Direct targeting of αvβ3 integrin on tumor cells with a monoclonal antibody, Abegrin™

The humanized monoclonal antibody Abegrin™, currently in phase II trials for treatment of solid tumors, specifically recognizes the integrin αvβ3. Due to its high expression on mature osteoclasts, angiogenic endothelial cells, and tumor cells, integrin αvβ3 functions in several pathologic processes important to tumor growth and metastasis. Targeting of this integrin with Abegrin™ results in antitumor, antiangiogenic, and antiosteolytic activities. Here, we exploit the species specificity of Abegrin™ to evaluate the effects of direct targeting of tumor cells (independent of targeting of endothelia or osteoclasts). Flow cytometry analysis of human tumor cell lines shows high levels of αvβ3 on many solid tumors, including cancers of the prostate, skin, ovary, kidney, lung, and breast. We also show that tumor growth of αvβ3-expressing tumor cells is inhibited by Abegrin™ in a dose-dependent manner. We present a novel finding that high-dose administration can actively impair the antitumor activity of Abegrin™. We also provide evidence that antibody-dependent cellular cytotoxicity contributes to in vitro and in vivo antitumor activity. Finally, it was observed that peak biological activity of Abegrin™ arises at serum levels that are consistent with those achieved in clinical trials. These results support a concept that Abegrin™ can be used to achieve selective targeting of the many tumor cells that express αvβ3 integrin. In combination with the well-established concept that αvβ3 plays a key role in cancer-associated angiogenesis and osteolytic activities, this triad of activity could provide new opportunities for therapeutic targeting of cancer. [Mol Cancer Ther 2006;5(12):3122–9]


Introduction
Biological therapy is providing new weapons for oncologists that allow the selective targeting of cancer. In particular, monoclonal antibodies represent a successful and validated approach for targeting cells within the tumor environment (1). Examples include direct targeting of tumor cells (e.g., cetuximab) or components of the microenvironment (e.g., bevacizumab). Importantly, monoclonal antibodies have proven to be generally safe and effective.
In the search for potential targets for monoclonal antibodies, one could seek to identify either direct or indirect tumor targets. Examples of direct tumor targets include HER-2, epidermal growth factor receptor, CD20, and other molecules that are preferentially overexpressed in tumor cells. Alternatively, indirect tumor targets could include molecules within the tumor environment that nonetheless critically control tumor growth or survival, the most well-known being targets on angiogenic blood vessels.
Abegrin TM (etaracizumab; previously known as VitaxinR or MEDI-522) is a monoclonal antibody that was chosen for its unique ability to selectively target multiple and different cell types, all of which are relevant to cancer pathophysiology. The target for Abegrin TM is a v h 3 , an integrin that is overexpressed on tumor cells, angiogenic blood vessels, and osteoclasts. Antagonists of a v h 3 (including antibodies and small molecules) have been studied most extensively for their antiangiogenic properties (2 -5). In addition, a v h 3 is expressed on tumor cells and osteoclasts and is believed to play an important role in bone metastasis and subsequent resorption (6). These findings suggest a potential role for a v h 3 in the pathology of osteolytic diseases, including breast cancer, prostate cancer, and multiple myeloma. Finally, a v h 3 is overexpressed on a variety of different tumor types. For example, a preponderance of data suggests that a v h 3 is found to be overexpressed in metastatic melanoma, glioma, multiple myeloma, ovarian, renal, and breast cancer (7 -13).
The murine monoclonal antibody, LM609, binds human a v h 3 and was humanized and affinity matured based on its abilities to mediate antiangiogenic effects in preclinical models (14). The resulting antibody Abegrin TM has been tested in clinical trials of various solid tumors where it appears to be without significant toxicity. The trials also indicated that Abegrin TM may have effects on tumor perfusion and may exhibit clinical activity in renal cell cancer (15). Based on these findings, Abegrin TM is presently being investigated in clinical trials of androgen-independent prostate cancer and metastatic melanoma. In our present report, we seek to understand further mechanisms by which Abegrin TM functions in cancer. In particular, we have returned to preclinical investigation to address the question of whether Abegrin TM could provide additional opportunities to directly target cancer cells that overexpress a v h 3 . In addition, our studies reveal new opportunities for cancer cell targeting and identify novel mechanisms and dosing properties that could provide a wider understanding for targeting of cancer with monoclonal antibodies.

