`
`Patent Owner’s Response
`
`Filed on behalf of Patent Owner Genentech, Inc. by:
`
`David L. Cavanaugh (Reg. No. 36,476)
`Robert J. Gunther, Jr. (Pro Hac Vice)
`Lisa J. Pirozzolo (Pro Hac Vice)
`Kevin S. Prussia (Pro Hac Vice)
`Andrew J. Danford (Pro Hac Vice)
`WILMER CUTLER PICKERING
`
`HALE AND DORR LLP
`1875 Pennsylvania Ave., NW
`Washington, DC 20006
`
`Adam R. Brausa (Reg. No.
`60,287)
`Daralyn J. Durie (Pro Hac
`Vice)
`DURIE TANGRI LLP
`217 Leidesdorff Street
`
`San Francisco, CA 9411 l
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`
`
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`
`
`CELLTRION, INC,
`Petitioner,
`
`V.
`
`GENENTECH, INC,
`Patent Owner.
`
`
`Case 1PR2017-01373
`
`Patent 6,407,213
`
`
`PATENT OWNER’S RESPONSE
`
`
`
`TABLE OF CONTENTS
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`Page
`
`I.
`
`INTRODUCTION ........................................................................................... 1
`
`II.
`
`TECHNOLOGY BACKGROUND ................................................................. 4
`
`A.
`
`Antibody “Variable” And “Constant” Domains ........................................4
`
`B.
`
`“Humanized” Antibodies ............................................................................ 6
`
`III.
`
`’213 PATENT .................................................................................................. 8
`
`A.
`
`Invention ..................................................................................................... 8
`
`B.
`
`C.
`
`Advantages Of ’213 Invention ................................................................... 9
`
`Prosecution History .................................................................................. 10
`
`IV. ASSERTED REFERENCES ......................................................................... 11
`
`A.
`
`Queen-1989 .............................................................................................. 11
`
`B.
`
`C.
`
`Queen-1990 .............................................................................................. 12
`
`Protein Data Bank ..................................................................................... 14
`
`D.
`
`Tramontano ............................................................................................... 1 5
`
`E.
`
`Kabat- I 987 ............................................................................................... 16
`
`F.
`
`Hudziak ..................................................................................................... 16
`
`V.
`
`PERSON OF ORDINARY SKILL ............................................................... 17
`
`VI.
`
`CLAIM CONSTRUCTION .......................................................................... 17
`
`VII. SUMMARY OF ARGUMENT ..................................................................... 18
`
`VIII. ARGUMENT ................................................................................................. 22
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`Grounds 2—4, 7: The Board Should Confirm The Patentability Of
`Claims 12, 42, 60, 65, 71, 73—74, And 79 Because Neither Queen-
`1990 Nor Tramontano Is Prior Art. .......................................................... 22
`
`The inventors made and tested HuMAb4D5-5 and HuMAb4D5-
`
`8 before July 26, 1990. ........................................................................ 23
`
`Consensus sequence ....................................................................... 23
`
`Humanized 4D5 antibody sequences ............................................. 25
`
`Production and testing of humanized 4D5 antibodies ................... 27
`
`(i)
`
`First humanized 4D5 variable domain fragment ...................... 28
`
`(ii)
`
`First humanized 4D5 full-length antibody ................................ 30
`
`(iii) Other humanized 4D5 variants ................................................. 32
`
`HuMAb4D5—5 and HuMAb4D5-8 demonstrate actual reduction
`
`to practice of claims 12, 42, 60, 65, 71, 73—74, and 79 before
`July 26, 1990. ...................................................................................... 34
`
`HuMAb4D5—5 and HuMAb4D5—8 embody claims 12, 42,
`60, 65, 71, 73-74, and 79 ................................................................ 35
`
`The inventors determined that HuMAb4D5-5 and
`HuMAb4D5-8 would work for the intended purpose of the
`claims before July 26, 1990. .......................................................... 39
`
`Contemporaneous records from non-inventors corroborate
`the inventor’s actual reduction to practice before July 26,
`1990. ............................................................................................... 40
`
`a)
`
`b)
`
`c)
`
`a)
`
`b)
`
`c)
`
`Queen—1990 and Tramontano are not § 102(1)) prior art. .................... 41
`
`Grounds 1, 3, 6: Petitioner’s obviousness theory for Queen—1989
`actually rests on Queen—1990, which is not prior art to claims 12,
`42, 60, 65, 71, 73-74, and 79. ................................................................... 44
`
`ii
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`C.
