0022-1767/91/14712-4366$02.00/0
`THE JOURNAL OF IMMUNOLOGY
`Copyright 0 1991 by The American Association of Immunologists
`
`VOl. 147. 4366-4373, No. 12, December 15, 1991
`Prlnted In U.S.A.
`
`CONSTRUCTION, EXPRESSION AND CHARACTERIZATION OF HUMANIZED
`ANTIBODIES DIRECTED AGAINST THE HUMAN alp T CELL RECEPTOR
`
`CLYDE W. SHEARMAN,'* DAN POLLOCK,* GARY WHITE,* KATHY HEHIR,"
`GORDON P. MOORE,2" E. J. KANZY,+AND ROLAND KURRLE'
`From the %enzyme Corporation, Framingham, MA 01 701; and 'Behringwerke Aktiengesellschaft, 0-3550
`Marburg, West Germany
`
`Completely humanized antibodies with specificity
`tive mAb therapy may require frequent multiple treat-
`for the human alp TCR have been produced by ge-
`ments with large amounts of murine antibody. Second,
`netic engineering. The L and H chain V region exons
`administration of murine IgG elicits a brisk HAMA3 re-
`encoding the murine mAb BMA 031 CD regions and
`
`sponse that that can further reduce the circulating half-
`human EU framework regions were synthesized and
`life of the mAb and produce allergic reactions including
`replaced into previously isolated genomic frag-
`anaphylaxis (8-10).
`ments. These fragments were inserted into mam-
`Almost all of the murine mAb currently being used
`malian expression vectors containing the human K
`clinically provoke HAMA responses in patients. These
`and y 1 C region exons. Two variants were con-
`include HAMA against both the C region and the V region
`structed each containing selected BMA 031 amino
`(1 1). HAMA responses lead to altered pharmacokinetics
`acids within the human frameworks. The human-
`of the injected mAb. The antibody is rapidly cleared from
`ized genes were transfected into Sp2/0 hybridoma
`the serum and reduced antibody levels are attained (1 2).
`cells by electroporation and transfectomas secret-
`Although severe side effects are rare in patients with
`ing humanized antibody were
`isolated. Levels of
`HAMA after retreatment with antibody, it is clear that if
`antibody expression up to 7 pg/ce11/24 h were ob-
`mAb are to be used therapeutically, reliable methods
`tained. The humanized antibody, BMA 031-EUCIV2,
`must be devised to reduce immune mediated complica-
`competed poorly with murine BMA 031 for binding
`tions or adverse reactions (13).
`to T cells. BMA 031-EUCIV3, however, bound specif-
`One approach to better immunotherapies currently
`ically to T cells and competed effectively with both
`being explored is to produce a truly human antibody.
`the murine BMA 031 antibody and a previously
`constructed chimeric BMA 031 antibody for binding
`Unfortunately, human mAb technology has lagged far
`to these cells. The relative affinity of BMA 031-
`behind that of murine-based monoclonal technology. Hu-
`EUCIV3 was about 2.5 times lower than BMA 031.
`man hybridomas are difficult to prepare, are often unsta-
`The ability to promote antibody dependent cell-me-
`ble, and secrete antibody at low levels (14, 15). The mAb
`diated cytolysis was significantly enhanced with the
`generated are usually of the IgM class and of low affinity.
`engineered antibodies as compared to murine BMA
`An attractive and viable strategy is to produce "human-
`031. Humanized BMA 031 is a clinically relevant,
`ized" versions of murine mAb through genetic engineer-
`
`genetically engineered antibody with potential uses
`ing. Methods have been devised to replace all regions of
`in transplantation, graft vs host disease, and auto-
`a murine antibody with analogous human regions (1 6-
`immunity.
`18). Chimeric antibody technology has been applied to
`
`several therapeutically important antibodies (1 9-24) and
`has been useful in class switching and the production of
`mAb are emerging as a major modality for therapy of
`isotypes with specific effector functions (25, 26). A chi-
`various pathologic conditions including malignant dis-
`meric antibody composed of the V regions of murine mAb
`ease, cardiovascular disease, and autoimmune diseases.
`17 1 A and the human y 1 C region has recently been used
`Some of these have demonstrated efficacy in treating
`in patients with colon cancer. Whereas murine 171A has
`colon carcinoma (1). B cell lymphomas (2). neuroblastoma
`been used extensively in clinical trials and elicits a very
`(3). and in preventing transplant rejection (4, 5).
`pronounced HAMA response that alters its pharmacoki-
`Clinical trials with murine antibodies, although en-
`netics, antibody responses to chimeric 171A have been
`couraging, have indicated at least two fundamental prob-
`dramatically reduced. Moreover, the circulating half-life
`lems of antibody therapy. First, murine IgG has a much
`was increased relative to murine 17 1 A and higher serum
`shorter circulating half-life in man compared to what has
`levels could be maintained at lower infused doses (27).
