Monoclonal antibodies to fibroblast growth factor receptor 2

Abstract

To provide methods for producing hybridoma cells, the hybridoma cells, and method for producing FGFR2IIIb antibodies.SOLUTION: A method for producing hybridoma cells that produce antibodies is provided, the method including the steps of: administering a fusion protein containing the extracellular domain of FGFR2IIIb fused to a human Fc domain to a rodent; extracting spleen cells or lymph node cells from the rodent; and fusing the spleen cells or lymph node cells with an immortalized cell line to produce hybridomas.SELECTED DRAWING: Figure 9

Description

[Technical Field]

[0001] Cross-Reference to Related Applications This application claims the benefit under 35 U.S.C. §119(e) of U.S. patent application Ser. No. 61 / 112,686, filed Nov. 7, 2008, and U.S. patent application Ser. No. 61 / 164,870, filed Mar. 30, 2009, which are incorporated herein in their entirety for all purposes.

[0002] Statement of Government Interest The invention described in this application was made in part with funding provided by the National Institutes of Health under Grant 5R44 CA 101283-03. The U.S. Government has certain rights in this invention.

[0003] The present invention relates generally to the combination of monoclonal antibody (mAb) and recombinant DNA technology to develop novel biologics, and more particularly to the production of monoclonal antibodies that bind to and neutralize, for example, fibroblast growth factor receptor 2.

[0004] Background of the Invention The fibroblast growth factor (FGF) family has 22 known members, ranging in size from 17 to 34 kDa and sharing a similar internal core region, which can be classified into seven subfamilies based on activity and sequence similarity (Ornitz et al., Genome Biol. 2:3005.1, 2001). For example, the FGF1 subgroup consists of multiple prototype FGFs, FGF1 (acidic FGF) and FGF2 (basic FGF); the FGF4 subgroup consists of FGF4, FGF5, and FGF6; and the FGF7 subfamily consists of FGF3, FGF7, FGF10, and FGF22 (Zhang et al., J. Biol. Chem. 281:15694, 2006).

[0005] One form of FGF2 is an 18-kDa nonglycosylated polypeptide consisting of 146 amino acids derived from a 155-amino acid precursor (Ornitz et al., Genome Biol. 2:3005.1, 2001; Okada-Ban et al., Int. J. Biochem. Cell. Biol. 32:263, 2000). An exemplary sequence of human 146-amino acid FGF2 is provided in SEQ ID NO: 4 of US20020115603. Unlike most other FGFs, FGF2 does not encode a signal sequence for secretion, but the 18-kDa form can be secreted by a nonconventional energy-dependent pathway independent of the ER-Golgi complex. Another member of the FGF1 subfamily is FGF1 itself, which has a similar size and structure to FGF2 and, in addition, lacks a signal sequence but is capable of secretion. Another FGF of interest here is FGF7 (also called keratinocyte growth factor (KGF)), which is produced by cells of mesenchymal origin and stimulates epithelial cell proliferation (Finch et al., Adv. Cancer Res. 91:69, 2004; Finch et al., J. Natl. Cancer Inst. 98:812, 2006). KGF is expressed in many organs, including the lung, prostate, breast, gastrointestinal tract, and skin, and is involved in organ development and skin injury repair (Cho et al., Am. J. Pathol. 170:1964, 2007).

[0006] Members of the FGF family bind to only four known tyrosine kinase receptors, fibroblast growth factor receptors 1-4 (FGFR1-4) and their isoforms, and various FGFs bind to different FGFRs to varying degrees (Zhang et al., J. Biol. Chem. 281:15694, 2006). The protein sequence of human FGFR2 is provided, for example, in GenBank Locus AF487553. Each FGFR consists of an extracellular domain (ECD) containing three immunoglobulin (Ig)-like domains (D1, D2, and D3), a single transmembrane helix, and an intracellular catalytic kinase domain, as illustrated in Figure 1 (Mohammadi et al., Cytokine Growth Factor Revs, 16:107, 2005). A region of consecutive acidic amino acids is located in the linker between D1 and D2, termed the "acid box" (AB). The region containing D1 and AB is thought to be involved in receptor autoinhibition and is exposed upon ligand binding. FGFRs are characterized by multiple alternative splicing of their mRNAs, resulting in various isoforms (Ornitz et al., J. Biol. Chem. 271:15292, 1996; see also Swiss-Prot P21802 and isoforms P21802-1 to -20 for the sequences of FGFR2 and its isoforms). In particular, there are forms containing all three Ig domains (α isoforms) and forms containing only two Ig domains (D2 and D3 domains, excluding D1) (β isoforms). Of particular importance for FGFR1-FGFR3 is that all forms contain the first half of D3, representing exons IIIa, while two exons containing the second half of D3 result in forms IIIb and IIIc. In the case of FGFR2, these refer to FGFR2IIIb and FGFR2IIIc (or FGFR2b and FGFR2c), respectively, and the corresponding beta forms are referred to as FGFR2(beta)IIIb and FGFR2(beta)IIIc.The FGFR2IIIb type (also designated K-sam-II {Swiss-Prot calls it K-sam-IIC1?}) of FGFR2 (see Swiss-Prot P21802-18) is a high-affinity receptor for FGF1 and KGF, whereas FGFR2IIIc (also designated K-sam-I) (see Swiss-Prot P21820-5) binds well to FGF1 and FGF2 but not to KGF (Miki et al., Proc. Natl. Acad. Sci. USA 89:246, 1992). In fact, FGFR2IIIb is also called KGFR because it is the only receptor for KGF (Ornitz et al., 1996, op. cit.).

[0007] FGFRs and their isoforms are differentially expressed in various tissues. In particular, FGFR2IIIb (and the IIIb forms of FGFR1 and FGFR3) are expressed in epithelial tissues, whereas FGFRIIIc is expressed in mesenchymal tissues (Duan et al., J. Biol. Chem. 267:16076, 1992; Ornitz et al., 1996, op. cit.). Certain FGF ligands for these receptors show the opposite pattern of expression. Thus, members of the FGF3 subfamily, including FGF7 (KGF), bind only to FGFRIIIb (Zhang et al., op. cit.) and are expressed in mesenchymal tissues, suggesting that they may be paracrine effectors of epithelial cells (Ornitz et al., 1996, op. cit.). Conversely, FGF4-6, a member of the FGF4 subfamily, binds to FGFR2IIIc and is expressed in both epithelial and mesenchymal lineages, and may therefore have autocrine or paracrine functions. Regarding the expression patterns of FGFR2 isoforms and their ligands, FGFR2 plays a role in epithelial-mesenchymal interactions (Finch et al., Dev. Dyn. 203: 223, 1995), so it is not surprising that knockout of FGFR2IIIb in mice leads to severe embryonic abnormalities and lethality (De Moerlooze et al., Development 127:483, 2000).

[0008] In addition to binding to FGFR1-4 with high affinity, FGFs also bind to heparin sulfate proteoglycans (HSPGs) with lower affinity. Indeed, FGF binding to heparin / heparin sulfate (HS) on the cell surface requires signaling through FGFRs. The interaction of FGFs (especially FGF2) with FGFRs and heparin has been extensively studied by X-ray crystallography and mutational analysis, and it is now believed that heparin / HS is involved in the formation of a symmetric 2:2 FGF-FGFR dimer (Mohammadi et al., 2005), which induces receptor activation, autophosphorylation, and signal transduction.

[0009] FGFs mediate various responses in various cell types, including proliferation, migration, and differentiation (particularly during embryonic development) (Ornitz et al., op. cit.), and in adults, they are involved in tissue homeostasis and repair. For example, FGF2 stimulates proliferation (i.e., is mitogenic) of certain cells, including fibroblasts and endothelial cells, and is an anti-apoptotic survival factor for certain cells (e.g., neurons). FGF2 stimulates differentiation (morphogenesis) and migration (motility) of endothelial cells (Dow et al., Urology 55:800, 2000). Some FGFs, particularly FGF1 and FGF2, are potent angiogenic factors (Presta et al., Cytokine and Growth Factor Rev. 16:159, 2005).

[0010] The importance of the FGF system in development has been reinforced by the discovery of numerous mutations in FGFR1-3 associated with human congenital bone disorders, including craniosynostosis syndromes (premature fusion of cranial sutures) (Wilkie et al., Cytokine Growth Factor Revs 16:187, 2005). These genetic disorders are usually dominant because the associated mutations result in a gain of function, often by promoting receptor dimerization. In particular, the severe craniosynostosis disorder Apert syndrome (AS) is associated with one of two mutations in the conserved D2-D3 linker region of FGFR2 (Ser-252 → Trp or Pro-253 → Arg), which increases ligand binding affinity (Ibrahimi et al., Proc. Natl. Acad. Sci USA 98:7182, 2001).

[0011] FGF2 and other FGFs have been reported to contribute to cancer by directly stimulating both angiogenesis and tumor cell proliferation (Grose et al., Cytokine Growth Factor Revs. 16:179, 2005; Presta et al., op cit.). FGFR2IIIb RNA is expressed in many types of tumors (Finch et al., J. Natl. Cancer Inst. 98:812, 2006), mostly as a result of its expression in corresponding normal tissues (Orr-Urtreger et al., Dev. Biol. 158:475, 1993). KGF (FGF7) and KGFR (FGFR2IIIb) are overexpressed in many pancreatic cancers (Ishiwata et al., Am. J. Pathol. 153: 213, 1998), and their coexpression correlates with poor prognosis (Cho et al., Am. J. Pathol. 170: 1964, 2007). Somatic mutations in the FGFR2 gene were found in 12% of a large panel of endometrial (uterine) carcinomas, and in some cases, the mutations were required for tumor cell survival (Dutt et al., Proc. Natl. Acad. Sci. USA 105:8713, 2008). In two tumors, the FGFR2 mutation was found to be identical to the S252W substitution associated with Apert syndrome. Amplification and overexpression of FGFR2 are strongly associated with undifferentiated and diffuse-type gastric cancer, which is associated with a particularly poor prognosis. Inhibition of FGFR2 activity by small molecule compounds significantly inhibits the proliferation of such cancer cells (Kunii et al., Cancer Res. 68:2340, 2008; Nakamura et al., Gastroenterol. 131:1530, 2006).FGFR2IIIb RNA was expressed in 16 / 20 epithelial ovarian cancers (EOC) but not in normal ovarian surface epithelium (Steele et al., Oncogene 20:5878, 2001); and FGFR2IIIb ligands FGF1, FGF7, and FGF10 induced proliferation, motility, and protection from cell death in EOC cell lines (Steele et al., Growth Factors 24:45, 2006), suggesting that FGFR2IIIb may be involved in the malignant phenotype of ovarian cancer.

