`
`
`
`
`
`
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`
`
`
`Celltrion, Inc.
`
`
`
`Petitioner,
`
`v.
`
`Genentech, Inc.
`
`Patent Owner
`
`Patent No. 6,407,213
`
`
`
`
`
`
`Inter Partes Review No. IPR2017-01374
`
`
`
`EXPERT DECLARATION OF LUTZ RIECHMANN, PH.D.
`IN SUPPORT OF PETITIONER’S REPLY TO PATENT OWNER’S
`RESPONSE
`
`
`
`
`
`
`
`1 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`I, Lutz Riechmann, Ph.D. declare as follows:
`
`I.
`
`
`Introduction
`
`1.
`
`I am the same Lutz Riechmann who submitted a declaration in
`
`support of Celltrion’s Petition for Inter Partes Review of U.S. Patent 6,407,213 (the
`
`’213 patent) in May 2017. A detailed description of my background and
`
`qualifications may be found in that declaration, which I refer to as my “first
`
`declaration.” In addition to the materials listed in Exhibit 1003B to my first
`
`declaration, I have also considered the materials set forth in the Addendum to this
`
`declaration.
`
`2.
`
`I am being compensated at my standard rate for my time spent
`
`preparing this declaration, and my compensation is not contingent on the outcome
`
`of any matter or on any of the opinions provided below. I have no financial
`
`interest in the outcome of this proceeding.
`
`3.
`
`I provided my understanding of legal concepts as they relate to this
`
`proceeding in my first declaration. My understanding of those concepts has not
`
`changed since I submitted my first declaration.
`
` Opinions
`II.
`A. A Person of Ordinary Skill In the Art Would Understand that The
`Disclosure of the Patent Is Limited
`
`4.
`
`A person of ordinary skill would understand that the patent does not
`
`teach that all of the claimed back mutations and possible combinations of back
`
`1
`
`
`
`2 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`mutations will improve the binding of all antibodies that could be covered by the
`
`claims, or that any specific back mutation will work for a specific antibody other
`
`than those in the examples that were tested for binding. A person of ordinary skill
`
`would have understood from the prior art disclosing the specific structures of
`
`humanized antibodies, as well as the examples in the patent, that the back
`
`mutations that will increase binding will vary from antibody to antibody. The
`
`patent requires a person of ordinary skill to construct models to determine which of
`
`the described back mutations might possibly improve binding in a given project,
`
`and then use mutagenesis, which is the making of antibodies with different back
`
`mutations and then testing them, to confirm which precise combination of
`
`mutations works best to increase binding. A person of ordinary skill would have
`
`understood that there is no significant difference between this method and the
`
`methods in Queen 1989 and 1990 and the other prior art I have cited.
`
`B. A Person of Ordinary Skill in the Art Would Have Been Motivated to
`Combine Queen 1989 and/or Queen 1990 with the PDB
`
`5.
`
`Dr. Wilson argues that a person of ordinary skill in the art would not
`
`have combined the PDB with either of Queen 1989 or Queen 1990 in the manner I
`
`described in my first declaration. (Ex. 2041 (Wilson Decl. for IPR2017-01373) at
`
`¶¶ 177-84.) I do not agree with Dr. Wilson’s criticisms. As I explained in my first
`
`declaration, a person of ordinary skill would have used the publically available x-
`
`ray crystallographic studies of antibodies in the PDB during the humanization
`
`
`
`
`2
`
`
`
`3 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`process, because the PDB database provided a collection of the available antibody
`
`structures at that time. In fact, a person of ordinary skill in the art would have
`
`known it would not have been possible to create an accurate molecular model
`
`without using the information put forth in the PDB (or similar database). As I
`
`discussed, a person of ordinary skill could have processed the information either
`
`using one of a number of commercially available software packages such as
`
`CONTAX, PAIRS, or MIDAS, or could have done the calculations using a
`
`spreadsheet program such as Microsoft Excel. Regardless of the mechanism used
`
`to do these calculations, the results would have been the same.
