-mer oligonucleotide containing 39 mismatches to
`the template directed 24 simultaneous amino acid changes
`required to humanize mumAb4D5 VL. Humanization of
`mumAb4D5 VH required 32 amino acid changes, which were
`installed with a 361-mer containing 59 mismatches to the
`mumAb4D5 template. Two of eight clones sequenced precisely
`encode humAb4D5-5, although one of these clones contained a
`single nucleotide imperfection. The six other clones were es-
`sentially humanized but contained a small number of errors: <3
`nucleotide changes and <1 single nucleotide deletion per kilo-
`base. Additional humanized variants (Table 1) were constructed
`by site-directed mutagenesis of humAb4D5-5.
`Expression levels of humAb4D5 variants were 7—15 ug/ml
`as judged by ELISA using immobilized pISSHERZ ECD.
`Successive harvests of five 10-cm plates allowed 200—500 ug
`of each variant to be produced in a week. Antibodies affinity
`purified on protein A gave a single band on a Coomassie
`blue-stained SDS/polyacrylamide gel of mobility consistent
`with the expected mass of 2150 kDa. Electrophoresis under
`reducing conditions gave two bands consistent with the
`expected mass of free heavy (48 kDa) and light (23 kDa)
`chains (data not shown). N-terminal sequence analysis (10
`cycles) gave the mixed sequence expected (see Fig. 1) from
`an equimolar combination of light and heavy chains.
`humAb4D5 Variants. In general, FR residues were chosen
`from consensus human sequences (26) and CDR residues
`were chosen from mumAb4D5. Additional variants were
`constructed by replacing selected human residues in
`humAb4D5-1 with their mumAb4D5 counterparts. These are
`VH residues 71, 73, 78, 93, plus 102 and VL residues 55 plus
`66. VH residue 71 has previously been proposed by others
`(24) to be critical to the conformation of VH—CDRZ. Amino
`acid sequence differences between humAb4D5 variant mol-
`ecules are shown in Table 1 together with their p18SHER2 ECD
`
`Table 1.
`
`mAb4D5
`variant
`humAb4D5-1
`humAb4D5-2
`humAb4D5-3
`humAb4D5—4
`humAb4D5-5
`humAb4D5-6
`humAb4D5-7
`humAb4D5-8
`humAb4D5
`
`71
`(FR3)
`R
`Ala
`Ala
`Ala
`Ala
`Ala
`Ala
`Ala
`Ala
`
`73
`(FR3)
`D
`D
`Thr
`Thr
`Thr
`Thr
`Thr
`Thr
`Thr
`
`p185"ER2 ECD binding affinity and anti-proliferative activities of mAb4D5 variants
`VH residue
`VL residue
`78
`(FR3)
`L
`L
`Ala
`L
`Ala
`Ala
`Ala
`Ala
`Ala
`
`93
`(FR3)
`A
`A
`Ser
`Ser
`Ser
`Ser
`Ser
`Ser
`Ser
`
`102
`(CDR3)
`V
`V
`V
`V
`V
`V
`Tyr
`Tyr
`Tyr
`
`55
`(CDR2)
`E
`E
`E
`E
`E
`Tyr
`E
`Tyr
`Tyr
`
`66
`(FR3)
`G
`G
`G
`Arg
`Arg
`Arg
`Arg
`Arg
`Arg
`
`.
`
`Kd,
`nM
`25
`4.7
`4.4
`0.82
`1. 1
`0.22
`0.62
`0.10
`0.30
`
`Relative cell
`proliferation
`102
`101
`66
`56
`48
`51
`53
`54
`37
`
`Human and murine residues are shown in one-letter and three-letter amino acid codes, respectively. Kd values for the p185HERZ ECD were
`determined by the method of Friguet et al. (41) and the standard error of each estimate is :10%. Proliferation of SK-BR-3 cells incubated for
`96 hr with mAb4D5 variants is shown as a percentage of the untreated control as described (7). Data represent the maximal antiproliferative
`effect for each variant (see Fig. 2) calculated as the mean of triplicate determinations at a mAb4D5 concentration of 8 ug/ml. Data are all taken
`from the same experiment and the estimated standard error is t15%.
`
`

`

`4288
`
`100
`
`00O
`
`QC
`
`Percentofcontrolcell
`proliferatlon
`
`Immunology: Carter et a1.
`
`
`huMAb4D5-1
`
`
`
`
`
`
`huMAb4D5-8
`
`Proc. Natl. Acad. Sci. USA 89 (1992)
`
`placement ofR71 in humAb4D5-l with the corresponding murine
`residue, A71 (hurnAb4D5-2). In contrast, replacing VH L78 in
`humAb4D5-4 with the murine residue A78 (humAb4D5-5) does
`not significantly change the affinity for the p185HERZ ECD or
`change antiproliferative activity, suggesting that residue 78 is not
`of critical functional significance to humAb4D5 in interacting
`with p185HERZ ECD.