`
`D.
`
`Grounds 1—4, 6—7: The Queen references combined with the PDB
`database would not have led to the invention of claims 12, 42, 60,
`65-67, and 71-79 with a reasonable expectation of success ..................... 45
`
`Grounds 1-2, 5, 7: Claims 4, 33, 62, 64, And 69 Would Not Have
`Been Obvious Because The Asserted References Do Not Teach
`
`The “Consensus” Sequence Limitations. ................................................. 52
`
`l.
`
`The asserted references do not teach the “consensus” sequence
`limitation. ............................................................................................ 52
`
`a)
`
`b)
`
`Queen-1990 (Grounds 2, 7) ........................................................... 52
`
`Queen-1989 (Ground 5) ................................................................. 54
`
`2.
`
`The asserted references do not teach any antibody with the
`framework substitutions of claims 4, 33, 62, and 69 that
`incorporates non-human CDRs that bind antigen. .............................. 56
`
`E.
`
`Grounds 2-4: The Asserted References Do Not Render Obvious
`
`The “Up To 3-Fold More” Binding Affinity Limitation Of Claim
`65. ............................................................................................................. 57
`
`F.
`
`Grounds 1—2: Queen—1989 And Queen-1990 Do Not Render
`Obvious The “Lacks Immunogenicity” Limitation Of Claim 63. ........... 60
`
`G.
`
`Grounds 6—7: Petitioner Has Not Shown That It Would Have Been
`
`Obvious That A Humanized Antibody With The Framework
`Substitutions Recited In Claims 30-31, 33, 42, And 60 Would Bind
`p 1 85‘1””. ................................................................................................... 61
`
`H.
`
`Objective Indicia Of Non-Obviousness Confirm The Patentability
`Of The Challenged Claims. ...................................................................... 63
`
`l.
`
`2.
`
`Unexpected results .............................................................................. 63
`
`Commercial success ............................................................................ 66
`
`I.
`
`Inter Parfes Review Is Unconstitutional. ................................................. 67
`
`IX.
`
`CONCLUSION .............................................................................................. 67
`
`iii
`
`
`
`TABLE OF AUTHORITIES
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`Page(s)
`
`Federal Cases
`
`Brown & Williamson Tobacco Corp. v. I’hitip Morris Inc,
`229 F.3d 1120 (Fed. Cir. 2000) .......................................................................... 66
`
`In re Clarke,
`
`356 F.2d 987 (C.C.P.A. 1966) ............................................................................ 38
`
`Cooper v. Goldfarb,
`154 F.3d 1321 (Fed. Cir. 1998) .......................................................................... 40
`
`KSR International (To. v. Teleflex Inc,
`550 US. 398 (2007) ............................................................................................ 50
`
`Leo Pharm. Prods, Ltd. v. Rea,
`
`726 F.3d 1346 (Fed. Cir. 2013) .......................................................................... 51
`
`Markman v. Westview Instruments, Inc,
`517 U.S. 370 (1996) ............................................................................................ 67
`
`McCormick Harvesting Mach. (‘0. v. C. Aultman & Co,
`169 US. 606 (1898) ............................................................................................ 67
`
`Medichem, SA. v. RoIabo, 8L,
`437 F.3d 1157 (Fed. Cir. 2006) .......................................................................... 40
`
`In re Merchant,
`575 F.2d 865 (C.C.P.A. 1978) ............................................................................ 65
`
`NFC Tech, LLC v. MataI,
`871 F.3d 1367 (Fed. Cir. 2017) .................................................................... 34, 35
`
`In re NTP, Inc,
`
`654 F.3d 1279 (Fed. Cir. 201 l) .......................................................................... 34
`
`Oil States Energy Services, LLC v. Greene ’5' Energy Group, IJIJC',
`No. 16-712 .......................................................................................................... 67
`
`iv
`
`
`
`lPR2017-01373
`
`Patent Owner’s Response
`
`()rtho—McNei! Pharm, Inc. v. Mylar: Labs, Inc,
