`
`Humanization of a mouse anti-human IgE antibody: a potential
`therapeutic for IgE-mediated allergies
`
`Frank Kolblnger"3, José Saldanha2, Norman Hardman'
`and Mary M.Bendlg2'4
`
`'CIBA-GEIGY AG, CH-4002, Base!, Switzerland and 2Medical Rescarth
`Council Collaborative Centre, 1-3 Buttonhole Lane, Mill Hill,
`London NW7 lAD, UK
`3Presa,t address: EnrWicklUngsla!SJT für Immunoassays, Obere Hardonrasse 18,
`Pofach 1050, 7800 Freiburg, Germany
`4To whom correspondence should be addressed
`
`Mouse mAb TES-C21(C21) recognizes an epitope on human
`IgE and, therefore, has potential as a therapeutic agent In
`patients with IgE-mediated allergies such as hay fever, food
`and drug allergies and extrinsic asthma. The clinical
`usefulness of mouse antibodies is limited, however, due to
`their immunogenicity In humans. Mouse C21 antibody was
`humanized by complementarity determining region (CDR)
`grafting with the aim of developing an effective and safe
`therapeutic for the treatment of IgE-mediated allergies. The
`CDR-grafted, or reshaped human, C21 variable regions were
`carefully designed using a specially constructed molecular
`model of the mouse C21 variable regions. A key step in the
`design of reshaped human variable regions Is the selection
`of the human framework regions (FRs) to serve as the
`backbones of the reshaped human variable regions. Two
`approaches to the selection of human FRs were tested: (1)
`selection from human consensus sequences and (II) selection
`from individual human antibodies. The reshaped human and
`mouse C21 antibodies were tested and compared using a
`biosensor to measure the kinetics of binding to human IgE.
`Surprisingly, a few of the reshaped human C21 antibodies
`exhibited patterns of binding and affinities that were
`essentially identical to those of mouse C21 antibody.
`Key srds: antibody/biosensor/CDR grafting/human IgE/
`molecular modelling
`
`Introduction
`Mouse mAb TES-C21 (C21) was isolated from mice immunized
`with polyclonal IgE purified from human serum (Davis el al.,
`1993). Mouse mAb C21 binds to secreted human IgE circulating
`in plasma and to the membrane-anchored IgE present on the
`surface of IgE-expressing B cells, but does not interact with IgE
`when it is bound to its low- or high-affinity receptors (FcdU and
`FceRIL respectively) on mast cells, basophils and other cells.
`Mouse mAb C2 1, therefore, does not induce mast cells and
`basophils to release histamine and other mediators that cause
`allergic symptoms. Antibodies that recognize this very defined
`region of human IgE may be useful and safe for clearing
`circulating IgE from the blood and for specifically targeting IgE-
`secreting B cells, but not other cells bearing IgE. These
`antibodies, therefore, may have therapeutic applications in the
`treatment of IgE-mediated allergies (Chang et al., 1990).
`The development and use of mouse mAbs as therapeutic agents
`have been hindered by the human anti-mouse antibody response
`
`(HAMA) which reduces the half-life and, therefore, the efficacy
`of the mouse antibody in patients (see review by Adair et a).,
`1990). in addition, there are risks of adverse side-effects
`associated with repealed administrations of highly immunogenic
`foreign protein to patients. Many of the problems associated with
`the use of mouse mAbs as therapeutic agents could be overcome
`with the use of human mAbs. It has proven technically difficult,
`however, to isolate the latter. In addition, it would be particularly
`difficult to isolate high-affinity human antibodies recognizing self-
`antigens such as human IgE. In order to make mouse monoclonal
`antibodies more acceptable as therapeutic agents, a variety of
`approaches have been developed for rendering mouse mAbs less
`immunogenic in humans by making them resemble human anti-
`bodies. Most approaches focus on replacing parts of the mouse
`antibody with parts of human antibodies (see review by Presta,
`1992). The most complete method for 'humanization' consists
`of taking only the complementarity determining regions (CDRs)
`from the mouse antibody variable regions and grafting these
`mouse CDRs into human variable regions (Jones etal., 1986).
`The CDR-grafted variable regions, or reshaped human variable
`regions, are then joined to human constant regions to create a
`reshaped human antibody. This study describes the successful
`humanization of mouse C21 antibody by CDR grafting.