`VL residue 66 is usually a glycine in human and murine
`K-chain sequences (26) but an arginine occupies this position
`in the mumAb4D5 K light chain. The side chain of residue 66
`is likely to affect the conformation of VL—CDRI and VL—
`CDR2 and the hairpin turn at residues 68-69 (Fig. 3). Con-
`sistent with the importance of this residue, the mutation VL
`666R (humAb4D5-3 —> humAb4D5-5) increases the affinity
`for the p18SHER2 ECD by 4—fold with a concomitant increase
`in antiproliferative activity.
`From molecular modeling, it appears that the side chain of
`mumAb4D5 VL Y55 may either stabilize the conformation of
`VH—CDR3 or provide an interaction at the VL—VH interface.
`The latter function may be dependent on the presence of VH
`Y102. In the context of humAb4D5-5 the mutations VL E55Y
`(humAb4D5-6) and V" V102Y (humAb4D5-7) individually
`increase the affinity for p185"ERZ ECD by 5-fold and 2-fold,
`respectively, whereas together (humAb4D5-8) they increase
`the affinity by ll-fold. This is consistent with either proposed
`role of VL Y55 and V" Y102.
`Secondary Immune Function of humAMDS-s. humAb4D5-8
`efficiently mediates ADCC against SK-BR-3 breast carcinoma
`cells, which overexpress p185HERZ at high levels as anticipated
`from its IgGl isotype (Table 2). In contrast, humAb4D5—8 is
`very ineflicient in mediating ADCC against the normal lung
`epithelium cell line WI-38, which expresses p18SHER2 at 100-
`fold lower levels than SK—BR—3 cells (Table 2). The murine
`parent antibody is not very effective in mediating ADCC against
`either SK-BR—3 or WI-38 cells.
`
`DISCUSSION
`
`mumAb4D5 is potentially useful for human therapy since it is
`cytostatic toward human breast and ovarian tumor lines over-
`expressing p18SHER2. Here we have humanized mumAb4D5 in
`an attempt to improve its potential clinical efficacy by reducing
`its immunogenicity and tailoring the Fc region to support ADCC
`and possibly CDC.
`Rapid humanization of humAb4D5 was facilitated by the
`gene conversion mutagenesis strategy developed here using
`long preassembled oligonucleotides. This method uses less
`
`4
`
`12
`8
`[MAb4D5 variant] ttg/ml
`
`16
`
`Inhibition of SK—BR-3 proliferation by mAb4D5 variants.
`FIG. 2.
`Relative cell proliferation was determined as described (7) and data
`(average of triplicate determinations) are presented as a percentage
`of results with untreated cultures for mumAb4D5, humAb4D5-8, and
`humAb4D5-1.
`
`binding affinity and maximal antiproliferative activities
`against SK-BR-3 cells. Very similar Kd values were obtained
`for binding mAb4D5 variants to either SK-BR—3 cells (GK.
`and N. Dua, unpublished data) or to p18SHER2 ECD (Table 1).
`The most potent humanized variant designed by molecular
`modeling, humAb4D5-8, contains five FR residues from
`mumAb4D5. This antibody binds the p18SHER2 ECD 3-fold
`more tightly than does mumAb4D5 itself (Table l) and has
`comparable antiproliferative activity with SK-BR—3 cells
`(Fig. 2). In contrast, humAb4D5-1 is the most humanized but
`least potent mumAb4D5 variant, created by simply installing
`the mumAb4D5 CDRs into the consensus human sequences.
`humAb4D5-1 binds the p185HERZ ECD 80—fold less tightly
`than does the murine antibody and has no detectable antipro-
`liferative activity at the highest antibody concentration in-
`vestigated (16 pg/ml).
`The antiproliferative activity of humAb4D5 variants
`against p185HER2 overexpressing SK-BR-3 cells is not simply
`correlated with their binding affinity for the p185HERZ ECD-—
`e.g. , installation of three murine residues into the VH domain
`of humAb4D5-2 (D73T, L78A, and A938) to create
`humAb4D5-3 does not change the antigen binding affinity but
`does confer significant antiproliferative activity (Table 1).
`The importance of VH residue 71 (24) is supported by the
`observed 5-fold increase in affinity for p185HERZ ECD on re-
`
`
`
`FIG. 3. Stereoview of a-car-
`bon tracing for model of hum-
`Ab4DS-8 VL and V“. The CDR
`residues (26) are shown in boldface
`and side chains of VH residues
`A71, T73, A78, S93, and Y102 and
`VL residues Y55 and R66 (see Ta-
`ble 1) are shown.
`
`

`

`Immunology: Carter et al.
