Downloaded from
`
`http://www.jimmunol.org/
`
` at Ebling Library, University of Wisconsin-Madison on December 29, 2014
`
`Humanization and Pharmacokinetics of a Monoclonal
`Antibody with Specificity for Both E- and P-Selectin
`
`Xing-Yue He, Zhenghai Xu, Jennifer Melrose, Alison Mullowney, Maximiliano Vasquez,
`Cary Queen, Vladimir Vexler, Corine Klingbeil, Man Sung Co, and Ellen L. Berg1
`
`E- and P-selectin (CD62E and CD62P) are cell adhesion molecules that mediate leukocyte-endothelial cell and leukocyte-platelet
`interactions and are involved in leukocyte recruitment during inflammation. We previously developed a murine mAb, EP-5C7
`(or mEP-5C7), that binds and blocks both E- and P-selectin. When used in humans, murine mAbs have short circulating half-lives
`and generally induce potent human anti-mouse Ab responses. We therefore engineered a humanized, complementarity deter-
`mining region-grafted version of mEP-5C7 incorporating human g4 heavy and k light chain constant regions (HuEP5C7.g4).
`HuEP5C7.g4 retains the specificity and avidity of mEP-5C7, binding to human E- and P-selectin but not to human L-selectin, and
`blocking E- and P-selectin-mediated adhesion. Surprisingly, when administered to rhesus monkeys, HuEP5C7.g4 was eliminated
`from the circulation very rapidly, even faster than the original murine Ab. To isolate the cause of the short serum half-life of
`HuEP5C7.g4, several Ab variants were constructed. A chimeric IgG4 Ab was made by replacing the humanized V regions with
`murine V regions. A humanized IgG2 Ab, HuEP5C7.g2, was also made by replacing the human g4 with a g2 constant region.
`Results from pharmacokinetic studies in rhesus monkeys demonstrated that the chimeric IgG4 is also rapidly eliminated rapidly
`from serum, similar to the humanized IgG4 Ab, while the humanized IgG2 Ab displays a long circulation half-life, typical of
`human Abs. TheJournalofImmunology, 1998, 160: 1029 –1035.
`
`T he selectin family consists of the L-, E-, and P-selectins
`
`expressed on leukocytes, activated endothelial cells, and
`activated platelets. The three selectins act in concert with
`other cell adhesion molecules to permit recruitment of leukocytes
`from the blood into inflamed tissues (1, 2). Current models suggest
`that the selectins mediate initial adhesive rolling of leukocytes on
`the endothelial vessel wall, a process followed by firm attachment
`and transmigration of leukocytes, which are mediated by other
`adhesion molecules.
`Abs that block E- or P-selectin inhibit the inflammatory se-
`quelae in several animal models of inflammation, including leu-
`kocyte-mediated tissue damage due to ischemia/reperfusion injury
`(3–9). Results from studies in which Abs to E- and P-selectin are
`used in combination and related studies in knockout mice have
`suggested that E- and P-selectin can work cooperatively together
`in some inflammatory conditions, and the absence of one may be
`compensated by the other (10 –13). For therapeutic treatment of
`these inflammatory conditions, blockade of both selectins is pre-
`ferred and could be provided by a cross-reactive mAb, such as
`mEP-5C7, that binds and blocks both E- and P-selectin (14).
`Direct use of murine mAbs for human therapy is limited by the
`development of an immune response by the recipient patient
`against mouse-specific antigenic determinants, i.e., human anti-
`mouse Ab responses, as well as the presence of unwanted or the
`lack of desired human effector functions on mouse Abs. The hu-
`man anti-mouse Ab response can result in potentially harmful re-
`
`Protein Design Labs, Inc., Mountain View, CA 94043
`Received for publication June 5, 1997. Accepted for publication September
`26, 1997.
`The costs of publication of this article were defrayed in part by the payment of
`page charges. This article must therefore be hereby marked advertisement in
`accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
`1 Address correspondence and reprint requests to Dr. Ellen L. Berg, Protein De-
`sign Labs, Inc., 2375 Garcia Ave., Mountain View, CA 94043. E-mail address:
`eberg@pdl.com
`
`actions and the rapid clearance of circulating Ab (15–17). To re-
`duce the immunogenicity of mouse Abs for human therapy,
`chimeric Abs were initially developed. Chimeric Abs retain the
`mouse V regions, and hence the specificity, of the parental rodent
`Abs, but replace the constant regions with the appropriate human
`heavy and light chain sequences. The use of chimeric Abs clearly
`addresses the issue of immunogenicity of the constant region
`caused by species differences, but V region differences remain a
`potential source of immunogenicity (18, 19).