`
`Materials and methods
`Molecular modelling of the mouse C21 tunable regions
`The DNA sequences of the variable regions of mouse mAb C21
`were provided by Tanox Biosystems Inc. (Houston, TX). A 3-D
`model of the variable regions was built based on protein sequences
`derived from the DNA sequences. The model was developed on
`a Silicon Graphics IRIS 4D workstation using the molecular
`modelling package QUANTA (Polygen Corporation, Waltham,
`MA). The light chain variable region (V1) was modelled on the
`structure of the mouse anti-lysozyme antibody HyHEL-10 as
`solved by X-ray crystallography (Padlan et al., 1989). The heavy
`chain variable region (Vi,.) was modelled on the structure of the
`mouse anti-lysozyme antibody HyHEL-5 (Sheriff et at., 1987).
`The VL and VH regions of mouse C21 antibody have 79 and
`80% identity, respectively, to mouse HyHEL-10 and HyHEL-5
`antibodies. Identical residues in the framework regions (FRs)
`were retained and non-identical residues were substituted using
`QUANTA. CDRI, CDR2 and CDR3 of the V1 region and
`CDR and CDR2 of the VH region from mouse C21 antibody
`corresponded well to the canonical forms postulated by Chothia
`et al. (1989). Minor variations from the canonical sequences were
`seen, however, at residue 33 in CDR1 of the VL region and
`residue 55 in CDR2 of the V, region. The main chain torsion
`angles of these loops were the same as those of the original
`antibody structures (HyHEL-IO for CDR1, CDR2 and CDR3
`of the VL region; HyHEL-5 for CDR1 and CDR2 of the VH
`region). Because there are no canonical structures for CDR3s
`of VH regions, CDR3 of the VH region of mouse (21 antibody
`was modelled on a loop selected from 91 high-resolution protein
`structures in the Brookhaven Databank (Bernstein et al., 1977).
`
`© Oxford University Press
`
`82140721
`on 20 April
`by
`'
`
`Brenda
`2018
`
`EXHIBIT 1194 Ian A. Wilson, D.Phil.
`4/21/18 Planet Depos -Tncia Rosate, ROR, CRR, CSR No. 10891
`
`971
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01488
`PFIZER EX. 1194, Page 1
`
`
`
`F.Kolblngeo- ci al.
`
`Thirty candidate loops were extracted using the algorithm of Jones
`and Thirup (1986) as implemented in QUANTA. The best loops
`were selected by eye. The loops were anchored on three amino
`acid residues in the FRs on either side of the CDR3 in the mouse
`C21 VH region. The CDR3 in the mouse C21 VH region was
`modelled on residues 87-106 of the Bence-Jones protein RHE
`(Furey et al., 1983). This region of RHE corresponds
`approximately to the CDR3 of a VL region. The model was
`subjected to steepest descents and conjugate gradients energy
`minimization using the CHARMM potential (Brooks et al.,
`1983), as implemented in QUANTA, to relieve unfavourable
`atomic contacts and to optimize van der Waals and electrostatic
`interactions.
`Construction of the reshaped human C21 VL and VH regions
`The first versions of reshaped human C21 VL and VH regions
`(Li and HI) were constructed by gene synthesis using six
`overlapping synthetic DNA oligonucleotides for each construction
`(Table I, panel A; Table H, panel A). In each case, the six
`5'-phosphorylated and PAGE-purified oligonucleotides (Genosys
`Biot&hnologies, Houston, TX) were assembled using a PCR-
`based protocol. Aliquots of each oligonucleotide (5 pmol) were
`annealed and extended in a 100 cl reaction containing 10 mM
`Tris—HCI (pH 8.3), 1.5 mM M902, 50 mM KCI, 10 mM $-