`520 F.3d 1358 (Fed. Cir. 2008) .......................................................................... 50
`
`Sinorgehem (70. v. 11217 Trade Comm ’n,
`511 F.3d 1132 (Fed. Cir. 2007) .......................................................................... 17
`
`In re Sam,
`
`54 F.3d 746 (Fed. Cir. 1995) .............................................................................. 64
`
`In re Steed,
`
`802 F.3d 1311 (Fed- Cir. 2015) .......................................................................... 34
`
`Tokai (7(er. v. Easter: Balers, Inc,
`632 F.3d 1358 (Fed. Cir. 2011) .......................................................................... 66
`
`Patent Trial and Appeal Board Cases
`
`Greer: (.‘ross Corp. v. Shire Human Generic Therapies,
`lPR2016-00258, Paper 89 (Mar. 22, 2017) ........................................................ 40
`
`Nintendo ()fAm., Inc. v. iLife Tech, Inez,
`lPR2015-00109, Paper 40 (Apr. 28, 2016) ......................................................... 40
`
`Federal Statutes
`
`35 U.S.C. § 102(a) ............................................................................................. 34, 41
`
`35 U.S.C. § 102(b) ................................................................................................... 41
`
`35 U.S.C. § 120 ........................................................................................................ 41
`
`Constitutional Provisions
`
`U.S. Const. Amendment VII .................................................................................... 67
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`I.
`
`INTRODUCTION
`
`US. Patent No. 6,407,213 claims humanized antibodies with amino acid
`
`substitutions at specific positions. Unlike prior art humanized antibodies—which
`
`required handpicking a unique human framework sequence for each antibody—the
`
`claimed antibodies could be produced from a single human “consensus” sequence,
`
`which is a composite of all human antibody framework sequences of a particular
`
`subclass or subtype. The ’2] 3 invention thus provides a broadly-applicable
`
`humanization platform, which has produced numerous successful drugs, including
`
`treatments for cancer, asthma, and macular degeneration.
`
`In its preliminary response, Patent Owner identified several deficiencies in
`
`Petitioner’s proof for all challenged claims. However, to narrow the issues, Patent
`
`Owner now focuses on a subset of the challenged claims and presents specific
`
`reasons why Petitioner has failed to carry its burden for those claims. Patent
`
`Owner’s response is supported by new evidence obtained from cross-examination
`
`of Petitioner’s declarants Dr- Lutz Riechmann (Ex—2039) and Dr. Robert Leonard
`
`(EX-2040), as well as the declaration of Dr. Ian Wilson (Ex-2041) submitted
`
`herewith.
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`First, the Board should confirm the patentability of claims 12, 42, 60, 65,
`
`71, 73-74, and 791 because the inventors conceived and actually reduced to
`
`practice those claims prior to the publication of Queen-1990 and Tramontano.
`
`That prior reduction to practice is corroborated by several non-inventors whose
`
`contemporaneous notebooks confirm that the inventors made humanized
`
`antibodies embodying the claims and verified that they would work for their
`
`intended purpose before July 26, 1990.
`
`In addition, although Petitioner purport to
`
`rely upon Queen-1989 for Grounds l, 3, and 6, Petitioner’s obviousness theory in
`
`those grounds actually rests on Queen—1990 (which Petitioner’s expert explicitly
`
`cites as the basis for his analysis). If the Board finds Queen-1990 antedated, then it
`
`should also reject Petitioner’s challenge to those claims in Grounds l, 3, and 6.
`
`Second, the Board should confirm the patentability of claims 12, 42, 60, 65-
`
`67, and 71-79, Petitioner’s analysis of the Queen references combined with the
`
`PDB database discloses numerous potential framework substitutions. In fact, the
`
`record now shows that applying Petitioner’s own analysis of the PDB structures
`
`encompasses many more framework substitutions than the selective subset that
`
`1 Many claims have been challenged in multiple grounds. Patent Owner explains
`
`below (§VII) how the issues summarized in this introductory section correspond
`
`with the instituted grounds.