Materials and Methods
Cell Lines All cell lines used in this study were obtained from the American Type Culture Collection (Manassas, VA) and cultured according to specifications.
Radioligand Binding Assay For the concentration dependence of binding, [ 125 I]Abebegrin TM was added at various concentrations to M21 cells at 10 7 cells/mL and incubated with cells at 4jC for 30 min with shaking every 3 to 5 min. Duplicate 75 AL aliquots were layered over 200 AL phthalate oil mixture [Dibutylphthalate/bis(2-ethylhexyl)phthalate = 60:40; both from Sigma, St. Louis, MO] and immediately centrifuged for 1 min at 13,000 rpm in a microfuge. Two to 50 AL of the supernatant were removed and measured for 125 I radioactivity to determine the concentration of free ligand. The tubes were then snap frozen in dry ice, and the tips containing the pellet were cut and counted for radioactivity. Duplicate samples were within 10% of the mean for the data reported. Nonspecific binding was measured in samples, in which >20-fold excess unlabeled Abegrin TM was added 20 min before adding [ 125 I]Abegrin TM ; it was typically V10% of the total bound. Average K D and binding sites per cell are calculated from nine independent experiments.
Flow Cytometry Adherent cells were harvested using enzyme-free dissociation buffer (Invitrogen, Carlsbad, CA), washed once, and incubated with Abegrin TM (1 Ag/10 6 cells) at 4jC for 1 h in fluorescence-activated cell sorting buffer (PBS + 2% fetal bovine serum). Cells were then washed in fluorescenceactivated cell sorting buffer and stained with a goat F(ab ¶) 2 anti-human IgG R-phycoerythrin secondary antibody (used at 1:250 dilution; Southern Biotech, Birmingham, AL) at 4jC for 45 min. Subsequently, cells were washed twice and resuspended in 500 AL fluorescence-activated cell sorting buffer for analysis. Live whole cells were gated using propidium iodide (Roche, Indianapolis, IN) staining, and 10,000 events were acquired using a BD FACSCalibur flow cytometer (BD Biosciences, San Jose, CA). Data were analyzed using BD CellQuest software (BD Biosciences).
Immunohistochemistry Placental tissue, a rich source of a v h 3 , was collected either immediately after parturition or from freshly sacrificed, late-term pregnant animals. Tissue fragments, f1 cm 3 , were frozen in optimal cutting temperature, and thin sections were stained with Abegrin TM (10 Ag/mL) at Pathology Associates, Inc. (Frederick, MD).
Epitope Mapping Human, rat, mouse, and rabbit h 3 integrin sequences corresponding to residues 164 to 202 were aligned. Specific residues within human h 3 integrin were mutated to the corresponding residue from the rat sequence using QuikChange Site-Directed Mutagenesis kit from Stratagene (La Jolla, CA) as per the manufacturer's instructions. The resulting mutant human h 3 integrin-containing plasmids were transfected into HEK-293 cells, using the endogenous human a v chain for dimerization of the integrin receptor. The mutant h 3 clones were analyzed for binding to Abegrin TM by flow cytometry as described above. A human h 3 integrin-specific antibody, clone PM6/13, from Chemicon (Temecula, CA) was used as a control to compare transient expression levels of the mutants.
In vivo Tumor Xenograft Studies Athymic nu/nu (Harlan, Somerville, NJ), scid (Harlan), or scid/nod (Taconic, Germantown, MD) female mice 4 to 6 weeks of age were injected s.c. with 5 Â 10 6 tumor cells. Treatments of PBS, IgG control, bevacizumab (McKesson Bioservices Corp., Rockville, MD), or Abegrin TM were injected two or three times weekly in the i.p. cavity as indicated in the figure legends and continued throughout the study. Only scid and scid/nod mice with <100 ng/mL of mouse IgG as assessed by ELISA (Pierce Biotechnology, Rockford, IL) were used for the study.