`
`Proc. Natl. Acad. Sci. USA 89 (1992)
`
`4289
`
`Table 2. Selectivity of ADCC mediated by mAb4D5 variants
`Efl‘ector/
`SK-BR-3
`WI-38
`target ——
`ratio
`mumAb4D5 humAb4D5-8 mumAb4D5 humAb4D5-8
`
`25:1
`12.5:1
`6.25:1
`3.13:1
`
`25:1
`12.521
`6.25:1
`3.13:1
`
`Antibody concentration, 100 ng/ml
`<1.0
`9.3
`7.5
`<1.0
`11.1
`4.7
`<1.0
`8.9
`0.9
`<1.0
`8.5
`4.6
`Antibody concentration, 10 ng/ml
`<1.0
`3.1
`6.1
`<1.0
`1.7
`5.5
`1.3
`2.2
`2.0
`<1.0
`0.8
`2.4
`
`40.6
`36.8
`35.2
`19.6
`
`33.4
`26.2
`21.0
`13.4
`
`Sensitivity to ADCC of human cell lines WI—38 (normal lung
`epithelium) and SK-BR-3 (breast tumor), which express 0.6 and 64
`pg of p185mam per pg of cell protein, respectively, as determined by
`ELISA (40). ADCC assays were carried out as described (15) using
`interleukin 2 activated human peripheral blood mononuclear cells as
`effector cells and either WI-38 or SK-BR—3 target cells in 96-well
`microtiter plates for 4 hr at 37°C at different antibody concentrations.
`Values given represent percentage specific cell lysis as determined
`by “Cr release. The estimated standard error in these quadruplicate
`determinations was :10%.
`
`than half the amount of synthetic DNA, as does total gene
`synthesis, and does not require convenient restriction sites in
`the target DNA. Our method appears to be simpler and more
`reliable than a similar protocol recently reported (42). Tran-
`sient expression of humAb4D5 in human embryonic kidney
`293 cells permitted the isolation of 0.2- to 0.5-mg humAb4D5
`variants for rapid characterization by growth inhibition and
`antigen binding affinity assays. Furthermore, different com-
`binations of light and heavy chain were readily tested by
`cotransfection of corresponding cDNA expression vectors.
`The crucial role of molecular modeling in the humanization
`of mumAb4D5 is illustrated by the designed variant
`humAb4D5-8, which binds the p185flERZ ECD 250-fold more
`tightly than the simple CDR loop swap variant humAb4D5-1.
`It has previously been shown that the antigen binding affinity
`of a humanized antibody can be increased by mutagenesis
`based on molecular modeling (17, 20). Here we have designed
`a humanized antibody that binds its antigen 3-fold more
`tightly than the parent antibody and is almost as potent in
`blocking the proliferation of SK-BR-3 cells. While this result
`is gratifying, assessment of the success of molecular model-
`ing must await the outcome of ongoing x-ray crystallographic
`structure determination.
`humAb4D5-8 also supports cytotoxicity via ADCC against
`SK-BR-3 tumor cells in the presence ofhuman effector cells but
`is not effective in directing the killing of normal (WI-38) cells,
`which express p185HERZ at much lower levels. This augurs well
`for the ongoing treatment of human cancers overexpressing
`p18SHER2 by using humAb4D5-8.
`
`We thank Bill Henzel for N-terminal sequence analysis of mAb4D5
`variants; Nancy Simpson for sequencing the cDNAs for mumAb4D5
`V-region genes; Maria Yang for providing the CL-containing clone;
`Susie Wong for performing amino acid composition analysis; Irene
`Figari for performing the ADCC assays; Mark Vasser, Parkash
`Jhurani, Peter Ng, and Leonie Meima for synthesizing oligonucleo—
`tides; Bob Kelley for helpful discussions; and Tony Kossiakoff for
`support.
`
`1. Coussens, L., Yang-Feng, T. L., Liao, Y.-C., Chen, E., Gray, A.,
`McGrath, 1., Seeburg, P. H., Libennann, T. A., Schlessinger, J.,
`Francke, U., Levinson, A. & Ullrich, A. (1985) Science 230, 1132—1139.
`2. Yamamoto, T., Ikawa, S., Akiyarna, T., Semba, K., Nomura, N.,
`Miyajima, N., Saito, T. & Toyoshima, K. (1986) Nature (London) 319,
`230-234.
`3. King,C. R., Klaus, M. H. & Aaronson, S. A. (1985) Science 229,974-976.
`
`Slamon, D. J., Clark, G. M., Wong, S. 0., Levin, W. 1., Ullrich, A. &
`McGuire, W. L. (1987) Science 235, 177—182.
`Slamon, D. 1., Godolphin, W., Jones, L. A., Holt, J. A., Wong, S. 6.,
`Keith, D. E., Levin, W. 1., Stuart, S. G., Udove,1., Ullrich, A. & Press.