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`Petitioner cited in its petition. That broad disclosure does not render obvious
`
`claims 12, 42, 60, 65—67, and 71—79, which narrowly require at least one and up to
`
`five specific framework substitutions. Nor would those specific claimed
`
`framework substitutions have been obvious to try. What Petitioner cites is not a
`
`“small” or “easily traversed” number of possibilities in the context of antibody
`
`humanization, particularly as of 1991 when the field was still nascent. And the
`
`record also confirms that the high degree of unpredictability of making framework
`
`substitutions, where even a single substitution can affect antigen binding in
`
`unpredictable ways.
`
`Third, Petitioner has failed to show that Queen-1990, or Queen-1989 alone
`
`or combined with Kabat-1987, teaches the “consensus” sequence limitations of
`
`claims 4, 33, 62, 64, and 69. As the Board recognized in its institution decision,
`
`the ”213 patent expressly defines “consensus” sequence, as a sequence generated
`
`from “all human immunoglobulins of any particular subclass or subunit structure.”
`
`Queen-1990, however, describes “a consensus framework from many human
`
`antibodies,” not “all.” Dr. Wilson explains that a skilled artisan would understand
`
`that Queen-1990’s “consensus framework” is referring to a sequence generated
`
`from a subset of antibodies, which differs from what the ’2 1 3 patent requires.
`
`Queen-1989 does not even mention a consensus sequence, and Dr. Wilson explains
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`that Petitioner’s proposed combination with Kabat—l987 in Ground 5 would not
`
`have led to the consensus sequence of the ’2 l 3 patent.
`
`Fourth, claims 30—3 1, 33, 42, and 60 require an antibody with the recited
`
`substitutions that binds a specific antigen called “pl 8511532.» Petitioner has not
`
`shown that such an antibody would have been obvious. Petitioner merely cites the
`
`general disclosure of references involving humanized antibodies for different
`
`antigens and presents no evidence that those general techniques would result in the
`
`claimed substitutions when applied to an antibody that binds pl 85H“?
`
`Finally, claims 63 and 65 contain additional limitations requiring that the
`
`antibody “lacks immunogenicity” or has “up to 3-fold more” binding affinity as
`
`compared with the parent non-human antibody. Petitioner presented no evidence
`
`of any antibody disclosed in the asserted references that has those properties. And
`
`the record now confirms that these properties are highly unpredictable and that a
`
`skilled artisan would not have had a reasonable expectation of success in achieving
`
`those specific claim limitations.
`
`[1.
`
`TECHNOLOGY BACKGROUND
`
`A.
`
`Antibody “Variable” And “Constant” Domains
`
`The immune system defends against foreign substances called “antigens,” by
`
`producing antibodies. Antibodies are proteins that bind to antigens. (Ex-2041 '|]33;
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`Ex-1082 at 160.) A typical antibody, or “immunoglobulin,” has two identical
`
`heavy chains and two identical light chains:
`
`
`.9 a. L——————————_
` ———————-l p 7.}
`
`r-- I I I
`
`$31-..’
`
`(Ex—2041 183; EX-2023 at 10 (annotated); Ex—lOOl, 1:17-20.) Each chain contains
`
`a “variable” domain (red box above) and “constant” domains (green box above).
`
`(Ex—2041 fil35; Ex—IOOI, 1:20—27.) The heavy chain (VH) and light chain (V1,)
`
`variable domains are illustrated above in blue and pink, respectively.
`
`Variable domains directly bind to the antigen. (Ex-2041 1137; Ex-l 001,
`
`1:35-37.) Each variable domain contains three “complementarity determining
`
`regions,” or “CDRs,” (Ex—204] fl38; Ex—lOOl, 1:35—50), shown as CDRl, CDRZ,
`
`and CDR3 in the enlarged portion above. Variable domains also contain four
`
`“framework regions,” or “FRs”—one on either side of each CDR—shown as FRI,
`
`FRZ, FR3, and FR4 in the same enlarged portion. The framework regions form a
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`core structure from which the CDRs extend and form a binding site for the antigen.