`mercaptoethanol, 0.05% (w/v) Tween-20, 0.05% NP40, 200
`M dNTPs and 5 U Vent DNA polymerase (New England
`Biolabs, Beverly, MA). Following one cycle at 95°C for 1 mu,
`50°C for min and 72°C for 4 min in aTechne PHC-2 tempera-
`ture cycler, 50 pmol of oligonucleotide primers, designed to
`hybridize at the 5'- and 3'-ends of the full-length DNA fragment,
`
`Table I. Oligonudeotides used for the construction of the reshaped human
`C21 VL regions
`
`Panel A. 0ligonucleotidei for the synthesis ofvei-nion LI of reshaped human C21 V region
`
`Oligo I: C2ILI
`5' -PGAACXAAGC TTGCCOCCAC CATGOAGACC CCCGCCCAGC TGCTCTTCC'F
`Gc'rccrccin TGGCTGCCCG ACACCACcOc CCACATCCTG C'reAccCAGA Gcccc
`
`Oligo 2: C211_2
`5 ' -GATGTTGGTG cCGATGcTC'r GGCTGGCCCr GCAGCFCACG CPCCCCCTCT
`C&CCGGGG-CT CAGGCTCAGG GTQCC000GC TCTGGCTCAG CACCA
`
`Oligo 3: C21L3
`5' -CAGACCATCG GCACCA.ACA'T CCAcrGG'rAC CAGCACAAGC CCGGCCAGGC
`CCCCAGGCTG CTGATCA.AGT ACGCC
`
`Oligo 4: C21L4
`5' -AGGGTGAAG'r CCC'FCCCCCT GCCGCI'GCcG CTGAACCTGC TGGCGATGCC
`GCTGATGCTC Tcccroncor ACT'rOATCAG CAGCC'FC
`
`Oligo5:C2ILS
`S -GCCGCACCGA CTTCACCCTG ACCATCACCA CGCTGGAGCC CGACCAC1FC
`GCCATCTACD ACTCCCAGCA CACCCACACC TCGC
`
`Oligo 6: C21L6
`5' -rl'rGGATecr TCTAGAATAC TCACG1I'GA TCTCCACCTT CCTOCCCTGC
`CcGAAGGTCG TCCCCCAGCT GTCGCTCTGC TO
`
`5' Primer. C2I-5'
`5' -TGAAGAAAGC rrcCcsccAc C
`
`3' Primer C21-L3'
`5' -DS'l'GGATCCT TCTAGAATAC TCAC
`
`were added (C21-5', and C21-L3' or C21-H3', Tables land II,
`panels A). Then the full-length DNA fragment was amplified
`in 20 cycles at 95°C for 1 min, 60°C for 2 min and 72°C for
`2 mm. Next the reaction was chloroform-extracted. The DNA
`was ethanol-precipitated, digested with Hindlll and BamHT, and
`fragments of the correct size purified from an agarose gel. The
`Hindlll—BamHl DNA fragments were cloned into a pBluescrip
`KS+ vector (Stratagene, La Jolla, CA) and sequenced using
`Sequenase (United States Biochemical Corporation, Cleveland,
`OH). Point mutations and/or deletions within the DNA sequence
`were corrected by exchanging DNA restriction enzyme fragments
`between different clones and/or using PCR-based mutagenesis
`methods (Kammann ci a!,, 1989). Hindl]1—Bam}{I fragments
`exhibiting the correct DNA sequences were then subcloned into
`
`Table El. Ohgooude*ides used for the consiniction of the reshaped human
`C21 VH regions
`
`Panel A. Olgonucleoodcs for the synthesis of version HI of reshaped human C2I V
`reg=.
`
`Oligo I: C2IHI
`5' -TCAACAAAGC TTCCCCCCAC CATCCACTCG ACCTGGAGGG TGTTcrGCcF
`GCTCGCCCTG CCCCCCCGCC CCCACAGCCA GCTGCAGC'l'G GPGCAGA
`
`Oligo 2: c2IH2
`5' -CAGCCAGTAC ATOCTG5)GC TGTAGCCCCT GGCCTFGCAC C'rCACCTrCA
`cGCTGGCGCC CCGCI'TCTTC ACC'TCCCcCC CCCrcVCCAC CACCTGCACC ICC
`
`Oligo 3: (IH3
`5' -CACCTI'CAGC ATGTACTGCC TOGACTOGGI GAAGCAGAGG CCCCGCCACG
`GCCTGGAGTG GGT000CGAG ATCAGCCCCG GCACCCAC CACAACIAC AACGA
`
`Oligo 4. C21144
`5' -CTCCTQGCI'G CTCAGGCTGC TCAGCTCCAT GTAGGCGCTC TTGGTGC'TGG
`PCTCCCCCGT CAAGGIGGCC PTCGCCTI'GA AC'rrcl'CGTT G'PAGTrGGTG GTCAAGG
`
`Olugo 5: C2IH5
`5' -ACCAGCCTGA CCAGCCAGCA CACCfOCCII'rG TACTACTGCC CCAGGrTCAG
`CCACTFCAGC GGCAGCAACT ACCACTACTT CGA
`
`Oligo 6: C21H6
`5' -TTTGGAPCC'T TCTAGAACTC ACCTGAGC'I'C ACGGTCACCA GGGTGCCCTC
`GCCCCAGTAG TcGAAGTAGT CGTAGTTGCT CCC
`
`5' Primer. C21-5'
`5' -TGAAGAAAGC TTGCCGCCAC C
`
`3' Primer: C21-H3'
`3' -TrFGGATCCT TCTAGAAC'i'C ACC
`
`Pane! B. Oligonuclontides for the anban)uat caunatnicuion of versions H3, Hay], and Hay3
`of the reshaped human C21 Vu region.
