Therapeutics , Targets , and Chemical Biology Dual Functional Monoclonal Antibody PF-04605412 Targets Integrin a 5 b 1 and Elicits Potent Antibody-Dependent Cellular Cytotoxicity

Integrin a5b1 is overexpressed in tumor-associated stroma and cancer cells, and has been implicated in angiogenesis, tumor survival, and metastasis. Antibody-dependent cellular cytotoxicity (ADCC) by immune effector cells has been shown to contribute to clinical efficacy for several IgG1 monoclonal antibody (mAb) therapeutics. Taking advantage of these two mechanisms, we generated a fully human, fragment crystalizable (Fc)-engineered IgG1mAb, PF-04605412 (PF-5412), which specifically neutralizes a5 and binds the Fcg receptors (FcgR) with enhanced affinity. In vitro, PF-5412 potently inhibited a5b1-mediated intracellular signaling, cell adhesion, migration, and endothelial cell (EC) tubulogenesis. PF-5412 induced significantly greater ADCC in a5expressing tumor cells and ECs compared with a wild-type IgG1 (IgG1/wt) or IgG2 of identical antigen specificity. The degree of ADCC correlated with the abundance of natural killer (NK) cells in the peripheral blood mononuclear cells but was independent of donor FcgRIIIa polymorphism. In animal studies, PF-5412 displayed robust and dose-dependent antitumor efficacy superior to that observed with IgG1/wt, IgG2, or IgG4 of identical antigen specificity. The degree of efficacy correlated with a5 expression, macrophage and NK cell infiltration, and NK activity in the tumor. Depletion of host macrophages abrogated antitumor activity, suggesting a critical contribution of macrophage-mediated antitumor activity of PF-5412. Combination of PF-5412 with sunitinib significantly improved antitumor efficacy compared with either agent alone. The dual mechanism of action and robust antitumor efficacy of PF-5412 support its clinical development for the treatment of a broad spectrum of human malignancies. Cancer Res; 70(24); 10243–54. 2010 AACR. Introduction Integrin a5b1 is a classic fibronectin (Fn) receptor that plays important roles in vascular biology including embryonic vasculogenesis, vascular remodeling, pathologic angiogenesis, and inflammation (1–3). Its involvement in the function of tumor-associated endothelium and stroma cells has also been reported (4). In response to proangiogenic factors, a5b1 promotes endothelial cell (EC) adhesion, migration, proliferation, survival, and differentiation (5, 6). a5b1 is frequently overexpressed in tumor cells and has been associated with hypoxia, survival, epithelial-mesenchymal transition, invasion, and metastasis (7–9). Various studies have linked a5b1 upregulation with the progression of several cancers (10–13). The overall survival in patients with lung cancers and breast cancers was found to be inversely correlated with the degree of tumor a5 expression (8, 14), providing a basis for targeting a5b1 for cancer therapy. Among the several inhibitors against a5 is volociximab (M200; ref. 15). In the clinic, M200 generated disease stabilization and a number of partial responses in previously heavily treated metastatic renal cell carcinoma patients (16). Accumulating evidence suggests that antibody-dependent cellular cytotoxicity (ADCC) may play a significant role in anticancer therapy (17). ADCC is a major immune effector mechanism in which target cell–bound antibody–antigen complexes engage innate immune effector cells via fragment crystalizable g receptors (FcgR), resulting in the killing of antigen-expressing target cells (18–21). Clinical proof of principle for ADCC was shown for rituximab in non-Hodgkin lymphoma patients and trastuzumab in breast cancer Authors' Affiliations: Oncology Research Unit, Pfizer Inc., San Diego, California; and Biotherapeutics Pharmaceutical Sciences, Pfizer Inc., St. Louis, Missouri Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). Current addresses for Xin Jiang: Sova Pharmaceuticals, La Jolla, CA; Grant Wickman: The Beatson Institute for Cancer Research, Glasgow, UK; and Karin Amundson: Tocagen Inc., San Diego, CA. Corresponding Author: Dana D. Hu-Lowe, The Oncology Research Unit, Pfizer Inc., 10646 Science Center Drive, San Diego, CA 92121. Phone: 858-622-6019; Fax: 858-622-599. E-mail: dana.hu-lowe@pfizer.com doi: 10.1158/0008-5472.CAN-10-1996 2010 American Association for Cancer Research. Cancer Research www.aacrjournals.org 10243 Research. on April 13, 2017. © 2010 American Association for Cancer cancerres.aacrjournals.org Downloaded from patients; in both cases, better therapeutic outcomes were associated with high-affinity allele of FcgRs, FcgRIIIa/158V, or FcgRIIa/131H (22–25). Additionally, in metastatic colorectal cancer patients treated with cetuximab, the ones harboring the high-affinity FcgRIIIa/158V allele survived significantly longer than those with 158F/F (26). Importantly, this correlation was independent of mutant K-ras status, underlining a significant contribution of ADCC to the clinical benefit from cetuximab. Several technologies have been developed to