`been reported for human antibodies (6, 7). so that effec-
`Thus, with judicious genetic engineering, it is possible to
`manipulate antibody pharmacokinetics to minimize toxic
`side effects.
`Chimeric antibodies may be effective in lowering the
`HAMA response in patients and increasing serum half-
`lives, but these properties
`are still inferior to human
`
`Received for publication March 18, 199 1.
`Accepted for publication September 16, 1991.
`The costs of publication of this article were defrayed in part by the
`payment of page charges. This article must therefore be hereby marked
`aduertlsement in accordance with 18 U.S.C. Section 1734 solely to indi-
`cate this fact.
`Address correspondence and reprint requests to Dr. C. W. Shearman.
`Sterling Drug, Inc.. 512 Elmwood Avenue. Sharon Hill. PA 19079.
`Present address Department of Molecular Genetics. SmithKline Bee-
`cham, 709 Swedeland Road. King of Prussia. PA 19406.
`
`Abbreviations used in this paper: HAMA. human anti-murine anti-
`body; ADCC. antibody dependent cell mediated cytolysis; FR. framework
`region.
`4366
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1072 Page 1 of 8
`
`

`
`antibodies. Inasmuch as chimeric antibodies are still 30%
`murine, enhanced efficacy may be obtained by human-
`izing the V regions. New technologies have recently been
`advanced to produce totally humanized antibodies by
`grafting the CDR of murine antibodies onto human FR
`(17, 28-30). The resulting antibodies when expressed
`with human C regions should be essentially human. This
`technology, although technically straightforward, is not
`always totally successful. Selected FR amino acids ap-
`pear to be involved in Ag binding. Identification of im-
`portant FR amino acids has been achieved, up to now.
`
`by the use of x-ray crystallographic data (1 7) and sophis-
`ticated computer modeling (30) and several totally hu-
`manized antibodies have been produced with affinities
`close to those of their parental antibodies (17, 28-30).
`We report here the production of a humanized anti-
`body, without the use of sophisticated structural data,
`which retains the affinity and specificity of BMA 03 1, a
`murine mAb directed against the human (Y/P TCR. More-
`over, humanized BMA 031 displays enhanced ADCC ac-
`tivity. BMA 031 has been used successfully in preventing
`organ transplant rejection (5) and may have potential
`efficacy in other T cell-related disorders.
`
`4367
`competitive immunofluorescence assays were carried out. PBMC
`were separated by Ficoll-Hypaque density gradient centrifugation
`in the dark with mAb at various
`and incubated on ice for 1 h
`concentrations (0.05-50 pg/ml) premixed with either FITC-BMA031
`or FITC-BMAEUCIV3 (2 pg/ml). Unbound antibodies were removed
`by two washing steps. Cells from all experiments were analyzed
`either on an Ortho (Raritan, NJ) Cytofluorograph 50H/2150 Com-
`puter System or on a Becton Dickinson (Mountain View, CA) FACStar
`Plus a s described elsewhere (32). The intensity of fluorescence was
`calculated by modified Ortho or standard FACStar Plus software and
`is expressed as mean channel number.
`Cytotoxicity assays. To measure the cytolytic capacity of the
`BMA 031 antibody preparations, a 20 h [%r] release assay was
`performed to measure ADCC and NK activity. ["Crl-labeled HPB-
`ALL target cells were incubated with (ADCC) or without (NK activity]
`various concentrations of antibodies for 20 h in
`the presence of
`Ficoll-separated PBL (effector cells). a/D TCR negative CEM cells
`were used as control target cells. The antibodies were allowed to
`bind first to target cells (30 min) before the effector cells were added.
`The E:T cell ratio varied from 1: 1 to 50: 1. Cytolysis in the absence
`of antibodies was considered to be due to NK activity. The percentage
`of specific lysis was calculated as described earlier (33). Spontaneous
`[51Cr] release in the absence of effector cells and in the presence of
`the antibodies being tested was always less than 5%. All samples
`were analyzed in triplicate.
`
`HUMANIZED ANTI-T CELL ANTIBODIES
`
`RESULTS
`Designing humanized BMA 031 antibodies. To deter-
`mine the optimal human sequence with which to human-
`ize the murine BMA 031 antibody, the murine BMA 031
`MATERIALS AND METHODS
`amino acid sequence was used to search the NBRF data
`Cell culture. The BMA 031 and Sp2/O-Ag14 hybridomas were
`base for the most homologous human antibody. Inas-
`cultured in DMEM media supplemented with 10% FCS, 2 mM L-
`glutamine. 10 mM HEPES. pH 7.3, 10 mM nonessential amino acids
`much as molecular models of antibodies show strong
`(GIBCO, Gaithersburg, MD). and 10 mM pyruvate. Chimeric and
`interactions between the H and L chains, we decided to
`humanized BMA 031 transfectomas were grown in the above media
`use the H and L chain from the same human antibody.
