`REPORTS IN
`MEDICINAL
`CHEMISTRY
`Volume 29
`
`Sponsored by the Division of Medicinal Chemistry
`of the American Chemical Society
`
`EDITOR-IN-CHIEF
`JAMES A. BRISTOL
`PARKE-DAVIS PHARMACEUTICAL RESEARCH
`DIVISION OF WARNER-LAMBERT COMPANY
`ANN ARBOR, MICHIGAN
`
`SECTION EDITORS
`WILLIAM K HAGMANN • JOHN C. LEE • JOHN M. MCCALL
`JACOB J. PLATTNER • DAVID W. ROBERTSON • MICHAEL C. VENUTI
`
`EDITORIAL ASSISTANT
`LISA GREGORY
`
`(k
`
`ACADEMIC PRESS, INC.
`San Diego New York Boston London Sydney Tokyo Toronto
`
`eonard Presta, Ph.Dj
`/1/2O18
`
`j
`
`Repored iy Chrote Lacey
`APR, GSA S14224
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01489
`PFIZER EX. 1696, Page 1
`
`
`
`Academic Press Rapid Manuscript Reproduction
`
`This book is printed on acid-free paper.
`
`Copyright © 1994 by ACADEMIC PRESS, INC.
`
`All Rights Reserved.
`No part of this publication may be reproduced or transmitted in any form or by any
`means, electronic or mechanical, including photocopy, recording, or any informat4on
`storage and retrieval system, without permission in writing from the publisher.
`
`Academic Press, Inc.
`A Division of Harcourt Brace & Company
`525 B Street, Suite 1900, San Diego, California 9210J-4495
`
`United Kingdom Edition published h
`Academic Press Limited
`24-28 Oval Road, London NW 7DX
`
`International Standard Serial Number: 0065-7743
`
`International Standard Book Number: 0-12-040529-6
`
`PRINTED IN THE UNITED STATES OF AMERICA
`96 97 98 99 Q 987654321
`94
`95
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01489
`PFIZER EX. 1696, Page 2
`
`
`
`Chapter 32. Humanized Monoclonal Antibodies
`
`Leonard Presta
`Genentech, Inc., 460 Point San Bruno Blvd.
`South San Francisco, CA 94080
`
`-
`
`Introduction - Almost two decades ago, the development of monoclonal hybridoma
`technology -
`the ability to make cultured cells produce antibodies of predefined specificity
`promised a new weapon in the arsenal of molecules able to combat disease (1).
`Antibodies against a specific antigen (or target molecule) could be generated, incorporated
`into a hybridoma for production, and then used in diagnosis or therapy. However, only
`rodent monoclonal antibodies could be made due to technological limitations. In most
`clinical applications, these rodent monoclonal antibodies exhibit properties which severly
`limit their utility. First, they may induce an immunogenic response in humans, referred to as
`HAMA (human anti-mouse antibodies), when the human immune system recognizes them as
`foreign substances. Second, the therapeutic efficacy of the rodent antibody may also be
`reduced because a rodent antibody is cleared from serum more rapidly than human ones.
`Third, in humans, rodent antibodies generally exhibit only weak recruitment of effector
`Junctions, such as antibody-dependent cell-mediated cytotoxicity and complement fixation,
`which may be requisite for the function of the antibody. Use of a human antibody would
`circumvent these three problems.
`
`With advances in molecular biology and mammalian tissue culture, it became
`possible to obtain useful amounts of any antibody, thereby surmounting the limitations of
`hybridoma technology and use of only rodent antibodies. But one problem still existed --
`how does one obtain human antibodies against a particular human antigen? Even if one
`could ethically use a human subject as the biological factory, the human immune system, in
`general, does not produce antibodies against human proteins. Two techiques have been
`developed to address this problem. Earliest of these was the construction of chimeric
`antibodies in which entire antigen-binding domains from a rodent antibody are substituted
`for those of a human antibody (Fig. 1) (2). In many cases, these molecules still exhibited
`some, albeit reduced, HAMA. The next step was generation of humanized antibodies in
`which a significantly reduced number of rodent residues are incorporated into a human
`antibody such that the humanized antibody has the same binding characteristics and
`specificity as the original rodent antibody (3-5). In order to understand the technique of
`humanization and the difference between a chimeric and humanized antibody, the structure
`of antibodies must be appreciated.