improve the binding affinitybetween theFc regionand theFcgRsoneffector cells (27, 28). Products through glycoengineering, GA101 (antiCD20), GA201 (anti-EGFR), and MDX-1401 (anti-CD30) have shown increased ADCC and superior in vivo antitumor activity in preclinical models compared with rituximab, cetuximab, and MDX-060, respectively (29–31). XmAb5574, a humanized anti-CD19 IgG1 monclonal antibody (mAb) with enhanced ADCC via Fc mutagenesis, has entered clinical development as a potential therapy for B-cell malignancies (32). We developed a fully human, ADCC-enhanced anti-a5 IgG1 mAb PF-5412 via amino acid mutations in the Fc region (33). In this report, we provide evidence of its robust in vitro and in vivo activities through ADCC and antibody-dependent phagocytosis (ADPC). Materials and Methods Antibodies and compounds The following mAbs were made at Pfizer: PF-5412 (an IgG1 with DLE mutations in the Fc region), wild-type (wt) mAbs with identical variable domains but different constant regions (a5-IgG1/wt, a5-IgG2, and a5-IgG4), negative control antibodies keyhole limpet hemocyanin (KLH)-IgG1 and KLHIgG1/DLE, and M200. Also generated at Pfizer are the following compounds: sunitinib [ref.34; a multi-targeted inhibitor of vascular endothelial growth factor receptors (VEGFR), platelet-derived growth factor receptors, Kit and Flt3], axitinib (35), and PF-0337210 (ref. 36; both are potent and selective small molecule inhibitors of VEGFRs). Bevacizumab (anti-VEGF mAb by Genentech) was purchased from Northeast Medical Products, Inc. All antibodies were dosed subcutaneously once a week (QW) at the indicated doses. Sunitinib and PF-0337210 were given orally once daily (QD), and axitinib was given orally twice daily (BID). Cells Human umbilical vein endothelial cells (HUVEC) and normal human dermal fibroblasts (NHDF) were purchased from Lonza. U87MG and A549 were obtained from American Type Culture Collection. M24met was described elsewhere (37). A549-luc-C8 was purchased from Xenogen. Other luciferaseexpressing cells weremade by transfection with vector pLPCX: Luc-SH. To overexpress a5, parental MV522 cells were infected with retroviral particles made in 293T cells with vector pBabe-puro/a5. Human or monkey peripheral blood mononuclear cells (PBMC) were isolated and maintained as described previously (38). FcgRIIIa polymorphism genotyping and natural killer (NK)% measurements by flow cytometry were performed according to published protocols (39). The generation of mouse, rabbit, and monkey a5-transfected cells is described in Supplementary Materials. Generation of fully human anti-a5 hybridoma and antibodies from transgenic mice Human transchromosomal/transgenic KM mice from Medarex, Inc. were immunized with 1 10 NIH3T3 cells overexpressing human a5. Sera from the immunized mice were screened by flow cytometry for binding to a5 in a5/ NIH3T3 cells and to the parental NIH3T3 cells. Hybridomas were generated according to the standard procedure (40). Hybridomas expressing antibodies that bound a5/NIH3T3 cells, but not the parental cells, were cloned twice by limiting dilution. Integrin a5b1 and FcgR binding assays and affinity determination Antibody affinity for integrin a5b1 was measured by fluorescence-activated cell sorting (FACS) analysis and surface plasmon resonance (SPR) using a Biacore 3000 (GEHealthcare; see Supplementary Methods for details). Human FcgRI was purchased from R&D Systems. The other human and murine FcgRs were produced by transient expression in HEK293 cells, and purified using HisTrap FF columns (GE Healthcare). PF5412was immobilized on aCM5 sensor chip using the standard primary amine coupling protocol. FcgRs were infused followed by a 2to 3-minute dissociation phase. Data were fit to a 1:1 binding model (Langmuir) using Scrubber2 data analysis software (BioLogic software). Kinetic variables were used to calculate the equilibrium dissociation constant (KD). In vitro ADCC assay A total of 1 10 target cells were preincubated with mAbs at indicated concentrations. Human or monkey PBMCs were added at effector:target (E:T) cell ratio of 50 100:1. Assay plates were incubated at 37 C for 4 hours. Cytolysis was determined using either the LDH Cytotoxicity Detection Kit (Roche) or the ToxiLight BioAssay Kit (Cambrex). In vitro ADPC assay Human donor monocytes were isolated using CD14 magnetic beads (Miltenyi Biotech) and differentiated into macrophages with 10 ng/mL granulocyte macrophage colony stimulating factor (R&D Systems) for 5 days. Macrophages and U87MG target cells were labeled with PKH26 (Sigma) and CMFDA (Invitrogen), respectively. Labeled U87MG cells were incubated with mAbs for 30 minutes and then added to the labeled macrophages. Four hours later, phagocytosis was determined by counting double-labeled cells by FACS analysis. In vivo angiogenesis models To establish the high-density (HD) Matrigel-based human angiogenesis model, a mixture of HUVEC (3 10) and NHDF cells (1 10) in HDMatrigel (BD Biosciences, 10 mg/mL) was implanted subcutaneously (0