`
`(EX—2041 1140; Ex-lOOl, 1:47—50.) Unlike the CDRs, which generally contain
`
`unique amino acids (or “residues”) for a particular antigen, the framework regions
`
`typically share more amino acid sequences in common (126., the same amino acids
`
`at the same positions) across other antibodies. [Ex-2041 1139; Ex-lOOl, 1:37-44.)
`
`The constant domains are not directly involved in antigen binding and
`
`typically have similar amino acid sequences across all antibodies within a subclass.
`
`(Ex—2041 1136; Bit-2016 1115.)
`
`B.
`
`“Humanized” Antibodies
`
`Before the ’213 patent, antibodies targeting a specific antigen could be
`
`obtained from animals (eg, mice).
`
`(EX—2041 1148; Ex—lOOl, 1:52—58.) Those non—
`
`human antibodies, however, had limited use therapeutically because the human
`
`immune system would overtime identify them as antigens and attack them—
`
`known as an “immunogenic” response. (Ex-2039, 159:5-11; Ex-2041 1150; Ex-
`
`1001, 1:55-58.) An immunogenic response had adverse clinical consequences,
`
`including diminished efficacy and allergic reactions.
`
`(Ex—2041 1151.)
`
`Scientists developed several techniques seeking to address immunogenicity.
`
`One involved “chimeric” antibodies that combined a non-human variable domain
`
`with a human constant domain.
`
`(Ex—2041 1153; Ex—lOOl, 1:59-2:19.) However,
`
`immunogenicity could still result because chimeric antibodies retained a significant
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`portion of the non—human antibody sequence.
`
`(Ex—2039, 242:3—20; Ex—204l 1154;
`
`Ex-1001, 2:12-19; Ex—2022 at 2156.)
`
`Scientists also created “humanized” antibodies containing a human variable
`
`domain substituted with the amino acid sequence of the non-human CDRs. (Ex-
`
`2041 1155; Ex-lOOl, 2:20-52.) But that approach could reduce the antibody’s
`
`ability to bind to specific antigens.
`
`(Ex—2041 1161; Ex—1034 at 10033.)
`
`Scientists pursued techniques for making humanized antibodies that
`
`balanced strong binding with low immunogenicity. (Ex-2041 1161.) For example,
`
`Queen—1989 (Ex-1034) chose an existing human framework that was “as
`
`homologous as possible to the original mouse antibody to reduce any deformation
`
`of the mouse CDRs.” (Ex-1034 at 10033.) The humanized sequence was then
`
`fiirther refined using computer modeling “to identify several framework amino
`
`acids in the mouse antibody that might interact with the C DRs or directly with
`
`antigen, and these amino acids were transferred to the human framework along
`
`with the CDRs.” (Id) That technique became known as the “best-fit” approach
`
`because it started from an existing human sequence with the closest match to the
`
`non-human antibody. (Ex-2041 111156-60; Ex-2024 at 4184.)
`
`Even using the best—fit approach, however, it still was difficult to produce an
`
`antibody with both strong binding and low immunogenicity. (Ex-2041 1168; Ex-
`
`1001, 3:50-52.) The best-fit approach also was inefficient because it required
`
`
`
`identifying a new human framework sequence for each different humanized
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`antibody. (Ex-2041 111185, 261—62.)
`
`[11.
`
`’21 3 PATENT
`
`A.
`
`Invention
`
`Beginning in the late 19805, the inventors of the ’213 patent—Drs. Paul
`
`Carter and Leonard Presta at Genentech—developed a new approach to
`
`humanizing antibodies that solved the prior art binding and immunogenicity
`
`problems. Rather than starting from the most homologous human sequence of an
`
`actual antibody, the inventors developed an artificial “consensus human
`
`sequence”—i.e., “an amino acid sequence which comprises the most frequently
`
`occurring amino acid residues at each location in all human immunoglobulins of
`
`any particular subclass or subunit structure.” (Ex—1001, 11:32—38.) That
`
`“consensus” sequence provided a single human sequence for any humanized
`
`antibody of a particular subclass or subunit structure (e.g., light chain Kl).
