US 7,087,409 B2Grant

Humanization of murine antibody

Issue Date:2006-08-08

•12 Claims

•14 Drawing Sheets

Abstract

A humanized murine antibody is provided. The amino acid sequences of a light chain complementarity determining region from a mouse antibody are grafted onto a human light chain, and a heavy chain complementarity determining region from a mouse antibody are grafted onto a human antibody heavy chain to produce libraries from which a humanized murine antibody having the desired specificity is selected.

Metadata

Assignee

The Scripps Research Institute

Inventors

Carlos F. Barbas, III
Christoph Rader

Application Information

Application Number:US 10/078,757

Filing Date:2002-02-19

Priority Date:1997-12-05

Art Unit:1643

Classifications

IPC:

C12P21/04C12Q1/70C12P21/08C07K16/00C07H21/04

Patent Drawings (14 sheets)

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application is a continuation of U.S. patent application Ser. No. 08/986,016 filed on Dec. 5, 1997, now abandoned.

GOVERNMENTAL RIGHTS

[0002] This invention was made with government support under Contract No. AI 37470 by the National Institutes of Health. The government has certain rights in the invention.

TECHNICAL FIELD

[0003] This invention relates to humanization of murine antibodies.

BACKGROUND OF THE INVENTION

[0004] Antibodies typically comprise two heavy chains linked together by disulfide bonds and two light chains linked to a respective heavy chain by a disulfide bond. Beginning at one end of each heavy chain there is a variable domain followed by several constant domains. Similarly, each light chain has a variable domain at one end, but only a single constant domain at its other end. There are two types of light chain, which are termed lambda (λ) and kappa (κ) chains. No functional difference has been found between antibodies having λ or κ light chains. The ratio of the two types of light chain varies from species to species, however. In mice, the κ:λ ratio is 20:1, whereas in humans it is 2:1.

[0005] The variable domains of the light and heavy chains are aligned, as are the constant domain of the light chain and the first constant domain of the heavy chain. The constant domains in the light and heavy chains are not involved directly in binding the antibody to antigen.

[0006] It is the variable domains that form the antigen binding site of antibodies. The general structure of each light and heavy chain domain comprises a framework of four regions, whose sequences are relatively conserved, connected by three complementarity determining regions (CDRs). The four framework regions employ a beta-sheet conformation and the CDRs form loops connecting, and in some cases forming part of, the beta-sheet structure. The CDRs are held in close proximity by the framework regions and, with the CDRs from the other domain, contribute to the formation of the antigen binding site.

[0007] While cell surface antigens of tumor cells are the traditional targets for antibody-guided cancer therapy, one of the major limitations for the therapy of solid tumors is the low accessibility of tumor antigens to antibodies circulating in the blood stream. The dense packing of tumor cells and the elevated interstitial pressure in the tumor core present formidable physical barriers.

[0008] A solution to the problem of poor penetration of antibodies into solid tumors would be to attack the endothelial cells lining the blood vessels of the tumor rather than the tumor cells themselves. While it may be difficult to target the mature tumor vasculature specifically, i.e., without destroying healthy tissue, promising strategies aim at the inhibition of neovascularization.

[0009] Neovascularization, also termed angiogenesis, is induced by cytokines that are secreted from tumor cells and depends on vascular cell migration and invasion, processes regulated by cell adhesion molecules (CAM) and protease. These molecules are currently considered potential targets for angiogenic inhibitors. In this regard, the vascular integrin α_Vβ₃has recently been identified as a marker of angiogenic blood vessels. See Brooks, P. C., et al. (1994), REQUIREMENT OF INTEGRIN α_Vβ₃FOR ANGIOGENESIS, Science 264, 569–571. Moreover, it was shown that the mouse monoclonal antibody (Mab) LM609 directed to integrin α_Vβ₃was able to suppress angiogenesis, indicating than integrin α_Vβ₃has a critical role in angiogenesis.

[0010] It has been further demonstrated that LM609 selectively promotes apoptosis of vascular cells that have been stimulated to undergo angiogenesis. See Brooks, P. C., et al. (1994), INTEGRIN α_Vβ₃ANTAGONISTS PROMOTE TUMOR REGRESSION BY INDUCING APOPTOSIS OF ANGIOGENIC BLOOD VESSELS, Cell 79, 1157–1164. These findings suggest that integrin α_Vβ₃may be a target and LM609 a tool for cancer diagnosis and therapy.

[0011] Indeed, LM609 not only prevented the growth of histologically distinct human tumors implanted on the chorioallantoic membranes of chicken embryos, but also induced their regression. See, Cell 79, 1157–1164. Using a more clinically relevant model of tumor growth, it was found that LM609 blocked human breast cancer growth in a SCID mouse/human chimeric model. Importantly, not only did LM609 block tumor growth, but it also inhibited metastasis of the breast carcinomas examined. See Brooks, et al. (1995) ANTI-INTEGRIN α_Vβ₃BLOCKS HUMAN BREAST CANCER GROWTH AND ANGIOGENESIS IN HUMAN SKIN, J. Clin. Invest. 96, 1815–1822.

[0012] The Brooks et al. results are consistent with previous studies that have suggested that angiogenesis contributes to the metastatic spread of breast tumor cells. See Weidner, N., et al. (1991) TUMOR ANGIOGENESIS AND METASTASIS: CORRELATION IN INVASIVE BREAST CARCINOMA, N. Engl. J. Med. 324, 1–8; and Weidner, N., et al. (1992) TUMOR ANGIOGENESIS: A NEW SIGNIFICANT AND INDEPENDENT PROGNOSTIC INDICATOR IN EARLY-STAGE BREAST CARCINOMA, J.Natl. Cancer Inst. 84, 1875–1887.

[0013] Within the last few years evidence has been presented that two cytokine-dependent pathways of angiogenesis exist and that these are defined by their dependency on distinct vascular integrins. See Friedlander, M., et al. (1995) DEFINITION OF TWO ANGIOGENIC PATHWAYS BY DISTINCT AV INTEGRINS, Science 270, 1500–1502. The results of the Friedlander et al. studies show that anti-α_Vβ₃antibody LM609 blocked angiogenesis in response to bFGF and TNFα, yet have little effect on angiogenesis induced by VEGF, TGFα, or phorbol ester PMA. In contrast, the anti-α_Vβ₅antibody P1F6 blocks angiogenesis induced by VEGF, TGFα, and phorbol ester PMA, while having minimal effects on that induced by bFGF or TNFα.

[0014] It is conceivable, thus, that tumors showing less susceptibility to anti-α_Vβ₃antibodies might secrete cytokines that promote angiogenesis in an α_Vβ₅-dependent manner. Taken together, both anti-α_Vβ₃and anti-α_Vβ₅antibodies are promising tools for diagnosis and therapy of cancer.

[0015] Mouse monoclonal antibodies such as LM609, however, are highly immunogenic in humans, thus limiting their potential use for cancer therapy, especially when repeated administration is necessary. To reduce the immunogenicity of mouse monoclonal antibodies, chimeric monoclonal antibodies were generated, with the variable Ig domains of a mouse monoclonal antibody being fused to human constant Ig domains. See Morrison, S. L., et al. (1984) CHIMERIC HUMAN ANTIBODY MOLECULES; MOUSE ANTIGEN-BINDING DOMAINS WITH HUMAN CONSTANT REGION DOMAINS, Proc. Natl. Acad. Sci. USA 81, 6841–6855; and, Boulianne, G. L., et al. (1984) PRODUCTION OF A FUNCTIONAL CHIMAERIC MOUSE/HUMAN ANTIBODY, Nature 312, 643–646. This process is commonly referred to as “humanization” of an antibody.

[0016] In general, the chimeric monoclonal antibodies retain the binding specificity of the mouse monoclonal antibody and exhibit improved interactions with human effector cells. This results in an improved antibody-dependent cellular cytotoxicity which is presumed to be one of the ways of eliminating tumor cells using monoclonal antibodies. See Morrison, S. L. (1992) IN VITRO ANTIBODIES: STRATEGIES FOR PRODUCTION AND APPLICATION, Ann. Rev. Immunol. 10, 239–265. Though some chimeric monoclonal antibodies have proved less immunogenic in humans, the mouse variable Ig domains can still lead to a significant human anti-mouse response. See Bruggemann, M., et al. (1989) THE IMMUNOGENICITY OF CHIMERIC ANTIBODIES, J. Exp. Med. 170, 2153–2157. Therefore, for therapeutic purposes it may be necessary to fully humanize a murine monoclonal antibody by altering both the variable and the constant Ig domains.

[0017] Full humanization is feasible by introducing the six CDRs from the mouse heavy and light chain variable Ig domains into the appropriate framework regions of human variable Ig domains. This CDR grafting technique (Riechmann, L., et al. (1988) RESHAPING HUMAN ANTIBODIES FOR THERAPY, Nature 332, 323) takes advantage of the conserved structure of the variable Ig domains, with the four framework regions (FR1–FR4) serving as a scaffold to support the CDR loops which are the primary contacts with antigen. U.S. Pat. No. 5,502,167 to Waldmann, et al. discloses a “humanised antibody” having the CDR loops LCDR1 through LCDR3 and HCDR1 through HCDR3 from YTH 655(5)6, a rat IgG2b monoclonal antibody, grafted onto a human T cell antibody.

[0018] A drawback, however, of the CDR grafting technique is the fact that amino acids of the framework regions can contribute to antigen binding, as well as amino acids of the CDR loops can influence the association of the two variable Ig domains. To maintain the affinity of the humanized monoclonal antibody, the CDR grafting technique relies on the proper choice of the human framework regions and site-directed mutagenesis of single amino acids aided by computer modeling of the antigen binding site (e.g., Co, M. S., et al. (1994) A HUMANIZED ANTIBODY SPECIFIC FOR THE PLATELET INTEGRIN gpllb/lla, J. Immunol. 152, 2968–2976). A number of successful humanizations of mouse monoclonal antibodies by rational design have been reported. Among them are several monoclonal antibodies that are directed to human integrins and have potential clinical application. See, J. Immunol. 152, 2968–2976; Hsiao, K. C., et al. (1994) HUMANIZATION OF 60.3, AN ANTI-CD18 ANTIBODY; IMPORTANCE OF THE L2 LOOP, Protein Eng. 7, 815–822; and, Poul, M. A., et al. (1995) INHIBITION OF T CELL ACTIVATION WITH A HUMANIZED ANTI-BETA 1 INTEGRIN CHAIN mAb, Mol. Immunol. 32, 101–116.