`Humanized Abs are generated by grafting complementarity-de-
`termining regions (CDRs)2 of mouse Abs into a background of
`human framework regions and constant regions by recombinant
`DNA technology (20). Humanized Abs contain only ;10% mouse
`sequences and 90% human sequences, and thus further reduce or
`eliminate the immunogenicity while retaining the Ab specificities
`(21, 22). Using a computer-aided CDR-grafting method, we have
`successfully humanized several mouse Abs for potential treatment
`of various human diseases (23–27). Several humanized Abs, in-
`cluding humanized M195 (anti-CD33, a myeloid leukemia surface
`Ag) and humanized anti-Tac (anti-IL-2 receptor a subunit), have
`been evaluated in human clinical trials, demonstrating long half-
`lives and no evidence of immunogenicity when administered in
`multiple doses (24, 28). In this report we describe an unexpected
`outcome of the humanization of mEP-5C7; when injected into rhe-
`sus monkeys, the humanized Ab, HuEP5C7.g4, displayed a very
`short circulating half-life. A greatly lengthened circulating half-life
`was obtained, however, by replacement of the human IgG4 heavy
`chain constant region with a human IgG2 constant region.
`
`2 Abbreviations used in this paper: CDR, complementarity-determining region;
`the concentration-time curve; HRP, horseradish
`AUC0-tlast, area under
`peroxidase.
`
`Copyright © 1998 by The American Association of Immunologists
`
`0022-1767/98/$02.00
`
`IMMUNOGEN 2170, pg. 1
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`
`Downloaded from
`
`http://www.jimmunol.org/
`
` at Ebling Library, University of Wisconsin-Madison on December 29, 2014
`
`1030
`
`Materials and Methods
`Absandcelllines
`
`Mouse monoclonal anti-human E/P-selectin EP-5C7, anti-human P-selec-
`tin WAPS 12.2, anti-human E-selectin E-1E4, and anti-human CD18 NA8
`Abs have been described previously (14). Mouse (IgG1 k) myeloma
`MOPC 21 and purified human (IgG4 k) were obtained from Sigma Chem-
`ical Co. (St. Louis, MO). SP2/0-Ag14, a mouse myeloma cell line, and
`HL60, a human promyelocytic cell line, were obtained from American
`Type Culture Collection (CRL 1581, and CCL 240, Rockville, MD). A
`mouse pre-B cell
`line, L1-2,
`transfected with human P-selectin,
`L1-2P-selectin, and Chinese hamster ovary (CHO) cell lines transfected with
`human E- or P-selectin gene (CHOE-selectin or CHOP-selectin cells) were de-
`scribed previously (14). HuDREG-200, is a humanized IgG4 Ab (29), and
`binds human, but not rhesus, L-selectin (E. Berg, unpublished observa-
`tions). D200Id and 2H9 are mouse monoclonal anti-Id Abs generated
`against HuDREG-200 and HuEP5C7.g4, respectively. The hybridoma pro-
`ducing 2H9 was generated by injecting BALB/c mice in the hind footpads
`with 10 mg of purified HuEP5C7.g4 prepared in monophosphoryl lipid A
`plus trehalose dicorynomycolate adjuvant (Sigma Chemical Co.), then 4
`and 7 days later with Ab in PBS. Three days following the last immuni-
`zation, fusion with P3X cells was performed, generally as previously de-
`scribed (14). Hybridoma supernatants were screened by ELISA on plates
`coated with HuEP5C7.g4 in the presence of 10% normal human serum
`(Sigma Chemical Co.). 2H9 binds HuEP5C7.g4 and mEP-5C7, but fails to
`bind a panel of other humanized Abs or purified human IgG4.
`CloningofVregioncDNAs
`
`The V domain cDNAs of the light and heavy chains of mEP-5C7 were
`cloned by the anchored PCR method using 39 primers that hybridize with
`the C regions and 59 primers that hybridize with the G tails attached to the
`cDNA using terminal deoxytransferase (30). The sequences were deter-
`mined using the dideoxy termination method with an Applied Biosystems
`373A automated sequencer (Foster City, CA).
`Computeranalysis
`
`Three-dimensional models of the variable domains of the mEP-5C7 Ab
`were built with the help of the ABMOD and ENCAD programs created by
`Levitt and co-workers (31, 32). Sequence analysis, homology searches, and
`structural analysis of three-dimensional models were conducted with the
`program DELTA developed in our laboratory (33).