`containing 1 pg/ml mycophenolic acid, 50 pglml xanthine, and 500
`pg/ml Geneticin (GIBCO). All lines were maintained at 37°C in 7%
`The human EU antibody turned out to be the best overall
`CO*.
`choice. The homology between the BMA 031 and EU FR
`Computer analysis. Sequences were manipulated and homology
`(nos. 1-3) was 79% (67% identical) for the H chain and
`searches were performed with the Genetics Computer Group Se-
`8 1 % (63% identical) for the L chain. The BMA 03 1 anti-
`quence Analysis Software Package (University of Wisconsin Biotech-
`nology Center, Madison, WI)) using the National Biomedical Research
`body uses J H 3 and JK5. These are most homologous to
`Foundation databases.
`human JH4 and JK4. A first generation humanized BMA
`Synthesls of VH
`and VL regions. The VH and VL exons were
`031 antibody would contain BMA 031 CDR, EU FR, and
`synthesized on an Applied Biosystems (Foster City, CA) model 380A
`DNA synthesizer. Each region was synthesized completely as EcoRI-
`homologous human J regions. We refer to this antibody
`HindIII fragments consisting of overlapping (1 0- 15 nucleotide over-
`as BMA 031-EUCIV1 (Fig. 1).
`lap) oligomers (75- 1 10 nucleotides). The oligomers were deprotected
`A refinement to this basic humanized version can be
`and purified by electroelution from polyacrylamide gels. The oligo-
`made in the sequence immediately before and after the
`mers were then mixed in equimolar amounts (30 pmol), phosphory-
`lated. annealed, and ligated into pUC 19 previously digested with
`CDR. The CDR are assigned based on sequence homology
`EcoRI and HindIII.
`data (34). Molecular models of antibodies have shown
`Nucleotide sequencing. DNA sequencing of the synthesized VH
`that the actual CDR loops can contain amino acids up to
`and VL regions was performed directly on pUC subclones using
`universal forward and reverse primers (31).
`five amino acids away from the "Kabat" CDR (36). Also,
`Construction of humanized genes. To ensure efficient expres-
`Reichmann et al. (1 7) have shown the functional impor-
`sion, the synthesized V regions were inserted into previously isolated
`tance of a FR amino acid four residues from a CDR.
`genomic fragments (24) in place of the murine V regions. The re-
`sulting 5.6-kb EcoRl VH fragment was cloned into a mammalian
`Therefore, maintaining at least the major amino acid
`expression vector containing the human y l C region and the gpt
`differences (in size or charge) within four amino acids of
`gene for selection. The 3.0-kb Hind111 VL fragment was cloned into
`the CDR as murine may be beneficial. We refer to the
`a vector containing the human K C region and the neo gene (see Fig.
`antibody containing these changes as BMA 03 1 -EUCIV2
`5).
`Transfection of DNA into Sp2/0 cells by electroporation. DNA
`(Fig. 1). Additionally, all differences within four amino
`was introduced into murine hybridoma Sp2/O-Agl4 cells by electro-
`poration. The 1 to 2 x lo7 actively growing Sp2/O-Ag14 cells were
`acids of the CDR could be maintained murine. We refer
`washed and resuspended in 1 .O ml of sterile PBS. A total of 15 pg of
`to this antibody as BMA 03 1 -EUCIV3.
`each humanized, IgK and IgGl, plasmid (linearized with BamHI) was
`Further refinements can be made, but, without com-
`added to the cell suspension. The DNA/celIs were transferred to a
`plex computer modeling, it is difficult to prioritize their
`ice at least 5 min and then
`precooled shocking cuvette, incubated on
`importance. For example, several amino acids are either
`a 0.5 kv/cm electric pulse was delivered for 10 ms (Transfector 300.
`BTX. San Diego, CAI. After shocking, the DNA/cell mixture was
`BMA 031 specific or EU specific (i.e., different from the
`returned to ice for 10 min and then diluted in 40 ml of supplemented
`consensus sequence within their subgroups). Inasmuch
`DMEM and incubated at room temperature for 10 min. Finally, the
`as these amino acids presumably arose through somatic
`cells were transferred to a 37°C incubator with 7% COz for 48 h
`mutation to enhance their respective activities, it would
`before plating in selective medium, containing 1 pg/ml mycophenolic
`acid, 50 pg/ml xanthine, and 1 mg/ml Geneticin. Cells were plated
`seem logical to maintain the BMA 031-specific amino
`in 96-well plates at 3 x lo4 cells/well.