`
`Primer H/R38K-A40R-L (introducts R38K and A40R into HI, complementary strand)
`5' -ccG000_cr GCçTCACCCA CTCCAGCC
`
`Prima HJR38K-A40R-SL (introduces RiSK and A40R into HI, coding strand)
`CCCGGCCACG GCCTGGAGS'
`S • -GAGCAG
`
`Prima H/R66K-L (introduces R66K into HI, complementary strand)
`3' -GAAGGPGCCC TCCCcrrGA ACI'PCTCG'PF CISC
`
`Primer }{/R66K-SL (Introduces; R66K into HI, coding strand)
`5' -CAACCCCAG GCCACCri'CA CCOCCCAC
`
`Prima H1R83T-L (introduces R83T into HI, complementary grand)
`3' -G1'CCTCCCT
`reAGGC'rcC TCACCI'CCAT C
`
`Printer HJR83T-S (introduces R83T into HI gerte, coding strand)
`5 1-CAGCCTGAr& AGCGAGGACA C
`
`Pinel B. Olgonucleoiiden for the mjbauumu construction of vcniuxu L2 and U of the
`reshaped human C2I V,, region
`
`Primer HayFR2 (FR2 changes from HI toyI,
`5' -CCACCACTC CAGQCT5'CG CCCGGCC
`
`complementary strand)
`
`Prima LID60S-L (Introduces 0605 into LI, complemmuaty strand)
`5' - CTGAACCTGI QGGGCATGCC GCTGATCCTC
`
`Printer L/D605-SL (unroduces 0605 into LI, coding strand)
`5' -CCCG_C.AGG'F TCAGCGGCAG CCGCA
`
`Primer HayFR2.L (FR2 changes from H3 to
`y3, complementary strand)
`5' -CCACCACTC CAGCTTGG CCCGGGCT GC
`
`Primer HayFL2.5 (FR2 changes from HI to Hayl and H3 to Hay), coding strand)
`5' -CAGcroG AGTGGATCGG CCAGATC
`
`Primer UEI D-V3L (introduces El I) and V3L into LI, cmmplemaltasy au -and)
`5' -GG'TCAGCAQG APcFCGCCGG TO
`
`Prima HayFRi (FR) changes from HI to Hay! and H) to Hay3, complementary strand)
`5' -GTCçTGGQGC TCGTGTCGGC
`
`Primer UEID-V3L-SI., (introduces EID and V31 into LI, coding strand)
`5' GAATegTGC TCACCCAGAG CCCCGGC
`
`Primer HayFR3-S (FR) changes from HI to Hay) and H) to Hay), coding strand)
`3' -ACCAGcCCA QCACCGCCTA C
`
`972
`
`Downloaded from https://academic.oup.com/peds/article-abstract/6/8/971/1559740
`by 82140721, Brenda Burton
`on 20 April 2018
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01488
`PFIZER EX. 1194, Page 2
`
`
`
`vectors designed to express human x light chains or human y-1
`heavy chains in mammalian cells (Maeda et at., 1991).
`Additional versions of reshaped human C21 VL regions (L2
`and L3) were generated from version Li by oligonucleotide-
`directed PCR mutagenesis. The PCR primers used to create
`versions U and L3 from version Li are listed in Table I, panel
`B. Similarly, additional versions of reshaped human C21 V
`regions (H3, Hay 1 and Hay3) were generated from version Hi
`by oligonucleotide-directed PCR mutagenesis. The PCR primers
`used to create the new versions of reshaped human C21 VH
`regions are
`listed
`in Table II, panel B. The resulting
`Hindffl —BamHI fragments were cloned, sequenced and
`subcloned into the expression vectors as described previously.
`Expression of reshaped human C21 antibodies in cos cells
`The cos cells were co-transfected by electroporation with the
`plasmid DNAs designed to express the reshaped human C21 light
`and heavy chains (Maeda et al., 1991). After a 10 min recovery
`period, the cells were plated out in 10 ml Dulbecco's minimal
`essential medium containing 5% y globulin-free, heat-inactivated
`fetal calf serum. After 72 h incubation, the medium was collected
`and centrifuged to remove cells and cellular debris. The
`supernatant was filtered through a 0.45 lim membrane and
`analysed by ELISA for assembled antibody with human x light
`chains and human y heavy chains.