`ConstructionofchimericandhumanizedAbs
`See Figure 3 for a description of the Abs constructed. Plasmids pVk, con-
`taining the human Ck region (25, 34), and pVg4, containing the human g4
`constant region, were used for the construction of chimeric and humanized
`versions of mEP-5C7 of the human IgG4 isotype. pVg4 is identical with
`pVg1 except for containing heavy chain sequences from human g4 instead
`of g1 (25). The XbaI/BamHI fragment containing the human g4 constant
`region in the pVg4 expression vector was derived from phage provided by
`L. Hood (35). For the construction of humanized HuEP5C7.g2, the plasmid
`pVg4 was replaced by pVg2.M3.D.Tt (36). pVg2.M3.D.Tt incorporates
`two mutations into a human g2 constant region (with Ala replacing Val and
`Gly at positions 234 and 237, using the EU numbering system) to reduce
`interaction of IgG2 with Fc receptors and minimize Ab effector activity.
`Heavy and light chain expression plasmids were constructed by inserting
`the appropriate V region genes into the XbaI site of the respective vectors.
`For the construction of chimeric Ab expression vectors, XbaI fragments
`containing the murine VL and VH genes (including the 59 signal peptide
`sequence and the 39 splice donor signal) were constructed by PCR from
`their respective murine V region cDNAs and then inserted into the XbaI
`site of the respective pVk and pVg4 vectors. For the construction of hu-
`manized Ab expression vectors, XbaI fragments containing the humanized
`VL and VH genes (including the 59 signal peptide sequence and the 39
`splice donor signal) were constructed from assembling eight synthetic
`oligonucleotides.
`Nucleotide sequences were designed to encode the protein sequences of
`the humanized EP5C7 VL and VH, generally using codons found in the
`mouse sequence. Several degenerate codons were changed to create con-
`venient restriction sites or remove undesired ones. To synthesize each V
`gene, four pairs of oligonucleotides (;80 bases in length) were synthesized
`(380B DNA Synthesizer, Applied Biosystems) with overlapping stretches
`of 20 nucleotides. Assembly and amplification of the genes were conducted
`in four steps: 1) the four pairs of complementary oligonucleotides were
`annealed and extended with Klenow fragment in separate reactions; 2) the
`resulting four dsDNA fragments were mixed pairwise, denatured, rean-
`nealed, and extended in two separate reactions; 3) the resulting two dsDNA
`
`HUMANIZATION OF AN ANTI-E/P-SELECTIN mAb
`
`fragments were mixed, denatured, reannealed, and extended to create the
`final full-length dsDNA; and 4) the resulting DNA was used for PCR
`amplification with two short primers, which correspond to the 59 and 39
`ends of the template DNA and contain an XbaI site for subsequent cloning.
`The synthesized XbaI fragments were then inserted into the XbaI site of
`the respective pVk, pVg4, or pVg2.M3.D.Tt vectors. Hybrid Abs contain-
`ing a humanized g4 heavy chain and chimeric light chain (HuH/ChL) or a
`chimeric g4 heavy chain and a humanized light chain (ChH/HuL) were
`also constructed for evaluation.
`
`ProductionandpurificationofhumanizedandchimericAbs
`
`Cells producing humanized Abs (HuEP5C7.g4 or HuEP5C7.g2), chimeric
`Abs (ChEP5C7.g4), or hybrid Abs (Hybrid HuL/ChH, or Hybrid HuH/
`ChL) were generated by transfection of Sp2/0 cells with the appropriate
`plasmids by electroporation using a Gene Pulser apparatus (Bio-Rad, Her-
`cules, CA) at 360 V and 25 mF capacitance according to the manufacturer’s
`instructions. Before transfections, plasmids were linearized using PvuII.
`All transfections were performed using 20 mg of plasmid DNA and 107
`cells in PBS. The cells from each transfection were plated into four 96-well
`tissue culture plates. After 48 h, selective medium was applied. Cells were
`selected in DMEM/10% FBS/hypoxanthine/thymidine media supplement
`(Sigma Chemical Co.) and 1 mg/ml mycophenolic acid. Ab-producing
`clones were screened by assaying for the presence of Ab in the culture
`supernatant by ELISA. Abs from transfectants as well as hybridomas were
`purified from serum-free culture supernatants (Hybridoma-SFM, Life
`Technologies, Gaithersburg, MD) by passage over a column of staphylo-
`coccal protein A-Sepharose CL-4B (Pharmacia, Piscataway, NJ). The
`bound Abs were eluted with 0.2 M glycine-HCl, pH 3.0, and neutralized
`with 1 M Tris-HCl, pH 8.0. The buffer was exchanged for PBS by passing
`over a PD10 column (Pharmacia).