[0012] Only a limited number of monoclonal antibodies against FGFR2 have been reported. Fortin et al. (J. Neurosci. 25:7470, 2005) reported a blocking antibody against FGFR2, and Wei et al. (Hybridoma 25:115, 2006) developed two mAbs specific to type IIIb of FGFR2 (i.e., KGFR) that inhibited KGF-induced cell proliferation. Yayon et al. (WO2007 / 144893, 2006) disclosed an inhibitory mAb that binds to both FGFR2 and FGFR3. R&D Systems has marketed an anti-FGFR2 mAb since 2005 that neutralizes activity in their assay, favoring type IIIb. However, there have been no reports on the antitumor activity of antibodies against FGFR2 in vivo. Summary of the Invention

[0013] In one embodiment, the present invention provides a monoclonal antibody (mAb) against human fibroblast growth factor receptor 2 (FGFR2) that inhibits the growth of human tumor xenografts in mice. The mAb can inhibit at least one, and preferably some or all, of the receptor's biological activities (including binding of FGF2 to the receptor). The mAb can bind to either or both of the FGFR2IIIb and FGFRIIIc receptors (e.g., bind to FGFR2IIIb but not to FGFR2IIIc). Preferably, the mAb of the present invention is genetically engineered (e.g., chimeric, humanized, or human). Exemplary antibodies are GAL-FR21, GAL-FR22, and GAL-FR23 and their chimeric and humanized forms, and mAbs that share the same epitope as or compete for binding with one of these mAbs. In another embodiment, a pharmaceutical composition is provided comprising a genetically engineered anti-FGFR2 antibody (e.g., chimeric or humanized GAL-FR21, GAL-FR22, and GAL-FR23). In a third embodiment, the pharmaceutical composition is administered to a patient to treat cancer or other diseases, such as gastric cancer.

[0014] Exemplary humanized antibodies have: a humanized light chain with CDRs from the sequence in Figure 13A (GAL-FR21) and a humanized heavy chain with CDRs from the sequence in Figure 13B (GAL-FR21), or a humanized light chain with CDRs from the sequence in Figure 16A (GAL-FR22) and a humanized heavy chain with CDRs from the sequence in Figure 16B (GAL-FR22). Some humanized antibodies have the three light chain CDRs shown in Figure 13A (GAL-FR21) and the three heavy chain CDRs shown in Figure 13B (GAL-FR21), or the three light chain CDRs shown in Figure 16A (GAL-FR22) and the three heavy chain CDRs shown in Figure 16B (GAL-FR22). Optionally, the light chain variable region has at least 95% sequence identity to the sequence shown in Figure 13A (HuGAL-FR21), and the heavy chain variable region has at least 95% sequence identity to the sequence shown in Figure 13B (HuGAL-FR21). In some such antibodies, residues H27, H28, H30, H48, and H67 according to the Kabat numbering are occupied by residues occupying the corresponding positions in the heavy chain shown in Figure 13B (HuGAL-FR21). A preferred humanized antibody comprises a light chain variable region having the sequence shown in Figure 13A (HuGAL-FR21) and a heavy chain variable region having the sequence shown in Figure 13B (HuGAL-FR21). [Brief explanation of the drawings]

[0015] [Figure 1] Figure 1 is a schematic diagram of the structure of FGFR2, showing the three Ig-like domains (D1, D2, and D3), the transmembrane domain (black box), and the intracellular kinase domain. The heparin-binding site (HBS), acid box (AB), and partial domains IIIb and IIIc are shown. N = amino terminus, C = carboxy terminus. [Figure 2] FIG. 2 is a summary of the properties of the anti-FGFR2 mAbs GAL-FR21, GAL-FG22, and GAL-FR23 described in the Examples. [Figure 3] FIG. 3 shows the binding ELISA of mAbs GAL-FR21, GAL-FR22, and GAL-FR23 against FGFR2IIIb, and the negative control mLGG. [Figure 4] Figure 4 shows the binding ELISA results of mAbs GAL-FR21, GAL-FR22, and GAL-FR23, and the negative control mouse mAb 5G8, against four types of FGFR2, which are fusion proteins with Fc (FGFR2IIIb, FGFR2(beta)IIIb, FGFR2(IIIc), and FGFR2(beta)IIIc). Each mAb was used at a fixed concentration in the assay. [Figure 5] Figure 5 shows competitive binding ELISA results for FGFR2IIIb using biotinylated mAbs GAL-FR21, GAL-FR22, and GAL-FR23 and the negative control mouse mAb 5G8, respectively. The ratio of biotinylated to unlabeled mAb was 100:1. [Figure 6] FIG. 6 shows flow cytometry analysis of the binding of mAbs GAL-FR21, GAL-FR22, and GAL-FR23 to FGFR2IIIb on SNU-16 and KATO III cells, as well as a negative control mAb. [Figure 7] FIG. 7 shows flow cytometry of mAbs GAL-FR21, GAL-FR22, and GAL-FR23 and a negative control mAb binding to 293F cells transfected with FGFR2IIIc or FGFR2IIIb (S252W). [Figure 8A] FIG. 8(A) is an ELISA assay measuring the inhibition of binding of FGF1 (upper panel) and FGF2 (lower panel) to FGFR2IIIb by mAbs GAL-FR21, GAL-FR22, and GAL-FR23. [Figure 8B] FIG. 8(B) is an ELISA assay measuring the inhibition of binding of FGF7 (upper panel) and FGF10 (lower panel) to FGFR2IIIb by mAbs GAL-FR21 and GAL-FR22. [Figure 9]Figure 9 shows the growth of SNU-16 human gastric tumor xenografts in mice treated with PBS alone, GAL-FR21, GAL-FR23, or FR2bC 54.8.11 (upper panels), or PBS, GAL-FR22, or FR2bC 54.8.11 (lower panels). The mAbs were administered at 20 μg twice weekly to approximately five mice per group. [Figure 10] FIG. 10 shows the growth of SNU-16 (A, upper panel) or OCUM-2M (B, lower panel) human gastric tumor xenografts in mice treated with PBS alone, GAL-FR21, or GAL-FR22. [Figure 11] FIG. 11 shows the binding ELISA of GAL-FR21 (upper panel) or GAL-FR22 (lower panel) to human and mouse FGFR2IIIb. [Figure 12] FIG. 12 shows the binding ELISA of GAL-FR21 (upper panel) or GAL-FR22 (lower panel) to human and cynomolgus monkey FGFR2IIIb. [Figure 13] Figure 13 shows the amino acid sequences of the mature variable regions of the light chain (A) and heavy chain (B) of HuGAL-FR21, along with the mouse GAL-FR21 and human acceptor V regions. CDRs in the GAL-FR21 sequence are underlined, and amino acids substituted by mouse amino acids in the HuGAL-FR21 sequence are double underlined. The one-letter amino acid code and Kabat numbering system are used for both the light and heavy chains. [Figure 14] Figure 14 shows the amino acid sequences of the light chain (A) and heavy chain (B) of the fully matured HuGAL-FR21 antibody. The first amino acid of each line is numbered consecutively. In the light chain, the first amino acid of the Ck region is underlined, and in the heavy chain, the first amino acids of the CH1, hinge, CH2, and CH3 regions are underlined. [Figure 15]FIG. 15 shows competitive binding of humanized HuGAL-FR21 and mouse GAL-FR21 mAbs with the control human antibody hlgG, performed as described in the specification. [Figure 16] Figure 16 shows the mature variable regions of the light chain (A) and heavy chain (B) of HuGAL-FR22, with the CDRs underlined. The one-letter amino acid code and Kabat numbering system are used for both the light and heavy chains. DETAILED DESCRIPTION OF THE INVENTION

[0016] The present invention provides anti-FGFR2 monoclonal antibodies (mAbs) that inhibit the biological activity of FGFR2 and / or inhibit the growth of FGFR2-expressing tumor xenografts in mice, pharmaceutical compositions containing such mAbs, and methods of using them to treat diseases.

[0017] 1. Antibodies Antibodies are very large, complex molecules (molecular weight of ∼150,000 or approximately 1320 amino acids) with complex internal structures. Natural antibody molecules contain two identical pairs of polypeptide chains, each pair having one light and one heavy chain. Each light and heavy chain consists of two regions: a variable ("V") region, which is involved in binding to a target antigen, and a constant ("C") region, which interacts with other components of the immune system. The variable regions of the light and heavy chains fit together in three dimensions to form the variable region that binds to antigens (e.g., cell surface receptors). Within each light or heavy chain variable region, there are three short segments (averaging 10 amino acids in length) called complementarity-determining regions ("CDRs"). The six CDRs in an antibody variable domain (three from the light chain and three from the heavy chain) fold together in three dimensions to form the actual antibody binding site that binds to the target antigen. The location and length of the CDRs are precisely defined by Kabat, E. et al., Sequences of Proteins of Immunological Interest, US Department of Health and Human Services, 1983, 1987. The portions of the variable regions not included in the CDRs are called the framework and form the environment for the CDRs.