`Protein A purification of the reshaped human CII antibodies
`Reshaped human C21 antibodies were purified from the cos cell
`supernatants by affinity chromatography on I ml immobilized
`protein A (Prosep A, Bioprocessing Ltd, Durham, UK) packed
`into HR 5/5 FPLC columns (Pharmacia, Uppsala, Sweden). The
`columns were run at constant flow rates of 2 ml/min on an FPLC
`system (Pharmacia). Eluted protein was detected in a flow cell
`(UV absorbency at 280 run). The columns were prepared by
`washing in 10 column volumes of PBS (20 mM sodium
`phosphate, 150 mM NaCl, pH 8.0), pre-elution with 10 column
`volumes of 100 mM sodium citrate buffer (pH 3.0) and re-
`equilibration with 10 column volumes of PBS (pH 8.0). The cos
`cell supernatants (20-50 ml) were clarified by filtration through
`a 0.45 an membrane and then loaded directly onto the column
`with a peristaltic pump. The column was washed with PBS
`(pH 8.0) until the UV absorbency returned to baseline. Bovine
`IgG was then eluted by washing with 100 mM sodium citrate
`buffer (pH 5.0) until the baseline returned to zero. Finally,
`reshaped human antibodies were eluted with 100 mM sodium
`citrate buffer (pH 3.0). The pH was adjusted immediately to
`pH 7.0 with 1 M Tris. The neutralized eluates containing the
`reshaped human antibodies were concentrated in a Centricon-10
`microconcentrator (Amicon, Stonehouse, UK) and the buffer
`changed to PBS (pH 7.2). Purity of the reshaped human anti-
`bodies was analysed by SDS—PAGE and Coomassie blue
`staining (L.aemmli, 1970). Protein concentration was determined
`by UV absorption at 280 rim and by ELISA.
`
`EUSA for human -yfx antibody
`Microtiter 96-well plates were coated with goat anti-human IgG
`(Fc specific) (Dianova). After washing, the plates were blocked
`with 1% bovine serum albumin in PBS (pH 7.2) plus 0.05%
`Tween. Sample and sample dilutions were added and, after
`incubation and washing, bound human IgGIx antibody was
`detected using affinity-purified goat anti-human x light chain
`polyclonal antibody conjugated with horseradish peroxidase
`(Sigma, Poole, UK). A purified recombinant human antibody
`(human IgG lIx) of known concentration was used as a standard.
`
`Humanization of an anti-IgE antibody
`
`Analysis of the mouse, chimeric and reshaped human CII
`antibodies by Biospecific Interaction Analysis (MA)
`A biosensor-based analytical system (Pharmacia BlAcore) was
`used to analyse the kinetics of interaction between the C21
`antibodies and their antigen, human IgE. Mouse C21 antibody
`(TES-C21), chimeric C21 antibody (TESC-2) and mouse—human
`chimeric IgE antibody (SE44, Sun er al., 1991) were provided
`by Tanox Biosystems Inc. As capture antibodies, — 11,000
`resonance units (RU) (11 ng/mm2) of polyclonal rabbit anti-
`mouse IgGi (Pharmacia Biosensor AB, Freiburg, Germany) or
`rabbit anti-human IgG (kin(ly donated by Dr U.Roder, Pharmacia
`Biosensor AB) were immobilized to the CMS sensor chip surface
`via their amino groups (Jönsson et al., 1991). For each C21 test
`antibody, four experimental cycles were performed. Each cycle
`consisted of binding a constant amount of test C21 antibody to
`the respective capture antibody followed by the interaction of
`this test C21 antibody with fixed concentrations of human IgE.
`The assays were carried out at 25°C. Test C21 antibody was
`diluted in HBS (10 mM HEPES, 3.4 mM EDTA, 150 mM
`NaCl, 0.05% BlAsurfactant, pH 7.4) to a final concentration of
`5-10 14g/ml and bound to catching antibody to obtain
`1300-2200 RU (1.3-2.2 ng/mm2) of bound test antibody.
`Human IgE, at concentrations of 3.125, 6.25, 12.5 and 25 nM,
`was passed over the bound test C21 antibody at a flow rate of
`5 Allmin for 9 mm. An aliquot of 11 of 40 mM HC1 was used
`to remove antibody - antigen complexes and to prepare the
`surface for the next cycle. The surface plasmon resonance (SPR)
`signals were measured and illustrated as a sensorgram. The
`rates of association for the antibody - antigen interactions were
`calculated using computer programs implemented in the BlAcore
`system.
`For the determination of the rates of dissociation, a similar
`protocol was used. Test C21 antibodies were first bound to the
`sensor chip surface via the immobilized capture antibodies.
`Human IgE (25 nM) was allowed to bind to the C21 test antibody.
`Then 1-lBS buffer was passed over the sensor chip surface at a
`constant flow rate of 5 p1/min and the decrease in resonance signal
`monitored over a period of 15-25 mm. The sensor chip surface
`was later regenerated with 4 ul 40 mM HO. Because the
`dissociation of antibody—antigen complexes is a first order
`reaction, the linear parts of the sensorgrams were used to calculate
`the rates of dissociation using computer programs implemented
`in the BlAcore system.