`acids and change the EU-specific amino acids
`to the
`Cytofluorornetric assay for affinity. To analyze the relative affin-
`human consensus. But this can have potential adverse
`ities of murine, chimeric, and humanized
`BMA 031 antibodies,
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1072 Page 2 of 8
`
`

`
`4368
`
`70
`110
`90
`BMA NE K K A L S K S S
`EU AOKPQGRVTITADESTNTAYnBLSSLRSBDTAPYPCAGG.YGIYSPEEY. .NGGLVTVS:
`T S V E Y R S Y D D G P V W G O T
`S YD DGPV WGQ T
`CIV-1 NE K
`CIV-2 NE K A
`R S YD DGPV WGO T
`CIV-3 NE K KA L
`VEY R S YD DGPV WGO T
`\--CDR-3--/
`""-/
`
`B
`
`A
`A
`
`90
`70
`EU RPIGSGSGTBPTLTISSLOPDDPATYYCOO~SDS~PGQG~~~
`L L
`WS NPLT A
`HEAB A
`BMA
`SYS
`S
`WS NPLT G
`I
`CIV-1
`WS NPLT G
`I
`CIV-2
`WS NPLT G
`I
`CIV-3
`\-CDR-3-/
`Figure 1 . Amino acid sequences of EU, BMA 031, and humanized
`BMA 031 V regions. A. The VH region and E , the VL region. The positions
`of the CDR are indicated.
`
`TABLE I
`Amino acid (AAJ dlfferences between EMA 031 and EU and their
`consensus seauences
`
`
`
`Val
`
`Ile
`
`
`
`Ile
`
`
`
`Leu
`b
`b
`
`b
`
`TYr
`'4%
`
`Leu
`Ser
`LY s
`
`'4%
`
`
`
`Ser Ile
`
`
`
`Ile
`
`Ser
`
`ASP
`Glu
`
`Ser Ser
`Ser
`
`Val
`LYS
`Val
`TY
`
`Glu
`
`HUMANIZED ANTI-T CELL ANTIBODIES
`BMA-031-
`addressed in
`
`981, and these are
`70, 95, and
`
`
`
`
`30
`10
`50
`EU O V O L V O S G A E V K K P G S S ~ V S C K A S G G T F S R S A I I V V R Y
`I y N ~ D V T K EUCIV3. In one position (no. 93) the human consensus
`BHA E 0
`YK TSYVME K K
`P LV
`A H
`SYVME
`sequence is the same as BMA 031. Moreover, the Phe,,
`CIV-1
`YNDVTK
`Y N YNDVTK
`YK TSYVME
`CIV-2
`acid is only found in
`in EU is highly unusual: this amino
`CIV-3
`I Y N YNDVTK
`YKTSYWE K
`this position in one other human antibody in subgroup
`\-----CDR-2
`\CDR/
`1
`VH-111. One could rationalize changing
`this from EU to
`
`the human consensus, so we incorporated this change
`into BMA 031-EUCIV3. For the two remaining positions
`(nos. 72 and 74), there is no clear human consensus so
`
`we maintained the
`
`
`EU seauence. The
`L chain had
`five
`
`EU-specific amino acids. One is within four amino acids
`of the CDR (no. 48) and is maintained as BMA 031 in
`BMA 031-EUCIV3. In two positions (nos. 63 and 81) the
`human consensus is the same as BMA 03 1 and therefore
`could be changed to the human consensus. We decided
`not to make these changes at this time. The other two
`10
`30
`positions (nos. 10 and 70) were also not changed to limit
`50
`EU D I O H T O S P S T L S A S V G D R V T I T C R A S O S I ~ W ~ W Y O O ~ G ~ P K L ~ ~ S S L E S G V P S
`the number of substitutions. There are eight BMA 031
`S TS RWI DT K A
`EK M S TS V.SYME
`BHA 0 VL AIM P
`A
`S TS V.SYHE
`D T K A
`CIV-1
`specific amino acids in the H chain. In two positions (nos.
`R D T K A
`S TS V.SYME
`CIV-2
`7 and 82) the BMA 03 1 sequence is the same as EU. His,,
`RWI DT K A
`H S TS V.SYME
`CIV-3
`\CDR-Z/
`\--CDR-l--/
`is unique to BMA 03 1. This position is considered "invar-
`iant" with Tyrg4 occurring more than 98% of the time.