[0018] A humanized antibody is a genetically engineered antibody in which CDRs from a mouse antibody (the "donor antibody," which may be a rat, hamster, or other non-human species) are grafted onto a human antibody (the "acceptor antibody"). The acceptor antibody sequence may be, for example, the sequence of a mature human antibody, a consensus sequence of human antibody sequences, or a germline region sequence. Thus, a humanized antibody is an antibody having CDRs from the donor antibody and variable region framework and constant regions from a human antibody. In addition, at least one of two additional structural elements may be employed to maintain high binding affinity. U.S. Patent Nos. 5,530,101 and 5,585,089 (incorporated herein by reference) provide detailed instructions for constructing humanized antibodies. Humanized antibodies often incorporate all six CDRs from a murine antibody (preferably as defined by Kabat, or other definitions, e.g., Chothia), but can also be generated with less than the complete CDRs from a murine antibody (e.g., Pascalis et al., J. Immunol. 169:3076, 2002); Vajdos et al., Journal of Molecular Biology, 320: 415-428 (2002); Iwahashi et al., Mol. Immunol. 36:1079- 1091, (1999); Tamura et al., Journal of Immunology, 164:1432-1441 (2000).

[0019] Similarly, it may be necessary to incorporate only a portion of the CDRs (i.e., a subset of CDR residues required for binding, referred to as SDRs) into the humanized antibody. CDR residues that do not contact antigen and are not in the SDRs can be identified based on previous studies (e.g., residues H60-H65 of CDRH2 are often not required), from regions of the Kabat CDRs located outside the Chothia hypervariable loops (Chothia, J. Mol. Biol. 196:901, 1987), by molecular modeling and / or experience, or as described in Gonzales et al., Mol. Immunol. 41:863, 2004. In humanized antibodies having positions where one or more donor CDR residues are missing, the amino acids occupying those positions may be the amino acids occupying the corresponding positions (according to the Kabat numbering) in the acceptor antibody sequence. The number of substitutions included reflects a balance of several competing considerations. Such substitutions are potentially advantageous in that they reduce the number of mouse amino acids present in the humanized antibody, thereby reducing potential immunogenicity. However, substitutions may result in changes in affinity, and significant decreases in affinity are preferably avoided. The positions of substitutions in the CDRs and the amino acids to be substituted can also be selected empirically.

[0020] Thus, typically, a humanized antibody has (i) a light chain comprising CDRs (often three CDRs) derived from a mouse antibody (e.g., GAL-FR21), a human variable region framework, and a human constant region; and (ii) a heavy chain comprising CDRs (often three CDRs) derived from a mouse antibody (e.g., GAL-FR21), a human variable region framework, and a human constant region. The light and heavy chain variable region frameworks may be sequences of a mature human antibody, a consensus sequence of human antibody sequences, or germline regions, respectively.

[0021] In the first structural element, the framework of the heavy chain variable region of the humanized antibody is selected by appropriately selecting an acceptor antibody from among many known human antibodies so as to have maximum sequence identity (between 65% and 95%) with the framework of the heavy chain variable region of the donor antibody. In the second structural element, during construction of the humanized antibody, selected amino acids in the framework of the human acceptor antibody (outside the CDRs) are replaced with corresponding amino acids from the donor antibody according to predetermined rules. Specifically, the amino acids to be substituted in the framework are selected based on their ability to interact with the CDRs. For example, the substituted amino acids can be adjacent to the CDRs in the donor antibody sequence or can be adjacent to the CDRs in the humanized antibody within 4-6 angstroms measured in three-dimensional space.

[0022] Chimeric antibodies, which combine the variable regions of a mouse (or other rodent) antibody with the constant regions of a human antibody, are well known for their genetic engineering construction. While retaining the binding specificity of the mouse antibody, these antibodies are approximately two-thirds human. The proportion of non-human sequences present in mouse, chimeric, and humanized antibodies indicates that the immunogenicity of chimeric antibodies is intermediate between that of mouse and humanized antibodies. Other types of genetically engineered antibodies that may have reduced immunogenicity compared to murine antibodies include human antibodies produced using phage display technology (Dower et al., WO91 / 17271; McCafferty et al., WO92 / 001047; Winter, WO92 / 20791; and Winter, FEBS Lett. 23:92, 1998, each of which is incorporated herein by reference) or using transgenic animals (Lonberg et al., WO93 / 12227; Kucherlapati WO91 / 10741, each of which is incorporated herein by reference).

[0023] As used herein, the term "human-like" antibody refers to a mAb in which a significant portion of the amino acid sequence (e.g., greater than about 50%) of one or both chains originates from human immunoglobulin genes. Thus, human-like antibodies include, but are not limited to, chimeric, humanized, and human antibodies. As used herein, a "reduced immunogenic" mAb is one that is expected to be significantly less immunogenic than a murine antibody when administered to a human patient. Such antibodies include chimeric, humanized, and human mAbs, as well as mAbs created by substituting specific amino acids in a murine antibody that may contribute to B-cell or T-cell epitopes (e.g., exposed residues) (Padlan, Mol. Immunol. 28:489, 1991). As used herein, a "genetically engineered" mAb is one whose genes have been constructed or placed into a non-native environment (e.g., a human gene in a mouse or on a bacteriophage) with the aid of recombinant DNA technology. Thus, for example, murine mAbs produced by conventional hybridoma technology would not be included.

[0024] Other methods for engineering humanized antibodies can be used to achieve the same results as those described in the aforementioned U.S. Patent Nos. 5,530,101 and 5,585,089 (e.g., "superhumanization" (see Tan et al. J. Immunol. 169: 1119, 2002, and U.S. Patent No. 6,881,557)) or the method of Studnicak et al., Protein Eng. 7:805, 1994. Additionally, other methods for producing genetically engineered mAbs with reduced immunogenicity include "reshaping," "hyperchimerization," and veneering / resurfacing (e.g., those described in Vaswami et al., Annals of Allergy, Asthma and Immunology 81:105, 1998; Roguska et al. Protein Eng. 9:895, 1996, and U.S. Patent Nos. 6,072,035 and 5,639,641).

[0025] The epitope of a mAb is the region of its antigen to which the mAb binds. Two antibodies bind to the same or overlapping epitope if they competitively inhibit (block) the binding of one to the antigen. That is, one antibody in 1x, 5x, 10x, 20x, or 100x excess inhibits binding of the other by at least 50%, but preferably 75%, 90%, or even 99% (as measured in a competitive binding assay (see, e.g., Junghans et al., Cancer Res. 50:1495, 1990)). Alternatively, two antibodies have the same epitope if substantially all amino acid mutations in the antigen that reduce or eliminate binding of one antibody also reduce or eliminate binding of the other. Two antibodies have overlapping epitopes if some amino acid mutations that reduce or eliminate binding of one antibody also reduce or eliminate binding of the other.

[0026] 2. Anti-FGFR2 antibody A monoclonal antibody (mAb) that binds to FGFR2 (i.e., an anti-FGFR2 mAb) is said to neutralize (or inhibit or be an antagonist of) FGFR2 if such binding partially or completely inhibits one or more biological activities of FGFR2. The biological activity of FGFR2 that a neutralizing antibody can inhibit or block is the ability of FGFR2 to bind to one or more or all FGF ligands (e.g., FGF1 and / or FGF2). In the case of FGFRIIIb, these ligands include FGF1, FGF7 (KGF), and other members of the FGF7 subfamily, FGF3, FGF10, and FGF22. In the case of FGFRIIIc, these ligands include FGF1 and FGF2; FGF4 and other members of the FGF4 subfamily, FGF5 and FGF6; FGF8 and other members of the FGF8 subfamily, FGF17 and FGF18; and FGF9 and other members of the FGF9 subfamily, FGF16 and FGF20. Another important activity of FGFR2 that can be inhibited by neutralizing anti-FGFR2 mAbs is the stimulation of cell proliferation (e.g., epithelial or endothelial cells, fibroblasts, cells such as Ba / F3 cells transfected with FGFR2, and various human tumor cells). Other activities that can be inhibited by neutralizing anti-FGFR2 mAbs include the stimulation of cell (e.g., endothelial cell) differentiation and migration and the induction of angiogenesis, as measured, for example, by stimulating the proliferation or tube formation of human vascular endothelial cells (HUVECs) or by inducing blood vessels when performed on chick embryo chorioallantoic membrane (CAM). Typically, neutralizing mAbs inhibit these activities induced by one or more of the FGFs listed above. Similarly, the mAbs preferably inhibit all or part of the signal transduction pathways (e.g., phosphorylation of FGFR2 and downstream MAP kinases) stimulated by binding of FGF ligands to FGFR2 (Dailey et al., Cytokine Growth Factor Revs 16:233, 2005).

[0027] For example, a neutralizing mAb of the present invention at a concentration of 0.01, 0.1, 0.5, 1, 2, 5, 10, 20, or 50 μg / ml inhibits the biological function of FGFR2 by at least about 50%, but preferably 75%, more preferably 90%, 95%, or even 99%, and most preferably nearly 100% (substantially completely or indistinguishable from a negative control lacking FGFR2) when assayed by the methods described in the Examples or known in the art. Typically, the degree of inhibition is measured when the amount of FGF ligand used is just sufficient to fully stimulate biological activity, or when the amount is 1, 2, or 5 μg / ml, or 0.01, 0.02, 0.05, 0.1, 0.5, 1, 3, or 10 μg / ml. Preferably, the mAb is neutralizing (i.e., inhibits biological activity) when used as a single agent, although, optionally, two mAbs can be used together to achieve inhibition. Most preferably, the mAb neutralizes not just one, but two, three, or several of the above biological activities. For purposes herein, an anti-FGFR2 mAb used as a single agent is said to "fully neutralize" all biological activities of FGFR2, and such mAbs are most preferred.