a5b1 is frequently overexpressed in tumor cells and has been associated with hypoxia, survival, epithelial-mesenchymal transition, invasion, and metastasis (7)(8)(9).Various studies have linked a5b1 upregulation with the progression of several cancers (10)(11)(12)(13).The overall survival in patients with lung cancers and breast cancers was found to be inversely correlated with the degree of tumor a5 expression (8,14), providing a basis for targeting a5b1 for cancer therapy.Among the several inhibitors against a5 is volociximab (M200; ref. 15).In the clinic, M200 generated disease stabilization and a number of partial responses in previously heavily treated metastatic renal cell carcinoma patients (16).
Accumulating evidence suggests that antibody-dependent cellular cytotoxicity (ADCC) may play a significant role in anticancer therapy (17).ADCC is a major immune effector mechanism in which target cell-bound antibody-antigen complexes engage innate immune effector cells via fragment crystalizable g receptors (FcgR), resulting in the killing of antigen-expressing target cells (18)(19)(20)(21).Clinical proof of principle for ADCC was shown for rituximab in non-Hodgkin lymphoma patients and trastuzumab in breast cancer patients; in both cases, better therapeutic outcomes were associated with high-affinity allele of FcgRs, FcgRIIIa/158V, or FcgRIIa/131H (22)(23)(24)(25).Additionally, in metastatic colorectal cancer patients treated with cetuximab, the ones harboring the high-affinity FcgRIIIa/158V allele survived significantly longer than those with 158F/F (26).Importantly, this correlation was independent of mutant K-ras status, underlining a significant contribution of ADCC to the clinical benefit from cetuximab.
We developed a fully human, ADCC-enhanced anti-a5 IgG1 mAb PF-5412 via amino acid mutations in the Fc region (33).In this report, we provide evidence of its robust in vitro and in vivo activities through ADCC and antibody-dependent phagocytosis (ADPC).