`
`two identical light
`Antibody Structure - An antibody consists of four peptide chains --
`which form a 'V' shape. The functions of the
`chains and two identical heavy chains -
`antibody reside in different domains. Antigen binding occurs at the ends of the arms of the
`'V', each arm being referred to as a Fab or antigen binding fragment. Hence each antibody
`can bind two antigen molecules. The effector functions reside in the base of the 'Y',
`referred to as the Fc portion. Each light chain consists of one variable domain and one
`constant domain; each heavy chain consists of one variable domain and three constant
`domains (Fig. 1). The constant' notation refers to the fact that for a particular species and
`immunog!obulin class (e.g. lgG, IgE, lgA) the amino acid sequence for the constant domain
`
`ANNUAL REPORTS IN MEDICINAL CHEMISTRY-29
`
`M
`
`Copyright (D 1994 by Academic Press, Inc.
`All rghLs of reproduction In any form reserved.
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01489
`PFIZER EX. 1696, Page 3
`
`
`
`Z51 8
`
`Section Vi-Topos ua Drug Design and Discovery
`
`Venuti, Ed.
`
`is, for our purposes, invariant. The 'variable' notation refers to the fact that certain portions
`of variable domains differ extensively in sequence among antibodies and are used in the
`binding and specificity of each particular antibody for its particular antigen. These portions
`are referred to as complementarity-determining regions (CDRs). Both light and heavy chain
`variable domains contain three CDRs to give a total of six CDRs involved in binding antigen.
`Each CDR is a contiguous sequence of amino acids which form a loop connecting two 3-
`strands of the framework. Note though that the amino acid sequence of the non-CDR
`portion of the 'variable' domain is relatively invariant and is used to categorize these
`domains into subgroups (6).
`
`o
`
`Mouse
`
`Chimera
`
`\,
`
`U
`Humanized
`Human
`Figure 1. Chimeric and Humanized Antibodies. CDRs are represented by 'fingers'
`on variable domains of the light chain (VO and heavy chain (VH). Note that only the mouse,
`chimera and humanized antibodies bind antigen. Disulfide bonds between the two heavy
`chains and between the light and heavy chains are denoted by thicker, dark lines.
`
`In one type of chimeric antibody all of the variable domains from a rodent antibody
`are fused onto the constant domains of a human antibody (Fig. 1). This type of chimeric
`antibody is comprised of four rodent domains and eight human domains and consequently
`retains approximately 33% rodent residues. In the corresponding humanized antibody, only
`the six rodent CDRs replace the six human CDRs (Fig. 1). The resulting antibody contains
`only about 5-10% rodent residues and is still chimeric in that it has residues derived from
`different species. From the perspective of the human immune system, the humanized
`antibody is the more human of the two types as it contains fewer rodent residues.
`Moreover one must appreciate that the amino acids in CDRs are highly variable and this
`variance is independent of species. The nature of the CDRs is dependent upon the target,
`whereas the framework is species dependent. Hence even though the humanized antibody
`still retains about 5-10% rodent residues, these residues would also be variant in human as
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01489
`PFIZER EX. 1696, Page 4
`
`
`
`Chap. 32
`
`Humanized Monoclonal Antibodies
`
`Presta
`
`319
`
`well as rodent and thus the humanized antibody could well be very similar to a human
`antibody directed against the same site on the same target as the original rodent antibody.
`
`Humanized Antibody Construction - Al first glance constructing a humanized antibody is
`rather simple and straightforward. One chooses a human antibody, clips off the six human
`CDRs and replaces them with their structurally analogous rodent CORs. Indeed some
`humanized antibodies have approached this simplicity in method (7,8). However, there are
`both theoretical and technical difficulties. Among the former is the exact definition of a
`CDR, i.e. exactly which residues comprise the CDRs and which comprise the framework.