[0028] The present invention provides neutralizing mAbs that bind to FGFR2IIIb but bind poorly or not detectably to FGFRIIIc, or alternatively, bind to FGFRIIIc but bind poorly or not detectably to FGFR2IIIb, or, as a third option, bind to both FGFR2IIIb and FGFR2IIIc. The present invention also provides the use of any of these types of antibodies in pharmaceutical compositions, particularly pharmaceutical compositions for treating cancer or other diseases. The present invention provides mAbs (either neutralizing or non-neutralizing) that bind to one or more forms of FGFR2 and inhibit, preferably completely, the growth of FGFR2-expressing tumor xenografts (e.g., SNU-16 or OCUM-2M xenografts). Such mAbs may inhibit tumor growth, for example, by transmitting negative growth signals or pro-apoptotic signals through FGFR2. The mAbs of the invention are preferably specific for or preferentially bind to FGFR2, i.e., do not bind or bind to a lesser extent (e.g., at least 10-fold less) to proteins related to FGFR2 (e.g., other FGF receptors FGFR1, FGFR3, and FGFR4, or other membrane receptor tyrosine kinases). On the other hand, in some cases, mAbs that bind to one or more other FGF receptors in addition to FGFR2 are preferred. The mAbs of the invention typically bind to at least 10-fold less FGF receptors. 7 M -1 , but preferably 10 8 M -1 or more, and most preferably 10 9 M -1 More than or equal to 10 10 M -1 The binding affinity to FGFR2 (association constant K a) mAbs that exhibit differential or preferential binding to one form of FGFR or FGFR2 over another preferably exhibit at least a 5-, 10-, or 100-fold preference between the forms, as measured, for example, by Ka. Lack of binding between an antibody and an antigen (i.e., the antibody does not bind to the antigen) means that any signal from an attempted binding reaction between the two is indistinguishable from that from a negative control (e.g., no antibody or antigen, or one substituted with an inactive substance).

[0029] Some mAbs of the invention bind to both human and mouse FGFR2, or to human FGFR2 and one, two, more, or all mouse, rat, rabbit, chicken, dog, and / or monkey (e.g., cynomolgus) FGFR2. In some cases, the mAbs bind to human FGFR2 with affinities (i.e., K) within 2, 10, or 100-fold of their affinity for human FGFR2. a Similarly, the mAb may bind to cynomolgus monkey and / or chimpanzee FGFR2 (e.g., FGFR2IIIb) with an affinity that is within 2, 10, or 100-fold of the affinity for human FGFR2, or with an affinity that is substantially the same as or indistinguishable (i.e., within experimental error) from the binding to human FGFR2. Other mAbs are specific only for human FGFR2.

[0030] The mAbs of the present invention include anti-FGFR2 antibodies that are native tetramers (two light chains and two heavy chains) and may be of any of the known isoforms IgG, IgA, IgM, IgD, and IgE and their subtypes (i.e., human IgG1, IgG2, IgG3, and IgG4 and mouse IgG1, IgG2a, IgG2b, and IgG3). mAbs of the present invention also include antibody fragments (e.g., Fv, Fab, and F(ab')2), bifunctional hybrid antibodies (e.g., Lanzavecchia et al., Eur. J. Immunol. 17:105, 1987), single-chain antibodies (Huston et al., Proc. Natl. Acad. Sci. USA 85:5879, 1988; Bird et al., Science 242:423, 1988), single-arm antibodies (Nguyen et al., Cancer Gene Ther. 10:840, 2003), and antibodies with altered constant regions (e.g., U.S. Pat. No. 5,624,821). mAbs may be of animal origin (e.g., mouse, rat, hamster, or chicken) or may be genetically engineered. Rodent mAbs are produced by standard methods. This method involves multiple immunizations with FGF2 in a suitable adjuvant, either ip, iv, or intra-footpad, followed by extraction of spleen or lymph node cells, fusion with a suitable immortalized cell line, and selection of hybridomas producing antibodies that bind to FGF2 (see, e.g., Examples). Chimeric and humanized mAbs produced by methods known in the art as described above are preferred embodiments of the present invention. Human antibodies produced, for example, by phage display or transgenic mouse technology are also suitable (see, e.g., Dower et al., McCafferty et al., Winter, Lonberg et al., Kucherlapati, supra).

[0031] The anti-FGFR2 mAbs GAL-FR21, GAL-FR22, and GAL-FR23 described below are illustrative of the present invention. Once a single, prototypical anti-human FGFR2 mAb (e.g., GAL-FR21) that neutralizes FGFR2 and has the desired properties described herein is isolated, other mAbs with similar properties (e.g., having the same epitope) can be readily generated using known methods. For example, hybridomas can be generated as described above, and mice can be immunized with FGFR2 to obtain mAbs by screening for their ability to compete with the prototypical mAb for binding to FGFR2. Mice can also be immunized with a smaller FGFR2 fragment that contains the epitope bound by the prototypical mAb. The epitope can be identified, for example, by screening for binding to a series of overlapping peptides spanning FGFR2. Alternatively, the method described by Jespers et al., Biotechnology 12:899, 1994, can be used to guide the selection of mAbs that share the same epitope and, therefore, have properties similar to those of a prototypical mAb (e.g., GAL-FR21). Using phage display, the heavy chain of the prototypical antibody is first paired with a population of (preferably human) light chains to select for FGFR2-binding mAbs. The novel light chain is then paired with a population of (preferably human) heavy chains to select for (preferably human) FGFR2-binding mAbs that share the same epitope as the prototypical mAb. Alternatively, for example, GAL-FR21 variants can be obtained by mutagenesis of cDNAs encoding the heavy and light chains of GAL-FR21.

[0032] mAbs having the same or overlapping epitopes as GAL-FR21, GAL-FG22, or GAL-FR23 (e.g., mAbs that compete with each other in binding to FGFR2) are provided as other examples of the present invention. Chimeric or humanized forms of GAL-FR21, GAL-FG22, or GAL-FR23 are particularly preferred embodiments. Also included in the present invention are mAbs that are 90%, 95%, or 99% identical to the amino acid sequence of the heavy and / or light chain variable regions of GAL-FR21, GAL-FG22, or GAL-FR23 (sequences excluding signal sequences) and retain their functional properties, and / or mAbs that are distinguished by a small number of functionally insignificant amino acid substitutions (e.g., conservative substitutions), deletions, or insertions. Also included are mAbs having at least one, and preferably all six, CDRs that are 90%, 95%, 99%, or 100% identical to the corresponding CDRs of GAL-FR21, GAL-FG22, or GAL-FR23. Here, as elsewhere in this application, percent sequence identity is determined with maximally aligned antibody sequences according to the Kabat numbering convention. After alignment, when comparing a region of a subject antibody (e.g., the entire mature variable region of a heavy or light chain) with the same region of a reference antibody, the percent sequence identity between the region of the subject antibody and the reference antibody region is converted to a percentage by dividing the number of positions occupied by the same amino acid in both the region of the subject antibody and the reference antibody region by the total number of aligned positions in the two regions (not counting gaps) and multiplying by 100.

[0033] For the purposes of classifying amino acid substitutions as conservative or non-conservative, amino acids may be grouped as follows: Group I (hydrophobic side chains): met, ala, val, leu, ile; Group II (neutral hydrophilic side chains): cys, ser, thr; Group III (acidic side chains): asp, glu; Group IV (basic side chains): asn, gln, his, lys, arg; Group V (residues that affect chain conformation): gly, pro; Group VI (aromatic side chains): trp, tyr, phe. Conservative substitutions involve substitutions between amino acids in the same group. Non-conservative substitutions consist of exchanging a member of one of the above groups for a member of another group.

[0034] Natural mAbs of the present invention may be produced from hybridomas. Genetically engineered mAbs (e.g., chimeric or humanized mAbs) may be expressed by various methods known in the art. For example, genes encoding their light and heavy chain V regions may be synthesized from overlapping oligonucleotides and inserted, along with available C regions, into an expression vector (e.g., commercially available from Invitrogen) that provides the necessary regulatory regions (e.g., promoter, enhancer, polyA site, etc.). The use of a CMV promoter-enhancer is preferred. Expression vectors may be transfected into various mammalian cell lines (e.g., CHO or non-producing myelomas, including Sp2 / 0 and NS0) using various well-known methods (e.g., lipofection or electroporation), and antibody-expressing cells selected by selection with an appropriate antibiotic. See U.S. Patent No. 5,530,101. Larger quantities of antibody may also be produced by growing cells in commercially available bioreactors.

[0035] Once expressed, the mAb or other antibodies of the invention may be purified according to standard technical methods (e.g., microfiltration, ultrafiltration, protein A or G affinity chromatography, size exclusion chromatography, anion exchange chromatography, cation exchange chromatography, and / or affinity chromatography based on organic dyes or the like). Substantially pure antibodies having at least about 90 or 95% w / w homogeneity are preferred, with 98% or 99% w / w or more homogeneity being most preferred for pharmaceutical uses.

[0036] 3. Treatment Method The present invention provides therapeutic methods in which mAbs of the invention (e.g., anti-FGF2 MAbs) are administered to patients suffering from a disease (therapeutic treatment) or at risk for developing or recurring a disease (prophylactic treatment). The term "patient" includes human patients, veterinary patients (e.g., cats, dogs, and horses), livestock (e.g., cattle, sheep, and pigs), and laboratory animals used for testing purposes (e.g., mice and rats). The above methods are particularly suited to treating human patients. The mAbs used in the methods for treating human patients bind to the human FGFR2 protein (sequence provided by GenBank locus AF487553). Citations for other FGFRs or FGFs referenced in this disclosure are provided in the Background section. mAbs against human proteins can also be used in other species whose species homologs have antigenic cross-reactivity with the human protein. In species lacking such cross-reactivity, antibodies are used with appropriate specificity for species homologs present in that species. However, in xenotransplantation experiments in laboratory animals, mAbs with specificity against human proteins expressed by the xenograft are commonly used.