Antibody-Dependent Cellular Cytotoxicity Assay Antibody-dependent cellular cytotoxicity (ADCC) was assayed in a 4-h nonradioactive lactate dehydrogenase release assay (Promega Corp., Madison, WI). Briefly, target cells were resuspended in RPMI 1640 containing 5% fetal bovine serum at 2 Â 10 5 per mL and distributed into 96-well U-bottomed plates (1 Â 10 4 per well). The target cells were preincubated with serial dilution of antibodies (50 AL/well) for 20 min on ice before the human effector cells (100 AL/well) were added. Human effector cells were peripheral blood mononuclear cells purified from healthy donors using lymphocyte separation medium (MP Biomedicals, Solon, OH). After a 4-h incubation at 37jC in 5% CO 2 , plates were centrifuged. Fifty microliter supernatant was transferred to a 96-well flat-bottomed plate and incubated with 50 AL lactate dehydrogenase substrate for 30 min at room temperature. The reaction was stopped by adding a 50-AL stop solution. The samples were measured at 490 nm, and the percentage cytotoxicity was calculated according to the formula: % specific lysis = 100 Â (E X À E spon À T spon ) / (T max À T spon ), where E X represents the release from experimental wells, E spon is the spontaneous release of effector cells alone, T spon is spontaneous release of target cells alone, and T max is the maximum release from lysed target cells.
Human IgG1Quantitation ELISA Three mouse serum samples per treatment group were collected 4 days following the final treatment of the A498 tumor xenograft studies. They were diluted in PBS/T-BSA [1Â PBS (pH 7.4) + 0.1% Tween 20 + 0.5% bovine serum albumin] and added in duplicate to wells of BD BioCoat anti-human IgG ELISA plates (BD Biosciences). Abegrin TM serially diluted in PBS/T-BSA served as the standard. Plates were incubated at room temperature for 60 min and then washed thrice with PBS/T [1Â PBS (pH 7.4) + 0.1% Tween 20]. Goat anti-human IgG horseradish peroxidase conjugate (Pierce Biotechnology) diluted 1:60,000 in PBS/T-BSA was added to each well and incubated at room temperature for 60 min. Subsequently, plates were washed thrice with PBS/T and the binding of human IgG was detected using TMB substrate (BioFX Laboratories, Owing Mills, MD). Color development was analyzed by adding 0.18 mol/L (1%) H 2 SO 4 to each well and reading the plate at 450 nm.

Results
AntibodyTargeting of A v B 3 onTumor Cells Based on evidence that a v h 3 is overexpressed on many tumor types, we asked if a v h 3 might provide a role in direct tumor targeting independent of its established role in angiogenesis. The successful accomplishment of these studies required further characterization of a v h 3 expression and Abegrin TM binding to a v h 3 . RIAs revealed that Abegrin TM bound to cell surface a v h 3 with high affinity (K D of 2.29 nmol/L; Fig. 1A). Scatchard analysis on M21 melanoma cells revealed an average of 223,000 available binding sites per cell. Flow cytometric analyses revealed strong Abegrin TM immunoreactivity on multiple cell models that were representative of many different tumor types, including carcinomas of the breast, prostate, kidney, and skin (Fig. 1B). Melanoma models were emphasized based on a wealth of published evidence that has linked a v h 3 overexpression in clinical specimens of metastatic melanoma. Flow cytometric analyses revealed that the estimated density of a v h 3 receptors on melanoma cells ranged from 83,000 (SK-MEL-5) to >250,000 (A375) molecules per cell.
The studies above confirmed that a v h 3 is present on the surface of many tumor-derived cell lines. To focus on the role of a v h 3 on tumor cells, it would be necessary to identify antibodies that would not interact with cells of the host microenvironment (e.g., endothelial cells). For this, the species specificity of Abegrin TM was determined via immunohistochemical staining of placental trophoblasts. This cell type proved to be particularly useful because a v h 3 is ubiquitously expressed on mammalian placental trophoblasts. Immunohistochemical analyses revealed strong Abegrin TM immunoreactivity with human and moderate to weak binding to guinea pig, hamster, or rabbit a v h 3 (Fig. 1C). In contrast, Abegrin TM was completely unable to react with either mouse or rat a v h 3 .