`Two definitions have been proposed. The first relies solely on analyses of the amino acid
`sequences of a multitude of antibodies (6). The variance of amino acid types at each
`residue position of antibodies was evaluated and those positions which exhibited a
`relatively large variance were included in a CDR as long as they were contiguous to other
`positions with high variance. Residue positions showing relatively low variance were termed
`framework. The second definition was derived from structures of antibodies (9). When the
`two definitions are compared some striking differences are apparent. While the sequence-
`based CDRs are generally larger than, and encompass, the structure-based CDRs, a
`notable discrepancy occurs in the definition of CDR-H1 (i.e. the first CDR loop in the heavy
`chain sequence) where the two definitions overlap by only two residues. Generally, the
`sequence-based definition of CDRs has been used when an antibody is humanized.
`
`The technical difficulties fall into three categories. First, the choice of constant
`domains must be made. To date almost all antibodies humanized have been those of the
`lgG class. In humans there are four recognized subclasses, denoted lgG 1, lgG2, lgG3 and
`lgG4, the classification of which is based on the sequence of their constant domains. Each
`subclass also exhibits differences in effector, or biological, functions mediated by the
`constant domains. The intended therapeutic use of the antibody must be considered when
`choosing the lgG subclass, e.g. is complement fixation an advantage or disadvantage for
`the particular application?
`
`Second, the choice of human variable domains, both light and heavy, must be
`made. In the homology or 'best-fit' method, the sequence of the rodent variable domain is
`screened against the entire library of known human variable domain sequences. The human
`sequence which is closest to that of the rodent is then accepted as the human framework
`(8,10). Another method uses a particular framework derived from the consensus sequence
`of all human antibodies of a particular subgroup of light or heavy chains. The same
`framework may be used for several different humanized antibodies (11,12). Remembering
`that the purpose of humanization is to trick the human immune system into recognizing the
`humanized antibody as one of its own, is the 'best-fit' or the consensus framework better for
`this trickery? Unfortunately there are very limited in vivo data on the reaction of the human
`immune system to humanized antibodies and no comparative study has been performed on
`the two types. In the final analysis both may function well with regard to acceptance by the
`human immune system, with perhaps an occasional aberration.
`
`designing the humanized
`Third is the most imposing technical difficulty --
`antibody such that it has the same binding affinity and specificity as the original rodent
`antibody. When the rodent CDRs are simply grafted onto a human framework the resultant
`humanized antibody may exhibit signficiantly reduced binding compared to the original
`rodent antibody (11,12). This reduction in binding may be caused by certain buried
`framework residues. Each CDR is a loop and these loops may be anchored to the
`framework not only covalently at their ends but through noncovalent interactions with the
`sidechains of buried framework residues. For example, in CDR-L1 the sidechain at position
`29 (residue numbering is according to ref. 6) is usually hydrophobic, buried and contacts
`hydrophobic sidechains at framework positions 2, 25, 33, 71 and 90 (13). If an incorrect
`amino acid is chosen for any of these framework positions the packing of CDR-L1 against
`the protein might be altered, resulting in an incorrect presentation of the exposed CDR-L1
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01489
`PFIZER EX. 1696, Page 5
`
`
`
`Section VI—Topics in Drug Design and Discovery
`
`Venuti, E4.
`
`sidechains and consequent reduction in binding to antigen. Such a case may occur when
`the sidechains at position 29 in the rodent CDR and positions 2, 25, 33, 71 and 90 in the
`rodent framework differ from those in the human antibody. In order to retain proper binding
`the offending framework residues must be altered from human to murine so that the CDR
`packs against the protein properly. Alternatively, the buried sidechain in the CDR itself,
`such as at position 29 in CDR-L1, might be changed from murine to human to evaluate
`whether proper CDR conformation can be attained. This regimen must be evaluated for
`each of the six CDRs because if just one CDR has an improper conformation it may reduce
`the binding of the humanized antibody to the point that it is useless as a therapeutic
`replacement for the rodent antibody. While buried framework residues can affect the
`binding of the humanized antibody through an indirect mechanism, i.e. by influencing the
`conformation of a CDR, occassionally an exposed framework residue is involved in direct
`binding to the antigen and must be included in the humanized form (11,14).