`
`6.
`
`Further, Dr. Wilson’s assertion that Queen 1989 and Queen 1990
`
`would have taught a POSA to model structure of the parent murine antibody and
`
`not the human framework is immaterial to my analysis. Regardless of whether the
`
`starting point was the murine structure or a humanized structure, a person of
`
`ordinary skill in the art would have used the data from the three-dimensional model
`
`to locate the framework residues that could, because of their positions within the
`
`heavy and light chains of the variable domain, either alter the structure of the
`
`CDRs or interact directly with the antigen during binding. A person of ordinary
`
`skill also would have regarded the Queen methodology as a reliable means of
`
`identifying framework substitutions that could be used to improve binding affinity.
`
`Dr. Wilson has not explained any reason why the outcome of the identification of
`
`
`
`
`3
`
`
`
`4 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`possible framework residues for substitution would be different if the person of
`
`ordinary skill modeled the human framework with the murine CDRs inserted
`
`within it, and I am not aware of any reason.
`
`7.
`
`Also, Queen 1989 considered “other antibody V domains with known
`
`crystal structure” when constructing the antibody models used in Queen’s method,
`
`(Ex. 1034 (Queen 1989) at 3), and Queen 1990 disclosed using “known structures”
`
`of the mAbs in the PDB as rough models that could be used to help construct the
`
`antibody models used in Queen’s method. (Ex. 1050 (Queen 1990) at 16, 14:32-
`
`36.)
`
`C. Dr. Wilson Ignores the Impact of Interspecies Homology
`
`8.
`
`Dr. Wilson suggests that my analysis is flawed because the references
`
`on which I rely disclose additional potential residue back mutations that I did not
`
`specifically address and asserts that a person of ordinary skill in the art would not
`
`have been able to select the relevant back-mutations for a given humanized
`
`antibody from this larger group. (Ex. 2041 (Wilson Decl.) at ¶¶ 227-34.) I
`
`disagree.
`
`9.
`
`As the ’213 patent acknowledges, while the number of potential back
`
`mutations in any humanization project can be relatively large, the process of
`
`identifying which residues from this set of potential back mutations are more likely
`
`than others to improve the binding of a given humanized antibody would have
`
`
`
`
`4
`
`
`
`5 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`been routine to a person of ordinary skill in the art in 1990. For example, a person
`
`of ordinary skill would have known that in the case of most framework residues,
`
`simply a shorter spatial distance to the CDRs suggests a higher likelihood of a
`
`positive effect of its back mutation on antigen affinity. Furthermore, the degree of
`
`physical differences in size, charge and/or shape between residues, which are
`
`candidates for back mutations, makes an effect of a back mutation more or less
`
`likely. Also, Dr. Wilson’s criticism ignores that there is a high degree of
`
`interspecies homology for antibodies. (Ex. 1050 (Queen 1990) at 5:22-25 (“The
`
`human immunoglobulin framework sequence will typically have about 65 to 70%
`
`homology or more to the donor immunoglobulin framework sequences.”)) That
`
`renders many potentially relevant substitutions irrelevant because of overlap
`
`between the human framework and murine target.
`
`10. Take, for example, the humanized 4D5 antibody disclosed in the ’213
`
`patent. While the prior art would have taught a person of ordinary skill to look at a
`
`number of residues for potential back mutation, many of these residues are the
`
`same in both the consensus sequence constructed based on Kabat 1987 and the
`
`murine 4D5. Therefore, there would be no potential for back mutation and it
`
`would not be considered in a person of ordinary skill’s analysis. (Ex. 1001(the
`
`’213 patent) at Fig 1A & 1B.)
`
`
`
`
`5
`
`
`
`6 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`11. Similarly, Dr. Wilson’s argument that Queen 1989 would have taught
`
`a person of ordinary skill to analyze additional residues ignores the interspecies
`
`homology between the murine anti-Tac antibody and the Eu human framework
`
`used in Queen’s specific project. The majority of the residues listed in claims 12,
`
`42, 60, 65-67, and 71-79 are identical in the Eu and anti-Tac antibodies, so Queen
`
`1989 would not have contemplated back mutation at those positions. (Ex. 1034
`
`(Queen 1989) at 10031.)