`The specificities of mouse, chimeric and reshaped human C21
`antibodies for human IgE were also tested using the Pharmacia
`BlAcore machine. The test C21 antibodies were bound to
`immobilized capture antibodies on the sensor chip surface as
`described previously. Human Igs with x light chains and a variety
`of heavy chains (1gM, IgD; Serotec, Oxford, UK) (IgAl, IgA2;
`Calbiochem, Nottingham, UK) (IgG4, IgG3, IgG2, IgG I; Sigma)
`(IgE; Tanox Biosystems Inc.) were passed over the surface at
`concentrations of 5 pg/mI. The SPR signals were measured and
`illustrated as sensorgranis.
`
`Results
`Molecular model of the structure of the mouse C21 variable
`regions
`To design reshaped human variable regions that recreate as
`closely as possible the antigen binding site in the original mouse
`antibody, it would be useful to know the structure of the mouse
`Ig variable regions (Verhoeyen et at., 1988). In most cases,
`however, the structure of the mouse antibody to be humanized
`
`973
`
`Downloaded from https//acadeeicnup.com/peds/artio1e-abstract/6/8/97l/l559740
`by 82140721, Brenda Burton
`On 20 April 2018
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01488
`PFIZER EX. 1194, Page 3
`
`
`
`F.Kolbinger el at.
`
`A
`
`Fig. 1. A view of the molecular model of the variable regions of mouse C21 antibody. (A) The Ca trace of the variable regions with the FRs in yellow, the
`CDRS in the VL region in blue, the CDRS in the V14 region in green, residues of special interest in the FRs of the VL region in purple, and residues of
`special interest in the FRs of the VH region in red. (B) A line drawing of (A) with the residues of special interest labelled.
`
`has not, as yet, been determined. In these cases, a molecular
`model of the mouse antibody can be constructed as a guide to
`the design of the reshaped human variable regions (Kettleborough
`
`el al., 1991). In preparation for the design of the reshaped human
`C21 variable regions, a molecular model of the VL and VH
`regions of mouse C21 antibody was built. Details of the
`
`974
`
`Downloaded from h1tps//academicoupcom/peds/artic1e-abstract/6/8/971/1559740
`by 82040721, Brenda Burton
`on 20 April 2018
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01488
`PFIZER EX. 1194, Page 4
`
`
`
`
`
`
`Humanization of an anti-IgE antibody
`
`CDR2
`FR2
`CDR1
`FRi
`5
`4
`3
`2
`1
`12345678901234567890123 45678901234 567890123456789 0123456
`***
`*
`*********
`*
`DILLTQSPAILSVSPGERVSFSC RASQSIGTNIH WYQQRTDGSPRLLIK YASESIS
`
`C21
`
`80111 E IVLTQSPGTLSLSPGERATLSC
`
`WYQQKPGQAPRLLIY
`
`KAY
`
`E IVLTQSPGTLSLSPGERATLSC
`
`WYQQKPGQAPRLLIS
`
`Li
`
`1.2
`
`D—L --------------------
`
`RASQSIGTNIH --------------K YASESIS
`
`D—L --------------------
`
`----------- -------------- K -------
`
`L3-----------------------
`
`-----------
`
`--------------K
`
`FR4
`CDR3
`FR)
`10
`9
`8
`7
`6
`78901234567890123456789012345678 90123456 78901234567
`*
`*
`GIPSRFSGSGSGTEFTLNIHSVESEDIADYYC QQSDSWPTT FGCGTKLEIK
`
`C21
`
`80111 OIPDRFSGSGSGTDFTLTISRLEPED?AVYYC
`
`FGQGTXVEIK
`
`KAY
`
`GI PDRFSGSGSGTDFTLTI SRLEPEDFYYC
`
`FGQGTKVE Il
`
`Li
`
`L2
`
`L3
`
`---S
`
`---8
`
`QQSDSWPTT ----------
`
`Hg. 2. Comparisons of the amino acid sequences of moose and reshaped human C21 VL regions. C21 shows the FRs and CDRs of the mouse C21 VL
`region. The amino acid residues that are pan of the canonical sequences for the CDR loop nnscturcs are marked with an asterisk (Cb*hia ci at., 1989). The
`numbering is according to Kabat er W. (1987). SGffl shows the FRs of the consensus sequence for human S VL regions of subgroup ifi (Kahat et a!, 1987).
`KAF shows the FRs from the VL region of human KAF antibody (Newkirk etal., 1988). The residues in the FRs of KAF that differ from those in the
`consensus sequence are underlined. LI, U and U are the versions of reshaped human C21 VL reason. Thc residues in the FRs that differ from the KAF
`sequence are shown in bold.
`
`construction of the model are described in Materials and methods.
`A view of the model highlighting the amino acid residues that
`were of particular interest is shown in Figure 1.