`
`While performing this regimen for six CDRs may seem a daunting task, especially
`since crystal structures of the rodent and human antibodies are usually not available,
`nature has provided some guidance. Even though CDRs vary widely in sequence and to a
`lesser extent in size, analysis of crystal structures of Fab fragments shows that for each
`CDR (except CDR-H3) only a limited number of conformations may be utilized (13). These
`canonical conformations are dependent on size and sequence of the CDR. Hence if one
`inspects the sequence and size of the particular rodent CDR the canonical conformation
`can then be assigned. This provides information as to which CDR residues will be exposed,
`which will be buried and which framework residues may play a role in stabilizing the CDR
`conformation.
`
`How has this technique fared regarding the ability of a humanized antibody to
`reproduce the binding affinity of its parent rodent antibody? The first humanization involved
`transferring only the CDRs of the heavy chain from a murine to human antibody. When
`expressed with the murine light chain, the humanized heavy chain exhibited binding less
`than 2-fold down from the parent murine (15). Two years later another two humanizations
`were published in which both the light and heavy chain CDRs were transferred. In the first, a
`humanized anti-lysozyme antibody, binding was reduced by 10-fold (16). The second was
`more interesting since it was potentially clinically relevant, underscored the importance of a
`single buried residue for maintaining proper CDR conformation, and noted the difference
`between the sequence-based and structure-based CDR definitions (17). Directed against
`the CAMPATH-1 antigen on human lymphocytes and monocytes, the initial humanized
`version was 40-fold down in binding. Changing the human Ser27 to murine Phe27 in the
`heavy chain resulted in a version which exhibited only a 3-fold reduction in affinity. Was
`this a framework or a CDR residue? According to the structure-based definition of CDRs
`used to design the humanized anti-CAMPATH-1 antibody, heavy chain position 27 was part
`of the framework while according to the sequence-based definition position 27 was part of
`the CDR. Regardless of the definition, the marked improvement in binding when the murine
`Phe was substituted for the human Ser emphasized the need to pay attention to framework,
`as well as CDR, sequence in future humanizations.
`
`Since 1988 the technique has developed and become increasingly utilized for
`clinically relevant antibodies. The 'best-tit method, used first in 1989 (10), has remained
`the more popular method for designing the sequence of the humanized antibody than the
`later consensus method (11). More recently, in the design of an anti-IgE humanized
`antibody, both methods were evaluated (14). The study concluded that although the
`consensus sequence showed the best overall binding, no clear advantage in binding was
`evident for the consensus antibody versus the best-fit' antibody. Another humanization of
`an anti-IgE antibody, utilizing only the consensus method, was reported earlier that year
`(12).
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01489
`PFIZER EX. 1696, Page 6
`
`
`
`Chap. 32
`
`Humanized Monoclonal Arntnbodiee
`
`Prosta
`
`N a
`
`Both of the anti-IgE humanizations used another recent technology in the design
`of the antibody - molecular modeling. As the database of antibody crystal structures
`grows (which consists predominanly of structures of Fab fragments), information
`accumulates on the three-dimensional structure of variable domains and the relationship
`between CORs and framework. Since the structure of the rodent antibody of interest will
`usually not be available, a computer graphical model based on crystal structures can be
`generated and used to determine which framework residues might influence CDR
`conformation. Many of the humanizations published of late include use of molecular
`modeling or crystal structure information (7,12,14,18,19). Only recently, however, have
`crystal structures of humanized antibodies been determined and compared to the molecular
`model used to design the humanized antibody (20,21). In this case, the model correctly
`predicted the conformation of the framework and five of the CDR loops. The conformation of
`the sixth CDR in the model differed from that in the crystal structure, but this same CDR also
`varied in conformation among the crystal structures of three closely related humanized
`variants (21).