`
`12. The fact that the prior art discloses framework residue back mutations
`
`that are not in the ’213 patent shows that the list of back mutations disclosed in the
`
`patent that are likely to improve binding is not complete. For the most part, the
`
`claims of the ’213 patent cover antibodies of any specificity and a wide range of
`
`structures. A person of ordinary skill would have understood that residues
`
`disclosed in the Queen and Kurrle references improved the binding in the specific
`
`antibodies that were described in those papers, and that residues described in
`
`Chothia and Lesk, Tramontano, and Chothia 1989 have a potential effect on CDR
`
`structures and these residues fit the selection criteria disclosed in the ’213 patent.
`
`When following the guidance in the patent to humanize an antibody, a person of
`
`ordinary skill would not have limited back mutations to those described in the
`
`patent, but would have considered any back mutation that appeared likely to
`
`increase binding. The ’213 patent does not provide any reason why any of the
`
`
`
`
`6
`
`
`
`7 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`residue back mutations it lists are more likely that any other to improve binding
`
`affinity for humanized antibodies generally, or any specific antibody beyond the
`
`handful of examples in the patent.
`
`D. Queen 1990 Discloses the Use of a Human Consensus Sequence
`Framework
`
`13. Contrary to Dr. Wilson’s opinion, Queen 1990 discloses the use of
`
`human consensus sequences as the human framework.
`
`14. Queen 1990 lays out specific Criteria for humanizing antibodies. (Ex.
`
`1050 (Queen 1990) at 12:17-14:31.) Criterion I provides two options for the
`
`starting framework: “[a]s acceptor, use a framework from a particular human
`
`immunoglobulin that is unusually homologous to the donor immunoglobulin to be
`
`humanized, or use a consensus framework from many human antibodies.” (Ex.
`
`1050 (Queen 1990) at 12:17-21 (emphasis added).) The concept of a “consensus”
`
`sequence was well understood in the field and would have been readily understood
`
`by a person of ordinary skill in the art. Based on Criterion I, a person of ordinary
`
`skill in the art would have been motivated to create a consensus sequence to use as
`
`an antibody framework. Queen 1990 states that the consensus is from “many
`
`human antibodies” so a person of ordinary skill in the art looking to make such a
`
`consensus sequence would have been motivated to use all available data regarding
`
`sequences within a particular subclass or subunit because he or she would know
`
`that additional data would improve the “consensus” nature of the sequence.
`
`
`
`
`7
`
`
`
`8 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`15. Dr. Wilson also takes portions of Queen 1990 out of context in an
`
`attempt to strength his argument. (Ex. 2041 (Wilson Decl. for IPR2017-01373) at
`
`¶ 208, citing Ex. 1050 (Queen 1990) at 13:3-11.) The quoted portion of Queen
`
`1990 does not refer to the use of a consensus sequence, but to the use of the human
`
`antibody that is most homologous to the murine donor, which is an alternative
`
`framework that Queen 1990 describes. (Ex. 1050 (Queen 1990) at 13:3-11.) In
`
`full, that section of Queen 1990 states: “Typically, one of the 3-5 most homologous
`
`heavy chain variable region sequences in a representative collection of at least
`
`about 10 to 20 distinct human heavy chains will be chosen as acceptor to provide
`
`the heavy chain framework, and similarly for the light chain. Preferably, one of the
`
`1-3 most homologous variable regions will be used. The selected acceptor
`
`immunoglobulin chain will most preferably have at least about 65% homology in
`
`the framework region to the donor immunoglobulin.” (Ex. 1050 (Queen 1990) at
`
`13:3-11.) When taken in context, it is unambiguous that this statement refers to the
`
`portion of Criterion I regarding the most homologous framework, not the
`
`consensus sequence.