`
`(1d,,
`
`54:66-56:57.)
`
`The ’213 inventors developed a multi-step process for their approach. First,
`
`they added the non—human CDRs to the human consensus sequence. (Id, 20: 12—
`
`31.) Next, they evaluated the differences between the framework regions of the
`
`non-human antibody and the human consensus sequence to determine whether
`
`further modifications to the consensus sequence were needed. (Id., 20:32—40.)
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`Where the non—human antibody framework sequence differed from the
`
`human consensus sequence, the inventors used computer modeling to identify
`
`whether the different non—human amino acid (i) “non—covalently binds antigen
`
`directly”; (ii) “interacts with a CDR”; (iii) “participates in the VL-VH interface,”
`
`£.e., the interface between variable domains of the heavy and light chains, or (iv) is
`
`a glycosylation site outside the CDRs that is likely to affect “antigen binding
`
`and/or biological activity.” (Id, 20:32—21 :36, 54:64—56:57.) The inventors
`
`believed that those positions were important to maintaining binding affinity. (1d,
`
`20:32-35.) If any of those requirements was met, that position in the consensus
`
`sequence could be substituted with the amino acid at the same position in the non-
`
`human antibody. Otherwise, the sequence of the human consensus sequence was
`
`retained. (Id, 20:66—2 l :8.)
`
`The ’213 claims reflect the inventors’ novel consensus sequence approach.
`
`They require a “humanized” antibody or variable domain that contains non-human
`
`CDRs that bind antigen when incorporated into the human framework sequence
`
`and certain specified framework substitutions that the inventors determined were
`
`important to antibody binding in their consensus sequence. (Ex-2016 1131.)
`
`B.
`
`Advantages Of “213 Invention
`
`Antibodies containing the ’213 patent’s consensus sequence were a
`
`significant advance over the prior art.
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`First, the ’213 patent’s consensus sequence addressed the immunogenicity
`
`problems of other humanization techniques.
`
`(EX—1002 at 456-58, 'fl'fl2—9; EX—204l
`
`1B3.) At the same time, humanized antibodies embodying the “213 invention
`
`retained strong binding affinity, or even have improved binding over the original
`
`non-human antibody. (Ex-100] , 4:24-28, 51:50-53; Ex-2041 1]83.)
`
`Second, unlike the prior art best—fit approach that used a unique human
`
`sequence for each antibody, the ’213 patent provided a single human sequence that
`
`could be applied to a wide variety of antibodies. (Ex-1002 at 456—58, 11112-9; Ex-
`
`2041 1185.) That broadly-applicable platform is reflected in the ’213 patent’s
`
`claims that specifically require a consensus sequence or that recite framework
`
`substitutions derived from that consensus sequence. (Ex-2041 1185.) Genentech
`
`has used the ’213 invention to develop numerous drugs, including Herceptin®
`
`(breast and gastric cancer), Perjeta® (breast cancer), Avastin® (colon, lung, ovarian,
`
`cervical, kidney, and brain cancer), Lucentis® (macular degeneration), and Xolair®
`
`(asthma).
`
`(EX-2017 1T4; Ex-2016 'fl5.)
`
`C.
`
`Prosecution History
`
`The ’213 patent is a continuation-in—part of an application filed on June 14,
`
`1991. (Ex-1001 at 1.) The challenged claims issued over hundreds of references
`
`considered during prosecution, including every reference in the instituted grounds.
`
`(Ex—1001 at 1-6.)
`
`10
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`During prosecution, the applicants successfully antedated U.S. Patent No.
`
`5,693,762, which had a filing date of September 28, 1990.
`
`(Ex—1002 at 710-11,
`
`721.) As detailed below, the record in this proceeding further confirms that certain
`
`challenged claims were also invented before the publication of either Queen-1990
`
`(July 26, 1990) or Tramontano (September 5, 1990).
`
`IV. ASSERTED REFERENCES
`
`A.
`
`Queen-1989
`
`Queen-1989 describes the humanization of a murine anti-Tac antibody.