`
`Aviditymeasurement
`
`Binding avidities of mEP-5C7 and HuEP5C7.g4 with E- and P-selectin
`were determined by competitive binding of
`radiolabeled Ab to
`CHOE-selectin and L1-2P-selectin cells, respectively. Purified mEP-5C7 Ab
`was labeled with Na125I (Amersham, Arlington Heights, IL) using lac-
`toperoxidase at 4 mCi/mg of protein (37). Increasing amounts of cold com-
`petitor Ab (mEP-5C7 or HuEP5C7.g4) were added to 2 ng of radiolabeled
`tracer mEP-5C7 Ab and incubated with 4 3 105 CHOE-selectin or
`L1-2P-selectin cells in 0.2 ml of binding buffer (PBS with 2% FCS and 0.1%
`sodium azide) for 2 h at 4°C with constant shaking. Cells were washed and
`pelleted, and the radioactivity associated with the cell pellets was mea-
`sured. The ratio of bound and and free tracer Ab was calculated, and the
`avidities were calculated according to the formula: [X] 2 [EP5C7] 5
`(1/Kx) 2 (1/Ka), where Ka is the avidity of EP5C7, Kx is the avidity of the
`competitor X, brackets indicate the concentration of competitor Ab at
`which bound/free tracer binding is R0/2, and R0 is maximal bound/free
`tracer binding (38).
`
`E-andP-selectinadhesionassays
`HL-60 cell binding to CHOE-selectin or CHOP-selectin transfectant cell lines
`was performed as described previously (14). Confluent cultures of
`CHOE-selectin or CHOP-selectin transfectant cells, grown in 96-well plates,
`were washed and incubated for 15 min with 50 ml of assay buffer (10%
`adult
`bovine
`serum/10% normal
`rabbit
`serum/10 mM HEPES
`(pH 7.2)/RPMI) containing various test and control Abs (Sigma Chemical
`Co.) at 10 mg/ml. Fluorescently labeled HL-60 cells were prepared as pre-
`viously described (14) and resuspended in assay buffer containing 0.25
`mg/ml anti-CD18 Ab, NA8, at 2 3 106 cells/ml. Assays were initiated by
`addition of 50 ml of HL-60 cells to CHOE-selectin or CHOP-selectin cells for
`a final volume of 0.1 ml while plates were rotated at 40 rpm (Innova 200
`orbital shaker, New Brunswick, Inc., Edison, NJ). After 15 min at room
`temperature, unbound cells were removed by washing plates four times
`with 0.2 ml RPMI/well. Bound cells were fixed to the CHOE-selectin or
`CHOP-selectin cells by addition of 100 ml 1% paraformaldehyde (Sigma) in
`PBS. Plates were analyzed using a Microplate Fluorometer (model 7620,
`Cambridge Technology, Inc., Watertown, MA), and the relative number of
`cells bound per well was calculated from the total amount of fluorescence
`measured at 530 nm, using an excitation at 485 nm.
`
`PharmacokineticsofmurineandhumanizedAbsinrhesus
`monkeys
`
`Pharmacokinetic studies were performed at the California Primate Re-
`search Center (Davis, CA). Rhesus monkeys (3–5 kg; three per group)
`were injected with 2 mg/kg of the indicated Abs. Serum samples were
`collected relative to Ab injection at 27 days, 0 (predose), 0.5, 1.5, 2, 6, 12,
`
`IMMUNOGEN 2170, pg. 2
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`
`The Journal of Immunology
`
`1031
`
`Downloaded from
`
`http://www.jimmunol.org/
`
` at Ebling Library, University of Wisconsin-Madison on December 29, 2014
`
`FIGURE 1.
`cDNA and translated amino acid sequences of the heavy
`chain (A) and light chain (B) V regions of the murine EP-5C7 Ab. The
`three CDRs in each chain are underlined. The first amino acid of the
`mature light and heavy chains are double underlined, preceded by
`their respective signal peptide sequences.