`
`16. Dr. Wilson also argues that the discussion about rare residues in
`
`Criterion II somehow impacts the consensus sequence in Criterion I. (Ex. 2041
`
`(Wilson Decl. for IPR2017-01373) at ¶ 209.) Dr. Wilson does not account for the
`
`fact that Queen 1990 explicitly states that the Criteria may be used singly. (Ex.
`
`
`
`
`8
`
`
`
`9 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`1050 (Queen 1990) at 12:12-15 (“These criteria may be used singly, or when
`
`necessary in combination, to achieve the desired affinity or other
`
`characteristics.”).) A person of ordinary skill would not have understood the
`
`language in Queen 1990’s Criterion II as applying to or limiting the consensus
`
`sequence in Criterion I.
`
`E. Queen 1989, in Combination with Kabat 1987, Teaches a Human
`Framework Consensus Sequence
`
`17. Contrary to Dr. Wilson’s opinion, Queen 1989, in combination of
`
`Kabat 1987, discloses the use of human consensus sequences as the human
`
`framework.
`
`18. Queen 1989 teaches humanizing an antibody by replacing “unusual
`
`residues” with “residue[s] much more typical of human sequences . . . to make the
`
`antibody more generically human.” (Ex. 1034 (Queen 1989) at 10032.) A person
`
`of ordinary skill would have understood this to mean that the framework should be
`
`modified to make it more like a human consensus sequence. A person of ordinary
`
`skill in the art, who was motivated to make a framework that was generically
`
`human, would have used the information in Kabat 1987 to determine what residues
`
`to substitute. As a person of ordinary skill in the art made these substitutions, the
`
`antibody would contain consensus sequences.
`
`19. Dr. Wilson asserts that Kabat 1987 would have only been used by a
`
`person of ordinary skill in the art to check whether a particular reference in a
`
`
`
`
`9
`
`
`
`10 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`sequence was correct. (Ex. 2041 (Wilson Decl. for IPR2017-01373) at ¶ 157.)
`
`This is incorrect. Based on my first-hand knowledge, a person of ordinary skill
`
`would have used Kabat 1987 in a number of ways during the humanization
`
`process, including to (1) locate one or more homologous sequences, (2) provide
`
`information regarding transferring the CDRs, and (3) look for unusual residues at
`
`specific positions.
`
`F.
`
`Humanized Antibodies Bound Antigens and It Was a Routine Part of
`the Optimization Process to Test Binding Affinity
`
`20. While the precise binding characteristics of a specific humanized
`
`antibody are not predictable, a person of ordinary skill would have expected that
`
`humanized antibodies would bind antigens. A person of ordinary skill would have
`
`also expected that humanized antibodies that had back mutations in the framework
`
`region identified using the methods in the Queen references and elsewhere would
`
`be more likely to bind antigens with higher affinities, and that the precise back
`
`mutations required to optimize binding would vary from antibody to antibody. An
`
`example of this increase in affinity was disclosed in Kurrle. (Ex. 1071 (Kurrle) at
`
`9:26-31.) As exemplified in Figures 5 and 6 and Table 3 of the ’213 patent, a
`
`person of ordinary skill in the art would have undertaken mutagenesis to optimize
`
`binding, which would have included making a series of antibodies with different
`
`back mutations and testing them to see which mutations led to higher binding
`
`affinity. Optimizing the binding of an antibody in this way was a normal and
`
`
`
`
`10
`
`
`
`11 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`routine part of the humanization process, and would have been well within the
`
`skills of a person of ordinary skill in the art.