[0037] In a preferred embodiment, the present invention provides a pharmaceutical formulation comprising an antibody described herein. The pharmaceutical formulation comprises a mAb in a physiologically acceptable carrier. The formulation may optionally contain excipients or stabilizers and may be in the form of a lyophilized or aqueous solution. Acceptable carriers, excipients, or stabilizers are non-toxic to recipients at the dosages and concentrations employed and may include buffers (e.g., phosphate, citrate, or acetate, typically at pH 5.0-8.0, most often pH 6.0-7.0), salts (e.g., sodium chloride, potassium chloride, etc., for isotonicity), antioxidants, preservatives, low molecular weight polypeptides, proteins, hydrophilic polymers (e.g., polysorbate 80), amino acids (e.g., glycine), carbohydrates, chelating agents, sugars, and other standard components known to those skilled in the art (see Remington's Pharmaceutical Sciences 16). th edition, Osol, A. Ed. 1980). The mAb is typically at a concentration of 0.1-100 mg / ml (e.g., 1-10 mg / ml or 10-50 mg / ml, e.g., 5, 10, 20, 30, 40, 50, or 60 mg / ml).

[0038] In another preferred embodiment, the present invention provides a method for treating a patient with a disease using a pharmaceutical formulation containing an anti-FGFR2 mAb. The mAb, once formulated in the pharmaceutical formulation, can be administered to the patient by any suitable route, particularly parenteral, intravenous, or bolus administration, intramuscular, or subcutaneous routes. Intravenous administration can be over a period of at least 15 minutes, but more often over 30 minutes, or over 1, 2, or even 3 hours. The mAb can be injected directly into the disease site (e.g., tumor) or encapsulated in a carrier drug (e.g., liposome). The dose administered is sufficient to at least partially alleviate the condition being treated (a "therapeutically effective dose"), optionally between 0.1 and 5 mg / kg of body weight, e.g., 1, 2, 3, or 4 mg / kg, but can also be as much as 0.1 or 1-10 mg / kg, or even between 1-15, 20, or 30 mg / kg. They may be administered in fixed unit doses (e.g., 100, 200, 500, 1000, or 2000 mg), or these doses may be spread over a patient's surface area (e.g., 1000 mg / m 2 ) is administered. Typically, 1 to 8 doses (e.g., 1, 2, 3, 4, 5, 6, 7, or 8) are administered for cancer treatment, although 10, 20, or more doses may be administered. The mAb may be administered daily, twice weekly, weekly, biweekly, or monthly. It may also be administered at some other interval (e.g., 1 week, 2 weeks, 4 weeks, 8 weeks, 3-6 months, or more), depending, for example, on the half-life of the mAb. Repeated courses of treatment are also possible, as is chronic administration.

[0039] A therapeutically effective regimen is a combination of a dose, frequency, and route of administration that is effective to at least partially alleviate the disease from which the patient being treated suffers. A prophylactically effective regimen is a combination of a dose, frequency, and route of administration that is effective to inhibit or delay the onset of the disease in the patient.

[0040] Diseases particularly amenable to treatment with the anti-FGFR2 mAbs of the present invention include solid tumors (e.g., ovarian cancer, endometrial cancer, breast cancer, lung cancer (small cell or non-small cell), colon cancer, prostate cancer, cervical cancer, pancreatic cancer, gastric cancer, esophageal cancer, hepatocellular carcinoma (liver cancer), renal cancer (kidney cancer), head and neck tumors, mesothelioma, melanoma, sarcoma, and brain tumors (e.g., gliomas such as glioblastoma)) that are believed to require angiogenesis or are believed to be associated with detectable, preferably elevated, levels of FGFR2 and / or certain FGFs. Elevated levels can be measured by comparing protein or mRNA levels in cancer tissue with comparable levels of FGFR2 (e.g., FGFR2IIIb) or FGF (e.g., FGF2, FGF7, or FGF10) in normal tissue (equivalent to non-cancerous tissue, preferably from the same patient). Detectable levels can also be determined by measuring protein or mRNA levels in cancer tissue and comparing them to background levels in a control sample (absent the analyte (e.g., FGFR2 or FGF)) or a negative control (detection performed using an antibody, primer, or probe known not to bind to the analyte or nucleic acid encoding the analyte). Leukemia, lymphoma, multiple myeloma, and other hematopoietic tumors (particularly any of these cancers with increased expression of FGFR2 and / or FGF) are amenable to treatment with anti-FGFR2 mAbs. Other angiogenesis-related diseases for treatment with the anti-FGFR2 mAbs of the present invention include age-related macular degeneration (AMD), diabetic retinopathy, neovascular glaucoma, and other eye diseases; psoriasis and other skin diseases; rheumatoid arthritis; and genetic skeletal disorders associated with FGFR2 mutations (e.g., Apert syndrome).

[0041] In a preferred embodiment, the anti-FGFR2 mAb is administered in combination with (i.e., along with, or before, during, or after) another therapy. For example, for the treatment of cancer, an anti-FGFR2 mAb is administered. mAbs can be used in combination with known chemotherapeutic agents, such as alkylating agents (e.g., carmustine, chlorambucil, cisplatin, carboplatin, oxaliplatin, procarbazine, and cyclophosphamide); antimetabolites (e.g., fluorouracil, floxuridine, fludarabine, gemcitabine, methotrexate, and hydroxyurea); natural products, including plant alkaloids and antibiotics (e.g., bleomycin, doxorubicin, daunorubicin, idarubicin, etoposide, mitomycin, mitoxantrone, vinblastine, vincristine, and taxol (paclitaxel) or related compounds (e.g., Taxotere®); the topoisomerase 1 inhibitor irinotecan; agents specifically approved for brain tumors, including temozolomide and Gliadel® wafers, which contain carmustine; and tyrosine kinase inhibitors (e.g., Gleevec®, Sutent®). (sunitinib malate), Nexavar® (sorafenib), and Tarceva® (erlotinib) or Iressa® (gefitinib); angiogenesis inhibitors; and all approved and investigational anti-cancer agents listed in WO 2005 / 017107 A2 (incorporated herein by reference). Anti-FGFR2 mAbs may be used in combination with one, two, three, or more of these other agents used in standard chemotherapy regimens. Typically, the other agents are known to be effective against the specific type of cancer being treated. Anti-FGFR2 mAbs are particularly useful for overcoming resistance to chemotherapeutic agents, thereby increasing their effectiveness.

[0042] Other drugs that can be administered with anti-FGFR2 mAbs to treat cancer include monoclonal antibodies (e.g., Herceptin, directed against the HER2 antigen). TMincluding Avastin® against VEGF; or biologics such as antibodies against the epidermal growth factor (EGF) receptor (e.g., Erbitux® (cetuximab) and Vectibix® (panitumumab)). Antibodies against hepatocyte growth factor (HGF) are particularly preferred for use with anti-FGFR2 mAbs. The anti-FGFR2 mAbs include mAb L2G7 (Kim et al., Clin Cancer Res 12:1292, 2006 and U.S. Patent No. 7,220,410) and particularly its chimeric and humanized forms (e.g., HuL2G7 (WO 07115049 A2); the human anti-HGF mAb described in WO 2005 / 017107 A2 (particularly 2.12.1); and the HGF-binding proteins described in WO 07143090 A2 or WO 07143098 A2; and other neutralizing anti-HGF mAbs that compete for binding with any of the above-mentioned mAbs. mAbs that bind to the cMet receptor of HGF are also suitable (e.g., anti-cMet mAb OA-5D5 (Martens et al., Clin. Cancer Res. 12:6144, (2006) which are genetically engineered to have only one "arm" (i.e., binding domain). Antibodies to other FGFR receptors, FGFR1, 3, 4, or antibodies to various FGFs (e.g., FGF1, FGF2, and FGF7), may also be preferably used in combination with anti-FGFR2 mAbs. Furthermore, anti-FGFR2 mAbs may be used in conjunction with any form of surgery and / or radiation therapy, including external beam radiation therapy, intensity-modulated radiation therapy (IMRT), and any form of radiosurgery (e.g., gamma knife).

[0043] Treatments (e.g., standard chemotherapy) involving anti-FGFR2 mAb antibodies may reduce disease by increasing the median progression-free time or median overall survival of cancer patients by at least 30% or 40%, but preferably 50%, 60% to 70%, or even 100% or more, compared to the same treatment (e.g., chemotherapy) except for the absence of the anti-FGFR2 mAb. Additionally or alternatively, treatments (e.g., standard chemotherapy) involving anti-FGFR2 mAbs may increase the complete remission rate, partial remission rate, or objective remission rate (complete + partial) in patients with these tumors (e.g., ovarian, endometrial, splenic, breast, lung, colon, and glioblastoma, especially when relapsed or refractory) by at least 30% or 40%, but preferably 50%, 60% to 70%, or 100%, compared to the same treatment (e.g., chemotherapy) except for the absence of the anti-FGFR2 mAb.

[0044] Typically, the above-mentioned increases in median time to progression and / or remission rate in patients treated with chemotherapy plus anti-FGF2 mAb compared to a control group of patients receiving chemotherapy alone (or chemotherapy plus placebo) in a clinical trial (e.g., a phase 2, phase 2 / 3, or phase 3 trial) are statistically significant, e.g., at the p = 0.05 or 0.01 or even 0.001 level. Complete and partial remission rates are measured by objective criteria commonly used in cancer clinical trials (e.g., those listed or approved by the National Cancer Institute and / or the Food and Drug Administration).