`
`tion, and thus binding affinity. Amino acids from the murine
`EP5C7 sequence were used in those positions. This was performed
`at residues 69 and 70 in the VL domain, but was not required at any
`residues in the VH domain. Furthermore, in agreement with the
`protocol outlined previously (23), a number of amino acids in the
`III-3R human sequences that were unusual at their positions for
`their respective (human) subgroups were also identified. Those
`amino acids were changed to correspond to consensus human se-
`quences of human subgroup III (VH) or subgroup I (VL). This was
`performed at residues 61, 72, 82, and 99 of VL and at residues 1,
`75, and 78 of VH. The final amino acid sequences of the human-
`ized EP-5C7 heavy and light chain V regions are shown compared
`with the murine sequences in Figure 2.
`
`Aviditymeasurement
`Recombinant Abs containing mouse or humanized V regions and
`mouse or human constant regions, as shown in Figure 3, were
`produced as described in Materials and Methods. The avidities of
`the humanized Ab, HuEP5C7.g4, for E-selectin and P-selectin
`were determined and compared with those of mEP-5C7 by com-
`petition with the radioiodinated mEP-5C7 Ab. CHOE-selectin cells
`and L1-2P-selectin cells (14) were used as sources of E-selectin and
`P-selectin. The binding avidities were calculated as described in
`Materials and Methods. HuEP5C7.g4 had an avidity of 3.3 3 108
`M21 for E-selectin, identical with that measured for mEP-5C7,
`while for P-selectin, HuEP5C7.g4 had an avidity of 1.5 3 108
`M21, and mEP-5C7 had an avidity of 6.7 3 108 M21.
`
`and 18 h and 1, 2, 3, 4, and 7 days. Samples from some animals were also
`obtained on days 14, 15, and/or 21.
`
`Pharmacokineticanalysis
`
`Compartment-independent pharmacokinetic analysis (39) was performed
`using the KaleidaGraph program for Macintosh (Developed by Abeldeck
`Software, distributed by Synergy Software, Reading, PA). The area under
`the concentration-time curve, AUC0-tlast, was calculated by the trapezoidal
`rule. AUCtlast-‘ was determined by extrapolating to infinity from Ctlast as-
`5 Ctlast/k2, where
`suming exponential decay using the equation AUCtlast-‘
`Ctlast is the concentration at the last measured time point, and k2 is the
`elimination rate constant of the terminal portion of the concentration-time
`1 AUCtlast-‘. The parameter k2
`curve. AUC0-‘ was derived from AUC0-tlast
`was estimated by linear regression analysis of the terminal portion of the
`curve from a minimum of four last data points. The elimination half-life,
`t1/2, was derived from 0.693/k2.
`ELISAs
`
`Microtiter plates (Nunc Immunolon 1) were coated overnight with 100
`ml/well of capture Abs. These included polyclonal anti-mouse IgG (H1L)
`(human Ig absorbed, Biosource, Camarillo, CA), diluted 1/200, or anti-Id
`Abs against EP5C7 (2H9) or HuDREG-200 (D200Id) at 2 mg/ml in PBS.
`Plates were then blocked by incubating wells with 0.5% nonfat dry milk/
`PBS for 1 to 2 h and subsequently washed twice in 0.05% Tween-20/PBS.
`Serum samples diluted appropriately (1/10 to 1/2000) or standards diluted
`and prepared in 0.5% milk/PBS were added to wells, and plates were
`incubated overnight at 4°C. Following four washes in 0.05% Tween-20/
`PBS, secondary Abs were applied. These included horseradish peroxidase
`(HRP)-conjugated Fc-specific anti-mouse Ig (The Jackson Laboratory, Bar
`Harbor, ME) at a 1/4000 dilution for evaluation of mEP-5C7, HRP-con-
`jugated anti-human IgG4 (The Binding Site, Birmingham, U.K.) at a
`1/1000 dilution for evaluation of HuEP5C7.g4, and chimeric and hybrid
`IgG4 Abs and HRP-conjugated anti-human IgG2 (The Binding Site) at a
`1/400 dilution for evaluation of HuEP5C7.g2. After 1 h, plates were again
`washed in 0.05% Tween-20/PBS. Bound Abs were detected by addition of
`100 ml/well TMB substrate (Kirkegaard and Perry, Gaithersburg, MD),
`color development for 10 min, termination of the reaction by addition of
`100 ml/well 2 M H2SO4, and then measurement of absorbance at 450 nm.