[0019] Human immunoglobulin transgenic mice provide a promising alternative to the humanization of mouse monoclonal antibodies. See, e.g., Fishwild, D. M., et al. (1996) HIGH-AVIDITY HUMAN IgGK MONOCLONAL ANTIBODIES FROM A NOVEL STRAIN OF MINILOCUS TRANSGENIC MICE, Nature Biotechnology 14, 845–851. In response to immunization, these mice express human monoclonal antibodies, which can be accessed by conventional hybridoma technology.

[0020] Rational design strategies in protein engineering have been challenged by in vitro selection strategies that are mainly based on phage display libraries. See Clackson, T., and Wells, J. A. (1994) IN VITRO SELECTION FROM PROTEIN AND PEPTIDE LIBRARIES, TIBTECH 12, 173–184. In particular, in vitro selection and evolution of antibodies derived from phage display libraries has become a powerful tool. See Burton, D. R., and Barbas III, C. F. (1994) HUMAN ANTIBODIES FROM COMBINATORIAL LIBRARIES, Adv. Immunol. 57, 191–280; and, Winter, G., et al. (1994) MAKING ANTIBODIES BY PHAGE DISPLAY TECHNOLOGY, Annu. Rev. Immunol. 12, 433–455.

[0021] The development of technologies for making repertoires of human antibody genes, and the display of the encoded antibody fragments on the surface of filamentous bacteriophage, has provided a means for making human antibodies directly. The antibodies produced by phage technology are produced as antigen binding fragments—usually Fv or Fab fragments—in bacteria and thus lack effector functions. Effector functions can be introduced by one of two strategies: The fragments can be engineered either into complete antibodies for expression in mammalian cells, or into bispecific antibody fragments with a second binding site capable of triggering an effector function.

[0022] Typically, the Fd fragment (V_H-C_H1) and light chain (V_L-C_L) of antibodies are separately cloned by PCR and recombined randomly in combinatorial phage display libraries, which can then be selected for binding to a particular antigen. The Fab fragments are expressed on the phage surface, i.e., physically linked to the genes that encode them. Thus, selection of Fab by antigen binding co-selects for the Fab encoding sequences, which can be amplified subsequently. By several rounds of antigen binding and reamplification, a procedure termed panning, Fab specific for the antigen are enriched and finally isolated.

[0023] In 1994, an approach for the humanization of antibodies, called “guided selection”, was described. Guided selection utilizes the power of the phage display technique for the humanization of mouse monoclonal antibody. See Jespers, L. S., et al. (1994) GUIDING THE SELECTION OF HUMAN ANTIBODIES FROM PHAGE DISPLAY REPERTOIRES TO A SINGLE EPITOPE OF AN ANTIGEN, Bio/Technology 12, 899–903. For this, the Fd fragment of the mouse monoclonal antibody can be displayed in combination with a human light chain library, and the resulting hybrid Fab library may then be selected with antigen. The mouse Fd fragment thereby provides a template to guide the selection.

[0024] Subsequently, the selected human light chains are combined with a human Fd fragment library. Selection of the resulting library yields entirely human Fab.

[0025] For the full humanization of murine monoclonal antibodies, the present invention uses a unique combination of CDR grafting and guided selection. The anti-integrin antibody generated is useful for cancer diagnosis and therapy.

SUMMARY Of THE INVENTION

[0026] Humanization of a mouse monoclonal antibody is achieved by a combination of guided selection and CDR grafting. The term “humanized” as used herein and in the appended claims means that at least one chain of a mouse monoclonal antibody includes a region of a human monoclonal antibody.

[0027] A humanized mouse monoclonal antibody is produced by constructing a library of human antibody heavy chains or light chains in which each such chain includes a variable domain and has at least one complementarity determining region (CDR) amino acid sequence which is that of a corresponding mouse heavy or light antibody chain, and then combining the library so constructed with a complementary chain from an antibody which binds a preselected antigen. In this manner, the complementary chain together with a human chain present in the constructed library forms a heavy and light chain pair in a resulting library of humanized chain pairs. Thereafter a particular humanized heavy and light chain pair is selected from the humanized pair library using the aforementioned complementary chain.

[0028] In a particular embodiment, a mouse monoclonal antibody can be humanized by constructing a human light chain library in which each light chain includes at least the variable domain thereof and at least one CDR amino acid sequence of a mouse light chain, and a human heavy chain library in which each such heavy chain includes at least the variable domain thereof and at least one CDR amino acid sequence of a mouse heavy chain. The heavy chain usually is no more than about 200 amino acid residues in size.

[0029] A human light chain having a mouse CDR is selected from the constructed human light chain library using a heavy chain from an antibody which binds a preselected antigen. The constructed heavy chain library is combined with the selected human light having a mouse CDR to produce a humanized library of heavy and light chain pairs, each containing at least one mouse CDR. Thereafter, a heavy and light chain pair with mouse CDR is selected from the aforesaid humanized library using the selected human light chain with mouse CDR. The sequence of aforementioned library construction is not critical.

[0030] Preferably, only the light chain complementarity determining region three (LCDR3) loop of the monoclonal antibody is grafted onto the human light chain. Similarly, it is preferable that only the HCDR3 loop be grafted onto the human heavy chain (HC) fragment. The selection of either the human light chain or human heavy chain having the grafted mouse CDR is preferably made by using a chimeric mouse/human complementary chain as a template.

[0031] In CDR grafting onto a human light chain, the human light chain is cloned, then the clones are randomly recombined to form a library such as a combinatorial phage display library. The same method can be followed for grafting onto the human heavy chain.

BRIEF DESCRIPTION OF THE DRAWINGS

[0032] FIG. 1 is a schematic illustration showing the sequence of the steps in the combined CDR grafting technique and guided selection technique to form the humanized Fab fragment.

[0033] FIGS. 2a and 2b show the amino acid sequences of V_λ (SEQ ID NO: 44) and V_κ (SEQ ID NO: 45) , respectively, of mouse monoclonal antibody LM609. The N-terminal two amino acids (Leu)(Glu) of V and (Glu)(Leu) of V encoded by the vector cloning sites CTCGAG (Xhol) and GAGCTC (Sacl), respectively, are artificial. The CDR loops are underlined.

[0034] FIG. 3a through FIG. 3e show the amino acid sequence alignment of mouse LM609 V_κ (top full line of each sequence grouping) and six selected human V_κ's (#1–6). Framework regions (FR1–3) and CDR (CDR1–2) loops are separated. Lines (-) indicate identical amino acids. Note that due to the LCDR3 grafting parts of FR3, entire CDR3 and entire FR4 are identical in mouse LM609 V_κand the selected human V_κ's. Therefore, these two sequences are not shown.

[0035] In FIG. 3a, the top line (labeled “V_κ”) represents amino acid residues 5–23 of SEQ ID NO: 45. The lines labeled “#1” and “#3” each represent amino acid residues of human V_κ clones having Ser in alignment with Thr7 of SEQ ID NO: 45, Gly-Thr in alignment with Ala9-Thr10 of SEQ ID NO: 45, Leu-Ser in alignment with Val13-Thr14 of SEQ ID NO: 45, and Glu-Arg-Ala-Thr (SEQ ID NO: 57) in alignment with Asp17-Ser18-Val19-Ser20 of SEQ ID NO: 45. The line labeled “#2” represents amino acid residues of a human V_κ clone having Ser in alignment with Thr7 of SEQ ID NO: 45, Gly-Thr in alignment with Ala9-Thr10 of SEQ ID NO: 45, Leu-Ser in alignment with Val13-Thr14 of SEQ ID NO: 45, and Glu-Arg-Gly-Ser (SEQ ID NO: 58) in alignment with Asp17-Ser18-Val19-Ser20 of SEQ ID NO: 45. The lines labeled “#4” and “#6” each represent amino acid residues of a human V_κ clone having Ser in alignment with Thr7 of SEQ ID NO: 45, Ser-Ser in alignment with Ala9-Thr10 of SEQ ID NO: 45, Ala-Ser-Val in alignment with Val13-Tbr14-Pro15 of SEQ ID NO: 45, Arg in alignment with Ser18 of SEQ ID NO: 45, and Thr-Ile-Thr in alignment with Ser20-Leu21-Ser22 of SEQ ID NO: 45. The line labeled “#5” represents amino acid residues of a human V_κ clone having Ser in alignment with Thr7 of SEQ ID NO: 45, Ser-Ser in alignment with Ala9-Thr10 of SEQ ID NO: 45, Ala-Ser-Val in alignment with Val13-Thr14-Pro15 of SEQ ID NO: 45, and Thr-Ile-Thr in alignment with Ser20-Leu21-Ser22 of SEQ ID NO: 45.

[0036] In FIG. 3b, the top line (labeled “V_κ”) represents amino acid residues 24–34 of SEQ ID NO: 45. The line labeled“#1” represents amino acid residues of a human V_κ clone having Val in alignment with Ile29 of SEQ ID NO: 45, and Ser-Ser-Thr-Leu-Ala (SEQ ID NO: 59) in alignment with Asn31-His32-Leu33-His34 of SEQ ID NO: 45. The line labeled “#2” represents amino acid residues of a human V_κ clone having Val in alignment with Ile29 of SEQ ID NO: 45, and Ser-Ser-Phe-Leu-Ala (SEQ ID NO: 60) in alignment with Asn31-His32-Leu33-His34 of SEQ ID NO: 45. The line labeled “#3” represents amino acid residues of a human V_κ clone having Val-Thr-Ser-Ser-Tyr-Leu-Ala (SEQ ID NO: 61) in alignment with Ile29-Ser30-Asn31-His32-Leu33-His34 of SEQ ID NO: 45. The line labeled “#4” represents amino acid residues of a human V_κ clone having Thr-Phe in alignment with Asn31-His32 of SEQ ID NO: 45, and Asn in alignment with His34 of SEQ ID NO: 45. The lines labeled “#5” and “#6” each represent amino acid residues of human V_κ clones having Ser-Tyr in alignment with Asn31-His32 of SEQ ID NO: 45, and Asn in alignment with His34 of SEQ ID NO: 45.