Antibodies and compounds
The following mAbs were made at Pfizer: PF-5412 (an IgG1 with DLE mutations in the Fc region), wild-type (wt) mAbs with identical variable domains but different constant regions (a5-IgG1/wt, a5-IgG2, and a5-IgG4), negative control antibodies keyhole limpet hemocyanin (KLH)-IgG1 and KLH-IgG1/DLE, and M200.Also generated at Pfizer are the following compounds: sunitinib [ref.34; a multi-targeted inhibitor of vascular endothelial growth factor receptors (VEGFR), platelet-derived growth factor receptors, Kit and Flt3], axitinib (35), and PF-0337210 (ref.36; both are potent and selective small molecule inhibitors of VEGFRs).Bevacizumab (anti-VEGF mAb by Genentech) was purchased from Northeast Medical Products, Inc.All antibodies were dosed subcutaneously once a week (QW) at the indicated doses.Sunitinib and PF-0337210 were given orally once daily (QD), and axitinib was given orally twice daily (BID).

Cells
Human umbilical vein endothelial cells (HUVEC) and normal human dermal fibroblasts (NHDF) were purchased from Lonza.U87MG and A549 were obtained from American Type Culture Collection.M24met was described elsewhere (37).A549-luc-C8 was purchased from Xenogen.Other luciferaseexpressing cells were made by transfection with vector pLPCX: Luc-SH.To overexpress a5, parental MV522 cells were infected with retroviral particles made in 293T cells with vector pBabe-puro/a5.Human or monkey peripheral blood mononuclear cells (PBMC) were isolated and maintained as described previously (38).FcgRIIIa polymorphism genotyping and natural killer (NK)% measurements by flow cytometry were performed according to published protocols (39).The generation of mouse, rabbit, and monkey a5-transfected cells is described in Supplementary Materials.

Generation of fully human anti-a5 hybridoma and antibodies from transgenic mice
Human transchromosomal/transgenic KM mice from Medarex, Inc. were immunized with 1 Â 10 7 NIH3T3 cells overexpressing human a5.Sera from the immunized mice were screened by flow cytometry for binding to a5 in a5/ NIH3T3 cells and to the parental NIH3T3 cells.Hybridomas were generated according to the standard procedure (40).Hybridomas expressing antibodies that bound a5/NIH3T3 cells, but not the parental cells, were cloned twice by limiting dilution.

Integrin a5b1 and FcgR binding assays and affinity determination
Antibody affinity for integrin a5b1 was measured by fluorescence-activated cell sorting (FACS) analysis and surface plasmon resonance (SPR) using a Biacore 3000 (GE Healthcare; see Supplementary Methods for details).Human FcgRI was purchased from R&D Systems.The other human and murine FcgRs were produced by transient expression in HEK293 cells, and purified using HisTrap FF columns (GE Healthcare).PF-5412 was immobilized on a CM5 sensor chip using the standard primary amine coupling protocol.FcgRs were infused followed by a 2-to 3-minute dissociation phase.Data were fit to a 1:1 binding model (Langmuir) using Scrubber2 data analysis software (BioLogic software).Kinetic variables were used to calculate the equilibrium dissociation constant (K D ).

In vitro ADCC assay
A total of 1 Â 10 4 target cells were preincubated with mAbs at indicated concentrations.Human or monkey PBMCs were added at effector:target (E:T) cell ratio of 50$100:1.Assay plates were incubated at 37 C for 4 hours.Cytolysis was determined using either the LDH Cytotoxicity Detection Kit (Roche) or the ToxiLight BioAssay Kit (Cambrex).

In vitro ADPC assay
Human donor monocytes were isolated using CD14 magnetic beads (Miltenyi Biotech) and differentiated into macrophages with 10 ng/mL granulocyte macrophage colony stimulating factor (R&D Systems) for 5 days.Macrophages and U87MG target cells were labeled with PKH26 (Sigma) and CMFDA (Invitrogen), respectively.Labeled U87MG cells were incubated with mAbs for 30 minutes and then added to the labeled macrophages.Four hours later, phagocytosis was determined by counting double-labeled cells by FACS analysis.