`Therefore, we decided
`to incorporate this change into
`BMA 031-EUCIV3. The remaining five positions (nos. 1,
`9, 20, 40, and 72) were maintained EU to limit the num-
`ber of changes. There are no BMA 031-specific amino
`acids in the L chain. The sequence is identical to the
`subgroup VI consensus. The changes in the human EU
`framework sequence back to BMA 031 are summarized
`in Table 11. Twelve changes were made in the H chain: 5
`in BMA 03 1 -EUCIV2 and 7 more in BMA 031 -EUCIV3.
`Five changes were introduced into the L chain: two in
`BMA 031 -EUCIV2 and three more in BMA 031-EUCIV3.
`Determination of DNA sequence for humanized V
`regions. The amino acid sequence of the V regions were
`BMA 031 codons
`reverse translated using the actual
`wherever possible and BMA 03 1 codon preferences every-
`where else. To aid in future modifications, unique restric-
`tion enzyme sites were engineered into the sequence at
`approximately 60-bp intervals by making use of the de-
`generacy of the genetic code. Finally, convenient restric-
`tion enzyme sites 5' and 3' of the coding region of BMA
`031 were identified and this flanking sequence was in-
`corporated into the final humanized sequence to be syn-
`
`H chain, EU specific
`70
`Ile
`72
`Ala
`74
`Glu
`Val
`Phe
`93
`95
`
`Phe TYr
`GlY
`98
`
`
`
`
`
`L chain, EU specific
`10
`Thr
`48
`Met
`63
`Ile
`Glu Ser
`70
`
`
`Glu
`81
`ASP
`
`
`
`
`
`H chain, BMA specific
`Gln
`Gln
`1
`7
`Ser
`Ser
`Ala
`Ala
`9
`Ala
`Val
`20
`Val
`Ala
`40
`Ala
`72
`Ala
`b
`Gln
`82
`Glu
`Glu
`94
`TYr
`TYr
`L chain. BMA soeclflc: None
`a Numbers correspond to those in Figure 1
`Variable.
`
`
`
`Glu
`Ser
`Pro
`Met
`LYS
`Ser
`Glu
`His
`
`Gln
`Pro
`
`Leu
`
`Val
`
`TY r
`
`consequences. Changing an amino acid in one chain may
`cause changes in the interactions with other amino acids
`of that chain as well as with amino acids in the other
`chain. Therefore, extreme caution must be exercised to
`limit the number of changes. Table I outlines these po-
`tential changes. The residue numbers correspond to
`those in Figure 1. As can be seen, EU differs from the
`human VH-I subgroup consensus sequence in six posi-
`tions. Three are within four amino acids of the CDR (nos.
`
`TABLE I1
`Amino acid (AAJ changes in EU FR
`crvz
`AA
`
`BMA031
`AA
`
`EU AA
`
`AA Position"
`
`crv3
`AA
`
`H chain
`27
`28
`30
`38
`48
`67
`68
`70
`93
`94
`95
`98
`L chain
`Ile
`21
`Leu
`46
`Leu
`47
`TrP
`Ile
`Met
`48
`Ala
`Ser
`60
`Numbers correspond to those in Figure 1.
`
`Met
`
`Ile
`A%
`Leu
`Met
`Ala
`
`Met
`
`TrP
`Ile
`Ala
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1072 Page 3 of 8
`
`

`
`HUMANIZED ANTI-T
`
`CELL ANTIBODIES
`
`4369
`
`VH fragment was isolated from the pUC19 subclone and
`thesized. The final DNA sequences of BMA 031-EUCIV2
`cloned into pUCBMAVH-1.OHAN. Then, the
`1.0-kb
`VH and VL, excluding the EcoRI and HindIII cloning ends,
`HindIII fragment was isolated and cloned into
`are shown in Fig. 2.
`Synthesis of humanized BMA 031 V regions. The L
`pUCAHBMAVH-5.6RAH. Finally, the 5.6-kb EcoRI frag-
`ment was isolated and subcloned into the mammalian
`and H chain V region exons encoding the humanized
`expression vector containing the human y 1 C region and
`antibodies were synthesized completely as EcoRI-Hind111
`the gpt gene for selection (Fig. 5).
`fragments consisting of 10 to 15 overlapping (10-1 5
`The newly synthesized SauI-HinclI BMA 03 1 -EUCIV2
`nucleotide overlap) oligomers (75- 1 10 nucleotides). The
`VL fragment was isolated and cloned into pUCBMAVL-
`oligomers were phosphorylated, annealed and ligated into
`a pUC vector previously cut with EcoRI and HindIII. The
`1.4RH2. Then, the 1.4-kb EcoRI-HincII fragment was
`isolated and cloned into pUCARSBMAVL-3.0H. Finally,
`assembled fragments were sequenced to verify accuracy
`the 3.0 HindIII fragment was isolated and cloned into the
`of synthesis.