`Results
`CloningandsequencingofVregioncDNA
`The murine EP-5C7 Ab heavy and light chain V region cDNAs
`were amplified using an anchored PCR method (30), then cloned
`into pUC18 for sequence determination. Nucleotide sequences
`were obtained from several independent clones for both VL and VH
`cDNAs. For the heavy chain, a unique sequence was identified. For
`the light chain, two sequences were identified. One sequence had
`a nucleotide missing at the end of the V gene, causing a frame shift
`at the V-J junction, and was identified as the nonproductive allele
`from the fusion partner cells. The other sequence is typical of a
`mouse kchain V region. The nucleotide and translated amino acid
`sequences of the coding light and heavy chain V regions of mEP-
`5C7 are shown in Figure 1. Sequence analysis indicates that the
`light chain V region gene of mEP-5C7 belongs to mouse k-chain
`subgroup I, and the heavy chain gene belongs to mouse heavy
`chain subgroup III (40).
`
`Modelinganddesignofhumanizedsequences
`For humanization, the approach of Queen et al. (23) was followed.
`First, sequences of human V regions most similar to mEP-5C7
`were identified. Among the best VH sequences is III-3R (41) of
`Kabat’s subgroup III, with 72% identity in the framework region.
`The VL from III-3R was also used. This VL belongs to the Kabat
`subgroup I of k-chains and has 61% identity with mEP-5C7 in the
`framework region.
`With the help of the three-dimensional model, a number of
`framework positions were identified that differed between mEP-
`5C7 and the chosen human III-3R sequence and whose location in
`three-dimensional space relative to the hypervariable regions, or
`CDRs, makes it likely that they could influence CDR conforma-
`
`IMMUNOGEN 2170, pg. 3
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`
`1032
`
`HUMANIZATION OF AN ANTI-E/P-SELECTIN mAb
`
`Downloaded from
`
`http://www.jimmunol.org/
`
` at Ebling Library, University of Wisconsin-Madison on December 29, 2014
`
`FIGURE 2. Amino acid sequences of the mature heavy chain (A)
`and light chain (B) V regions of the humanized EP-5C7 (lower lines)
`and the mouse EP-5C7 (upper lines) Abs. The three CDRs in each
`chain are underlined once. Residues in the framework that have been
`replaced with mouse amino acids are also underlined once, and con-
`sensus human amino acids, different from those in III-3R, in the hu-
`manized Ab are doubly underlined.
`
`BlockingofE-selectin-andP-selectin-mediatedadhesion
`To compare the abilities of HuEP5C7.g4 and HuEP5C7.g2, the
`humanized EP5C7 Ab of IgG2 isotype, with that of mEP-5C7 to
`inhibit binding of E-selectin to its counter-receptor, these Abs were
`tested for their ability to block binding of HL-60 cells to E-selectin
`transfectant cells. Assays of the adhesion of HL-60 cells with
`CHOE-selectin cells were performed as previously described (14) in
`the presence of the various Abs at the indicated concentrations.
`Figure 4 (A and B) shows that both HuEP5C7.g2 and HuEP5C7.g4
`blocked binding of HL-60 cells to CHOE-selectin transfectants as
`well as or slightly better than mEP-5C7.
`To measure the abilities of HuEP-5C7.g2, HuEP5C7.g4, and
`mEP-5C7 to inhibit binding of P-selectin to its counter-receptor,
`the binding of HL-60 cells to CHOP-selectin transfectants in the
`presence of these Abs was determined. Figure 4 (C and D) shows
`that HuEP5C7.g2 and HuEP5C7.g4 block binding of HL-60 cells
`to P-selectin transfectants as well as mEP-5C7. Isotype-matched
`control Abs had no effect on binding in this assay (14) (data not
`shown).
`
`FIGURE 3. Humanized and murine Abs. The species, isotype, and
`type of V region heavy and light chains for each Ab are listed. A rep-
`resentational figure for each Ab is also shown. Regions containing
`murine sequences are shown in black, human k light chain constant
`regions (Ck) are gray with black outlines, humanized VL and VH are
`white, human IgG4 heavy chains are horizontally striped, and human
`IgG2 heavy chains are gray.