`
`G. A Person of Ordinary Skill in the Art Would Have Known that
`Kurrle’s BMA-EUCIV4 Would Bind an Antigen
`
`21. Dr. Wilson argues that the lack of binding data in Kurrle for BMA-
`
`EUCIV4 suggests that a person of ordinary skill in the art would not have known
`
`that a humanized antibody would bind an antigen. (Ex. 2041 (Wilson Decl. for
`
`IPR2017-01374) at ¶ 162-64.) In doing so, Dr. Wilson ignores the humanization
`
`process that Kurrle pursued and misstates the conclusions. In humanizing the
`
`murine BMA 031 antibody, Kurrle first synthesized BMA-EUCIV1, which
`
`incorporated only the murine CDRs into the human framework, with no back
`
`mutations. (Ex. 1071 (Kurrle) at 9:25-26.) Unsurprisingly, without any additional
`
`back mutations to help maintain the conformation of the CDRs, BMA-EUCIV1 did
`
`not bind well to the target antigen. (Ex. 1071 (Kurrle) at 9:25-26.) Kurrle then
`
`created a series of other antibodies (BMA-EUCIV2, BMA-EUCIV3, and BMA-
`
`EUCIV4), with murine CDRs and varying degrees of back mutations. (Ex. 1071
`
`(Kurrle) at 9:26-31.)
`
`22. Kurrle then tested the binding affinity for BMA-EUCIV2 and BMA-
`
`EUCIV3. (Ex. 1071 (Kurrle) at 9:26-31.) Despite Dr. Wilson’s assertions to the
`
`contrary, Kurrle states that BMA-EUCIV2 did “not bind well,” which suggests not
`
`that it did not bind at all, but rather that the binding was weak. Further, Kurrle
`
`
`
`
`11
`
`
`
`12 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`states BMA-EUCIV3 “does bind well.” (Ex. 1071 (Kurrle) at 9:26-31.) A person
`
`of ordinary skill in the art would take this information to mean that humanized
`
`antibodies with some degree of back mutations would bind the target antigen;
`
`therefore, a person of ordinary skill in the art would have a reasonable expectation
`
`that BMA-EUCIV4 would bind an antigen to some degree.
`
`H. A Person of Ordinary Skill in the Art Would Have Used a 3.3Å Cutoff
`
`23. As I described in my first declaration, a person of ordinary skill in the
`
`art, in addition to looking at the residues that were near the CDRs in the primary
`
`amino acid sequence, would have also looked at residues that were near the CDRs
`
`in three dimensional space. A person of ordinary skill in the art would have known
`
`that amino acids with atoms within about 3.3 Å of each other would be considered
`
`to be in contact. A person of ordinary skill in the art would have used 3.3 Å as a
`
`guide because it is approximately twice the interatomic distance for most protein
`
`atoms. (Ex. 1145 (Pauling) at 261.)
`
`I.
`
`A Person of Ordinary Skill in the Art Would Have Known that a
`Humanized Antibody Would Likely Be Less Immunogenic Than the
`Murine Antibody
`
`24. As I discussed in my first declaration, the entire reason that scientists
`
`embarked on the humanization process was to develop antibodies that had binding
`
`similar to murine antibodies, without generating the strong immunogenic response
`
`that murine antibodies did. A person of ordinary skill would have expected a more
`
`
`
`
`12
`
`
`
`13 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`generically human antibody to be less immunogenic than a murine antibody
`
`because the human body is less likely to treat the antibody as foreign. This
`
`expectation was well documented in many of the prior art references I cite. (Ex.
`
`1034 (Queen 1989) at 10032; Ex. 1050 (Queen 1990) at 6:21-29; Ex. 1071 (Kurrle)
`
`at 33-5.)
`
`J.
`
`The Prior Art Taught Substitutions at 71HI 73HI 78Ha and 93H
`
`25.
`
`To the extent that the patent identifies 71H, 73H, 78H, and 93H as
`
`back mutations to include in a humanized antibody, the prior art identifies these
`
`residues at those positions, as discussed in my first declaration. The following
`
`chart summarized those substitutions.
`
`0 Ex. 1049 (Chothia 1989) at 879; Ex. 1003 at 1] 143
`
`0 Ex. 1051 (Tramontano) at 175; Ex. 1003 at1] 132
`
`. Ex. 1063 (Chothia 1985) at660; Ex. 1003 at1[138
`
`0 Ex. 1071 Kurrle at 25; Ex. 1003 at
`
`| 113
`
`. Ex. 1071 Kurrle at25;Ex.1003 at
`
`l 113
`
`. Ex. 1063 Chothia 1985 at 660; Ex 1003 at
`
`I 104
`
`K.