`
`Though some humanizations have been successful without any alteration of the
`chosen human framework (7,8), other studies have found it necessary to alter some human
`framework residues to murine in order to attain binding equivalent to the parent murine
`antibody.
`In both of the aforementioned anti-lgE humanizations, murine framework
`In humanization of the Mikj1 antibody, eighteen
`residues were important (12,14).
`framework changes from the 'best-fit' sequence were incorporated (18). Another version of
`Mik1 with just six changes bound only about 24old less well than the version with eighteen
`changes. However the latter was superior in biological activity and it was noted that the
`bioactivities of these two versions could not be predicted solely from their binding affinities
`to the target molecule, in this case the 1L2 receptor n-chain. Such a lack of correlation
`between binding affinity and bioactivity had been noted previously for another humanized
`antibody (11). The importance of a particular framework residue may vary from antibody to
`antibody. For example, among six humanizations reported very recently, two retained the
`human amino acid at heavy chain position 71 (7,8), three changed to the murine amino acid
`(14,18,19) and in the last the murine and human consensus residues at position were
`identical and no change was required (12). This emphasizes the importance of considering
`each murine antibody as idiosyncratic, i.e. each antibody may possess its own unique CDR-
`framework interactions which are sensitive to changes in amino acid constitution. As more
`humanizations are performed and antibody crystal structures become increasing available,
`perhaps nature's rules for determining these characteristic packing interactions may
`become more evident so that they can be ascertained during the initial design of the
`humanized antibody. Some effort in this area has already been accomplished (13,22,23).
`
`While buried framework residues have been shown to be important, less attention
`has been focused on exposed (or semi-exposed) framework residues. Crystal structures of
`Fab-antigen complexes have shown that on occassion a framework residue will be involved
`in direct binding to the antigen (24). The first mention of an exposed framework residue
`important for retention of binding and bioactivity in a humanized antibody was for the anti-
`p185HER2 antibody (11). The murine kg at position 66 in the light chain, predominantly Gly
`in human kappa light chains, was shown to improve binding by 4-fold and improve
`bioactivity. More recently, having the murine amino acids at light chain residues 1 and 3
`were shown to enhance binding (14). Though the involvement of exposed framework
`residues in binding antigen may not be common, they deserve consideration when
`attempting to further improve the binding of a less than competent humanized antibody.
`
`The focus of humanized antibodies has been on the variable domains since these
`bind antigen, but some studies have begun to concentrate on the rest of the humanized
`molecule, namely the Fc portion which possesses the effector functions. While some
`humanized antibodies have exhibited the ability to effect antibody-dependent cell-mediated
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01489
`PFIZER EX. 1696, Page 7
`
`
`
`322
`
`Section VI—Toptos in Drug Design and Discovery
`
`Venuti, Ed
`
`cytolysis (11,18,25,26) others have not (8). in other studies the differences between the
`various classes of lgG have been evaluated (27) and, recently, the removal of conserved
`carbohydrate in the Fc has been investigated (28). As humanizations become more
`commonplace (or are replaced by human antibodies) and clinical data become available, the
`necessity for understanding and tailoring the Fc for a desired effector function will increase.
`
`Therapeutic Use of Chimeric and Humanized Antibodies - By the mid-1980's, it was
`recognized that rodent antibodies used as therapeutics would be problematic in that they
`may elicit a HAMA response, are more rapidly cleared from serum and less efficacious at
`eliciting effector functions (29). The HAMA response was also categorized into at least two
`types of anti-antibodies: those directed against the CDRs, referred to as anti-idiotypic
`antibodies, and those directed against the remainder of the molecule, referred to as non-
`anti-idiotypic antibodies (30). Initially chimeric antibodies were thought be the solution. in
`one interesting study the immunogenicity of chimeric antibodies was evaluated by
`introducing human antibodies and human/murine chimeric antibodies into mice (31). The
`murine response to fully human antibodies was robust, with 90% of the anti-human
`antibodies directed against the constant domains and 10% directed against the variable
`domains. The chimeric antibodies also elicited a response directed against the human
`variable domains, whereas fully murine antibodies gave no response or only an anti-
`ailotypic response. The latter arise from variation in the sequence of constant (and
`sometimes variable) domains among individuals and is akin to blood group antigens (A, B,
`AS, 0). In humans, chimeric antibodies have given mixed results. For example, the first
`chimeric antibody in a clinical trial, directed against a surface antigen on adenocarcinoma
`cells of gastrointestinal origin, elicited a detectable but modest HAMA response in only 1 of
`10 patients (32). in contrast, the same investigators later reported that a different chimeric
`antibody, directed against a different surface antigen on adenocarcinoma cells, elicited a
`HAMA response in 7 of 12 patients after only one infusion (33). For both antibodies the
`HAMA response was directed against the murine variable region.