[0037] In FIG. 3c, the top line (labeled “V_κ”) represents amino acid residues 35–49 of SEQ ID NO: 45. The lines labeled “#1”, “#2”, and “#3” each represent amino acid residues of human V_κ clones having Pro-Gly-Gln-Ala (SEQ ID NO: 62) in alignment with Ser40-His41-Glu42-Ser43 of SEQ ID NO: 45 and Tyr in alignment with Lys49 of SEQ ID NO: 45. The line labeled “#4” represents amino acid residues of a human V_κ clone having Pro-Gly-Lys-Ala (SEQ ID NO: 63) in alignment with Ser40-His41-Glu42-Ser43 of SEQ ID NO: 45, Lys-Phe in alignment with Arg45-Leu46 of SEQ ID NO: 45, and Tyr in alignment with Lys49 of SEQ ID NO: 45. The line labeled “#5” represents amino acid residues of a human V_κ clone having Arg in alignment with Gln38 of SEQ ID NO: 45, Pro-Gly-Lys-Ala (SEQ ID NO: 63) in alignment with Ser40-His41-Glu42-Ser43 of SEQ ID NO: 45, Lys-Leu in alignment with Arg45-Leu46 of SEQ ID NO: 45, and Tyr in alignment with Lys49 of SEQ ID NO: 45. The line labeled “#6” represents amino acid residues of a human V_κ clone having Pro-Gly-Lys-Ala (SEQ ID NO: 63) in alignment with Ser40-His41-Glu42-Ser43 of SEQ ID NO: 45, Lys-Leu in alignment with Arg45-Leu46 of SEQ ID NO: 45, and Tyr in alignment with Lys49 of SEQ ID NO: 45.

[0038] In FIG. 3d, the top line (labeled “V_κ”) represents amino acid residues 50–56 of SEQ ID NO: 45. The lines labeled “#1”, “#2”, and “#3” each represent amino acid residues of human V_κ clones having Gly in alignment with Tyr50 of SEQ ID NO: 45, and Ser-Arg-Ala-Thr (SEQ ID NO: 64) in alignment with Gln53-Ser54-Ile55-Ser56 of SEQ ID NO: 45. The lines labeled “#4”, “#5” and “#6” each represent amino acid residues of human V_κ clones having Ala in alignment with Tyr50 of SEQ ID NO: 45, and Thr-Leu-Gln in alignment with Gln53-Ser54-Ile55 of SEQ ID NO: 45.

[0039] In FIG. 3e, the top 2 lines (labeled “V_κ”) represent amino acid residues 57–88 of SEQ ID NO: 45. The lines labeled “#1” represent amino acid residues of a human V_κ clone having Asp in alignment with Ser60 of SEQ ID NO: 45, Ile in alignment with Thr72 of SEQ ID NO: 45, Thr in alignment with Ser74 of SEQ ID NO: 45, Ser-Arg-Leu in alignment with Asn76-Ser77-Val78 of SEQ ID NO: 45, Pro in alignment with Thr80 of SEQ ID NO: 45, Ala-Val in alignment with Gly84-Met85 of SEQ ID NO: 45, and Tyr in alignment with Phe87 of SEQ ID NO: 45. The lines labeled “#2” represent amino acid residues of a human V_κ clone having Asp in alignment with Ser60 of SEQ ID NO: 45, Val in alignment with Asp70 of SEQ ID NO: 45, Thr in alignment with Ser74 of SEQ ID NO: 45, Ser-Arg-Leu in alignment with Asn76-Ser77-Val78 of SEQ ID NO: 45, Pro in alignment with Thr80 of SEQ ID NO: 45, Ala-Val in alignment with Gly84-Met85 of SEQ ID NO: 45, and Tyr in alignment with Phe87 of SEQ ID NO: 45. The lines labeled “#3” represent amino acid residues of a human V_κ clone having Asp in alignment with Ser60 of SEQ ID NO: 45, Ile-Phe-Thr in alignment with Thr72-Leu73-Ser74 of SEQ ID NO: 45, Ser-Arg-Leu in alignment with Asn76-Ser77-Val78 of SEQ ID NO: 45, Pro in alignment with Thr80 of SEQ ID NO; 45, Ala-Val in alignment with Gly84-Met85 of SEQ ID NO: 45, and Tyr in alignment with Phe87 of SEQ ID NO: 45. The lines labeled “#4” and “#6” each represent amino acid residues of human V_κ clones having Val in alignment with Ile58 of SEQ ID NO: 45, Thr in alignment with Ser74 of SEQ ID NO: 45, Ser in alignment with Asn76 of SEQ ID NO: 45, Leu-Gln-Pro in alignment with Val78-Glu79-Thr80 of SEQ ID NO: 45, Ala-Val in alignment with Gly84-Met85 of SEQ ID NO: 45, and Tyr in alignment with Phe87 of SEQ ID NO: 45. The lines labeled “#5” represent amino acid residues of a human V_κ clone having Val in alignment with Ile58 of SEQ ID NO: 45, Ala in alignment with Gly68 of SEQ ID NO: 45, Thr in alignment with Ser74 of SEQ ID NO: 45, Ser in alignment with Asn76 of SEQ ID NO: 45, Leu-Gln-Pro in alignment with Val78-Glu79-Thr80 of SEQ ID NO: 45, Ala-Val in alignment with Gly84-Met85 of SEQ ID NO: 45, and Tyr in alignment with Phe87 of SEQ ID NO: 45.

[0040] FIG. 4 shows a comparison of three selected human fragment sequences and four unselected human fragment sequences to the original sequences of the mouse LCDR1 and LCDR2 loops.

[0041] In FIG. 4, the top sequence line shows portions of the mouse LCDR1 and LCDR2 loops, i.e., Arg-Ala-Ser-GIn-Ser-Ile-Ser-Asn (SEQ ID NO: 65) and His-Leu-His in the LCDR1 loop, and Lys-Tyr-Ala-Ser-Gln-Ser-Ile-Ser (SEQ ID NO: 66) in the LCDR2 loop. The second sequence line shows a selected human sequence in which a first portion (SEQ ID NO: 67) is in alignment with SEQ ID NO: 65 of the mouse LCDR1 loop, a second portion (Ser-Leu-His) in alignment with His-Leu-His in the mouse LCDR1 loop, and the LCDR2 loop comprises SEQ ID NO: 68. The third sequence line shows a selected human sequence in which a first portion (SEQ ID NO: 69) is in alignment with SEQ ID NO: 65 of the mouse LCDR1 loop, a second portion (Ser-Leu-His) in alignment with His-Leu-His in the mouse LCDR1 loop, and the LCDR2 loop comprises SEQ ID NO: 68. The fourth sequence line shows a selected human sequence in which a first portion (SEQ ID NO:70) is in alignment with SEQ ID NO: 65 of the mouse LCDR2 loop, a second portion (Ser-Leu-His) in alignment with His-Leu-His in the mouse LCDR1 loop, and the LCDR2 loop comprises SEQ ID NO: 71. The fifth sequence line shows an unselected human sequence in which a first portion (SEQ ID NO: 72) is in alignment with SEQ ID NO: 65 of the mouse LCDR1 loop, a second portion (Thr-Leu-Ala) in alignment with His-Leu-His in the mouse LCDR1 loop, and the LCDR2 loop comprises SEQ ID NO: 73. The sixth sequence line shows an unselected human sequence in which a first portion (SEQ ID NO: 74) is in alignment with SEQ ID NO: 65 of the mouse LCDR1 loop, a second portion (Thr-Leu-Ala) in alignment with His-Leu-His in the mouse LCDR1 loop, and the LCDR2 loop comprises SEQ ID NO: 75. The seventh sequence line shows an unselected human sequence in which a first portion (SEQ ID NO: 76) is in alignment with SEQ ID NO: 65 of the mouse LCDR1 loop, a second portion (Thr-Leu-Ala) in alignment with His-Leu-His in the mouse LCDR1 loop, and the LCDR2 loop comprises SEQ ID NO: 77. The eighth sequence line shows an unselected human sequence in which a first portion (SEQ ID NO: 78) is in alignment with SEQ ID NO: 65 of the mouse LCDR1 loop, a second portion (Thr-Leu-Ala) in alignment with His-Leu-His in the mouse LCDR1 loop, and the LCDR2 loop comprises SEQ ID NO: 79.

[0042] FIG. 5a through FIG. 5f are line graphs which show the binding of human integrin α_Vβ₃on the cell surface by humanized LM609 clones 2, 4, 7, 11, 24, and control antibody, respectively. Line A indicates untransfected CS-1 hamster cells; Line B indicates human γ₅cDNA transfected CS-1 hamster cells (essentially the same line as line A in FIG. 5f); and line C indicates human β₃cDNA transfected CS-1 hamster cells.

[0043] FIG. 6 is a bar graph showing the cross-reactivity of the LM609 antibody, and clones 2, 4, 7, 11, 24, and control antibody, respectively. Columns represent the mean of triplicates, with the left columns indicating binding to human integrin α_Vβ₃, the central columns indicating binding to human integrin α_IIbβ₃, and the right columns indicating background binding. Error bars indicate standard deviations.