`Reconstruction of BMA 031 genomic fragments with
`mammalian expression vector containing the human K C
`region and the neo gene for selection (Fig. 5).
`humanized V exons. To increase the probability of effi-
`The BMA 03 1 -EUCIV3 constructs were prepared in the
`cient expression of the synthesized coding regions, the
`same manner as BMA 031-EUClV2. Replacement oligo-
`humanized sequences were replaced into the previously
`mers incorporating the coding changes for BMA 031-
`isolated 5.6-kb EcoRI VH and 3.0-kb Hind111 VL genomic
`EUCIV3 were synthesized and cloned into the pUCBMA-
`fragments of BMA 031 (Fig. 3). Due to the lack of unique
`EUCIV2 constructs. The final clone was sequenced to
`restriction enzyme sites, several subclonings were nec-
`ensure accuracy of the coding sequence. The BMA 031-
`essary. To achieve this goal, four vectors, each containing
`EUCIV3 V regions were replaced into the original BMA
`modified genomic subfragments. were constructed. The
`03 1 genomic fragments and these fragments were cloned
`first vector, pUCBMAVH-1 .OHAN was constructed by
`into the mammalian expression vectors described above.
`subcloning the 1 .O-kb HindIII BMA 03 1 VH fragment into
`Expression and purification of humanized BMA 031
`pUC19 with subsequent deletion of the 5'-NsfI site. The
`antibodies. The humanized genes were transfected into
`second vector, pUCAHBMAVH-5.6RAH, was derived by
`Sp2/0 hybridoma cells by electroporation and selected in
`cloning the 5.6-kb EcoRI BMA 031 VH fragment into a
`pUCl9 vector with a previously deleted Hind111 site. The
`media containing both mycophenolic acid and Geneticin.
`Transfectomas secreting humanized BMA 03 1 antibodies
`5'-HindIII site of the insert was then deleted to complete
`were identified by ELISA. Secretion levels up to 7 pg/cell/
`the construction. The third vector, pUCBMAVL- 1.4RH2,
`24 h were obtained. The best clone from each transfec-
`was constructed by subcloning the 1.4-kb EcoRI-HincII
`tion (CIV2 and CIV3), with respect to secretion level and
`BMA 031 VL fragment into pUC19. The fourth vector,
`growth characteristics, was expanded for further study.
`pUCARSBMAVL-3.0H. was made by cloning the 3.0-kb
`HindIII BMA 031 VL fragment into a pUCl9 vector that
`The BMA 031 -EUCIV2 and -EUCIV3 antibodies were
`partially purified by protein A-Sepharose column chro-
`had a previous deletion from the EcoRI site to the Sal1
`matography. Analysis of the antibodies by reducing and
`site in the polylinker.
`nonreducing SDS-PAGE showed a high degree of purity
`replace the humanized se-
`The cloning scheme to
`(data not shown). Analysis by a series of ELISA assays
`quences into the genomic fragments is outlined in Figure
`4. The newly synthesized S&I-NsiI
`BMA 03 1 -EUCIV2
`showed that the antibodies contained human K and y 1
`C
`
`FLgure 2. DNA sequences of the V re-
`gions of BMA 031-EUCIV2. A, The BMA
`031-EUCIV2 VH region and B. the BMA
`03 1 -EUCIV2 VL region.
`
`A
`
`1
`
`S."
`
`I
`
`B
`
`6 0
`
`1
`
`61 .............................+...................*.....
`.."+
`120
`T T G A T C ~ L I C * G A C c 1 f f i T ~ C C M T T ~ ~ A C C C T ~ A R W C T T C C A M f f i A ~
`L V O S G A E V Y K P C S S V K V S C K
`Bspn IT
`Dr. 111
`
`181
`
`121
`
`181
`
`241
`
`301
`
`361
`
`421
`
`................. I
`
`N s i I
`I
`..... ....+.........*......... 449
`CCTMCTTCTCCCAITCTAMTGUTGTT
`GCATIGMCACGGT-TlTACACAA
`
`241
`
`180
`
`301
`
`240
`
`3w
`
`360
`
`420
`
`361
`
`421
`
`681
`
`541
`
`-1
`
`601
`
`AMTGGAGG*U%CTCAITATCARTGAC
`..... ....+.........*...... ... 629
`TlTACCTCCTCCCGAGTMTAGTCM~
`
`180
`
`240
`
`3M)
`
`360
`
`420
`
`4EO
`
`540
`
`6W
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1072 Page 4 of 8
`
`

`
`E
`
`m s N E SS
`I I II
`
`P
`
`I
`
`R
`
`I
`
`
`
`
`
`1.0 kb
`
`HUMANIZED ANTI-T CELL ANTIBODIES
`says. The data shown in Figure 6 indicate that both the
`murine BMA 031 antibody and the previously con-
`structed chimeric BMA 031-G1 antibody block the bind-
`ing of BMA 031-FITC in the same dose-dependent man-
`ner. BMA 031-EUCIV3 was about 2.5 times less efficient
`than murine BMA 031. BMA 031-EUCIV2 was unable to
`totally block BMA 03 1 -FITC binding, even at concentra-
`tions as high as 50 pg/ml.