`
`The first humanized antibody used in humans, CAMPATH-1H, induced remission in
`two patients with non-Hodgkin lymphoma and showed no detectable HAMA (34). CAMPATH-
`iN was evaluated a few years later for rheumatoid arthritis (35). In this study, 7 of 8
`patients showed significant clinical benefit and all 8 patients exhibited no HAMA after one
`course of treatment. However upon retreatment, 3 of 4 patients did show a response. Of
`these, two had only an anti-idiotypic response while the third had both an anti-idiotypic and
`anti-allotypic response. Another antibody which was among the earliest humanized, the
`anti-Tac antibody (10), has been shown to be less immunogenic in cynomolgus monkeys
`than the parent murine antibody (36). As expected the response in the cynomolgus
`monkeys to the humanized antibody was anti-idiotypic while that against the murine was
`directed against the rest of the immunoglobulin. These anti-idiotypic antibodies have
`recently been shown to be directed primarily against combinations of CDRs Hi, H2 and L3
`rather than against a single CDR (37). In a similar study, murine and humanized OKT4A
`antibodies were evaluated in cynomolgus monkeys (38). All 17 monkeys which received
`murine OKT4A showed a response which was both idiotypic and non-idiotypic. While all 8
`monkeys which received the humanized OKT4A also showed a response, it was only anti-
`idiotypic, In addition, higher serum levels of the humanized antibody were maintained for a
`longer period following treatment than were serum levels of the murine OKT4A.
`
`Finally, during the past year a study was published in which mice injected with
`murine antibodies were used as a model for estimating potential patient sensitization
`against humanized antibodies (39). As in a similar earlier study (31), anti-idiotypic and anti-
`altotypic antibodies were elicited. That both types of anti-antibodies have been found
`underscores the need to determine the biological and therapeutic consequences of each.
`Anti-idiotypic antibodies directed against the CDRs may prove to be a necessary evil when
`humanized antibodies are used as therapeutics since the CDRs are the part of the
`immunoglobulin which bind antigen and therefore cannot be altered. Anti-idiotypic
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01489
`PFIZER EX. 1696, Page 8
`
`
`
`Chap. 32
`
`Humanized Monoclonal Antibodies
`
`Presta
`
`323
`
`antibodies could present a problem in that they would block the function of the humanized
`antibody after a repeated application, thereby requiring an increased dose and/or
`significantly reducing the efficacy of the humanized antibody. If this is the only problem
`caused by anti-idiotypic antibodies, then one possible solution would be to have several
`bioactive humanized antibodies which bind the same antigen but have different CDRs. After
`a course of therapy with the first, the second (or even third) could then be used if a chronic
`therapy was required. Anti-allotypic antibodies, directed against idiosyncratic sequences
`of an individual, will be more problematic. Chronic therapy may require matching of a
`recipient's allotype prior to therapy (39). If the allotype of the recipient differs from that of
`the available humanized antibody then treatment might have to be abandoned or,
`alternatively, the constant portion of the humanized antibody would have to be redesigned
`for the patient's allotype. While such a tailored-to-fit' antibody is technically feasible, it
`would undoubtedly significantly increase the cost of treatment. Since no studies have been
`published on the biological consequences of either anti-idiotypic or anti-allotypic antibodies
`in humans, these questions await further clinical trial.