The species-dependent binding of Abegrin TM provided an opportunity to further characterize the epitope. Based on knowledge that binding to LM609 is imparted by residues 164 to 202 of the h 3 chain (16), the homologues of h 3 integrin were aligned from multiple species. Using the species specificity information, we were able to focus on several residues in the h chain that distinguish human from mouse or rat a v h 3 . Each difference was introduced into human h 3 integrin by site-directed mutagenesis, and the resulting constructs were expressed in HEK-293 cells, which endogenously express the human a v chain. The human a v h 3 integrin that contained a mutation in residue 171 of the h 3 chain was not recognized by Abegrin TM , suggesting that this residue is critically important in the binding and species specificity of Abegrin TM (Fig. 1D). Equal cell surface expression of h 3 integrin was determined by staining transfected cells with a h 3 integrin-specific antibody (data not shown).
Abegrin TM Antitumor Activity and Mechanism of Action Based on the studies above, we considered that human tumors implanted within a murine background could provide the opportunity to evaluate direct targeting of tumor cells by Abegrin TM . The use of human tumor cells that expressed a v h 3 , in combination with the species specificity of Abegrin TM , provided an opportunity to ask if Abegrin TM could confer antitumor effects independent of its role in angiogenesis. Biodistribution studies with labeled Abegrin TM showed specific targeting of the Targeting of a v b 3 Integrin on Tumor Cells with Abegrin TM antibody to a v h 3 -expressing human tumor cells in mouse xenografts (17). Analysis of xenografts revealed that Abegrin TM targeting of tumor cells is sufficient to decrease tumor growth in vivo. For example, treatment of mice with Abegrin TM prevented the subsequent growth of tumors by M21 melanoma cells (Fig. 2A). These effects were not unique to any particular model as Abegrin TM decreased the growth of A375 melanoma cells (Fig. 2B) as well as experimental models of renal and prostate cancer.
Once we had established a proof of principle that Abegrin TM can inhibit tumor growth, we then sought to identify mechanisms by which Abegrin TM acts on tumor cells. Not surprisingly, in vitro assays revealed that Abegrin TM treatment of tumor cells prevented them from interacting with underlying extracellular matrix proteins (e.g., fibrinogen and other RGD-bearing extracellular matrix proteins; data not shown) as has been shown with other a v h 3 antagonists (14,18,19). This finding is consistent with the well-established concept that a v h 3 integrin confers an important cell adhesion function.
We also considered that host effector mechanisms (e.g., ADCC and complement-dependent cytotoxicity) could contribute to the antitumor activity observed in xenograft models. To examine the relative importance of ADCC, xenograft analyses were done using scid/nod mice, which have a deficiency in monocytes and natural killer cells, thereby compromising their ADCC activity. Bevacizumab (Avastin) provided a control as it binds a soluble ligand and thus does not efficiently mediate ADCC. As expected, Abegrin TM treatment of scid mice decreased A375 melanoma tumor growth by at least 70% relative to isotypetreated controls (P < 0.0001; Fig. 3A). In contrast, tumor inhibition was marginal (<25%) when similar studies were conducted within a scid/nod background (Fig. 3B). As expected, the antitumor effects of bevacizumab were not altered in the nod background. Altogether, these results suggest that host effector mechanisms, and ADCC in particular, critically contribute to the antitumor activity of Abegrin TM in vivo.

Dose-Limited Antitumor Activity
In the course of conducting xenograft analyses with Abegrin TM , we found that the tumor-inhibitory properties of Abegrin TM decreased when drug was given at relatively high doses (>20 mg/kg). We termed this behavior ''dose-limited antitumor activity.'' Whereas the antitumor activity of Abegrin TM steadily increased when animals were treated with 1 to 10 mg/kg, the biological activity of Abegrin TM decreased at doses >20 mg/kg (Fig. 4). This outcome was not limited to any particular cell line or tumor type as comparable findings were obtained using multiple and different models (melanoma, prostate, and renal; data not shown).