In vivo angiogenesis models
To establish the high-density (HD) Matrigel-based human angiogenesis model, a mixture of HUVEC (3 Â 10 6 ) and NHDF cells (1 Â 10 6 ) in HD Matrigel (BD Biosciences, 10 mg/mL) was implanted subcutaneously (0.8 mL/injection) in the flank of mice.After 4 to 6 weeks, the Matrigel plugs were removed and fixed in 1% paraformaldehyde at 4 C overnight, embedded in ornithine carbamyl transferase, and sectioned for histology analysis.
Another model is the human foreskin-severe combined immunodeficient (SCID) mouse chimera based on the report by Tahtis and colleagues (41).Briefly, a piece of 1 cm Â 2 cm human neonatal foreskin was sutured to the back of a BALB/c SCID mouse.After recovery (5-7 weeks), M24met cells (2 Â 10 6 ) were intradermally injected into the engrafted human skin.When tumor volume reached between 50 and 100 mm 3 , the mice were randomized to receive treatments.Human-specific CD31 staining was then used to detect human vessels (42).

Experimental metastasis model
A549-luc-C8 cells (3 Â 10 6 /100 mL) were injected intravenously into BALB/c SCID mice that had received a single predose 2 days prior.The animals were treated with indicated antibodies for 8 weeks.BLI was carried out once a week until the end of the study.

In situ macrophage depletion in mice
Clodronate-liposomes (5.5 mg/mL) or PBS-liposomes (25-200 nm in mean particle diameters) were purchased from Dr. Nico van Rooijen's laboratory (Vrije University, Amsterdam, the Netherlands).One day before tumor implant, 2 injections of the liposome particles were given to mice (2.5 mg/g body weight, intravenously and 50 mL, subcutaneously, near the site of tumor implantation).The particles were subsequently given intravenously twice a week until end of the study.

Immunohistochemistry staining and quantification of NK cells and macrophages
Murine NK cells were stained with antimouse NK1.1 (Biolegend), and macrophages were stained using a rat antimouse antibody for macrophage marker F4/80 (Abcam; ref. 43).Tissue immunohistochemistry (IHC) images were captured using an Olympus MicroFire digital camera and PictureFrame software.Percentage of positive staining was quantified using the ChromaVision Automated Cell Imaging System.

Statistical analysis
Statistical significance was determined by analysis of variance using Dunnett's multiple-comparison post-test with GraphPad Prism software unless otherwise noted.

Generation of PF-5412
The mAb PF-5412 was derived from a5-IgG1/wt by mutagenesis in the Fc region to introduce 3 mutations, S239D/ A330L/I322E (DLE), that have been reported to enhance ADCC activity (33), and by mutagenesis of the variable domains to return several framework residues to those encoded in the germ line (for details see Supplementary Methods).

In vitro antigen binding and neutralization activities of PF-5412
SPR-and FACS-based assays showed that PF-5412 bound to human a5b1 with nanomolar affinity (Table 1; Fig. 1A).The K D values (2.7 nmol/L for PF-5412 and 2.2 nmol/L for a5-IgG1/wt) indicated that Fc engineering did not alter antigen binding affinity.Competition binding experiments using SPR indicated that PF-5412 and M200 bound at the same binding site on a5b1 (Supplementary Fig. S1A and B).In cellular assays, PF-5412 blocked HUVEC cell adhesion to Fn with a similar potency as a5-IgG1/wt, a5-IgG2, and M200 (Fig. 1B).In addition, PF-5412 dose dependently blocked HUVEC migration (Supplementary Fig. S1C).In a coculture tubule formation assay, PF-5412 exhibited a dose-dependent activity similar to axitinib (Fig. 1C).PF-5412 also inhibited total and phosphorylated focal adhesion kinase (FAK; Fig. 1D), a downstream signaling molecule of a5b1.Consistent with a5b1's role in EC survival, PF-5412 induced caspase 3/7-mediated apoptosis of HUVEC with an EC 50 value of 15.3 nmol/L (Supplementary Fig. S1D).PF-5412 was found to bind to integrin a5 of cynomolgus monkey and rabbit, but not to that of rat or mouse (Table 1).Consistent with its binding affinity to cynomolgus a5, PF-5412 inhibited monkey monocyte adhesion to Fn in an ex vivo assay (Supplementary Fig. S1E).These data are summarized in Table 1.
When fully differentiated human macrophages were mixed with U87MG cells, phagocytosis was observed in the presence of PF-5412 and a5-IgG1/wt (0.1 mg/mL and 1 mg/mL), but not the KLH control mAbs.At either concentration, PF-5412 induced a greater degree of phagocytosis than a5-IgG1/wt (Fig. 2C).These data indicate that PF-5412 has a greater potential to kill target cells through ADPC than a5-IgG1/wt.
To investigate if ADCC was influenced by FcgRIIIa polymorphism, ADCC assays were performed with PBMCs from 14 donors representing all 3 FcgRIIIa genotypes.There was no correlation between ADCC max and FcgRIIIa polymorphism (Fig. 2D, left).However ADCC max positively correlated with the percentage of NK cells in PBMCs (Fig. 2D, right), indicating that NK cells played an important role in target cell cytolysis.