[0045] 4. Other methods The anti-FGFR2 mAbs of the present invention also find use in diagnostic, prophylactic, and testing methods. They may be used to measure FGFR2 levels in tumors or the circulatory system of tumor patients, which may then be used to monitor or guide tumor treatment. For example, tumors with high levels of FGFR2 (e.g., elevated levels compared with non-cancerous samples from comparable tissues from the same patient) are particularly susceptible to treatment with anti-FGFR2 mAbs. In certain embodiments, the mAbs can be used in ELISA or radioimmunoassays to measure FGFR2 levels (e.g., serum levels) or in immunohistochemistry to localize FGFR2 expression (e.g., in tumor biopsies). The use of two anti-FGFR2 mAbs that bind to different epitopes (i.e., do not compete for binding) is particularly useful for developing sensitive "sandwich" ELISAs for detecting FGFR2. For various assays, the mAbs may be labeled with fluorescent molecules, spin-labeled molecules, enzymes, or radioisotopes and provided in the form of kits containing all the necessary reagents to perform FGFR2 assays. In other applications, anti-FGFR2 mAbs are used to purify FGFR2, for example, by affinity chromatography. [Example]

[0046] Example 1: Reagents and Assays Preparation of Flag-FGF1, FLAG-FGF2, and FLAG-FGF7 The DNA sequences of human FGF1 (155 amino acid form; Chiu et al., Oncogene 5:755-1990) and human FGF2 (155 amino acid form; Sommer et al., Biochem. Biophys. Res. Comm. 144:543, 1987) were synthesized (GenScript, Inc.) and amplified by PCR to contain a Flag peptide tag at the N-terminus. They were then cloned into a derivative of the pET vector (Invitrogen) using standard molecular biology techniques. These plasmids were transformed into E. coli BL21(DE3) cells, and FGF1 or FGF2 expression was induced using 1 mM IPTG. FGF expression levels were measured using FGF1- or FGF2-specific ELISA kits (R&D Systems). FGF was purified using heparin-Sepharose CL-6B beads (Amersham Biosciences) as described by Wiedlocha et al., Mol. Cell. Biol., 16:270, 1996. Similarly, the gene for human FGF7 (a 194-amino acid precursor; Finch, PW et al., Science 245:752, 1989) was synthesized and amplified by PCR in the pCMV vector (a derivative of pDrive; Invitrogen) to contain a Flag tag at the N-terminus. Flag-FGF10 was also produced in a similar manner. Plasmid DNA was transfected into human 293F cells. Culture supernatants of transfected 293F cells were used for ligand-receptor binding assays.

[0047] Preparation of FGFR2 fusion protein The extracellular domains (ECDs) of human FGFR2IIIb and human FGFR2IIIc were expressed as immunoadhesin molecules. For the alpha form, a DNA fragment encoding the complete ECD of FGFR2IIIb (amino acids 1-378) or FGFR2IIIc (amino acids 1-377) was fused to human Fc (residues 216-446) via a polypeptide linker; for the beta form (lacking D1), FGFR2(beta)IIIb (amino acids 152-378) and FGFR2(beta)IIIc (amino acids 152-377) were used instead. These FGFR2-Fc molecules were expressed by the following procedure: 293F cells were transfected, and stable transfectants were selected in 293 expression medium (Invitrogen) containing G418 (1 mg / ml). FGFR2-Fc secreted from 293F transfected cells was purified using a protein A / G column. Similarly, cDNA for cynomolgus monkey (cyno) FGFR2 ECD was cloned from cyno liver mRNA by standard methods, and amino acids 1-378 were fused to human Fc to generate cynoFGFR2IIIb-Fc for expression. Chimpanzee FGFR2IIIb-Fc was constructed using in vitro mutagenesis to convert one amino acid in the human FGFR2IIIb ECD that differs from chimpanzee FGFR2IIIb to a chimpanzee amino acid (methionine at residue 186 to threonine, based on the known sequence in GenBank). Mouse FGFR2(beta)IIIb-Fc protein was purchased from R&D Systems (catalog # 708-MF).

[0048] ELISA assay for mAb binding to FGFR2 fusion protein ELISA plates were coated with goat anti-human IgG-Fc (2 μg / ml) overnight at 4°C. Nonspecific binding sites were then blocked with 2% BSA for 1 hour at RT. The plates were incubated with one of the above FGFR2 fusion proteins (1 μg / ml) for 1 hour. Subsequently, the plates were incubated with various concentrations of mAb or hybridoma culture medium for 1 hour. Bound mAb was detected using HRP-goat anti-mouse antibody. Detection was performed by washing, adding TMB substrate (Sigma), and reading at 450 nm. In all ELISA assays, the plates were washed three times between each step.

[0049] Flow cytometry After washing twice with cell sorting buffer (CSB: PBS / 1% FBS / 0.02% NaN3), 2 × 10 5 Cells were resuspended in 50 μl CSB in microtiter wells and incubated with 50 μl of anti-FGFR2 mAb (1 μg / 50 μl) for 1 hour on ice. Cells were washed twice with CSB, and bound antibody was detected by incubation with FITC-goat anti-mouse IgG (Jackson ImmunoResearch Laboratories) for 1 hour on ice. After washing twice with CSB, cells were analyzed on a FACScan (Becton Dickinson).

[0050] Example 2: Generation of monoclonal antibodies against FGFR2 Balb / c mice (5-6 week old females) were immunized in their hind footpads at weekly intervals with 20 or 22 injections of FGFR2(beta)IIIb-Fc (initial dose 10 μg / footpad, then 5 μg / footpad), or 17 injections of FGFR2IIIc-Fc (initial dose 10 μg / footpad, then 5 μg / footpad), followed by 5 injections of FGFR2(beta)IIIc-Fc (5 μg / footpad) suspended in MPL / TDM (Sigma-Aldrich). Three days after the final injection, popliteal lymphoid cells were extracted and fused with P3 / X63-Ag8U1 mouse myeloma cells at a 1:1 ratio using the Hybridimune electrofusion system (Cyto Pulse Sciences). Hybridomas were supplemented with 2x HAT (Sigma) and selected 24 hours later. Ten days after fusion, hybridoma culture supernatants were screened (using ELISA) for their ability to bind to FGFR2IIIb-Fc but not to human IgG. Selected mAbs were then screened for their ability to recognize FGFR2IIIb on the human gastric tumor cell line SNU-16 (Shin et al., J. Cancer Res. Clin. Oncol. 126: 519, 2000). Selected hybridomas were cloned twice using limiting dilution. Three mAbs were thus selected: GAL-FR21 and GAL-FR22 from the first immunization regimen, and GAL-FR23 from the second immunization regimen. The characteristics of these mAbs are shown in Figure 2, as described further below.

[0051] In addition, many of the other anti-FGFR2 mAbs were derived from multiple fusions (including FR2bB 100.12.9, FR2bC 54.8.11, FR2bC 100.7.9, FR2bC 101.8.2, FR2bC 115.1.5, FR2bC 149.8.8, FR2bB 11.5.3, and FR2bB 18.1.6).

[0052] Example 3: Properties of anti-FGFR2 mAb As shown in Figure 3, all three screened mAbs, GAL-FR21, GAL-FR22, and GAL-FR23, bind well to FGFR2IIIb in the ELISA assay described in Example 1. Using four different forms of FGFR2-Fc in ELISA, it was determined that each of these mAbs exhibits a different binding pattern, indicating that it has a different epitope (Figure 4). GAL-FR21 binds to both the alpha and beta forms of FGFR2IIIb (i.e., with and without D1) but not to FGFRIIIc. Therefore, the epitope must be contained in D3IIIb, not D1. Therefore, the epitope is likely contained in D3IIIb or D3. GAL-FR22 also binds to the alpha and beta forms but not to both the IIIb and IIIc configurations; therefore, the epitope is probably contained in D2-D3IIIa or definitely in D2-D3. Finally, because GAL-FR23 does not bind to any of the beta forms, its epitope must reside entirely or partially in D1. Thus, mAbs with epitopes in either D1 or D2-D3 or D3 are included in the present invention.

[0053] To confirm that mAbs GAL-FR21, GAL-FR22, and GAL-FR23 bind to different epitopes, each mAb was biotinylated in the above ELISA assay (but with the detection reagent HRP-streptavidin). 0.4 μg of biotinylated mAb was competed for binding to FGFR2IIIb-Fc with a 100:1 excess of each unlabeled mAb (or control mouse mAb 5G8). As shown in Figure 5, each mAb competed for binding with itself but not with the other mAbs, indicating that they have different epitopes. In addition, other mAbs FR2bB 100.12.9, FR2bC 54.8.11, FR2bC 100.7.9, FR2bC 101.8.2, FR2bC 115.1.5, and FR2bC 149.8.8 competed for binding with biotinylated GAL-FR21 in this assay, indicating that they have the same or overlapping epitopes as GAL-FR21, whereas mAbs FR2bB 11.5.3 and FR2bB 18.1.6 competed for binding with biotinylated GAL-FR22, indicating that they have the same or overlapping epitopes as GAL-FR22.

[0054] Flow cytometry was performed to confirm that the selected mAbs bound to the appropriate form of FGFR2 on the cell membrane. Binding tests for the appropriate form of the receptor were performed using KATO-III (ATCC HTB-103) and SNU-16 (ATCC CRL-5974) cells, which overexpress FGFR2IIIb. As shown in Figure 6, all three mAbs, GAL-FR21, GAL-FR22, and GAL-FR23, bound to both cell lines, as expected from their epitopes. After confirming that none of the mAbs bound to the host 293F cells themselves, binding tests for the appropriate form were performed using human 293F cells transfected with the FGFRIIIc gene. As shown in Figure 7, GAL-FR22 and GALFR23 bound to FGFRIIIb-transfected cells but not to GALFR21, as expected from their epitopes. Finally, because the S252W mutation of FGFR2 is found in some cancer cells, we tested the binding of the mAbs to 293F cells transfected with an FGFR2IIIb gene (FGFR2IIIb(S252W)) constructed to contain this mutation. As shown in Figure 7, all mAbs bound to FGFR2IIIb(S252W)-transfected cells. The ability to bind to FGFR2IIIb(S252W) is a desirable property of the mAbs of the present invention.