[0044] FIG. 7 is a schematic illustration of a stretch of four amino acids in a light chain complementarity determining region three (LCDR3) and a heavy chain complementarity determining region three (HCDR3) being optimized. In FIG. 7, QQSNSWPHT is SEQ ID NO: 2, HNYGSFAY is SEQ ID NO: 1, QQXXXXPHT is SEQ ID NO: 121, and XXXXSFAY is SEQ ID NO: 122.

[0045] FIGS. 8a and 8b are fragmented illustrations of the V_Lamino acid sequences of a mouse antibody compared to the amino acid sequences of five versions of humanized clones represented by group letters A (clones 10, 11, and 37), and B (clones 7, 8, and 22), C (clones 4, 31, and 36), D (clones 24, 34, 35, and 40), and E (clone 2) which are combined. In FIG. 8a, the mouse FR1 region is SEQ ID NO: 80; the human Group A and Group BCDE FR1 regions are each SEQ ID NO: 81; the mouse CDR1 region is SEQ ID NO: 82; the human Group A CDR1 region is SEQ ID NO: 83; the human Group BCDE CDR1 regions are each SEQ ID NO: 84; the mouse FR2 region is SEQ ID NO: 85; the human Group A and Group BCDE FR2 regions are each SEQ ID NO: 86; the mouse CDR2 region is SEQ ID NO: 87; and the human Group A and Group BCDE CDR2 regions are each SEQ ID NO: 88. In FIG. 8b, the mouse FR3 region is SEQ ID NO: 89; the human Group A FR3 region is SEQ ID NO: 90; the human Group BCDE FR3 regions are each SEQ ID NO: 91; the mouse CDR3 region, as well as the human Group A and human Group BCDE CDR3 regions, are each SEQ ID NO: 2; the mouse FR4 region is SEQ ID NO: 92; while the human Group A and human Group BCDE FR4 regions are each SEQ ID NO: 93.

[0046] FIGS. 8c through 8e are fragmented illustrations of the V_Hamino acid sequences of a mouse antibody compared to the amino acid sequences of five versions of humanized clones represented by group letters A (clones 10, 11, and 37), B (clones 7, 8, and 22), C (clones 4, 31, and 36), D (clones 24, 34, 35, and 40), and E (clone 2).

[0047] In FIG. 8c, the mouse FR1 region is SEQ ID NO: 94; the human Group A FR1 region is SEQ ID NO: 95; the human Group B FR1 region is SEQ ID NO: 96; the human Group C FR1 region is SEQ ID NO: 97; the human Group D FR1 region is SEQ ID NO: 98; the human Group E FR1 region is SEQ ID NO: 99; the mouse CDR1 region is SEQ ID NO: 100; the human Group A CDR1 region is SEQ ID NO: 101; the human Group B CDR1 region is SEQ ID NO: 102; the human Group C, D, and E CDR1 regions are each SEQ ID NO: 103; the mouse FR2 region is SEQ ID NO: 104; the human Group A FR2 region is SEQ ID NO: 105; the human Group B FR2 region is SEQ ID NO: 106; the human Group C and E FR2 regions are each SEQ ID NO: 107; and the human Group D FL2 region is SEQ ID NO: 108.

[0048] In FIG. 8d, the mouse CDR2 region is SEQ ID NO: 108; the human Group A CDR2 region is SEQ ID NO: 109; the human Group B CDR2 region is SEQ ID NO: 110; the human Group C CDR2 region is SEQ ID NO: 111; the human Group D and Group E CDR2 regions are both SEQ ID NO: 112; the mouse FR3 region is SEQ ID NO: 113; the human Group A FR3 region is SEQ ID NO: 114; the human Group B FR3 region is SEQ ID NO: 115; the human Group C FR3 region is SEQ ID NO: 116; the human Group D FR3 region is SEQ ID NO: 117; and the human Group E FR3 region is SEQ ID NO: 118.

[0049] In FIG. 8e, the mouse CDR3 region, as well as the human Group A, B, C, D and E CDR3 regions are each SEQ ID NO: 1; the mouse FR2 region is SEQ ID NO: 119, while the human Group A, B, C, D, and E FR2 regions are each SEQ ID NO: 120.

DESCRIPTION OF A PREFERRED EMBODIMENT

[0050] While the present invention is susceptible to embodiments in many different forms, a preferred embodiment of the invention is described below. It should be understood, however, that the present disclosure is to be considered as an exemplification of the principles of the invention and is not intended to limit the invention to the specific embodiments illustrated.

cDNA Cloning of Mouse Monoclonal Antibody LM609

[0051] Beginning with a LM609 hybridoma cell line (Deposited at and accepted by American Type Culture Collection, 10801 University Boulevard, Manassas. Va. 20110-2209. USA on Sep. 15, 1987; ATCC Designation HR 9537), total RNA was prepared from 10⁸LM609 hybridoma cells using an RNA Isolation Kit (Stratagene, La Jolla, Calif.); Reverse transcription and polymerase chain reaction (PCR) amplification of Fd fragment and κ chain encoding sequences were performed essentially as described in “Combinatorial immunoglobulin libraries in phage 1”, (Methods 2, 119 (1991)) by A. S. Kang, et al.

[0052] Fd fragment and κ chain encoding PCR products were cut with XhoI/SpeI and SacI/XbaI, respectively, and ligated sequentially into the appropriately digested phagemid vector pComb3H. The ligation products were introduced into E. coli strain XL1-Blue by electrotransformation and subsequent steps were as described in “Assembly of combinatorial antibody libraries on phage surfaces: the gene III site”, (Proc. Natl. Acad. Sci. USA 88, 7978–7982) by C. F. Barbas III, et al., to produce phage displaying Fab on their surface. Phage were selected by panning against immobilized integrin α_Vβ₃. After two panning rounds single clones were analyzed for LM609 Fab expression. Supernatant from IPTG-induced cultures was tested for binding to immobilized integrin α_Vβ₃by enzyme-linked immunosorbent assay (ELISA) using goat anti-mouse F(ab′)₂conjugated to alkaline phosphatase as a secondary antibody. The sequence of each Fd fragment and each κ chain encoding sequence of positive clones was determined by DNA sequencing.

Amplification of Human Light Chain and Fd Fragment Sequences

[0053] Total RNA was prepared from the bone marrow of five donors (Poietic Technologies; Germantown, Md.) using TRI Reagent (Molecular Research Center; Cincinnati, Ohio) and was further purified by lithium chloride precipitation. See Sambrook, J., et al. (1989) MOLECULAR CLONING: A LABORATORY MANUAL, Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. First-strand cDNA was synthesized using the ‘SUPERSCRIPT Preamplification System for First Strand cDNA Synthesis’ kit with oligo (dT) priming (Life Technologies; Gaithersburg, Md.). The five generated first-strand cDNAs were subjected to separate PCR amplifications.

[0054] V_κ sequences of each of the first-strand cDNAs were amplified in eight separate reactions by combining four sense and two antisense primers (see list below). V_λ sequences were amplified in nine separate reactions using nine sense and one antisense primer (see list below). The V_λ and V_κ amino acid sequences, including the underlining of the CDR loops, are shown in FIGS. 2a and 2b, respectively (See also, SEQ ID NO:44 and SEQ ID NO:45, respectively).

[0055] V_Hsequences (See SEQ ID NO:56) were amplified in four reactions using four sense and one antisense primer (see list below). All amplifications were performed under standard PCR conditions using Taq polymerase (Pharmacia; Uppsala, Sweden). While the sense primers hybridize to sequences that encode the N-terminal amino acids of the various V_κ, V_λ, and V_Hfamilies, the antisense primers hybridize to a sequence that encodes the C-terminal amino acids of FR3 of V_κ, V_λ, or V_H, respectively, which are highly conserved.

[0056] The list of primers used for the amplification of human antibody sequences include:

[0057]


V_ksense primers:
	HSCK1 - F	SEQ ID NO:3
	HSCK24 - F	SEQ ID NO:4
	HSCK3 - F	SEQ ID NO:5
	HSCK5 - 5	SEQ ID NO:6
V_Kantisense primers:
	BKFR3UN	SEQ ID NO:7
	BK2FR3UN	SEQ ID NO:8
V_λ sense primers:
	HSCLam1a	SEQ ID NO:9
	HSCLam1b	SEQ ID NO:10
	HSCLam2	SEQ ID NO:11
	HSCLam3	SEQ ID NO:12
	HSCLam4	SEQ ID NO:13
	HSCLam6	SEQ ID NO:14
	HSCLam70	SEQ ID NO:15
	HSCLam78	SEQ ID NO:16
	HSCLam9	SEQ ID NO:17
V_λ antisense primer:
	BLFR3UN	SEQ ID NO:18
V_Hsense primers:
	HFVH1-F	SEQ ID NO:19
	HFVH2-F	SEQ ID NO:20
	HFVH35-F	SEQ ID NO:21
	HFVH4-F	SEQ ID NO:22
V_Hantisense primers:
	BFR3UN	SEQ ID NO:23

Construction of a Chimeric Mouse/Human Fd Fragment by Fusing V_Hof LM609 to Human C_H1

[0058] The phagemid vector pComb3H containing the LM609 Fab sequences was used as a template for amplification of the sequence encoding the N-terminal FR1 through FR3 fragment of LM609's V_Hby the PCR primer pair PELSEQ (SEQ ID NO:24)/BFR3UN (SEQ ID NO:25). The sense primer PELSEQ hybridizes to the pe1B leader sequence upstream of the Fd fragment encoding sequence in pComb3H. The antisense primer BFR3UN hybridizes to a sequence that encodes eight C-terminal amino acids of FR3 of V_H, which are highly conserved (SEQ ID NO:26), and differ in one amino acid from the corresponding amino acid sequence of LM609's V_H(SEQ ID NO:27).