In vivo antiangiogenesis activity
Because PF-5412 does not cross-react with rodent a5b1, in vivo angiogenesis models containing human ECs were established.In the HD Matrigel-based human angiogenesis model, PF-5412 reduced human CD31 þ vessels compared with vehicle or the isotype-matched KLH mAbs, a5-IgG1/wt and bevacizumab (Fig. 3A; *P < 0.05 in Supplementary Fig. S3A).In a separate study, PF-5412 was more effective than M200 at 3 mg/kg, and the 2 agents showed similar antiangiogenic activity at 10 mg/kg (Supplementary Fig. S3B).
In the human foreskin-SCID mouse chimera model in which human vessels were present and expressed human a5 (44), single injections of PF-5412 dose dependently inhibited human CD31 þ vessel formation (Fig. 3B) with an in vivo ED 50 of 1.2 mg/kg and ED 90 of 4.6 mg/kg (Fig. 3B, inset).

Antitumor efficacy
In CB17.SCID mice with functional NK cells and macrophages, weekly administration of PF-5412 at either 1 mg/kg or 10 mg/kg significantly delayed the growth of the U87MG tumors (Fig. 3C, left).The degree of tumor growth inhibition (TGI) by the high dose of PF-5412 (86%) was greater than that by bevacizumab (69%).Antitumor efficacy of PF-5412, but not bevacizumab, was associated with increased tumor infiltration of NK cells and macrophages (Fig. 3C, right).We next assessed the TGI of PF-5412 in an immunoxenograft model in which U87MG-luc cells and human PBMCs were coimplanted in CB17.SCID/beige mice (deficient in host NK).In this model, the human PBMCs served as a source of neutrophils and NK cells, whereas the mouse supplied macrophages.PF-5412 treatment induced tumor regression at 3 and 10 mg/kg.M200 generated moderate TGI (40%; Fig. 3D, left; Supplementary Fig. S3C).Compared with a5-IgG1/wt and M200, PF-5412-treated tumors showed increased NK cell activity (Granzyme B), a greater number of macrophages, and a higher degree of tumor cell apoptosis (Fig. 3D, right; Supplementary Fig. S3D).A separate study showed that the extent of antitumor efficacy correlated with the number of human PBMCs coimplanted with the tumor cells (not shown).These results indicated that human NK cells and potentially host macro-phages contributed significantly to antitumor efficacy of PF-5412.

Host macrophages contribute significantly to efficacy
We then used tumor-bearing CB17.SCID/beige mice (NK deficient) to address whether host macrophages alone could play a significant role in the activity of PF-5412.We observed that PF-5412, but not M200, significantly and dose dependently inhibited U87MG-luc tumor growth (Fig. 4A); the TGI was associated with a dose-dependent increase in mouse macrophage infiltration in the tumor (Fig. 4B).When the host macrophages were depleted by clodronate-liposomes (45), the TGI was significantly compromised compared with mice treated with PBS-liposomes (Fig. 4C; Supplementary Fig. S4A).The effect was not because of any intrinsic activity of clodronate as clodronate-liposomes alone did not affect tumor growth (Fig. 4C).These data strongly suggest that the efficacy of PF-5412 is dependent on ADPC from mouse macrophages.The influx of macrophages was detectable 24 hours after a single dose of PF-5412 (10 mg/kg) and continued to increase through 72 hours postdose (Fig. 4D).In a separate study, elevated levels of macrophages in the tumor were observed for as long as 7 days following a single dose of PF-5412 (data not shown).A single dose of PF-5412 also induced a marked reduction of phosphorylated FAK and total FAK in the tumor (Supplementary Fig. S4B), suggesting a direct effect on a5b1 signaling in addition to or as a consequence of ADCC and ADPC by PF-5412.