[0055] An ELISA assay was used to measure the ability of mAbs to inhibit FGF ligand binding to FGFR2. ELISA wells were coated with 2 μg / ml goat anti-human IgG-Fc overnight at 4°C. After blocking with 2% BSA for 1 hour at RT, the wells were incubated with 0.5 μg / ml FGFR2IIIb-Fc for 1 hour, followed by incubation with Flag-FGF1 or FLAG-FGF2 (0.2 μg / ml) in the presence of various concentrations of mAb for 1 hour. Bound Flag-FGF1 was detected by adding HRP-anti-Flag M2 antibody (Sigma) and TMB substrate. As can be seen in Figure 8A, mAb GAL-FR21 weakly inhibited FGF1 binding to FGFR2IIIb in this assay, whereas GAL-FR22 and GAL-FR23 did not inhibit FGF1 binding. In contrast, GAL-FR21 strongly inhibited FGF2 binding to FGFR2IIIb, GAL-FR22 moderately inhibited FGF2 binding, but GAL-FR23 did not inhibit binding. In an identical assay using Flag-FGF7 and Flag-FGF10, GAL-FR21 and GAL-FR22 were also shown to inhibit FGF7 and FGF10 binding to FGFR2IIIb (Figure 8B). Indeed, advantageous mAbs of the present invention, such as GAL-FR21 and GAL-FR22, inhibit the binding of FGF2, FGF7, and FGF10 to FGFR2IIIb, preferably by 80%, 90%, 95%, completely, or essentially completely. Thus, it is demonstrated that GAL-FR21 and GAL-FR22, but not GAL-FR23, neutralize at least one biological activity of FGFR2.

[0056] Example 4: Xenograft model Xenograft experiments are performed as previously described (Kim et al., Nature 362:841, 1993). Human tumor cells, typically grown in complete DMEM medium, are harvested in HBSS. Female athymic nude or NIH-III Xid / Beige / nud mice (4-6 weeks old) are inoculated with 2-10 x 10 60.1 ml of HBSS containing the cells was injected subcutaneously on the back of the mouse. 3 Once tumor volume reaches 100 μg, mice are randomized to receive 5 mg / kg (100 μg total) or slightly different doses of mAb administered i.p. twice weekly in 0.1 ml volumes. Tumor size is measured twice weekly by measuring in two dimensions [length (a) and width (b)]. Tumor volume is calculated as V = ab 2 The mean tumor volume is calculated according to the mean t-test (SEM) and expressed as mean tumor volume ± SEM. The number of mice in each treatment group is typically 5-7 mice. Statistical analysis can be performed, for example, using the Student's t-test.

[0057] Figures 9 and 10A show that in various experiments, GAL-FR21, GAL-FR22, and GAL-FR23, all administered twice weekly at a dose level of 20 μg (1 mg / kg), strongly inhibited the growth of SNU-16 gastric tumor xenografts, with GAL-FR21 providing the most potent and complete inhibition of xenograft growth. The aforementioned mAb FR2bC 54.8.11, which competes with GAL-FR21 for binding, also inhibited xenograft growth. Figure 10B shows that GAL-FR21 and GAL-FR22, administered twice weekly at a dose level of 50 μg (2.5 mg / kg), also strongly inhibited the growth of xenografts of the OCUM-2M human gastric tumor cell line (described in Yashiro et al., Jpn J Cancer Res 85:883, 1994). The ability of the mAbs to inhibit xenografts of KATO III or other FGFR2-expressing cell lines is similarly demonstrated. The ability of the mAb to inhibit tumor growth additively or synergistically with other antitumor drugs is demonstrated by treatment of xenografted mice with the mAb alone, the other drug alone, and the mAb in combination with the other drug, demonstrating that treatment with the two drugs has a stronger inhibitory effect than either drug alone.

[0058] Example 5: Binding of mAbs to FGFR2 from other species An ELISA assay was used to measure the ability of mAbs to bind to FGFR2 from species other than human. ELISA wells were coated with 2 μg / ml goat anti-human IgG-Fc overnight at 4°C. After blocking with 2% BSA for 1 hour at RT, the wells were incubated with 0.2 μg / ml FGFR2IIIb-Fc for 1 hour. The FGFRIIIb in the fusion protein was either human, mouse, cynomolgus monkey, or chimpanzee FGFRIIIb. The wells were incubated with various concentrations of GAL-FR21 or GAL-FR22 mAb. Bound mAb was detected by adding HRP-conjugated goat anti-mouse IgG-Fc and TMB substrate. Figure 11 shows that GAL-FR21 binds to mouse FGFR2 almost as well as it binds to human FGFR2 (within 10-fold), while GAL-FR22 binds to mouse FGFR2 moderately (within about 100-fold of human FGFR2). Figure 12 shows that GAL-FR21 binds to cynomolgus monkey FGFR2 similarly as it binds to human FGFR2 (indistinguishable), while GAL-FR22 binds to cynomolgus monkey FGFR2 moderately (within about 100-fold of human FGFR2). Similar experiments using chimpanzee FGFR2 yielded similar results to those using cynomolgus monkey FGFR2: GAL-FR21 binds chimpanzee FGFR2 to the same extent (indistinguishable) as human FGFR2, while GAL-FR22 binds chimpanzee FGFR2 reasonably well (within about 100-fold of human FGFR2). Preferred mAbs of the invention, like GAL-FR21 and GAL-FR22, bind to all of mouse, monkey, chimpanzee, and human FGFR2, and most preferably bind to mouse FGFR2 within 2, 10, 100, or 1000-fold of human FGFR2, and / or bind to monkey and / or chimpanzee FGFR2 within 2, 10, or 100-fold of human FGFR2 or indistinguishable (within experimental variation) from human FGFR2. This is measured, for example, by Ka. Binding to FGFR2 in other species facilitates testing of the mAb in those animal species.

[0059] Example 6: Humanization of GAL-FR21 and GAL-FR22 Cloning of the light and heavy chain variable regions in GAL-FR21 mAb, construction and expression of chimeric mAbs, and design, construction, expression, and purification of humanized GAL-FR21 mAb were all performed using standard molecular biology methods (e.g., as described in US2008019974 for L2G7 mAb, which is incorporated herein by reference for all purposes). The amino acid sequences of the (mature) light and heavy chain variable (V) regions in GAL-FR21 are shown in Figures 13A and 13B, respectively, with GAL-FR21 labeled at the top. More specifically, to design humanized GAL-FR21 mAb, the method of Queen et al. (U.S. Patent Nos. 5,530,101 and 5,585,089) was generally followed. The human Vκ sequence CAG27369 and VH sequence AAB00780, shown at the bottom of Figures 13A and 13B, respectively, were selected to serve as acceptor sequences for the GAL-FR21 VL and VH sequences because they share particularly high framework homology (i.e., sequence identity). A computer-generated molecular model of the GAL-FR21 variable domain was used to determine amino acid positions in the GAL-FR21 framework that were close enough to the CDRs to potentially interact with them. To design the humanized GAL-FR21 light and heavy chain variable regions, the CDRs from the murine GAL-FR21 mAb were initially conceptually grafted into the acceptor framework regions. At framework positions where the computer model suggested important contacts with the CDRs (which may be necessary to maintain the CDR conformation), amino acids from the murine antibody were substituted with amino acids from the human framework. For the humanized GAL-FR21 mAb (designated HuGAL-FR21), residues 27, 28, and 30 (located in the Chotiah hypervariable loop H1) and 48 and 67 of the heavy chain, using Kabat numbering, were performed, but no residues of the light chain were performed.The sequences of the light and heavy chain V regions of HuGAL-FR21 are shown in Figures 13A and 13B, respectively, with HuGAL-FR21 displayed in the middle, positioned opposite the respective GAL-FR21 donor and human acceptor V regions, with the CDRs (as defined by Kabat) underlined and the substituted amino acids double-underlined.

[0060] The present invention provides not only HuGAL-FR21, a humanized GAL-FR21 mAb comprising the light and heavy chain V regions shown in Figure 13, but also variant humanized mAbs. These variant humanized mAbs differ from the HuGAL-FR21 sequence in the light and heavy chain variable regions by a small number (e.g., typically no more than 1, 2, 3, 5, or 10) of substitutions, deletions, or insertions, usually in the framework (and potentially in the CDRs). In particular, it is possible to make only a subset of the above substitutions in the acceptor framework, or to make additional substitutions. For example, mouse GAL-FR21 VH amino acid 69L can be substituted with acceptor amino acid 69I, and / or the mouse amino acid can be substituted with each amino acid in the humanized light chain at Kabat number positions 1, 3, 60, and 63 (positions slightly closer to the CDRs). Indeed, many of the framework residues that do not contact the CDRs of a humanized mAb are amenable to substitution with amino acids from the corresponding positions in the donor murine mAb or other murine or human antibodies, and even many residues that may contact the CDRs are amenable to substitution, or even amino acids in the CDRs can be changed. One example of a CDR substitution is to replace a residue in a CDR with the residue occupying the corresponding position in the human acceptor sequence used to provide the framework for the variable region.

[0061] In most cases, substitutions made in a variant humanized GAL-FR21 sequence are conservative relative to the substituted HuGAL-FR21 amino acid. Amino acids can be grouped as follows to determine conservative substitutions (i.e., intra-group substitutions): Group I (hydrophobic side chains): met, ala, val, leu, ile; Group II (neutral hydrophilic side chains): cys, ser, thr; Group III (acidic side chains): asp, glu; Group IV (basic side chains): asn, gln, his, lys, arg; Group V (residues affecting chain orientation): gly, pro; Group VI (aromatic side chains): trp, tyr, phe.