`
`Pharmacokineticsinrhesusmonkeys
`The elimination profiles for 1 wk following injection in rhesus
`monkeys of mEP-5C7, HuEP5C7.g4, HuEP5C7.g2, chimeric or
`hybrid IgG4 Abs, or a control humanized IgG4 Ab, HuDREG-200,
`that does not bind to any Ag in rhesus monkeys are shown in
`Figure 5. HuEP5C7.g4 and chimeric and hybrid IgG4 versions of
`EP5C7 all cleared rapidly from the circulation, so their serum lev-
`els were ,1 mg/ml after 72 h, while HuEP5C7.g2 displayed a long
`serum half-life, with a level of .10 mg/ml at 72 h. The calculated
`terminal elimination half-lives calculated from the last several time
`points and averaged for three rhesus monkeys are shown in Table
`I. The terminal elimination half-life calculated for HuEP5C7.g2
`(6.6 6 0.8 days) was similar to that for the control humanized
`IgG4 Ab, HuDREG-200 (6.5 6 1.2 days). However, while the
`half-life calculated for mEP-5C7 was as short as the half-lives
`calculated for the IgG4 versions of EP5C7 (2.5 6 0.7 vs 2.2–2.9
`days), inspection of Figure 5 shows that mEP-5C7 behaves more
`like HuEP5C7.g2 and the control humanized IgG4, HuDREG-200,
`up until day 4. The 2-wk elimination profiles for each of three
`individual rhesus monkeys given HuEP5C7.g2 at 2 mg/kg are
`shown in Figure 6.
`
`Discussion
`We have previously described the identification of a mAb that
`binds and blocks the adhesive functions of both E- and P-selectin
`(14). Blockade of these receptors may have clinical utility in sit-
`uations of leukocyte-mediated tissue damage, such as ischemia/
`
`IMMUNOGEN 2170, pg. 4
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`
`The Journal of Immunology
`
`1033
`
`Downloaded from
`
`http://www.jimmunol.org/
`
` at Ebling Library, University of Wisconsin-Madison on December 29, 2014
`
`Inhibition of HL-60 cell adhesion to CHOE-selectin or
`FIGURE 4.
`CHOP-selectin cells by mEP-5C7 (h), HuEP5C7.g2 (f), or HuEP5C7.g4
`((cid:130)) Abs. Fluorescently labeled HL-60 cells were incubated with
`CHOE-selectin (Aand B) or CHOP-selectin (Cand D) cells in the presence
`of mEP-5C7 (h) and HuEP5C7.g2 (f; Aand C) or HuEP5C7.g2 (f) and
`HuEP5C7.g4 ((cid:130); Band D) at the indicated concentrations for 15 min.
`After washing, the relative number of bound cells was determined as
`described in Materials and Methods. The results from representative
`experiments performed with each sample in quadruplicate (6 SD) are
`shown.
`
`reperfusion injury (myocardial infarction or stroke), trauma, and
`shock, as well as chronic inflammatory conditions, such as rheu-
`matoid arthritis or psoriasis. For therapeutic purposes, Abs have
`certain advantages over conventional small molecule drugs, e.g.,
`relatively long serum half-lives and often more attractive safety
`profiles. Abs are bivalent, and the relatively spacious Ag-combin-
`ing site naturally provides for a number of Ag contact points. Fur-
`thermore, the ability to “humanize” high affinity murine mAbs has
`resulted in improved therapeutic efficacy of Abs by reducing im-
`munogenicity and improving serum half-lives.
`The techniques of Ab humanization have evolved over the past
`9 yr, so that in our laboratory, humanization of murine Abs specific
`for single protein Ags is nearly always successful. The case de-
`scribed here, however, is the first report of the humanization of an
`Ab with reactivity against two Ags, the E- and P-selectins. Al-
`though these two proteins share homology, there are significant
`differences between them (the various domains share 34 – 62%
`amino acid identity) (42, 43). This additional constraint made the
`likelihood of success of the humanization process less predictable.
`Humanization of mEP-5C7 was successful in that the avidity of
`the humanized Ab, HuEP5C7.g4, for both E- and P-selectin was
`substantially retained. Interestingly, the murine Ab slightly favors
`P-selectin, while HuEP5C7.g4 slightly favors E-selectin. These
`slight differences in avidity are not likely to affect the efficacy of
`the humanized Ab in vivo, since they are not accompanied by
`changes in functional blocking activity (Fig. 4).
`
`FIGURE 5. One-week pharmacokinetic profiles of murine, human-
`ized, and chimeric Abs in rhesus monkeys. Rhesus monkeys (three per
`group) were given 2 mg/kg of each of the following Abs: mEP-5C7,
`HuEP5C7.g4, HuEP5C7.g2, HuDREG-200 (control HuIgG4), or chi-
`meric (ChEP5C7.g4) or hybrid (Hybrid HuL/ChH or Hybrid HuHChL)
`EP5C7 Abs with human IgG4 constant regions. At the indicated times,
`serum samples were prepared and analyzed for the presence of the
`indicated Ab as described in MaterialsandMethods. The mean and SD
`of the serum concentration of Ab measured in three animals are
`shown.