Macrophage-mediated ADPC and efficacy requires the presence of a5
PF-5412 produced a greater TGI with MV522/a5 tumors engineered to express a moderate level of a5 (clone 11), compared with clone 1 that expressed a lower level of a5 (Supplementary Fig. S5A), or the parental cells (do not have detectable a5; Fig. 5A).The TGI was associated with increased intratumoral macrophage staining in MV522/a5 clone 11, but not clone 1 (Fig. 5B).

Antimetastatic activity of PF-5412
In the lung metastasis model of A549-Luc-C8, an 8-week treatment with PF-5412 and a5-IgG2 exhibited significant antimetastatic efficacy with a TGI of 97% and 89% (week 8), respectively (Fig. 5C, left).After dosing was stopped, tumors in the PF-5412-treated group remained suppressed through week 13, whereas those in a5-IgG2-treated group began to regrow 2 weeks after dosing cessation.PF-5412 significantly extended the median time-to-progression to 20 weeks or more compared with 14 weeks and 10 weeks for a5-IgG2 and the vehicle group, respectively (Fig. 5C, right).

Combination of low-dose PF-5412 with sunitinib or bevacizumab significantly enhanced antitumor efficacy
We further assessed efficacy of combination of low-dose PF-5412 with sunitinib, an agent with a different antiangiogenic mechanism of action than PF-5412.In the U87MG-luc tumor model in CB17.SCID/beige mice, PF-5412 (1 mg/kg) and sunitinib (30 mg/kg, a suboptimal dose) produced TGI of 46% and 32%, respectively; the combination treatment produced an 83% TGI, significantly better than either monotherapy alone (P < 0.05 versus PF-5412 and P < 0.001 versus sunitinib; Fig. 5D; Supplementary Fig. S5B).In the same study, M200 (1 mg/kg) alone generated a 26% TGI, which was improved to 41% when combined with sunitinib.This latter TGI was significantly less (P < 0.01) compared with that of PF-5412 in combination with sunitinib (Supplementary Fig. S5B).