`
`(EX-
`
`1034 at 10029.) Unlike the ’213 patent, Queen—1989 does not disclose or suggest
`
`the use of a generalized “consensus” sequence. (Ex-2041 11109.) Instead, Queen-
`
`1989 used a best-fit approach, which involved (i) identifying a framework
`
`sequence of an actual human antibody that was “as homologous as possible to the
`
`original mouse antibody” (Ex-1034 at 1003 3); and (ii) incorporating the murine
`
`CDRs into that human sequence (Ex-1034 at 10033).
`
`(Ex—2041 fl107, 109.)
`
`Queen—1989 then identified additional locations in the human framework to
`
`substitute with murine residues. If the human framework contained “atypical”
`
`residues, Queen-1989 substituted them with more commonly-occurring amino
`
`acids from the murine antibody.
`
`(Ex—1034 at 10032.) Queen—1989 also used a
`
`computer model of the murine antibody “to identify several amino acids which,
`
`while outside the CDRs, are likely to interact with the CDRs or antigen.” (Ex-
`
`11
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`1034 at 10029.) Using those techniques, Queen-1989 made a humanized antibody
`
`with 15 framework substitutions—none of which fall within the scope of the
`
`challenged claims.
`
`(Ex—1034 at 10031—32; Err—2041 fl105.)
`
`B.
`
`Queen-1990
`
`Queen-1990 is a PCT application published July 26, 1990.
`
`It is not prior art
`
`to certain challenged claims. (Infra §VIII.A.)
`
`Queen-1990 used a best-fit approach to produce a humanized antibody. (Ex-
`
`1050, 265—3325; EX-2041 11111 10—1 1 .) Queen—1990 identified four general criteria
`
`for designing humanized antibodies. (Ex-2041 W1 1 [-19.)
`
`Criterion I: Queen—1990 emphasized the importance of choosing the human
`
`sequence most similar to the non—human antibody to reduce the possibility of
`
`distorting the binding site formed by the CDRs. (Ex-1050, 12: 17-35.) Queen-
`
`1990 mentioned “a consensus framework from many human antibodies” (id,
`
`12:19-20), but included no details of what that “consensus framework” might be or
`
`how it might be used to make a humanized antibody. (Ex-2041 WI 13.) Indeed,
`
`Petitioner’s expert, Dr. Riechmann conceded that the term as used in Queen—1990
`
`was “ambiguous.” (Ex—2039, 284:10-13.)
`
`Criterion H: After selecting a best-fit human framework sequence, Queen-
`
`1990 provided that “unusual” or “rare” amino acids could be replaced with more
`
`common amino acids from the non-human sequence.
`
`(Ex—1050, 13:22—32.) This
`
`12
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`step was intended to eliminate residues that may “disrupt the antibody structure”
`
`by replacing them with non—human residues commonly found in other human
`
`antibody sequences.
`
`(1d,, 13:32—37.)
`
`Criterion HI: Queen-1990 disclosed that non-human residues may be used
`
`immediately adjacent to CDRs to help maintain binding affinity. (Id, 14: 1-12.)
`
`But as Petitioner’s expert Dr. Riechmann confirmed, substituting residues at these
`
`positions is “optional, not mandatory.” (Ex—2039, 289120—22.) Queen-1990
`
`provides no guidance on which of these residues should be substituted for any
`
`given antibody.
`
`Indeed, as Dr. Riechmann noted, “[t]hat would not be a sensible
`
`thing to do” because substitutions would vary according to the particular antibody
`
`to be humanized, and “the structural components in each case are different.” (Ex-
`
`2039, 29122—292110.)
`
`Criterion IV: Queen-1990 used computer modeling, “typically of the
`
`original donor antibody,” to identify other residues that “have a good probability of
`
`interacting with amino acids in the CDR’s [sic] by hydrogen bonding, Van der
`
`Waals forces, hydrophobic interactions, etc.” (Ex-1050, 14:14-19.) Non-human
`
`residues “may [or] may not” be substituted at those positions that may interact with
`
`CDRs “depending on the particular antibody that you’re trying to humanize.” (Ex—
`
`2039, 29425-8; Ex-lOSO, 14:19-21.) Amino acids satisfying this criterion
`
`13
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`“generally have a side chain atom within about 3 angstrom units of some site in the
`
`CDR’s [sic].” (Ex—1050, 14:22—25.)