[0062] Preferably, substitutions (conservative or not) in HuGAL-FR21 do not substantially affect the binding affinity or potency of the humanized mAb, i.e., its ability to neutralize the biological activity of FGFR2. This potency means, for example, that the potency of a variant humanized GAL-FR21 mAb in some or all of the assays described herein is essentially the same as that of HuGAL-FR21 (i.e., within experimental error). Preferably, the sequences of the mature variant light and heavy chain V regions are at least 90%, more preferably at least 95%, and most preferably at least 98% identical to those of the HuGAL-FR21 mature light and heavy chain V regions, respectively. Alternatively, other human antibody variable regions with high sequence identity to the GAL-FR21 variable regions are also suitable for providing the humanized antibody framework, particularly kappa V regions from human subgroup I and heavy chain V regions from human subgroup I, or the consensus sequences of these subgroups.

[0063] In other humanized antibodies, at least one, two, three, four, or all five of the acceptor positions for donor substitutions mentioned for the exemplified antibodies (i.e., H27, H28, H30, H48, H67) are preferably occupied by residues occupying the corresponding positions in the heavy chain of the murine donor antibody. When the heavy chain acceptor sequence is other than AAB00780, the acceptor for donor substitutions may or may not require specific occupancies at specific variable framework region positions, depending on whether the residues occupying those positions are already identical between the acceptor and donor.

[0064] The exemplary mAb HuGAL-FR21 described herein is an IgG1 because it possesses human kappa and gamma 1 constant regions (e.g., as shown in US 2008019974). The complete sequences of the (mature) light and heavy chains of HuGAL-FR21 are shown in Figure 14. While these sequences are of the Km(3) and G1m(3) allotypes, respectively, it is understood that any (IgG1, kappa) allotype of an IgG1 mAb is encompassed by the designation HuGAL-FR21. When HuGAL-FR21 is produced by conventional procedures, one or several amino acids at the amino or carboxy termini of the light and / or heavy chains (e.g., the C-terminal lysine of the heavy chain) may be missing or derivatized in a proportion or in all molecules; such compositions would still be encompassed by the designation HuGAL-FR2 and considered a humanized mAb. Other isotypes of humanized mAbs (e.g., IgG2, IgG3, and IgG4) can be generated by combining the variable regions of HuGAL-FR21 with appropriate human constant regions. Substitutions can be made in the HuGAL-FR21 constant region to decrease or increase effector function (e.g., complement-mediated cytotoxicity or ADCC (see, e.g., Winter et al., U.S. Pat. No. 5,624,821; Tso et al., U.S. Pat. No. 5,834,597; and Lazar et al., Proc. Natl. Acad. Sci. USA 103:4005, 2006)) and to increase half-life in humans (see, e.g., Hinton et al., J. Biol. Chem. 279:6213, 2004). Specifically, but not by way of limitation, HuGAL-FR21 having mutations to Gln at position 250 and / or Leu at position 428 of the IgG constant region is an embodiment of the present invention.

[0065] To compare the binding affinity of HuGAL-FR21 with that of the mouse mAb GAL-FR21, a competitive binding experiment was performed using a standard ELISA. Specifically, ELISA wells were coated with 2 μg / ml goat anti-human IgG-Fc overnight at 4°C. After blocking with 2% BSA for 1 hour at RT, the wells were incubated with 0.5 μg / ml FGFR2IIIb-Fc. The wells were incubated with biotinylated GAL-FR21 mAb (0.05 μg / ml) in the presence of increasing concentrations of unlabeled GAL-FR21, HuGAL-FR21, or control human antibody hIgG. The binding level of biotinylated GAL-FR21 was measured by adding HRP-streptavidin and substrate. As shown in Figure 15, HuGAL-FR21 and GAL-FR21 competed equally well, with HuGAL-FR21 possibly being slightly better, indicating that the binding affinity of HuGAL-FR21 to FGFR2 is at least as good as that of the (murine) GAL-FR21 mAb. The concentration of HuGAL-FR21 required to inhibit 50% of the binding of the labeled mAb indicates that the binding affinity, Ka, of HuGAL-FR21 to FGFR2 is at least about 10. 9 M -1 HuGAL-FR21 may be tested in any of the biological assays for FGFR2 activity described herein (e.g., inhibition of FGF2 or FGF7 binding to FGFR2) and inhibits FGFR2 activity comparably to GAL-FR21.

[0066] Humanized GAL-FR22 mAbs can be designed, constructed, produced, and assayed in the same or similar manner as HuGAL-FR22. The amino acid sequences of the (mature) light and heavy chain variable regions of GAL-FR22 are shown in Figures 16A and 16B, respectively, and are represented by a line drawn through GAL-FR22. Humanized GAL-FR22 mAbs have a humanized light chain containing CDRs derived from the sequence in Figure 16A and a humanized heavy chain containing CDRs derived from the sequence in Figure 16B. In some cases, humanized GAL-FR22 mAbs contain the three light chain CDRs shown in Figure 16A and the three heavy chain CDRs shown in Figure 16B. Preferably, the humanized GAL-FR22 mAb has a binding affinity for FGFR2 that is within 2- or 3-fold of the affinity of murine GAL-FR22 mAb, and most preferably has a binding affinity that is indistinguishable from or greater than that of GAL-FR22 mAb, as measured, for example, by competitive ELISA, as described for HuGAL-FR21 and GAL-FR21.

[0067] Although the present invention has been described with respect to the presently preferred embodiments, it should be understood that various modifications can be made without departing from the invention. Unless otherwise apparent from the context, any step, element, embodiment, feature, or aspect of the invention can be used with any other. All publications, patents, patent applications, registration numbers, and similar references are incorporated herein by reference in their entirety for all purposes for the same context, as if each individual publication, patent, patent application, or registration number were specifically and individually indicated to be incorporated by reference in its entirety for all purposes. In the event that a nucleic acid sequence or protein sequence associated with a registration number changes, the version of the sequence associated with that registration number as of November 7, 2008 is intended.

[0068] Under the Budapest Treaty, on January 8, 2008, hybridomas producing monoclonal antibodies GAL-FR21, GAL-FR22, and GAL-FR23 were deposited with the American Type Culture Collection, P.O. Box 1549, Manassa, VA 20108, under ATCC No. PTA-9586 (November 6, 2008), ATCC No. PTA-9587 (November 6, 2008), and ATCC No. PTA-9408 (August 12, 2008), respectively. These deposits will be kept in an accredited depository and will be replaced in the event of mutation, non-viability, or destruction for at least five years after the most recent request for release of samples received by the depository, or for at least 30 years from the date of deposit, or for the term of any relevant patent, whichever is longer. All publicly disclosed restrictions on availability of these cell lines will be removed upon the issuance of a patent resulting from this application.

Claims

1. A humanized or human monoclonal antibody (mAb) that binds to fibroblast growth factor receptor 2 (FGFR2) and inhibits the growth of human tumor xenografts in mice.

2. The mAb of claim 1 , which inhibits binding of FGF2 to FGFR2.

3. The mAb of claim 1 , which binds to FGFR2IIIb but does not bind to FGFR2IIIc.

4. The mAb of claim 1 , which binds to both FGFR2IIIb and FGFR2IIIc.

5. The mAb may be a Fab or F(ab') 2 The mAb of claim 1 which is a fragment or single chain antibody.

6. A pharmaceutical composition comprising the mAb of claim 1.

7. A method for treating a patient with cancer, comprising administering to the patient a pharmaceutical composition comprising the mAb of claim 1.

8. A monoclonal antibody (mAb) whose binding to FGFR2 competes with an antibody selected from the group consisting of GAL-FR21, GAL-FR22 and GAL-FR23, said mAb being genetically engineered.

9. The mAb of claim 8, wherein the mAb inhibits the growth of SNU-16 human tumor xenografts in mice.

10. The mAb of claim 8 , wherein the mAb is a chimeric or humanized mAb.

11. The mAb of claim 8 , wherein the mAb is a humanized GAL-FR21 or GAL-FR22 mAb.

12. The mAb of claim 8 , wherein the mAb is a human mAb.

13. A pharmaceutical composition comprising the mAb of claim 8.

14. A method for treating a patient with cancer, comprising administering to the patient a pharmaceutical composition comprising the mAb of claim 8.

15. 15. The method of claim 14, wherein the cancer is gastric cancer.

16. A humanized antibody comprising a humanized light chain having a CDR derived from the sequence of Figure 13A (GAL-FR21) and a humanized heavy chain having a CDR derived from the sequence of Figure 13B (GAL-FR21), or a humanized light chain having a CDR derived from the sequence of Figure 16A (GAL-FR22) and a humanized heavy chain having a CDR derived from the sequence of Figure 16B (GAL-FR22).

17. 17. The humanized antibody of claim 16, comprising the three light chain CDRs shown in Figure 13A (GAL-FR21) and the three heavy chain CDRs shown in Figure 13B (GAL-FR21), or the three light chain CDRs shown in Figure 16A (GAL-FR22) and the three heavy chain CDRs shown in Figure 16B (GAL-FR22).

18. 18. The humanized antibody of claim 17, wherein the light chain variable region has at least 95% sequence identity with the sequence shown in Figure 13A (HuGAL-FR21) and the heavy chain variable region has at least 95% sequence identity with the sequence shown in Figure 13B (HuGAL-FR21).

19. 19. The humanized antibody of claim 18, wherein residues H27, H28, H30, H48 and H67 according to the Kabat numbering are occupied by residues occupying the corresponding positions in the heavy chain shown in Figure 13B (GAL-FR21).

20. 20. The humanized antibody of claim 19, wherein the light chain variable region has the sequence shown in Figure 13A (HuGAL-FR21) and the heavy chain variable region has the sequence shown in Figure 13B (HuGAL-FR21).

Images