`BMA 03 1 has been shown to be poor at mediating ADCC
`using human effector cells. To evaluate the ADCC capac-
`ity of the humanized antibodies, we compared them to
`rabbit anti-GH-1 antiserum. This antiserum was the best
`of eight rabbit anti-human T cell globulins in ADCC
`capacity. As shown in Figure 7, both the chimeric BMA
`031 antibody and the BMA 031-EUCIV3 antibody were
`very efficient at ADCC. Even at very low effector:target
`cell ratios (Fig. 7A) or extremely low antibody concentra-
`tions (Fig. 7, B and C ) , the engineered antibodies are
`highly potent at mediating killing of the HPB-ALL cells.
`
`4370
`
`A
`
`R
`
`B
`
`H
`
`1
`
`
`
`R
`I
`
`S P
`
`H 2 P
`
`I
`
`S N H
`
`I
`
`1.0 kb
`
`Ftgure 3. Partial restriction enzyme maps of BMA 031 V regions. A.
`The 5.6-kb EcoRI VH fragment containing the VDJJ exon. B, the 3.0-kb
`Hind111 VL fragment containing the VJs exons. H. HtndIII: HZ, Hincll; N.
`Nsll: P , PstI; R. EcoRI; S , SauI.
`
`regions. Moreover, the antibodies did not react with anti-
`murine antibodies (data not shown).
`Characterization of humanized BMA 031 antibodies.
`The BMA 03 1 -EUCIV2 antibody bound poorly to T cells.
`In contrast, BMA 031-EUCIV3 shows an identical speci-
`ficity as murine BMA 03 1. They both bind specifically to
`T cells and show no reactivity toward monocytes, E, or
`granulocytes (data not shown).
`The relative affinities of murine BMA 031, chimeric
`BMA 031 (human IgGl), and the humanized variants
`were compared by competitive immunofluorescence as-
`
`DISCUSSION
`We have joined the DNA segments containing the CDR
`from the BMA 031 mAb specific for the a l p TCR and the
`FR from the human EU antibody to the DNA segments
`encoding human 7-1 and K C regions. When the human-
`ized genes were introduced into non-Ig producing Sp2/0
`
`cells, functional humanized antibodies specific for T cells
`were assembled and secreted.
`Functional antibody, however, was dependent on sub-
`stitution of various murine FR amino acids into the hu-
`man FR. The identification of important FR amino acids
`in the absence of structural data or computer models is
`difficult but, by careful analysis of antibody sequence
`homologies, it is possible to generate a humanized se-
`quence with a high probability of maintaining Ag bind-
`ing. Our method consists of three parts. First, and pos-
`sibly most important, is starting with the human anti-
`body most homologous to the murine antibody under
`
`F t g u r e 4 . The cloning scheme to re-
`generate the BMA 0 3 1 genomic fragments
`with the humanized V regions. A, Substi-
`tuting the humanized VH region into the
`5.6-kb EcoRI VH fragment. BH, BMA 03 1
`VH exon: CH. humanized BMA 031 VH
`exon: p l , pUCBMAVH-I.OHAN: p2.
`p 3 .
`pUCBMACIVH:
`pUCBMACIVH-
`1.OHAN; p4. pUCAHBMAVH-5.6RAH; p 5 .
`pUCAHBMACIVH-5.6RAH: p6. pSV2gpt-
`hur 1: p7, pSVZgpt-BMACIVH-huyl. E.
`Substituting the humanized VL region
`into the 3.0 HlndIlI VL fragment. BL. BMA
`031 VL exon: CL. humanized BMA 031
`VL exon: p8. pUCBMAVL-1.4RH2: p9.
`pUCBMACIVL:
`p l 0 . PUCBMACIVL-
`1.4RH2: p l 1 . pUCARSBMAVL-3.0H; p12.
`pUCARSBMACIVL-3.0H; p13. pSV2neo-
`huK: p14. pSV2neo-BMACIVL-hux. Re-
`striction enzyme sites identified are: H .
`Hindlll; HZ. HtncII: N, NsiI: R, EcoRI: S.
`Saul.