`
`Conclusion - Although several successful humanizations of antibodies have been
`published, clinical data at this point in time are scarce due to the relatively recent
`development of the technique. What is available supports the contention that antibodies
`hold great promise as therapeutics (3,4). Most likely, however, humanization as a
`technique will be replaced by novel methods of generating human antibodies. Two such
`technologies are the production of human antibody repertoires in transgenic mice (40) and
`from phage display libraries (41 42). Though it is destined for replacement, humanization of
`antibodies has stimulated research on the structure-function of antibodies and provided an
`important intermediary step in utilization of monoclonal antibodies as therapeutics.
`
`1. G. Kohler and C. Milstein, Nature, 2a 52 (1975).
`2. S.L. Morrison, M.J. Johnson, LA. Hevzenberg and V.T. 01, P,oc.Natt.Acad.Sci.USA, 5t 6851 (1984).
`3. G. Winter and W.J. Harris, Trends Ptiarmacol.Sci., 14, 139 (1993).
`4. L.K. Jolliffe, lntRev.lmmunol., 10, 241 (1993).
`5. L.G. Presta, Carr.0p4nn Struct.Biol., 2, 593(1992).
`6. E.A. Kabat T.T. Wu, H.M. Pony, K.S. Gottesman and C. Froefler, 'Sequences of Proteins of
`Immunological Interest', Vol. 1, 51h Ed., NIH Publ. No. 91-3242 (1991)
`7. M.E. Verhoeyen, J.A. Saunders, M.R. Pnce, J.D. Marugg, S. Briggs, E.L. Brodenck S.J. Eida, A.T.A.
`Mooren and R.A. Badley, Immunology, Z. 364(1993).
`8. M.J. Sims, D.G. Hassat, S. Brett, W. Rowan, M.J. Lockyor, A. Angel, A.P. Lewis, G. Hale, H. Waldmann
`and S.J. Crowe, Jimmunol., 151, 2296 (1993).
`9. C. Chothia and AM, Lesk, J.MoLBiol.,.196,901 (1987).
`10. C. Queen, W.P. Schneider, H.E. Selick, P.W. Payne, N.F. Landolfi, J.F. Duncan, N.M. Avdalovic, M.
`Levitt, R.P. Junghans and T.A. Waldmann, Proc.Nati.Acad.Sci.USA, 86, 10029(1989).
`11 P. Carter, L. Presta, C.M. Gorman, J.B.B. Ridgway, D. Henner, W.L.T. Wong, A.M. Rowland, C. Kotts,
`M.E. Carver and H.M. Shepard, Pn,c.Natl.Acad.Sci.USA, 8, 4285(1992).
`12.L.G. Presta, S.J. Lahr, R.L. Shields, J.P. Porter, C.M. Gorman, B.M. Fendly and P.M. Jardleu,
`J.Immunol,, J-51,2623 (1993).
`13. C. Chothia, A.M. Lesk, A. Tremontano, M. Levitt, S.J. Smith-Gill, G. Air, S. Sheriff, E.A. Padlan, 0.
`Davies, W.A.Tulip, P.M. Colman, S. Spinelli, P.M. Aizari and R.J. Poljak, Nature, 3AZ 877(1989),
`14. F. Kolbinger, J. Saldanha, N. Hardman and M.M. Bendlg, Protein Eng., 1 971 (1993).
`15. P.T. Jones, P.H. Dear, J. Foote, M.S. Neuberger and G. Winter, Nature,
`j.. 522(1986).
`16, M. Verhooyen, C. Milstein and G. Winter, Science, 22. 1534 (1988).
`323(1988).
`17. L. Roichmann, M. Clark, H. Waldmann and G. Winter, Nature, =
`18. J. Haicimi, V.C. Ha, P. Lin, E. Campbell, M.K. Gately, M. Tsudo, P.W. Payne, T.A. Waldmann, A.J.
`Grant, W.-H. Ten and W.P. Schneider, J.Immunol., .11, 1075(1993).