To begin addressing the mechanistic basis for ''doselimited activity,'' we considered that ADCC critically contributes to Abegrin TM tumor inhibition. It was important to also determine whether ''dose-limited activity'' might simply represent an artifact of using a murine model system. To address both questions, ADCC assays were conducted ex vivo using human effector cells. The results of ADCC assays were consistent with the dose response from xenograft analyses. Specifically, Abegrin TMmediated ADCC of tumor cells increased with dose, with optimal killing observed at 1 Ag/mL (Fig. 5A). With further increases in dose, however, ADCC activity decreased dramatically, with much less killing at 100 Ag/mL. Comparable findings were observed at different ratios of effector to target cells (from 5:1 to 50:1; data not shown). The dose-limited activity was also observed regardless of whether peripheral blood lymphocytes or purified natural killer cells were used as effector cells or whether A498, A375, or M21 cells were used as target cells in the assay (data not shown). We then asked whether this ''dose-limited activity'' might be specific for Abegrin TM or could antibodies to other cell surface proteins elicit this same response. To test this, ADCC assays were done using 3F2, a human IgG1 antibody specific for human EphA2 (Fig. 5A). An identical pattern of dose-limited activity was observed, suggesting that this effect is not specific to Abegrin TM or a v h 3 integrin.
We then asked whether inclusion of high-dose Abegrin TM would confer a ''dominant-negative'' phenotype to prevent other human IgG1 antibodies from mediating ADCC. This could arise if Abegrin TM saturated and inhibited effector cell function. To test this, ADCC assays were done using a v h 3 -deficient A549 lung carcinoma cells. 3F2 was again used for ADCC assays in the presence or absence of various doses of Abegrin TM . The inclusion of Abegrin TM , even at concentrations as high as 100 Ag/mL, did not decrease 3F2-mediated ADCC (Fig. 5B). Controls confirmed that Abegrin TM itself also did not mediate ADCC, which is consistent with the absence of a v h 3 on the target cells. These results indicate that high-dose Abegrin TM does not function as a dominant inhibitor of ADCC by acting on effector cells and that free unbound effector cells are in excess even at high antibody concentrations.
For our final set of studies, we considered that high-dose antibody administration might alter the serum levels of Abegrin TM . For example, bolus administration might cause Figure 3. Natural killer cells and monocytes are required for Abegrin TM activity. scid (A) or scid/nod (B) mice were implanted with A375 melanoma tumor cells and treated with 10 mg/kg bevacizumab or Abegrin TM twice weekly. Abegrin TM antitumor activity was attenuated in scid/nod versus scid mice (treated/control: 0.79 and 0.29, respectively), whereas bevacizumab activity was unchanged (treated/control: 0.40 and 0.35, respectively). P < 0.0005 for bevacizumab (scid and scid/nod ) and P < 0.00009 for Abegrin TM (scid ), Student's t test. Targeting of a v b 3 Integrin on Tumor Cells with Abegrin TM antibody aggregation in vivo, thereby accelerating its clearance and limiting its availability. We determined that the concentrated Abegrin TM stocks used for these experiments were free of aggregates (data not shown) but could not exclude the possibility that such an event might occur in vivo. A greater understanding of the serum levels was also important as Abegrin TM is currently being investigated in clinical trials of melanoma and prostate cancer. To ask how ''dose-limited activity'' in xenograft models (at 30 mg/kg, given twice weekly) relates to the situation encountered in the clinic (at 8 mg/kg given weekly), the serum level of Abegrin TM in mice was compared with the known pharmacokinetic profile in humans. The trough serum concentration of human IgG (Abegrin TM ) is defined by the serum concentration just before the administration of the next dose and was measured by ELISA in mice during week 8 of a xenograft study using A498 tumor cells (Fig. 6). The circulating levels of Abegrin TM were proportional to the dose administered, with mean trough concentrations of 11 and 192 Ag/mL in animals that had been treated with 1 or 10 mg/kg. Trough levels of 574 Ag/mL were attained in animals treated with 30 mg/kg Abegrin TM , indicating that lack of antitumor activity at this dose was not due to increased clearance of the antibody. Tumor homogenates also showed a comparable increase in IgG levels with increasing dose, suggesting that diffusion into the tumor was not altered at the high dose (data not shown). Notably, the serum levels of Abegrin TM from human trials at a dose of 8 mg/kg (peak and trough of approximately 300 and 100 Ag/mL, respectively) coincided with doses associated with optimal antitumor activity in our in vivo tumor studies (15).