`
`Design of the reshaped human C21 variable regions
`The design of the reshaped human C21 VL and VH regions was
`based on either the consensus sequences for certain subgroups
`of human VL and VH regions (Kabat a al., 1987) or the
`sequences from individual human antibodies. The amino acid
`sequences of the mouse C21 VL and VH regions were most
`similar to the consensus sequence for human x VL subgroup ifi
`(69% identity within the FRs) and for human VH subgroup I
`(70% identity within the FRs). In the first step of the design
`process, the mouse C21 CDRs were linked to the FRs from these
`human consensus sequences. The preliminary designs were
`examined and certain amino acid residues in the human FRs were
`identified as possible key residues in determining binding to
`antigen. For example, the amino acid residues that were part of
`the canonical structures for CDR loop formation, as proposed
`by Chothia a al. (1989), were highlighted (see residues marked
`with an asterisk in Figures 2 and 3). Residues that were potentially
`involved in VL —VH packing, as described by Chothia a al.
`
`(1985), were examined. The rare occurrence of certain amino
`acids at specific positions was noted. With this information, and
`with reference to the model of the mouse C21 variable regions,
`decisions were made as to whether or not certain amino acid
`residues in the selected human FRs should be replaced with the
`amino acid residues that occurred at those positions in the mouse
`C21 variable regions.
`For the design of the first version of reshaped human C21 VL
`region (Li), changes in the human FRs were made at positions
`1, 3,49 and 60 (numbering according to Kabat etal., 1987)
`(Figure 2). The amino acids at positions 1 and 3 were considered
`important because the model showed that the N-terminus of the
`mouse C21 light chain lay between CDR and CDR3 of the VL
`region. Therefore, the N-terminus either could be directly
`involved in antigen binding or could alter the conformation of
`the CDRs. The amino acid at position 49 is lysine and is located
`between two amino acids that are part of the canonical structure
`for CDR2 of the VL region. It is in the binding pocket created
`by CDR2 of the VL region and may form an interaction with
`the glutamic acid at position 53 in CDR2 of the VL region. The
`serine at position 60 in mouse C21 VL region was located in the
`model at the edge of the binding site and could be influencing
`975
`
`Downloaded from httpa://academiroup.com/peda/artir1e-abstract/6/B/971/1559740
`by 82140721, Brenda Burton
`on 20 April 2018
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01488
`PFIZER EX. 1194, Page 5
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`F.Kolblngei- at at.
`
`C21
`
`SQl
`
`Hi
`
`H3
`
`FRi
`
`CDR1
`
`FR2
`6
`5
`4
`3
`1
`2
`123456789012345678901234567890 12345 67890123456789 012A3456789012345
`*****
`**
`*
`QVQLQQSGAEL14XPGASVKISCKTTGYTFS MYWLE WVKQRPGHGLEWVG EISPGTFTTNYNEKFKA
`
`CDR2
`
`QVQLVQSGAEVXKPGXSVXVSCKASGYTFS
`Q------ --------A--K----------- MYWLE
` A--K ----------- ----- ------- H ------ -----------------
`
`WVRQAPGXGLEWVG
`
`--K-H--H ------ EISPGTFTTNYNEKFKA
`
`Q --------------
`
`KAY
`
`QVQLVQSGAEVKKPGASVVSCKASGYTF
`
`WRQAPGQLEWI4G
`
`Hayl -----------------------------S MYWLE --K-H --------- EISPGTFTTNYNEKFKA
`-----------------------------S
`
`----- -------------- -----------------
`
`Hay3
`
`FR3
`
`cDR3
`11
`10
`8
`7
`9
`67890123456789012A8C345678901234 567890ABCDEF12 34567890123
`*
`*
`KATFTADTSSNTAYLQLSGLTSEDSAVYFCAR FSHFSGSNYDYFDY WGQGTSLTVSS
`
`FR4
`
`RV'rXTXDXSXNTAYMELSSLRSEDTAVYYCAR
`
`WGQGTLVTVSS
`
`XA-F-A-T-T ---------- - ----------- FSHFSGSNYDYFDY -----------
`-------------- -----------
`-A-F-A-T-T ----------------------
`
`Cal
`
`501
`
`Hi
`
`53
`
`HAY
`
`RVTITBDISASTAYMELS SLRSEDTAVYYCAR
`Hayl KA-F-A -------------- I ----------- FSHFSGSNYDYFDY -----------
`-------------- -------
`-A-F-A --------------------------
`Hay3
`FIg. 3. Comparisons of the amino acid sequences of the mouse and reshaped human C21 VH regions. C21 shows the FRs and CDRs of the mouse C21 VH
`region. The amino acid residues that are part of the canonical sequences for the CDR loop structures are marked with an asterisk (Chothia €1 al., 1989). The
`numbering is according to Kabat at al. (1987). SC! shows the FRs of the consensus sequence for human VH regions of subgroup I (Kabat a: al., 1987). HAY
`shows the FRs from the VH region of human HAY antibody (Dcrslmonian at al., 1987). The residues in the FRs of HAY that differ from those in the
`consensus sequence are underlined. H! and 113 are the versions of reshaped human C21 VH region that were designed based on the consensus sequence. The
`residues in the FRs that differ from the consensus sequence are shown in bold. Hay! and Hsy3 are the versions of reshaped human C21 V1
` region that were
`designed based on the HAY sequence. The residues in the FRs that differ from the HAY sequence are shows in bold,
`
`WGQGTLVTVSS
`
`antigen binding. Two further versions of reshaped human C21
`VL regions (U and 1-3) were made (Figure 2). Version L2
`contained only the changes at positions 1, 3 and 49. Version L3
`contained only the changes at positions 49 and 60.