[0059] By overlap extension PCR (See McArn Horton, R., and Readington Pease, L. (1991) RECOMBINATION AND MUTAGENESIS OF DNA SEQUENCES USING PCR IN DIRECTED MUTAGENESIS: A PRACTICAL APPROACH, ed. M. J. McPherson, IRL Press, Oxford, UK, pp. 217–247), the PELSEQ/BFR3UN product was fused to a PCR fragment encoding the HCDR3 (SEQ ID NO:1) of LM609 coupled to FR4 of V_Hand the entire C_H1 domain of the human anti-gp120 antibody b8. This fragment had been amplified by the PCR primer pair CR501 (SEQ ID NO:28)/CR301 (SEQ ID NO:29). The sense primer CR501 encodes a synthetic link of the nine C-terminal amino acids of FR3, the eight amino acids forming the HCDR3 (SEQ ID NO:1) of LM609, and the six N-terminal amino acids of FR4 of b8. The FR4 of b8 is a preferred choice here because it is identical to FR4 of LM609's V_Hwith the exception of the C-terminal amino acid, which is A for LM609 and S for b8. The 24-bp overlap of CR501 and BFR3UN allowed to fuse the corresponding PCT products by overlap extension PCR.

[0060] The sense primer CR301 hybridizes to a sequence that encodes the C terminus of C_H1 and introduces a SpeI site that allows the PCR product to link to the gene III ORF on pComb3H. The product of the overlap extension PCR was cut with XhoI/SpeI, ligated into the appropriately digested phagemid vector pComb3H, and the correct sequence was confirmed by DNA sequencing.

Substitution of the LM609 Light Chain by a Human Light Chain that Contains the LCDR3 of LM609

[0061] Using overlap extension PCR, the amplified human sequences encoding the N-terminal FR1 through FR3 fragment of V_κ and V_λ were fused to PCR fragments encoding the LCDR3 (SEQ ID NO:2) of LM609 coupled to FR4 of human V_κ or V_λ and the human C_κ or C_λ domain. Two κ fragments were generated by the PCR primer pairs CR503 (SEQ ID NO:30)/T7B (SEQ ID NO:31) and CR508 (SEQ ID NO:32)/T7B using the sequence of the anti-gp120 antibody b11 in pComb3 as a template.

[0062] The sense primers CR503 and CR508 encode a synthetic link of eight C-terminal amino acids of FR3 of human V_κ (SEQ ID NO:33 or SEQ ID NO:34), the nine amino acids forming the LCDR3 (SEQ ID NO:2) of LM609, and the seven N-terminal amino acids of FR4 of b11. FR4 of b11 is the preferred choice because it is identical to FR4 of LM609's V_κ with the exception of the third N-terminal and C-terminal amino acid, which are G and T in LM609 versus Q and A in b11. The 23-bp overlap of CR503 with BKFR3UN and CR508 with BK2FR3UN allowed to fuse the corresponding PCR products by overlap extension PCR.

[0063] The backward primer T7B hybridizes to a pComb3 sequence downstream of the light chain encoding sequence. A λ fragment was generated by the PCR primer pair CR510 (SEQ ID NO:35)/CLext (SEQ ID NO:36) using CLext primed first strand cDNA from human bone marrow as a template.

[0064] The sense primer CR510 encodes a synthetic link of seven C-terminal amino acids of FR3 of human V_λ (SEQ ID NO:37), the nine amino acids forming the LCDR3 of LM609, and the seven N-terminal amino acids of FR4 of human V_λ (SEQ ID NO:38). The 21-bp overlap of CR510 with BLFR3UN allowed to fuse the corresponding PCR products by overlap extension PCR. The backward primer CLext hybridizes to the 3′ end of the human C_λ encoding sequence and introduces a Xbai site.

[0065] The generated light chain encoding sequences were cut with SacI/XbaI and ligated into the appropriately digested phagemid vector pComb3H that contained the chimeric mouse/human Fd fragment. Electrotransformation of the ligation products into E. coli strain ER 2537 (New England Biolabs; Beverly, Mass.) resulted in a light chain library consisting of 1.5×10⁸independent transformants. DNA sequencing revealed the correct assembly of the fused fragments.

[0066] Four rounds of panning against immobilized human integrin α_Vβ₃were carried out essentially as described in “High-affinity self-reactive human antibodies by design and selection: targeting the integrin ligand binding site”, (Proc. Natl. Acad. Sci. USA 90, 10003–10007 (1993)) by C. F. Barbas, III, et al. using 200 ng protein in 25 μl metal buffer (25 mM Tris-HCl, pH 7.5; 137 mM NaCl; 1 mM KCl; 1 mM MgCl₂; 1 mM CaCl₂; 1 mM MnCl₂) for coating, 0.05 percent Tween 20 in TBS for washing, and 10 mg/ml trypsin in TBS for elution. The washing steps were increased from 5 in the first round, to 10 in the second round and 15 in the third and fourth rounds. The output phage pool of each round was monitored by phage ELISA.

[0067] After the fourth round phage were produced from single clones and tested for binding to immobilized integrin α_Vβ₃by phage ELISA. Light chain encoding sequences of positive clones were analyzed by DNA sequencing using the primer OMPSEQ (SEQ ID NO:39) that hybridizes to the ompA leader sequence upstream of the light chain encoding sequence in pComb3H.

Substitution of the LM609 Fd Fragment by a Human Fd Fragment that Contains the Heavy Complementarity Determining Region Three (HCDR3) of LM609

[0068] Three PCR fragments were fused in one step by overlap extension PCR. Using the selected phagemids from the light chain panning as a template, Fragment 1 was amplified with the PCR primer pair RSC-F (SEQ ID NO:40)/lead-B (SEQ ID NO:41). While the sense primer RSC-F hybridizes to a sequence upstream of the light chain encoding sequence, the antisense primer lead-B hybridizes to a sequence upstream of the Fd fragment encoding sequence. The amplified human sequences encoding the FR1 through FR3 fragment of VH (see above) were used as fragment 2.

[0069] Fragment 3 was amplified with the PCR primer pair CR501/HIgG1-B (SEQ ID 42) using the hybrid mouse/human Fd fragment (see above) as a template. The antisense primer HIgG1-B hybridizes to the 3′ end of the C_H1 encoding sequence. Using the 21-bp overlap of lead-B with the HFVH-F primers and the 24-bp overlap of BFR3UN with CR501, the three fragments were fused and amplified with the PCR primer pair RSC-F/RSC-B (SEQ ID NO:43). The antisense primer RSC-B overlaps with HIgG1-B. RSC-F and RSC-B introduce two asymmetric SfiI sites.

[0070] To maintain high complexity, separate PCR reactions were performed for each selected phagemid from the light chain panning (Fragment 1) and for each of the five V_Hfragment pools derived from the five first strand cDNA sources (Fragment 2). The generated fragments encoding the selected human light chains linked to human Fd fragments were cut with SfiI and ligated into the appropriately digested phagemid vector pComb3H generating a library of 3×10⁷independent transformants.

[0071] DNA sequencing revealed the correct assembly of the fused DNA fragments. Four rounds of panning against immobilized human integrin α_Vβ₃were carried out exactly as described for the light chain panning. The output phage pool of each round was monitored by phage ELISA. After the fourth round, light chain and Fd fragment encoding sequences were isolated from the selected phagemids by SfiI digestion and subcloned into the compatible expression vector pPhoA-H6HA.

[0072] Lysates of individual clones grown in phosphate-deprived medium were analyzed for binding to immobilized integrin α_Vβ₃by ELISA using goat anti-human F(ab′)₂conjugated to alkaline phosphatase (Pierce) as secondary antibody. Light chain and Fd fragment encoding sequences of positive clones were analyzed by DNA sequencing using the primers OMPSEQ and PELSEQ, respectively.

Results

cDNA Cloning of LM609

[0073] Starting from LM609 expressing hybridoma cells, cDNAs encoding λ chain Fd fragments and entire κ chains were cloned by PCR. The PCR products were cloned into the phagemid pComb3H, which is derived from pComb3, and engineered to facilitate the expression of Fab on the surface of M13 filamentous phage. Phage displaying LM609 Fab were selected by panning against integrin α_Vβ₃and the corresponding cDNA sequences were determined. Soluble LM609 Fab purified from E. coli was analyzed and found to bind specifically to integrin α_Vβ₃by ELISA.

[0074] The approach for the sequential humanization of LM609 by a combination of guided selection and CDR grafting is illustrated in FIG. 1. For the human light chain selection, the mouse Fd fragment is substituted by a chimeric Fd fragment composed of mouse V_Hlinked to human C_H1 to stabilize the hybrid Fab of the first selection step by the interaction of two matching human constant domains, C_κ and C_H1. A stabilization of the hybrid Fab also stabilizes the antigen binding site.

[0075] The guided selection is started by substituting the κ light chain of LM609 with a human κ and λ light chain library that contained the grafted LCDR3 loop of LM609. The corresponding phage libraries displaying hybrid Fab are then selected by four rounds of panning against immobilized integrin α_Vβ₃. Though the output number does not increase from round to round, analysis of the output phage pool from each round for binding to integrin α_Vβ₃by phage ELISA reveals an increasing signal. After the fourth round, phage are produced from clones and tested for binding to integrin α_Vβ₃by phage ELISA.

[0076] While the majority of these clones give signals that reveal some binding above background, in the present case, six clones gave very strong signals (See FIGS. 3a through 3e). DNA sequence analysis of these clones revealed 3 different light chain sequences. Two light chain sequences found in five out of six positive clones differ only in four amino acids, i.e., are 96 percent identical. The third light chain sequence shares about 80 percent identity with the other two. However, this sequence had two parts, each of which could be aligned to germ-lines of different V_κ families, and, thus is deemed likely to have arisen from PCR cross-over, an artifact that has been reported to occur frequently in the amplification antibody sequences.

[0077] Referring to FIGS. 3a through 3e, analysis of the six human V_κ revealed two groups of highly related κ chain sequences. In addition, the CDR1 loops of all six selected human V_κ, which are believed to play a role in the assembly of V_κ and V_λ, resemble the corresponding region of LM609 V_κ. This indicates that the template V_λ of LM609 together with the LM609 antigen, the human integrin α_Vβ₃, selected for human κ chains that are related to the LM609 κ chain. The fact that no repeated sequences were found may indicate that the grafted LCDR3 loop of LM609, which is identical in all selected human κ chains, is mainly responsible for the contribution of the LM609 κ chain to antigen binding.