Discussion
The major finding of our present study is that antibody targeting of a v h 3 integrin on tumor cells can provide usefulness for targeting cancer independent of, and in addition to, antibody targeting of a v h 3 on angiogenic blood vessels. We have also increased our understanding of the distribution of a v h 3 receptors and the mechanisms of action for Abegrin TM . In particular, we provide evidence that Abegrin TM invokes host defense mechanisms and that ADCC critically contributes to Abegrin TM antitumor activity. Finally, we present a novel finding that high-dose administration can actively impair the antitumor activity of Abegrin TM . This new information has important implications not only for the future development of a v h 3 integrin inhibitors but also for understanding the mechanism of action for monoclonal antibody-based therapies.
Our present findings are novel, in part, because they shed new light about potential therapeutic opportunities for antibody targeting of a v h 3 in cancer. Previously, much of the interest surrounding this integrin centered on the overexpression and functional relevance of a v h 3 on tumor-associated endothelial cells. This information triggered considerable interest in a v h 3 as a target for antiangiogenic cancer therapy and, indeed, this was the original rationale supporting Abegrin TM as a cancer therapeutic. a v h 3 is also known to be overexpressed on  osteoclasts and functionally relevant to bone resorption. This has stimulated considerable interest in the application of antagonists of a v h 3 , including Abegrin TM , to osteolytic tumor types, such as melanoma, multiple myeloma, and breast cancer. Our present study widens this knowledge by presenting evidence for Abegrin TM activity against tumor cells themselves. In some ways, the ''triad'' of Abegrin TM activity may be likened to combination targeted therapy through simultaneous blocking of tumor growth and survival, angiogenesis, and bone metastasis. All of these strategies can be achieved with a single agent that shows minimal toxicity. Consequently, future studies will be focused on the potential for combining Abegrin TM with known cytotoxic agents to evaluate potential opportunities for synergistic therapies and to ask if Abegrin TM might have applications for drug sensitization in treatment-refractory indications.
Another novel aspect of our present study is the demonstration that ADCC is a major mechanism that contributes to the activity of Abegrin TM directed against the tumor cells. Before this report, a v h 3 was generally understood to play an important role in cell adhesion and signaling. Indeed, a v h 3 has been shown to critically control many different aspects of malignant character, including cell migration, invasion, and survival (e.g., via anoikis) and antagonists of a v h 3 (both antibodies, such as Abegrin TM and small molecules) are presently being investigated in clinical trials (5). Although our present studies do not preclude that these mechanisms contribute to antitumor activity, they do identify an important role for ADCC for targeting a v h 3 . This information is consistent with findings with other antibodies (rituximab and trastuzumab) and also has potential importance for identifying improved means of a v h 3 targeting (20). Specifically, small-molecule antagonists of a v h 3 function are incapable of mediating ADCC and this may limit their therapeutic potential.
Perhaps, the most intriguing finding of our present report is the demonstration of ''dose-limited'' antitumor activity. The decreased activity of Abegrin TM at the highest dose levels has important implications for our understanding of basic mechanistic activities of monoclonal antibodies and conveys potentially important information that is highly relevant to clinical applications. Importantly, this dose-limited activity was not unique to Abegrin TM or a v h 3 as we have observed comparable effects with other antibodies and with different antigens. The dose-limited activity described herein is reminiscent of the prozone-like effect observed in passive immunization approaches against human pathogens (21). Several studies have shown that sera or purified IgG from patients or infected animals elicited a prozone-like response in ADCC (22 -25), complement-dependent cytotoxicity (26), and neutralization (27) assays in vitro. Furthermore, prozone-like effects have been observed in vivo following infection with various pathogens, such as Streptococcus pneumoniae and Cryptococcus neoformans (28,29). Much less is known about how this phenomenon affects passive immunization approaches in other disease areas, such as oncology.
Regardless of why dose-limited activity occurs, the outcome could have particular importance for designing and applying antibody-based therapeutics. Historically, biologics generally have relatively few side effects and many clinical trials fail to reach a maximum tolerated dose. Consequently, the dose selected for further clinical application may be defined by other considerations (cost and availability). If the findings herein have a more general application to other antibodies, then important changes will be necessary to identify optimal doses, particularly with biological agents that function via ADCC. Within this context, we have conducted extensive pharmacokinetic analyses with Abegrin TM and are fortunate in that the clinical doses presently under investigation in clinical trials seem to fall within the optimal zone of preclinical activity. Nonetheless, our present findings indicate a need to determine and apply similar criteria to other biologicsbased therapeutic agents.