`
`Table I. Terminaleliminationhalf-livesinrhesusofmurineand
`humanizedAbstesteda
`
`Antibody
`
`AUC0-‘
`(mg 3 h/ml),
`mean 6 SD
`
`t 1/2 elimination
`(days),
`mean 6 SD
`
`mEP-5C7
`HuEP5C7.g4
`HuEP5C7.g2
`ChEP5C7.g4
`Hybrid ChL/HuH
`Hybrid HuL/ChH
`HuDREG-200
`
`2.52 6 0.73
`1905 6 1018
`2.87 6 1.29
`708 6 260
`6.64 6 0.87
`4693 6 572
`2.40 6 0.58
`963 6 289
`2.22 6 0.99
`531 6 168
`2.57 6 0.28
`564 6 171
`6.54 6 1.25
`1901 6 723
`a The pharmacokinetic parameters shown were determined by analyzing the
`3-wk elimination profiles of HuEP5C7.g2 and HuDREG-200 and 1-wk elimina-
`tion profiles of other Abs in rhesus monkeys (three per group) given 2 mg/kg.
`Data were analyzed as described in MaterialsandMethods.
`
`Since mEP-5C7 and HuEP5C7.g4 bind to the E- and P-selectins
`of nonhuman primates, including rhesus and baboon (44) (data not
`shown), pharmacokinetic studies were performed in both species.
`Unexpectedly, HuEP5C7.g4 displayed a very short elimination
`half-life, clearing from the circulation in both rhesus and baboons
`even more rapidly than mEP-5C7, particularly in the first 4 days
`(Fig. 5) (V. Vexler, unpublished observations). Several possibili-
`ties could account for a more rapid elimination profile of a hu-
`manized Ab than the original murine Ab, including chemical or
`physical instability, or creation of a novel binding site for another
`Ag. Such a novel binding site might have specificity for a single
`widely expressed Ag or might be nonspecific, binding many Ags
`but with low affinity (as do some natural Abs) (45).
`HuEP5C7.g4 exhibited no obvious physical or chemical insta-
`bility (data not shown), nor did it bind nonspecifically to Ags (e.g.,
`
`IMMUNOGEN 2170, pg. 5
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`
`1034
`
`HUMANIZATION OF AN ANTI-E/P-SELECTIN mAb
`
`Downloaded from
`
`http://www.jimmunol.org/
`
` at Ebling Library, University of Wisconsin-Madison on December 29, 2014
`
`HuEP5C7.g4 cannot be due only to its V region either, because
`when the same humanized V region is combined with a human g2
`constant region in HUEP5C7.g2, a long half-life results. More-
`over, the murine Ab mEP-5C7, with a V region of the same spec-
`ificity and affinity as HuEP5C7.g4, has a half-life as long as can be
`expected for mouse Abs injected into primates. It is only the com-
`bination of the EP-5C7 V region binding site in either murine or
`humanized form together with the human g4 constant region,
`whether in HuEP5C7.g4, the chimeric ChEP5C7.g4 Ab, or the
`hybrid humanized/chimeric Abs, that leads to rapid elimination
`from the serum (Fig. 5).
`Possible explanations for these findings are 1) the mode of flex-
`ibility of the human g4 hinge region may allow the EP-5C7 V
`region to bind with some affinity to other Ags in rhesus monkeys
`(perhaps to other lectin-like domains), leading to rapid clearance
`from the serum, or 2) the ability of the g4 constant region to bind
`to certain Fc receptors on leukocytes together with the ability of
`the V region to bind to E- and P-selectin may create an interaction
`that leads to clearance of the Ab from the serum. The mutated g2
`region used here, which was engineered to eliminate binding to Fc
`receptors (36), would avoid this problem. These are but two of the
`number of potential explanations left after more obvious ones were
`ruled out by our experimental results. The rapid clearance of
`HuEP5C7.g4 from the serum is more likely due to Ab sequestra-
`tion and not elimination from the body, since the rate of clearance
`of HuEP5C7.g4 observed during the first several days after ad-
`ministration slows down significantly after 4 days (Fig. 5).
`Regardless of the reason for the short-half life in primates of
`HuEP5C7.g4, the availability of the humanized HuEP5C7.g2 Ab
`with its unique function of binding and blocking both E- and P-
`selectin and with a long half-life in the circulation presents a viable
`anti-inflammatory drug candidate for clinical evaluation.
`
`Acknowledgments
`Sincere thanks to Lisa Hernandez and Chuck Bullock for organizing the
`pharmacokinetic studies, t