[0078] This supposition is supported by two additional observations. First, the initial humanization approach was based on the original human κ chain libraries. Four rounds of panning selected a repeated human κ chain with a sequence related to the LM609 κ chain. However, the corresponding hybrid Fab appeared to bind only weakly to human integrin α_Vβ₃. Therefore, the LCDR3 loop of LM609 was grafted in the human κ chain libraries, and, though only roughly estimated from ELISA, the binding of the corresponding selected hybrid Fab to human integrin α_Vβ₃improved.

[0079] Second, the selected phage were selected by two further rounds of panning against immobilized human integrin α_Vβ₃. Again, soluble ones derived from selected phagemids were analyzed for binding to immobilized human integrin α_Vβ₃by ELISA. This time, all of the analyzed 20 clones were found to bind specifically. However, sequencing of 16 clones revealed no repeated sequences. It appears, thus, that a number of different human κ chain sequences that contain the LCDR3 loop of LM609 can substitute the LM609 κ chain without much difference in binding to human integrin α_Vβ₃. This finding is believed to be of importance for the therapeutic application of humanized LM609.

[0080] Due to allotypic sequence variability, humanized antibodies can be neutralized by the patient's immune system after repeated injections. This problem is avoided by using humanized antibodies with identical antigen binding properties but different amino acid sequences for repeated administrations.

[0081] As with the original LM609 light chain, the selected light chains are each κ light chains. Moreover, database screening revealed that they were derived from the same germ-line, namely DPK-26, belonging to the V_κ6 family. This speaks in favor of a strong selection towards these sequences because the V_κ6 family is not frequently found in human antibodies. An obvious reason for this strong selection is a relatively high sequence similarity of the selected human light chains with the original mouse light chain.

[0082] Referring to FIG. 4, for comparison, four clones from the un-selected library were picked randomly and their light chain sequences determined. Three selected human light chains used in the comparison consist of eleven LCDR1 amino acids, the length of the original mouse LCDR1, while only one out of four unselected human light chains shared the same LCDR1 length with the original mouse sequence (see FIG. 4).

[0083] Moreover, both LCDR1 and LCDR2 of the selected human light chains are highly similar to the corresponding mouse sequence. The C-terminal amino acid of framework region 2 of the original mouse light chain sequence, a lysine (Lys) shown in brackets in FIG. 4, is an unusual amino acid at this position and, thus may be involved in the formation of the antigen binding site. Interestingly, all the selected human light chain sequences contain a lysine at this position, while all the unselected sequences contain a tyrosine (Tyr) instead. As a matter of fact, the V_κ6 family is the only human V_κ family that contains a lysine at that position.

[0084] Taken together, this evidence shows that the mouse V_Htemplate and the antigen selected for unbiased human V_κ sequences are related to the original mouse V_κ sequence. Three clones from the light chain selection, revealing weaker binding to integrin α_Vβ₃than the six clones discussed above but still having significant binding above background, were also analyzed by DNA sequencing. These analyses revealed three unrelated V_λ sequences, together with selected V_κ sequences-except the one that stemmed from the PCR cross-over artefact—were used as templates in the humanization of the heavy chain of LM609.

[0085] Based on the aforedescribed humanization strategy, five humanized versions of the anti-human integrin α_Vβ₃monoclonal antibody LM609 were generated. The five version were revealed through the sequence analysis of 14 humanized clones that bind to α_Vβ₃. Referring to FIGS. 8a and 8b, the amino acid sequences of a mouse V_L(SEQ ID NO:45) are compared to the amino acid sequences of the version A, and the combined versions (or groups) B, C, D, and E.

[0086] Similarly FIGS. 8c through 8e compare the amino acid sequences of a mouse V_H(SEQ ID NO:56) to the amino acid sequences of the five versions (or groups) A–E. Four of these versions, represented by clones 7, 8, and 22 (group B); 4, 31, and 36 (group C); 24, 34, and 40 (group D); and 2 (group E), are highly related in amino acid sequence. The sequence group BCDE in FIGS. 8a and 8b represent the four versions that share an identical V_Ldomain (SEQ ID NO:49). The amino acid sequence identity of their V_Hdomains (SEQ ID NOS:50–53), which are all derived from the germ-line DP-65 or the highly related DP-78, is at least 90% for each version.

[0087] In contrast, version five includes clones 10, 11, and 37 (A), and represents a humanized version with a V_Hdomain (SEQ ID NO:54) that is derived from a different germ-line family. This humanized version also contains a different V_Ldomain (SEQ ID NO:55), which is, however, 95% identical and derived from the same germ-line. Germ-lines were determined by nucleic acid sequence alignment using DNAPLOT software provided by the VBASE Directory of Human V Gene Sequences from the MRC Centre for Protein Engineering.

[0088] By preserving the original complementarity determining region sequences such as the LCDR3 (SEQ ID NO:2) and HCDR3 (SEQ ID NO:1) sequences of LM609, the disclosed humanization strategy ensures epitope conservation. Epitope conservation is a critical demand in the humanization of antibodies, especially in the case of LM609. The function-blocking anti-human integrin α_Vβ₃mouse monoclonal antibody LM609 binds to a yet unidentified nonlinear epitope that involves both the α_Vand β₃polypeptide chains. Importantly, by binding to this epitope LM609 induces apoptosis in α_Vβ₃expressing vascular cells, a unique feature among a number of anti-human integrin α_Vβ₃mouse monoclonal antibodies. LM609 does not recognize the related human integrin α_IIbβ₃. Any cross-reactivity with human α_IIbβ₃, which is expressed on platelets, precludes the use of LM609 as a tool in cancer therapy.

[0089] The five humanized versions of LM609, clones 2, 4, 7, 11, and 24, which had been selected by binding to immobilized, thus potentially denatured human integrin α_Vβ₃, were tested for binding to native human integrin α_Vβ₃expressed on the cell surface. For this, binding of humanized LM609 to untransfected CS-1 hamster cells and CS-1 hamster cells transfected with either human β₃or β₅cDNA was analyzed by flow cytometry. Like mouse LM609, but in contrast to a control antibody, all five humanized versions of LM609 revealed specific binding to CS-1 hamster cells transfected with human β₃cDNA (See FIGS. 5a–5f).

[0090] Potential cross-reactivity of humanized LM609 with human integrin α_IIbβ₃was analyzed by ELISA. While antibody Fab-9 with known cross-reactivity bound to both immobilized human integrin α_Vβ₃and α_IIbβ₃, cross-reactivity was detected neither for mouse LM609 nor its five humanized versions (See FIG. 6).

[0091] Thus, all evidence speaks in favor of epitope conservation through the process of humanization of LM609.

[0092] Affinity maturation is a highly relevant step in engineering antibodies for therapeutic applications. By increasing the target affinity, the in vivo concentration of an antibody that must be reached to be effective for therapy is lowered. In addition to reducing the costs of antibody therapy, low effective in vivo concentrations will help to reduce the chance of immune response.

[0093] The CDR walking strategy for the affinity maturation of antibodies has been described elsewhere, and is known in the relevant art. For the affinity maturation of humanized LM609 a sequential optimization of LCDR3 and HCDR3 was chosen (See FIG. 7). The randomized region in both CDRs was confined to a stretch of four amino acids that revealed highest variability in human antibody sequences. See Kabat, E. A. et al. (1991) SEQUENCES OF PROTEINS OF IMMUNOLOGICAL INTEREST, U.S. Dept of Health and Human Services, Washington, D.C. Using NNK doping (Barbas, C. F. et al., (1994) IN VITRO EVOLUTION OF A NEUTRALIZING HUMAN ANTIBODY TO HIV-1 TO ENHANCE AFFINITY AND BROADEN STRAIN CROSS-REACTIVITY, Proc. Natl. Acad. Sci. USA 91, 3809–3813) the randomization of four codons results in 32⁴or 1×10⁶different nucleotide sequences. Based on this and assuming a Poisson distribution (Clackson, T., and Wells, J. A. (1994) IN VITRO SELECTION FROM PROTEIN AND PEPTIDE LIBRARIES, Trends in Biotechnol. 12, 173–184), 5×10⁶independent transformants are required for a complete library with 99 percent confidence. For each of the five humanized versions of LM609 independent LCDR3 randomized libraries were generated. The number of independent transformants in each library lay well above 5×10⁶(see Experimental Procedures, below). Ten completely different LCDR3 sequences were obtained when two clones of each of the five libraries were analyzed by DNA sequencing. The three libraries that were based on humanized LM609 clones 2, 4, and 24, which contain identical light chains and highly related heavy chains derived from germ-line DP-65, were pooled. Corresponding to their underlying germ-line sequences the three remaining libraries were named DPK-26ranLCDR3/DP-10, DPK-26ranLCDR3/DP-65, and DPK-26ranLCDR3/DP-78. These three libraries were selected in parallel.

[0094] To improve the selection of higher affinity LCDR3 mutants, a solid phase off-rate selection (Yang, W. P. et al. (1995) CDR WALKING MUTAGENESIS FOR THE AFFINITY MATURATION OF A POTENT HUMAN ANTI-HIV-1 ANTIBODY INTO THE PICOMOLAR RANGE, J. Mol Biol. 254, 392–403) was used. In five subsequent cycles of selection 20 μg LM609 IgG was added to the well with 200 or 50 ng immobilized human integrin α_Vβ₃following phage incubation and washing. After 24 hours at room temperature, the well was washed again and bound phage eluted with trypsin. This off-rate selection step is discussed in the following paragraph.