`
`Queen—1990 disclosed a humanized antibody sequence produced using its
`
`technique. (Id, Fig. 2.) That antibody contained 15 framework substitutions—
`
`none of which correspond with the “213 claims. (Ex-2041 fl 122.) Queen-1990
`
`states that the antibody produced using its technique had a binding affinity within
`
`about 3- to 4—fold of the parent murine antibody, but does not indicate any
`
`improvement in binding affinity for the humanized antibody.
`
`(Ex—2041 11123; Ex-
`
`1050, 31:33-37.) Queen—1990 does not describe or report any testing of
`
`immunogenicity for this humanized antibody.
`
`(EX-2041 11123.)
`
`C.
`
`Protein Data Bank
`
`The Protein Data Bank (“PDB”) “was established in 1971 as a computer-
`
`based archival file for macromolecular structures.” (Ex—1080 at 535.) As of 199],
`
`the PDB included structural information for only a small number of antibodies or
`
`antibody fragments, whose crystal structure had been solved—a process that at the
`
`time could take several years for a single antibody.
`
`(Ex—2041 $54.) As a
`
`database, the PDB does not describe the humanization of antibodies, let alone what
`
`substitutions may be relevant for any particular antibody. (Ex-2041 11155.)
`
`Petitioner cites data from nine antibody crystal structures available in the
`
`PDB database prior to August 1989.
`
`(EX—1003C, Riechmann Exs. D—L.) As
`
`14
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`discussed below, Petitioner contends that those crystal structures would have
`
`supposedly led to numerous possible framework substitutions.
`
`D.
`
`Tramontano
`
`Tramontano (Ex—1051) was published on September 5, 1990.
`
`(Ex—2027
`
`(showing date).) Tramontano is not prior art to certain claims. (Infra §VIII.A.)
`
`Tramontano analyzed several antibody structures and found that “the major
`
`determinant” of the position of one of the CDRs “is the size of the residue at
`
`[heavy chain] site 71-” (Ex—1051 at 175.) Tramontano discussed potential
`
`“applications to antibody engineering,” explaining that “[fjor the binding site of
`
`the synthetic product to be the same as that in the original antibody, the
`
`frameworks should have the same residues at those sites important for the positions
`
`and conformations of the hypervariable regions.” (Ex-1051 at 181.) Tramontano,
`
`however, never suggested that substitutions at position 71H were desirable.
`
`(Ex—
`
`2041 11149.)
`
`Instead, Tramontano discussed humanized antibodies reported in Jones (Ex-
`
`1033), Verhoeyen (Ex—1068), and Riechmann (Ex—1069).
`
`(Ex—1051 at 18]; Ex—
`
`2041 $50.) Jones had the same residue at 71H as the parent antibody and “had the
`
`same affinity
`
`as the original mouse antibody.” (Ex-1051 at 181; Ex-204l
`
`11150.) By contrast, Verhoeyen (Ex—1068) included a different residue at 71H than
`
`the murine antibody and saw a 10—fold reduction in binding affinity. (Ex-1051 at
`
`15
`
`
`
`IPR2017-01373
`
`Patent Owner’s Response
`
`181; Ex-2041 1150.) Finally, Riechmann (Ex-1069) had a different residue at 71H,
`
`but maintained a binding affinity “close to that of the rat original.” (Ex—1051 at
`
`181; Ex-2041 11151-) Tramontano had no explanation for those divergent results.
`
`(Ex-1051 at 181; Ex-2041 111150-51.)
`
`E.
`
`Kabat—1987
`
`Kabat-l987 (Ex-1052) is a reference book of antibody sequences that
`
`includes statistics on the most common amino acids for a given type of
`
`immunoglobulin.
`
`(1d. at 8; Ex—2041 11156.)
`
`Kabat- 1987 does not describe antibody humanization or discuss
`
`substitutions that may be beneficial when humanizing an antibody. (Ex-2041
`
`1157.) Rather, Kabat—l987’s tabulati