`
`The Data in the Patent Are Insufficient to Show that Back Mutations
`
`at 71H, 73H, 78H, and 93H Are Better than Any Others
`
`26.
`
`The data put forth in the patent does not show that the back mutations
`
`at 71H, 73H, 78H, and 93H provide any better or unexpected results as compared
`
`to other back mutations. The claims that describe antibodies with these back
`
`mutations, for example claim 79, cover any antibodies with these back mutations,
`
`13
`
`14 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`
`|PR2017—01374
`
`
`
`and do not preclude other back mutations. The mutations will not produce
`
`increased binding in every humanized antibody, but only those that happen to have
`
`structures in which those four residues are important to maintaining the murine
`
`conformations of the CDRs or interact with the antigen. While Table 3 in the
`
`patent provides information regarding a limited number of substitutions, there is no
`
`evidence that those substitutions are any better or more important to the binding
`
`affinity of humanized antibodies generally than any of the other potential
`
`substitutions identified in the patent or prior art. In fact, according to the ’213
`
`patent, substitution at 71H had no impact on the antiproliferative activity of the
`
`antibody, and substitution at 73H, 78H, and 93H resulted in no significant
`
`improvement in binding affinity of the antibody. (Ex. 1001 (’213 Patent), Table 3;
`
`
`
`27. The identification of any particular substitution, including the
`
`identification of substitutions at 71H, 73H, 78H, and 93H, as being important to
`
`the binding of a particular antibody would, however, involve nothing more than
`
`routine computer modeling and mutagenesis as described in the prior art and the
`
`patent itself. Ex. 1001 (’213 patent) at 10:28-34 (“Since it is not entirely possible
`
`to predict in advance what the exact impact of a given substitution will be it may
`
`be necessary to make the substitution and assay the candidate antibody for the
`
`desired characteristic. These steps, however, are per se routine and well within the
`
`
`
`
`14
`
`
`
`15 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`ordinary skill of the art.”);
`
`
`
`L. A Person of Ordinary Skill in the Art Would Have Been Motivated to
`Humanize 4D5
`
`28. Based on the work of Hudziak, a person of ordinary skill in the art
`
`would have known that the murine antibody 4D5 was shown to downregulate
`
`p185HER2 and reduce the proliferation of cancer cells. (Ex. 1021 (Hudziak), 1169.)
`
`The skilled person would have understood that in order for 4D5 to be an effective
`
`therapeutic for humans, it would need to be humanized. Therefore, a person of
`
`ordinary skill in the art would have been motivated to employ the teachings of the
`
`Queen references or Kurrle to humanize the murine 4D5 antibody.
`
`M. U.S. Patent No. 5,821,337 Discloses Herceptin
`
`29.
`
`I have reviewed U.S. Patent No. 5,821,337 (“Carter 1998”). (Ex.
`
`1144.) The face of Carter 1998 states that is a continuation in part to
`
`PCT/US92/05126, which is the PCT application from which the ’213 eventually
`
`issued. This patent was issued to Paul J. Carter and Leonard G. Presta, the named
`
`inventors of the ’213 patent, on October 13, 1998. It was assigned to Genentech,
`
`Inc. I understand that it expired in 2015. Claim 15 of Carter 1998 discloses the
`
`light and heavy chain sequences for Herceptin, which differ at positions 55L and
`
`102H from the sequences of HU4D5 that are disclosed in Figures 1A and 1B of
`
`both the ’213 patent and Carter 1998.