`19. K. Sato, M. Tsuchiya, J, Saldanha, V. Koishihara, V. ohsugi, T. Klshimoto and M.M. Bendig, Cancer
`, 851 (1993).
`Res.,
`, 969(1993).
`20. C. Eigenbrot, M. Randal, L Presta and A.A, Kossiakoff, J.Mot.Biot.,
`21. C. Ei9enbrot, T. Gonzalez, J. Mayeda, P. Carter, W. Werther, T. Hotaling, J. Fox and J. Kessler,
`Proteins, ,j.. 49(1994).
`22. J. Foote and G. Winter, J.Mol.8io1., '21487 (1992).
`23. A. Tramontano, C. Chothia and A.M. Lesk, J.Mol.Biol., 215,175 (1990).
`
`PFIZER and SAMSUNG v. GENENTECH
`IPR2017-01489
`PFIZER EX. 1696, Page 9
`
`
`
`M
`
`Section VI- Topics in Drug Design axici Discovery
`
`Venut,, Ed.
`
`24. I.S. Mian, A.R. Bradwell and A,J. Olson, J.Mol.BoL, 2E, 133(1991).
`25. R.P. Junans, T.A. Waldmann, N.F. Landolfi, N.M. Avdaks,nc, W.P. Schneider and C. Queen, Cancer
`Res., 5Q, 1495(1989).
`26. P.C. Caron, M.S. Co, M.K. Bull, N.M. Avdalovic, C. Queen and D.A. Stheinberg, Cancer Res., 2. 6761
`(1992).
`, 1098(1993).
`27. J. Greenwood, M. Clark and H. Waldmann, Eur. J.lmmunoi,
`28. S. Bolt, E. Routledgo, I. Lloyd, L Chatonoud, H. Pope, S.D. Gorman and H. Waldmann, Eur.J.lmmunol.,
`403(1993).
`29. R.J. Benjamin, S.P. Cobbold, M.R. Clark and H. Waldmann, J.Exp.Med., j, 1539(1986).
`30. G.J. Jailers, T.C. Fuller, A.B. Cosimi, P.S. Russell, H.J. Winn and R.B. Colvin, Transplantation, 41, 572
`(1986).
`31. M. Brueggemann, G. Winter, H. Waldmann and M.S. Neuberger, J.Exp.Med., IZQ, 2163(1989).
`32, A.F. LoBugilo, R.H. Wheeler, J. Trang, A Haynes, K Rogers, E.B. Harvey, L Sun, J. Ghrayeb and M.B.
`. 4220(1989).
`Khazaeh, Proc.NatLAcad.Sci.USA,
`33. M.B. Khazaeli, M.N. Saleh, T.P. Uu, R.F. Meredith, R.H. Wheeler, T.S. Baker, D. King, D. Secher, L
`5461 (1991).
`Allen, K Rogers, D. Colcher, J. Schiorn, D. Shochat and A.F. LoBugho, Cancer Res.,
`34. G. Hale, M.R. Clark, P. Marcus, G. Winter, M.J.S. Dyer, J.M. Philips, L. Reichmann and H. Waldmann,
`Lancet,2, 1394(1988).
`35, J.D. Isaacs, R.A. Watts, B.L. Hazleman, G. Hale, M.T. Keogan, S.P. Cobbold and H. Waldmann, Lancet
`74.8(1992).
`36. J. Hakimi, R. Chizzonite, D.R. Luke, P.C. Familleth, P. Bailori, J.A. Kondas, P.S. Pilson, P. Lin, D.V.
`Weber, C. Spence, U. Mondini, W.-H. Tsien, J.L Levin, V.H. Gallati, L Kom, T.A. Wakthiann, C.
`Queen and W.R. Benjamin, J.Immunol.,,147,1352 (1991).
`37. W.P. Schneider, S.M. Glaser, J.A. Kondas and J. Hakimi, J.lmmunot, 150,3086 (1993).
`38. F.L. Delmonico, A.B. Cosimi, T. Kawai, D. Cavender, W.