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1072 Page 5 of 8
`
`

`
`HUMANIZED ANTI-T CELL ANTIBODIES
`
`100 7
`
`A
`
`R
`
`B
`
`R
`
`BUA 031-EUCIVB
`
`EWAN G A M A 1
`
`pSV2-gpt
`
`-
`
`c
`e
`0
`a
`0
`
`1
`
`B
`
`B
`
`H
`
`A
`
`R
`
`I
`
`B
`
`H
`
`437 1
`
`._.._.._....".. ...
`_"""
`
`."
`..... -
`
`anti-GH 1
`BMA 031
`BMA 031-G1
`BMA-EUCIV3
`NK-activity
`
`* "*-
`... " ....
`" 0 \
`*
`
`pSV2-neo
`B W A N KAPPA
`BEVL 031-EUCIVL
`Ffgure 5. Expression vectors for humanized BMA 031 V regions. A,
`The H chain expression vector containing the humanized BMA 031 VH
`region, the human y 1 C region, and the guanine phosphoribosyl trans-
`ferase gene for selection. B, The L chain expression vector containing the
`humanized BMA 031 VL region, the human K C region and the neomycin
`resistance gene for selection.
`600 -
`500 -
`400 -
`300 -
`200 -
`100 -
`
`Media
`BMA 031
`BMA 031-G1
`BMA-EUCIVZ
`BMA-EUCIV3
`
`0 4 .
`.01
`
`....'.'I
`
`.1
`
`' " ' 4
`' ' ~ " ' " ' ~ ' " " ' ~
`100
`10
`
`1
`
`Antibody (ug/ml)
`Figure 6. Relative affinities of BMA 031 antibodies. Competitive im-
`munofluorescence assays with the BMA 031 antibodies, HPB mononu-
`clear cells and BMA 031-FITC (2 &ml) were performed as outlined in
`Materials and Methods. Intensity of fluorescence is expressed as mean
`channel number.
`
`study. This effectively limits the number of amino acid
`differences that must be addressed. Second, because the
`assignment of CDR is based on homology and not func-
`tion, the choice of maintaining the murine sequence on
`either side of the CDR is important. Evidence is emerging
`that the "functional" CDR loops can be displaced from
`the "Kabat" CDR by as many as five amino acids. Kabat
`et al. (34) places CDR-1 of the H chain V region from
`amino acids nos. 31 to 35 whereas crystal structure
`shows the loop to be from residues nos. 26 to 32 (35).
`Third, the identification of potentia1ly"Ag specific"amin0
`acids in both the human and murine antibody may be
`important. Although the identification may be straight-
`forward, prioritizing their importance is very difficult.
`Inasmuch as the goal is to produce the most human-like
`sequence, these changes must be kept at a minimum.
`Our decision to keep similar amino acids human and only
`change the more unusual amino acids turned out to be
`correct in this instance. However, in the event that the
`humanized antibody was not functional, this analysis
`
`O
`
`!
`
`0
`
`.
`
`,
`1 0
`
`.
`
`'
`
`.
`
`,
`,
`3 0
`20
`5 0
`E:T Ratio
`
`.
`
`l
`4 0
`
`
`
`i
`
`
`
`c;
`
`loo 1 1
`
`-
`
`c
`e
`n
`
`100 ng/rnl
`10 nglrnl
`1 nglrnl
`0.1 nglrnl
`NK-activity
`
`100 nglrnl
`10 nglrnl
`1 ng/rnl
`0.1 nglrnl
`NK-activity
`
`4 0
`
`W
`5 0
`
`O
`
`0
`
`1 0
`
`2 0
`3 0
`E:T Ratlo
`Figure 7. ADCC capacity of BMA 031 antibodies. The cytolytic capac-
`ity of the BMA 031 antibodies was determined in a 20 h [5'Cr] release
`assay as described in Materials and Methods. A. Lysis in the presence
`(ADCC) or absence (NK activity) of antibody ( 100 ng/ml). B. Lysis at
`various concentratlons of BMA 031-EUCIV3. C. Lysis at various concen-
`trations of chimeric BMA 03 1.
`
`that could be
`
`provides insight into those amino acids
`altered to regain activity.
`The T cell binding data with the humanized BMA 031
`antibodies show the importance of FR amino acids in Ag
`binding. Inclusion of only the BMA 031 CDR (BMA 031-
`EUCIV1) would, most likely, not have been sufficient to
`maintain affinity for Ag. Twelve amino acid substitutions
`were made in the H chain V region to regain binding
`affinity (nos. 27, 28, 30, 38, 48, 67, 68, 70. 93, 94, 95,
`and 98). Of these, six may be more important (nos. 38,
`48, 70, 93, 94, and 95) because they represent changes
`from BMA-EUCIVZ, which