`
`
`
`
`15
`
`
`
`16 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`N. CAMPATH-1 Used a Consensus Sequence
`
`30. As discussed in my first declaration, I engineered CAMPATH-1, the
`
`first fully humanized antibody. As noted in my paper on the subject, I used as the
`
`starting point for the construction of the humanized CAMPATH-1 antibody a light
`
`chain that was provided by my colleague, Dr. Jefferson Foote. (Ex. 1069
`
`(Riechmann) at 324). Dr. Foote had engineered this light chain to build a
`
`humanized anti-lysozyme antibody. The framework regions of this light chain
`
`were designed to match a consensus light chain of the human kappa subgroup I and
`
`were near identical to the light chain of the human antibody light chain dimer REI,
`
`for which there was a three dimensional structure available in the PDB database at
`
`the time. (Ex. 1193 (Foote) at 106.) As such, CAMPATH-1, which I finished in
`
`1988, was a fully humanized antibody that bound antigen, which included a
`
`consensus sequence.
`
`O.
`
`The ’272 Application Does Not Disclose the Subject Matter of Claims
`12, 42, 60, 65, 71, 73-74, and 79 of the ’213 Patent
`
`31.
`
`I understand that U.S. Patent Application 07/715,272 (“the ’272
`
`Application”), filed on June 14, 1991, was the first application that was filed in the
`
`chain of applications that led to the ’213 patent. (Ex. 2037.) I was asked to
`
`consider whether the information included in the ’272 Application would allow a
`
`person of ordinary skill in the art in 1989/90 to practice claims 12, 42, 60, 65, 71,
`
`73-74, and 79 of the ’213 patent.
`
`
`
`
`16
`
`
`
`17 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`32. The only examples included in the ’272 Application are the variants
`
`made in the humanization of 4D5. Ex. 2037, 81-99. To the extent Queen 1989
`
`and the PDB database does not render obvious the claims of the ’213 patent, a
`
`person of ordinary skill in the art would not have understood that the residue
`
`substitutions in the ’213 patent would work to improve binding in other specific
`
`antibody humanization projects from just the data included in the ’272 Application
`
`regarding the humanization of the 4D5 antibody. Therefore, a person of ordinary
`
`skill would not have understood that the inventors of ’213 patent had disclosed
`
`residue substitutions that would work in any humanization project beyond what
`
`was disclosed in the prior art.
`
`P.
`
`Herceptin, Perjeta, Avastin, Lucentis, and Xolair Have Residue
`Substitutions Not Disclosed in the ’213 Patent
`
`33.
`
`I have analyzed the sequence provided for Herceptin, which I
`
`understand to be variant HU4D5-8 (see Table 3 in the ’213 patent) of the sequence
`
`HU4D5 provided in Figures 1A and 1B of the ’213 patent with residue 55L (E,
`
`glutamate) in Figure 1A substituted by Y (tyrosine) and residue 102H (V, valine)
`
`in Figure 1B substituted by Y (tyrosine). Based
`
`
`
`my own analysis, the human sequence was derived from the information in
`
`Kabat 1987.
`
`
`
`34. Several editions of Kabat were published, including in 1983, 1987,
`
`and 1991. For human heavy chain subgroup 3, Kabat 1987 contained 31
`
`
`
`
`17
`
`
`
`18 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`sequences, whereas Kabat 1991 contained 84 sequences. This additional
`
`information impacted the identification of the most common residues at certain
`
`positions in the sequence. For example, as Dr. Wilson testified, residue 73 in the
`
`human heavy chain subgroup 3 was an aspartic acid in Kabat 1987, but was
`
`asparagine based on the information Kabat 1991. (Ex. 1138 (Wilson Depo.
`
`Transcript) at 213:10-217:22.) As this testimony indicates, and my own analysis
`
`confirms, Herceptin does not use a consensus sequence of “all human
`
`immunoglobulins of any particular subclass or subunit structure” as required by the
`
`claims. Kabat 1991 indicates that the most common residue in “all human
`
`immunoglobulins” at position 73H is not aspartic acid.
`
`35.
`
`
`
`
`
`As discussed above, that
`
`sequence was based on Kabat 1987, and, for the same reasons, does not have a
`
`consensus sequence as defined in the ’213 patent.