`M. F. (1989) Science 244, 707-712.
`Fendly, B. M., Winger, M., Hudu'ak, R. M., Lipari, M. T., Napier,
`M. A. & Ullrich, A. (1990) Cancer Res. 50, 1550—1558.
`Hudziak, R. M., Lewis, G. D., Winget. M., Fendly, B. M., Shepard,
`H. M. & Ullrich, A. (1989) Mol. Cell. Biol. 9, 1165—1172.
`Lupu, R., Colomer, R., Zugmaier, 0., Sarup, J., Shepard, M., Slamon,
`D. & Lippman, M. E. (1990) Science 249, 1552—1555.
`Shepard, H. M. & Lewis, G. D. (1988) J. Clin. Immunol. 8, 333—395.
`Miller, R. A., Oserofl‘, A. R., Stratte, P. T. & Levy, R. (1983) Blood 62,
`988-995.
`Schroff, R. W., Foon, K. A., Beatty, S. M., Oldham, R. K. & Morgan,
`A. C., Jr. (1985) Cancer Res. 45, 879-885.
`Morrison, S. 1..., Johnson, M. 1., Herzenberg, L. A. & Oi, V. T. (1984)
`Proc. Natl. Acad. Sci. USA 81, 6851—6855.
`Boulianne, G. L., Hozumi, N. &. Shulman, M. J. (1984) Nature (London)
`312, 643-646.
`Neuberger, M. S., Williams, G. T., Mitchell, E. B.,Jouhal, S. S., Flana-
`gan, J. G. & Rabbitts, T. H. (1985) Nature (London) 314, 268—270.
`Brfiggemann, M., Williams, G. T., Bindon, C. 1., Clark, M. R., Walker,
`M. R., Jefferis, R., Waldmann, H. & Neuberger, M. S. (1987) J. Exp.
`Med. 166, 1351-1361.
`Jones, P. T., Dear, P. H., Foote, 1., Neuberger, M. S. & Winter, G.
`(1986) Nature (London) 321, 522-525.
`Riechmann, L., Clark, M., Waldmann, H. & Winter, G. (1988) Nature
`(London) 332, 323-327.
`Verhoeyen, M., Milstein, C. & Winter, G. (1988) Science 7.39, 1534—1536.
`Hale, G., Dyer, M. J. S., Clark, M. R., Phillips, J. M., Marcus, R.,
`Riechmann, L., Winter, G. & Waldmann, H. (1988) Lancet 1, 1394—1399.
`Queen, C., Schneider, W. P., Selick, H. E., Payne, P. W., Landolfi,
`N. F., Duncan, J. F., Avdalovic, N. M., Levitt, M., Junghans, R. P. &
`Waldmann. T. A. (1989) Proc. Natl. Acad. Sci. USA 86, 10029-10033.
`Mian, l. S., Bradwell, A. R. & Olson, A. J. (1991) J. Mol. Biol. 217,
`133-151.
`Chothia. C. & Lesk, A. M. (1987) J. Mol. Biol. 196, 901—917.
`Chothia, C., Lesk, A. M., Tramontano, A., Levitt. M., Smith-Gill, S. 1.,
`Air, 6., Sheriff, S., Padlan, E. A., Davies, D., Tulip, W. R., Colman,
`P. M., Spinelli, S., Alzari, P. M. & Poljak, R. J. (1989) Nature (London)
`342, 877-883.
`Tramontano, A., Chothia, C. & Lesk, A. M. (1990) J. Mol. Biol. 215,
`175-182.
`Orlandi, R., Gfissow, D. H., Jones, P. T. & Winter, G. (1989) Proc. Natl.
`Acad. Sci. USA 86, 3833—3837.
`Kabat, E. A., Wu, T. T., Reid—Miller, M., Perry, H. M. & Gottesmann,
`K. S. (1987) Sequences ofProteins ofImmunological Interest (Natl. Inst.
`Health, Bethesda, MD).
`Vieira, J. & Messing, J. (1987) Methods Enzymol. 153, 3-11.
`Sanger, F., Nicklen, S. & Coulson, A. R. (1977) Proc. Natl. Acad. Sci.
`USA 74, 5463—5467.
`Bernstein, F. C., Koetzle, T. F., Williams, G. J. 13., Meyer, E. F..
`Brice, M. D., Rodgers. J. R., Kennard, 0., Shimanouchi, T. & Tasumi,
`M. (1977) J. Mol. Biol. 112, 535—542.
`Weiner, S. 1., Kollman, P. A., Case, D. A., Singh, U. C., Ghio, C.,
`Alagona, G., Profeta, S., Jr., & Winer, P. (1984) J. Am. Chem. Soc. 106,
`765—784.
`Ponder, J. W. &. Richards, F. M. (1987) J. Mol. Biol. 1