[0095] Protein interactions are characterized by thermodynamic and kinetic parameters. While the affinity constant (K_a=k_on/k_off) is an equilibrium constant, the association (k_on) and dissociation (k_off) rate constants are more relevant to in vivo processes which are beyond equilibrium. See Williams, A. F. (1991) OUT OF EQUILIBRIUM, Nature 352, 473–474. In fact, to occur in vivo, interactions with high affinity, i.e., high K_avalues, still rely on rapid association, i.e., high k_onvalues. Antibodies are subject to kinetic selection based on binding target antigens rapidly, in parallel with thermodynamic selection for high affinity binding in order to allow sufficient time for antigen clearance. See Foote, J., and Milstein, C. (1991) KINETIC MATURATION OF AN IMMUNE RESPONSE, Nature 352, 530–532. A typical antibody/antigen interaction with a K_avalue in the range of 10⁹M⁻¹associates rapidly with a k_onvalue in the range of 10⁵to 10⁶M⁻¹s⁻¹and dissociates slowly with a k_offvalue in the range of 10⁻³to 10⁻⁴s⁻¹. An off-rate selection for affinity maturation, i.e., decreasing k_off, requires consideration of the half-life of the antibody/antigen interactions that is given by t_1/2=1n2/k_off. An antibody/antigen interaction with k_off=1×10⁻⁴s⁻¹has a half-life of about 2 hours. A tenfold lower dissociation constant, i.e., k_off=1×10⁻⁵s⁻¹, results in a tenfold longer half-life, i.e., about 20 hours. These long half-lives limit the off-rate selection in our affinity maturation protocol. Using a reasonable time frame, antibodies with dissociation constants below 1×10⁻⁶s⁻¹can not be enriched even after multiple selection cycles. However, using a similar protocol, an antibody was selected against gp120 with a k_offvalue in the range of 10⁻⁶s⁻¹. The corresponding affinity constant was in the range of 10¹¹M⁻¹, a more than 400-fold improvement of the parental antibody.

[0096] Eight clones from each of the three independently selected libraries were analyzed by DNA sequencing. The LCDR3 sequences are shown in Table 1, below.

[0097]

TABLE 1

Selected LCDR3 Mutants
Kabat position¹	91	92	93	94

LM609	Ser	Asn	Ser	Trp
Library DPK-	Ser	Gln	Trp	Trp
26ranLCDR3/DP-10²	Ser	Gln	Trp	Trp
	Ser	Gln	Trp	Trp
	Ser	Gln	Trp	Trp
	Ser	Gln	Val	Trp
	Ser	Gln	Phe	Trp
	Ser	Gln	Phe	His
	Ser	Gln	Phe	His
Library DPK-	Ser	Gln	Phe	Trp
26ranLCDR3/DP-65³	Ser	Gln	Phe	Trp
	Ser	Gln	Phe	Trp
	Ser	Gln	Phe	Trp
	Ser	Gln	Phe	Trp
	Ser	Gln	Phe	Trp
	Ser	Gln	Phe	Trp
	Ser	Gln	Phe	Trp
Library DPK-	Ser	Gln	Trp	Trp
26ranLCDR3/DP-78⁴	Ser	Gln	Trp	Trp
	Ser	Gln	Val	Trp
	Ser	Gln	Val	Trp
	Ser	Gln	Phe	Trp
	Ser	Gln	Phe	Trp
	Ser	Gln	Tyr	Trp
	Ser	Gln	His	Trp

¹cf. Kabat, et al. (1991)
²based on humanized LM609 clone 11
³based on humanized LM609 clones 2, 4, and 24
⁴based on humanized LM609 clone 7

[0098] A strong selection towards a consensus sequence that is highly related to the original sequence took place. All 24 analyzed clones contain a serine (Ser) in Position 91 and a glutamine (Gln) in Position 92 of the randomized region. The corresponding amino acids in the parental LCDR3 are serine (Ser) and asparagine (Asn), respectively. Interestingly, all three serine codons of the NNK genetic code (TCT, TCG, and ACT) are found in Position 91. Position 94, a tryptophane (Trp) in the parental LCDR3, was re-selected in 22 out of 24 clones. Two clones contain a histidine (His) instead. Only Position 93 reveals greater diversity. The original serine (Ser) is substituted by an aromatic or hydrophobic amino acid, either phenylalanine (Phe-13/24), tryptophane (Trp-6/24), valine (Val-3/24), tyrosine (Tyr-1/24) or histidine (His-1/24). Analysis of the heavy chain sequences revealed that no cross-contamination between the three independently selected libraries took place. All eight clones selected from library DPK-26ranLCDR3/DP-65, which contained the pool of the three highly related heavy chains encoding sequences derived from germ-line DP-65, were identical and derived from humanized LM609 clone 24.

[0099] The conserved LCDR3 sequence speaks in favor of a highly defined epitope on human integrin α_Vβ₃. Though binding to native human integrin α_Vβ₃on the cell surface needs to be proved yet, an epitope shift towards denatured human integrin α_Vβ₃is unlikely. The selected phage pools were analyzed for binding to human integrin α_Vβ₃by phage ELISA and in competition with LM609 IgG. These analyses suggest a significantly lower dissociation constant of the selected clones in comparison with LM609 as well as humanized LM609. The substitution of the original serine in Position 3 by an aromatic residue may give rise to a new hydrophobic interaction with a strong impact on the overall affinity.

Experimental Procedures

Materials

[0100] LM609 IgG purified from hybridoma cultures was provided by Dr. David A. Cheresh. Human integrin α_Vβ₃and human integrin α_IIbβ₃were from sources described in HIGH-AFFINITY SELF-REACTIVE HUMAN ANTIBODIES BY DESIGN AND SELECTION: TARGETING THE INTEGRIN LIGAND BINDING SITE, Barbas III, C. F., et al. (1993) Proc. Natl. Acad. Sci. USA, 90, 10003–10007. Untransfected CS-1 hamster cells and CS-1 hamster cells transfected with either human β₃or β₅cDNA were obtained from Dr. David A. Cheresh (Filardo, E. J., et al. (1995) REQUIREMENT OF THE NPXY MOTIF IN THE INTEGRIN β₃SUBUNIT CYTOPLASMIC TAIL FOR MELANOMA CELL MIGRATION IN VITRO AND IN VIVO, J. Cell Biol. 130, 441–450) and maintained in RPMI medium supplemented with 5% fetal calf serum (FCS) at 37° C. in 7% CO₂.

E. Coli Expression of Soluble Humanized LM609 Fab

[0101] Following phage library panning, the SfiI insert of the selected humanized LM609 phagemid pool was cloned into the E. coli expression vector pPhoA-H6HA (See Rader, C., and Barbas III, C. F. (1997) PHAGE DISPLAY OF COMBINATORIAL ANTIBODY LIBRARIES, Curr. Opin. Biotechnol. 8, 503–508) for detection of α_Vβ₃binders. Sequence determination of 14 clones revealed five different humanized LM609 versions, represented by clones 2, 4, 7, 11, and 24. cDNAs of these clones were cut out by SacI/SpeI digestion and ligated into SacI/NheI cut pComb3H, thereby removing the gene III fragment encoding cDNA of pComb3H and allowing for production of soluble Fab (See Rader, C. (1997) Curr. Opin. Biotechnol. 8, 503–508). The ligation products were electrotrasformed into E. coli strain XL1-Blue. Fab production was induced by addition of isopropyl β-D-thiogalactopyransoside as described (Barbas III, C. F., et al. (1991) ASSEMBLY OF COMBINATORIAL ANTIBODY LIBRARIES ON PHAGE SURFACES: THE GENE III SITE, Proc. Natl. Acad. Sci. USA 88, 7978–7982).

ELISA

[0102] Human integrins α_Vβ₃and α_IIbβ₃were coated for 90 minutes at 37° C. on a 96-well plate (Costar #3690) at a concentration of 60 ng/25 μl metal buffer (25 mM Tris-HCl, pH 7.5; 137 mM NaCl; 1 mM KCl; 1 mM MgCl ₂; 1 mM CaCl₂; 1 mM MnCl₂) per well. Following blocking with 150 μl 3% BSA/TBS for 1 hour at 37° C., 25 μl crude supernatants from overnight cultures of E. coli strain XL1-Blue expressing soluble LM609 or humanized LM609 Fab were added to the well and incubated for 2 hours at 37° C. Binding of each of the supernatants to wells coated with human integrin α_Vβ₃and α_IIbβ₃as well as to uncoated but blocked wells was analyzed in triplicates. As a positive control, 25 μl of 50 ng/μl purified Fab-9, an antibody binding to both human integrin α_Vβ₃and α_IIbβ₃(Barbas III, et al. (1993) Proc. Natl. Acad. Sci. USA 90, 10003–10007), and as a negative control, 25 μl of plain bacterial culture medium were used. After extensive washing with tap water, 25 μl of a 1:2,000 dilution of goat anti-mouse F(ab′)₂or goat anti-human F(ab′)₂conjugated with alkaline phosphatase (Pierce #31324 or 31312, respectively) in 1% BSA/TBS was added to the well and incubated for 1 hour at 37° C. Following another extensive washing with tap water, 50 μl of alkaline phosphatase substrate (5 mg disodium p-nitrophenyl phosphate hexahydrate (Sigma #104–105) dissolved in 5 ml 10% diethanolamine, 1 mM MgCl₂, 3 mM NaN₃, pH 9.8) were added to the well. The plate was analyzed with an ELISA reader (Molecular Devices) after 15 minutes incubation at room temperature.

Flow Cytometry

[0103] Flow cytometry was performed using a Becton Dickinson FACScan instrument. For each determination, 5×10³untransfected hamster CS-1 cells or hamster CS-1 cells expressing either human β₃or β₅, were analyzed. Indirect immunofluorescence staining was achieved with crude lysates of E. coli strain XL1-Blue expressing soluble humanized LM609 Fab or, as a negative control, an unrelated human Fab. A 1:100 dilution of FITC-conjugated goat anti-human F(ab′)₂(Jackson #109-096-097) was used for detection.