`
`Q.
`
`36.
`
`The Opinions Set Forth in My Declarations Are My Own
`
`I understand that my opinions have been criticized because I relied, in
`
`part, on the Expert Declaration submitted by Dr. Eduardo Padlan from Case Nos.
`
`IPR2016-01693 and IPR2016-01694. My reliance on Dr. Padlan’s work was in no
`
`way a blind adoption of his work. The opinions set forth in my first declaration
`
`
`
`
`18
`
`
`
`19 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`and in this declaration are my own. Prior to reviewing Dr. Padlan’s declaration, I
`
`reviewed the ’213 patent and the key prior art references cited in my declaration. I
`
`was familiar with many of these references and with the state of the art generally
`
`from my own work on humanizing antibodies both before and after 1989. Based
`
`on my review of these references, I concluded that the references would have
`
`taught a skilled person how to humanize an antibody, including how to identify the
`
`framework residues that would need to be back-mutated using computer modeling
`
`and serial mutagenesis. Only after I had come to that determination was I provided
`
`with the Padlan declaration, which I reviewed in detail, along with the references it
`
`cites. Dr. Padlan’s opinions matched my own and explained the process of
`
`humanization as a person of ordinary skill in the art would have understood it. I
`
`chose for efficiency and expediency to re-use Dr. Padlan’s language where I
`
`agreed with it rather than needlessly re-casting the same opinion in different words.
`
` Conclusion
`III.
`37. For all of the reasons discussed above and stated in my first
`
`declaration, it is my opinion that the challenged claims of the ’213 patent are
`
`anticipated by and/or obvious in view of the prior art.
`
` hereby declare that all statements made herein of my own knowledge are true and
`
` I
`
`that all statements made on information and belief are believed to be true. I further
`
`
`
`
`19
`
`
`
`20 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`declare that all of my statements are made with the knowledge that willful false
`
`statements are punishable by fine or imprisonment, or both, under Section 1001 of
`
`Title 18 of the United States Code.
`
`
`
`Dated: May 25, 2018
`
`
`
`By:
` Lutz Riechmann, Ph.D.
`
`
`
`
`
`
`
`20
`
`
`
`21 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`
`
`Addendum
`
`Additional Materials Considered
`
`
`Exhibit/Paper No. Description
`
`38
`
`38
`
`1138
`
`1140
`
`1142
`
`1144
`
`1145
`
`1193
`
`1194
`
`2037
`
`2041
`
`2041
`
`IPR2017-01373 Patent Owner’s Response
`
`IPR2017-01374 Patent Owner’s Response
`
`Deposition Transcript of Ian A. Wilson (Apr. 21, 2018)
`
`Deposition Transcript of Paul J. Carter (Apr. 27, 2018)
`PROTECTIVE ORDER MATERIAL
`
`Deposition Transcript of Leonard George Presta (May 1,
`2018)
`PROTECTIVE ORDER MATERIAL
`
`U.S. Patent 5,821,337
`
`Pauling, The Nature of Chemical Bond and the Structure of
`Molecules and Crystals: An Introduction to Modern Structure
`Chemistry, Cornell University Press (3rd Ed. 1960).
`
`Foote, Humanized Antibodies, 61 Nova Acta Leopoldine,
`269, 103-110 (1989) [Carter Deposition Ex. 1193].
`
`Kolbinger et al., Humanization of a Mouse Anti-Human IgE
`Antibody: A Potential Therapeutic for IgE-Mediated
`Allergies, 6 Protein Engineering, 8, 971-980 (1993) [Wilson
`Deposition Ex. 1194].
`
`U.S. Patent Application 07/715,272
`
`IPR2017-01373 Expert Declaration of Dr. Ian A. Wilson
`
`IPR2017-01374 Expert Declaration of Dr. Ian A. Wilson
`
`
`
`
`
`
`22 of 22
`
`Celltrion, Inc. 1143
`Celltrion v. Genentech
`IPR2017-01374
`
`