`For the design of the first version of reshaped human C21 V11
`region (Hi), changes in the human FRs were made at positions
`38, 40, 66, 67 and 83 (shown in bold in Figure 3). In the model,
`the arginine at position 40 and the lysine at position 66 formed
`salt bridges with the glutarnic acid at position 85 and the aspartic
`acid at position 86, respectively. For this reason, the arginine
`(position 40) and the lysine (position 66) were used in version
`Hi of reshaped human C21 VH region. At positions 38 and 83,
`lysine and threonine were used in version Hi of reshaped human
`C21 V14 region so as not to create a too highly positively
`charged area that could disrupt the overall structure. In the model,
`the alanine at position 67 appeared to be packed under the CDR2
`loop of the V11 region. It was decided, therefore, to use alanine
`
`instead of valine because a size change at this position might
`disrupt the conformation of the CDR2 loop. One further version
`of reshaped human C21 V14 region was designed based on the
`human subgroup I consensus sequence (143 in Figure 3). Version
`H3 contained only the change at position 67.
`There was no single preferred amino acid at positions 1, 16,
`19, 43, 69, 71, 73 and 75 in the FRs of the consensus sequence
`for human subgroup I V11 regions (as indicated by an FJQ or
`an X in the SGI sequence in Figure 3). In the design of versions
`Hi and 1-13 of reshaped human C21 V11 regions, the amino acids
`that occurred at these positions in the mouse C21 V11 region
`were used because there were examples of human subgroup I
`VH regions that had those amino acids at the same positions.
`The one exception was position 75 where threonine was selected
`because it is the most frequent amino acid at this position in
`human subgroup I VH regions. Because the amino acids used
`at these eight positions were observed in examples of human VH
`
`976
`
`Downloaded from httpa://academio.oup.com/peds/artiole-abstraot/6/8/971/1559740
`by 82140721, Brenda Burton
`on 20 April 2018
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01488
`PFIZER EX. 1194, Page 6
`
`
`
`regions, they were considered to resemble human sequences.
`To determine how similar the first versions of reshaped human
`C21 variable regions were to variable regions present in individual
`human antibodies, the amino acid sequences of the first versions
`of reshaped human C21 VL and VH regions (Li and Hi) were
`compared with human variable regions in a database derived from
`the Leeds Database (OWL Composite Protein Sequence
`Database, version 15, University of Leeds, UK).
`The first version of reshaped human C21 VL region (LI)
`showed the geatest similarity (80% identity) to the VL region in
`human KAF antibody (hybridoma EV 1-15 'CL) (Newkirk a al.,
`1988). Within the FRs, KAF and the first version of reshaped
`human C21 VL region (LI) differed by only four amino acids.
`These four amino acid positions (1, 3,49 and 60) were the same
`four positions in the human FRs that had been highlighted during
`the design of the first version of reshaped human C21 VL
`region. Thus, in the design of the reshaped human C21 VL
`region, there would be no significant difference if FRs derived
`from a human x VL consensus sequence or those from an
`individual human x VL sequence were used. Therefore, there
`was no reason to design additional versions of reshaped human
`C21 VL regions based on human FRs from individual human
`antibodies.
`The first version of reshaped human C21 VH region (Hl)
`showed greatest similarity (68% identity) to the VH region of
`human HAY antibody (hybridoma 8E10'CL) (Dersimonian
`et al., 1987). Within the FRs, HAY and the first version of
`reshaped human C21 VH region (Hi) differed by 13 amino
`acids. Thus, in the case of the reshaped human C21 V11 regions,
`there would be a significant difference if FRs derived from a
`human VH consensus sequence or those from an individual
`human V11 sequence were used. Two additional versions of
`reshaped human C21 VH regions (Hay 1 and Hay3) were
`designed based on the FRs present in the V11 region of human
`