Construction of LCDR3 Libraries

[0104] Humanized LM609 clones 2, 4, 7, 11, and 24 in pPhoA-H6HA were separately utilized as templates for overlap extension PCR mutagenesis as described (Barbas III, et al. (1994) IN VITRO EVOLUTION OF A NEUTRALIZING HUMAN ANTIBODY TO HIV-1 TO ENHANCE AFFINITY AND BROADEN STRAIN CROSS-REACTIVITY, Proc. Natl. Acad. Sci. USA 91, 3809–3813). The two fragments required for this procedure were obtained with the PCR primer pairs OMPSEQ (SEQ ID NO:39)/CR320 (SEQ ID NO:46) and CR520 (SEQ ID NO:47)/DPSEQ (SEQ ID NO:48), respectively. The resulting five cDNAs with randomized LCDR3 were cut with SfiI, ligated into the appropriately digested phagemid vector pComb3H, and electrotransformed into E. coli strain ER 2537. Two clones of each of the five libraries were analyzed by DNA sequencing and revealed correct assembly as well as 10 different LCDR3 sequences. Prior to selection, libraries based on clones 2, 4, and 24 (V_Lgerm-line DP-26; V_Hgerm-line DP-65) were combined to give a complexity of 6×10⁷independent transformants. Libraries based on clone 11 (V_Lgerm-line DP-26; V_Hgerm-line DP-10) and 7 (V_Lgerm-line DP-26; V_Hgerm-line DP-78) were kept separate with a complexity of 3×10⁷and 4×10⁷independent transformants, respectively.

Selection of LCDR3 Libraries

[0105] The three LCDR3 libraries were separately selected by panning against immobilized integrin α_Vβ₃for six cycles. Panning was performed substantially as described hereinabove for the LM609 humanization. The concentration of human integrin α_Vβ₃for coating was 200 ng/25 μl in the first through fourth cycles and 50 ng/25 μl in the fifth and sixth cycles. Also, the input number of phage, in the range of 10¹²in the first through fourth cycles as usual, was decreased by a factor of 10 in the fifth cycle and by a factor of 100 in the sixth cycle. In the second through the sixth cycles of selection 20 μg LM609 IgG in 50 μl metal buffer (25 mM Tris-HCl, pH 7.5; 137 mM NaCl; 1 mM KCl; 1 mM MgCl₂; 1 mM CaCl₂; 1 mM MnCl₂) was added to the well following removal of unbound phage by five to ten washing steps. The plate was then incubated for 24 hours at room temperature (off-rate selection) prior to five additional washing steps and trypsin elution as described. After the sixth cycle phage were produced from single clones and tested for binding to immobilized human integrin α_Vβ₃by phage ELISA using a sheep anti-M13 antibody conjugated with horseradish peroxidase (Pharmacia #27-9411-01) as a secondary antibody. Light chain and heavy chain encoding sequences of positive clones were analyzed by DNA sequencing using the primers OMPSEQ (SEQ ID NO:39) and PELSEQ (SEQ ID NO:24), respectively.

Optimization of LM609 by CDR Randomization

[0106] In addition to the humanization, LM609 can be optimized in two respects: First, by increasing the affinity to α_Vβ₃and, second, by broadening the species cross-reactivity. Increased affinity of engineered LM609 increases the potency and decreases the cost of a potential cancer therapy.

[0107] The original mouse monoclonal antibody LM609 already has a broad species cross-reactivity. It binds to human, dog, cat, bovine, rabbit, and chick but not mouse α_Vβ₃. The fact that LM609 does not recognize host α_Vβ₃in the mouse models of human cancer is a major concern for the therapeutic applicability of LM609. Engineered LM609 binding to both human and mouse α_Vβ₃would be an important tool towards clinical trials. In vitro methods for the improvement of monoclonal antibody affinity include chain shuffling (See Marks, J. D., et al. (1992) BY-PASSING IMMUNIZATION: BUILDING HIGH HUMAN ANTIBODIES BY CHAIN SHUFFLING, Bio/Technology 10, 779–783). Binding to α_Vβ₃can be further improved by subsequent CDR randomization, an approach termed CDR walking (See Barbas III, C. F., et al. (1994) IN VITRO EVOLUTION OF A NEUTRALIZING HUMAN ANTIBODY TO HIV-1 TO ENHANCE AFFINITY AND BROADEN STRAIN CROSS-REACTIVITY, Proc. Natl. Acad. Sci. USA 91, 3809–3813). The in vitro strategies for humanization and affinity improvement of LM609 are likely to generate cross-reactivity with mouse α_Vβ₃concurrently. Directed selection for mouse α_Vβ₃recognition is complicated by the fact that mouse α_Vβ₃has not been purified yet. However, several mouse cell lines, e.g., NIH/3T3, are known to express α_Vβ₃and, thus, may be included in the screening procedure.

[0108] The foregoing discussion and the accompanying examples are presented as illustrative, and are not to be taken as limiting. Still other variations within the spirit and scope of this invention are possible and will readily present themselves to those skilled in the art.

Claims

Certificate of Correction present

The claims shown below may not include correction changes. Use the corrected PDF for authoritative text.

We claim:

1. A method of producing a humanized mouse monoclonal antibody which consists of the steps:

(a) constructing a first human antibody heavy or light chain library consisting of antibody heavy chains or antibody light chains in which each chain in the first human antibody light or heavy chain library has three complementarity determining region (CDR) loops in which the complementarity determining region three (CDR3) loop in each light or heavy chain has the amino acid sequence of a corresponding mouse antibody heavy or light chain CDR3 loop and is flanked by unaltered human framework residues;

(b) constructing a second human antibody heavy or light chain library consisting of antibody heavy chains or antibody light chains, in which each chain in the second human antibody heavy or light chain library has three complementarity determining region (CDR) loops in which the complementarity determining region three (CDR3) loop in each light or heavy chain has the amino acid sequence of a corresponding mouse antibody heavy or light chain CDR3 loop and is flanked by unaltered human framework residues; wherein the chains of said second human antibody heavy or light chain library are the complementary heavy or light chain of said first human antibody heavy or light chain library chains, such that one library is a light chain library and the other library is a heavy chain library;

(c) creating a library of heavy and light chain pairs by combining chains from said first human antibody heavy or light chain library of step (a) with a complementary chain from an antibody which binds a preselected antigen forming a library of heavy and light chain pairs;

(d) isolating antigen binding chains of step (a) by selecting from the library of step (c), a heavy and light chain pair which binds to said preselected antigen, isolating from the antigen binding heavy and light chain pair, the antigen binding chains of step (a);

(e) creating a humanized pair library by combining isolated antigen binding chains of step (d), with chains of said library of step (b), so that a first isolated chain of step (d), together with a complementary second chain in said library of step (b) combine to form a heavy and light chain humanized pair library;

(f) selecting from the humanized pair library of step (e) a humanized heavy and light chain pair that binds to said preselected antigen; and

(g) combining two of said selected humanized heavy and light chain pair of step (f) to form a whole antibody; wherein said humanized mouse monoclonal antibody binds the same antigen as the mouse monoclonal antibody.

2. The method of claim 1 wherein the first human antibody heavy or light chain library contains only light chains and the complementary chain is a heavy chain.

3. The method of claim 1 wherein the first human antibody heavy or light chain library contains only heavy chains and the complementary chain is a light chain.

4. The method of claim 1 wherein the heavy chain is a Fd fragment.

5. A method of producing a humanized mouse monoclonal antibody heavy and light chain pair which consists of the steps:

(a) constructing a human light chain library wherein each light chain of said light chain library has three complementarity determining region (CDR) loops in which the complementarity determining region three (CDR3) loop in each light chain has the amino acid sequence of a corresponding mouse antibody light chain CDR3 loop and is flanked by unaltered human framework residues;

(b) selecting a light chain from the light chain library of step (a), wherein the selection comprises combining a heavy chain from an antibody that binds a preselected antigen with a light chain from the library of step (a) to form a heavy and light chain pair library, screening said pair library for binding to said preselected antigen, and isolating light chains that bind to said preselected antigen;

(c) constructing a library of human heavy chains wherein each heavy chain of said library of human heavy chains has three CDR loops in which the complementarity determining region three (CDR3) loop in each heavy chain has the amino acid sequence of a corresponding mouse antibody heavy chain CDR3 loop and is flanked by unaltered human framework residues;

(d) combining a heavy chain from the heavy chain library of step (c) with a selected light chain of step (b) to produce a second library of humanized mouse monoclonal antibody heavy and light chain pairs;

(e) isolating an antigen binding humanized mouse monoclonal antibody heavy and light chain pair from the library of step (d) by screening said library of step (d) for binding with said preselected antigen and isolating those humanized mouse monoclonal antibody heavy and light chain pairs that bind said preselected antigen; wherein said preselected antigen is the same antigen recognized by the mouse monoclonal antibody.

6. The method of claim 5 further comprising converting the selected heavy and light chain pair to a whole antibody.

7. The method of claim 5 wherein the heavy chain of step (b) is a Fd fragment.

8. The method of claim 7 wherein the heavy chain Fd is a humanized mouse heavy chain fragment or a template mouse heavy chain fragment.

9. The method of claim 5 wherein in step (b) a humanized mouse heavy chain is used.

10. The method of claim 5 wherein the CDR3 loop in each light and heavy chain is a CDR3 loop from antibody LM609 produced from hybridoma cell line deposited with American Type Culture Collection under ATCC Accession No. HB 9537.

11. The method of claim 7 wherein the heavy chain Fd is a humanized mouse heavy chain.

12. The method of claim 7 wherein the heavy chain Fd is a template mouse heavy chain fragment.

Patent Citations (2)

Patent	Date	Inventor	Cited By
US5565332(A)	1996-10-01	Hoogenboom et al.	Examiner
US5859205(A)	1999-01-01	Adair et al.	Examiner

Non-Patent Literature (3)

Paul, Fundamental Immunology, Raven Press, NY, Chapter 8, p. 242, 1993.Examiner
Brooks et al J. Clin. Invest. 96:1815-1822, 1995.Examiner
Rader et al. Proc. Natl. Acad. Sci. USA, 95:8910-8915, Jul. 1998.Examiner