Discussion
In this report, we showed that PF-5412, a fully human dual functional mAb against integrin a5b1, potently and dose dependently blocked EC adhesion, migration, tubule formation, and survival (Fig. 1).The mAb also inhibited angiogenesis mediated by human endothelial cells/vessels in in vivo models developed to harbor the human angiogenesis compo-nents (Fig. 3A and B).More importantly, as a result of Fc engineering to improve binding affinities to FcgRs, PF-5412 elicited enhanced ADCC/ADPC activity through host immune effector cells resulting in superior target cell lysis (Fig. 2A and  C), robust antitumor and antimetastasis activities in a cohort of xenograft tumor models compared with a5-IgG1/wt, a5-IgG2, a5-IgG4, and M200 mAbs (Figs.3C and D, 4A, and 5C).These are consistent with reports for other antibodies carrying enhanced FcgR binding capacity (30,31,33,46,47).We further provided evidence on the molecular and cellular levels (Figs.3C and D, 4B, and 5A; Supplementary Fig. S4A), and pharmacologic level (Figs.3D and 4C) that NK-mediated ADCC and macrophage-mediated ADPC significantly and positively impacted efficacy of PF-5412.Importantly, we showed a rapid immobilization of macrophages in the tumor (as early as 24 hours postdose; Fig. 4D) prior to measurable TGI, suggesting an active role of macrophages in tumor elimination.In addition, we showed that ADCC/ADPCmediated activity is dependent on the presence and density of the antigen (a5b1) expressed on target cells (Figs. 2B and  5A).To our knowledge this is the first report with comprehensive in vitro and in vivo characterization of an ADCC/ ADPC-enhanced therapeutic antibody targeting solid tumors.
We observed that PF-5412 produced a similar ADCC max with donor PBMCs regardless of the allele variation of FcgRIIIa (Fig. 2D).Cancer patients with the low-affinity allele FcgRIIIa/158F have been associated with worse responses to rituximab (23,24), trastuzumab (25), and cetuximab (26).In PF-5412, the binding affinities for the low-affinity FcgRIIIa/ 158F and FcgRIIa/131R were more significantly enhanced than for the higher affinity counterparts (Table 1; Supplementary Fig. S2A).As a result, the impact of allelic variations of the FcgRs on the degree of ADCC/ADPC was reduced.Similar observations were also reported for a low-fucose variant of rituximab (48).In addition, our in vitro data suggest that effector function enhancement in PF-5412 lowered the target density threshold required to produce cytolytic and antitumor efficacy compared with a5-IgG1/wt (Figs.2A and B, 3D, 4A and B, and 5B and C).Given the prevalence of FcgRIIIa/158F allele (40% V/F and 40% F/F) in the general population (33) and the varying levels of a5b1 expression in patient tumors (unpublished observation), we believe that an ADCC/ADPCenhanced mAb has the potential to impact a much larger and heterogeneous patient population in the clinic than would a wild-type mAb.It is also worth noting that although U87MG cells were used in many of the experiments, the potential indication of PF-5412 should not be limited to glioblastoma.In this study, the selection of the preclinical models for proof-ofprinciple studies was mainly based on the expression of the target in a model.
One of the strategies for rational combination is to simultaneously target multiple processes involved in cancer progression to provide meaningful benefit and circumvent resistance.Indeed, in the aggressive U87MG-luc mode, where single-agent sunitinib has not shown robust activity (unpublished data and ref. 49), combining a low dose of PF-5412 (1 mg/kg), but not M200, with a low dose of sunitinib (half of the projected clinically equivalent dose) produced a robust and   D, efficacy assessment of low-dose PF-5412 or M200 plus sunitinib (PO, QD) in the C.17 SCID/beige mice bearing U87MG-luc tumors.A synergistic effect was observed for PF-5412 þ sunitinib but no efficacy enhancement for M200 þ sunitinib.The table of statistics is presented in Supplementary Figure S5B.
greater than additive antitumor efficacy (tumor stasis; Fig. 5D).One possible explanation for the synergistic antitumor activity may be that sunitinib treatment induced tumor hypoxia, which in turn upregulated a5b1 (7) and sensitized a5b1-mediated tumor survival pathway.PF-5412 would be able to disrupt these events by blocking the augmented a5b1 signaling and eliciting strong ADCC/ADPC.Another hypothesis may include phagocytic macrophage subclass sequestration by PF-5412 to mitigate the recruitment and activation of bone marrow-derived immunosuppressive cells (including monocytes and proinflammatory macrophages) known to contribute to resistance to antiangiogenic therapies (50).
Research is underway to gain further understanding of the observed synergism.In summary, our data imply that such a rational combination strategy may be a safe and robust approach for the treatment of aggressive tumors.From a clinical development prospective, M200, an a5b1 neutralizing mAb, has shown that inhibiting integrin a5b1 in the clinic is safe and may provide incremental benefit to some cancer patients.Given the observations from this study, we believe that a5b1 neutralization may be necessary, but not sufficient, to produce a robust and sustained antitumor efficacy.Thus PF-5412 may represent a new-generation integrin-targeting modality that as a single agent or in combination may deliver a meaningful benefit to a broader cancer patient population in the clinic.

Figure 1 . 4 Figure 2 .
Figure 1.In vitro potency and activity of PF-5412.A, dose-dependent binding of PF-5412 and a5-IgG1/wt to HUVEC measured by FACS.B, dosedependent inhibition of HUVECs adhesion to Fn by anti-a5 mAbs.Similar potency was observed for PF-5412 compared with a5-IgG1/wt, a5-IgG2, and M200.C, dose-dependent inhibition of tubule formation in a human ECs/fibroblast cell coculture assay.Axitinib is a VEGF receptor tyrosine kinase inhibitor used as a positive control.D, inhibition of both total and phospho-FAK in U87MG cells by PF-5412 in a Western blotting assay.

Table 1 .
Key in vitro attributes of PF-5412