Mouse placental integrin α5β1 was purified by affinity chromotagraphy using rat anti-mouse α5β1 antibody BMC5 (Chemicon) essentially as described 25. Briefly, antibody was coupled to CNBr-coupled Sepharose (Pharmacia) and mouse placenta was homogenized using a Polytron tissue homogenizer in lysis buffer [200 mM octyl-β-gluocopyranoside, 1 mM phenylmethylsulfonyl fluoride, 1 mM MgCl₂, 0.5 mM CaCl₂, 150 mM NaCl, 25 mM HEPES (pH 7.0)], followed by incubation on ice. After centrifugation, the cleared lysate was loaded onto a BMC5-Sepharose column and washed with 20 column volumes of lysis buffer with the octyl-β-gluocopyranoside substituted with 0.1% Nonidet P-40. Integrin was eluted at low pH in a buffer containing 10 mM EDTA and dialyzed against phosphate buffered saline.

Total RNA (1 μg) isolated from mouse tissue was solubilized in Trizol™ (Invitrogen) and cDNA was produced using SuperScript II (Invitrogen). Full-length α5 and β1 were amplified by PCR using gene-specific primers (Qiagen N.V.). PCR products were subcloned into the pCR4 TOPO vector (Invitrogen) and verified by sequencing. To generate the α5 and β1-Fc fusion genes, regions encoding the extracellular domains of the α5 and β1 integrin genes were isolated by restriction digest and individually subcloned into the expression vector DEF38 (ICOS) to create an in-frame fusion between each of the extracellular domains and the constant region of the human γ1 immunoglobulin heavy chain gene. A stable cell line was generated by co-transfecting 293 cells with a 50:50 mix of α5-Fc and β1-Fc plasmid DNAs using FuGene (Roche). Conditioned medium was harvested and Fc fusion protein was purified by standard methods using Protein G Sepharose.

Female Sprague-Dawley rats (Simonsen) were immunized intraperitoneally with α5β1 purified from mouse placenta or with a mouse α5β1-Fc fusion protein. Monoclonal antibodies were generated by standard techniques fusing spleen cells from immunized mice with an NSO-derived fusion partner (American Type Culture Collection). A panel of α5β1-specific antibodies was identified by ELISA using the Fc-fusion protein and by flow cytometry analysis of binding to mouse endothelial cells and cell lines.

Purified α5β1-Fc (100 ng) was plated into wells in 50 mM Na₂CO₃ (pH 8.6), 0.5 mM each of CaCl₂, MgCl₂, and MnCl₂ at 4°C overnight. Plates were blocked with 5% BSA in PBS for 1 hr with shaking. Following washes in PBS with 0.05% Tween-20, wells were incubated for 1 hr with serial dilutions of candidate antibodies or control rat IgG (Jackson ImmunoResearch) in 1% BSA in PBS with 0.05% Tween-20. Wells were then washed and incubated with 100 μL/well of the goat anti-rat IgG-HRP (1:5,000) for 1 hr. Washed plates were incubated with TMB substrate (Sigma), developed with 650 Stop Reagent (BioFX), and read at A₆₅₀.

Cellular mouse fibronectin-coated plates (250 ng in coating buffer overnight at 4°C) were washed with PBS containing 0.05% Tween-20, and blocked as described above. Plates were incubated with α5β1-Fc fusion protein in the presence of various concentrations of the indicated antibodies for 1 hr at room temperature. Following 2 washes, plates were further incubated with goat anti-human IgG Fc-HRP (1:5000; Jackson ImmunoResearch) at room temperature for 1 hr. Plates were then washed 5 times and incubated with TMB substrate, developed, and read at A₆₅₀.

Cells were removed with 20 mM EDTA in Tris-HCl (pH 8.0) and blocked by centrifugation in HBSS containing 3% heat-inactivated FBS, 1% normal goat serum (Sigma) and 1% BSA at 4°C for 5 min. Cells were incubated for 1 hr at 4°C with the indicated supernatant or antibody (10 μg/mL) in FACS buffer (PBS containing 0.1% BSA). Excess mAb was removed by centrifugation and cells were resuspended in FACS buffer containing anti-rat IgG-PE secondary antibody (Southern Biotech). After an additional wash, fluorescence intensity was measured on a FACSCalibur flow cytometer (Becton Dickinson).

BD HTS FluoroBlok 96-well plates (top plate) were coated with mouse plasma fibronectin (10 μg/ml; Upstate) in PBS and air-dried. 200 μl migration medium (MM; RPMI + 0.1% BSA) with or without antibody was dispensed into a BD Falcon 96-Square Well Flat Bottom Assay plate (bottom plates) and 10,000 cells/well in 50 μl was added to the top plate in MM. After lowering the top plate into the lower plate, cells were incubated at 37°C at 5% CO₂ for 4 hr. Cells were stained with 2 μM Calcien-AM (Invitrogen) in MM and visualized using the Discovery-1 High Content Screening System (Molecular Devices).

Affinities between α5β1-Fc and anti-integrin antibodies were analyzed using a BIAcore 3000 essentially as described 2627. Goat anti-rat Fc antibodies were immobilized on a Research Grade CM5 chip using an amine coupling kit (BIAcore). Rat-anti-mouse α5β1 antibodies were captured onto goat anti-rat Fc surfaces, followed by injection of mouse α5β1 in running buffer [(10 mM HBS, 2 mM CaCl₂, 1 mM MnCl₂, 700 mM NaCl (pH 7.4)] at a flow rate of 30 μL/min at 25°C. Association phase occurred over 3 min and dissociation over 1.5 hr. Kinetics of binding was calculated from data at 6 different concentrations of analyte (512 nM, 128 nM, 32 nM, 8 nM, 2 nM, 0.5 nM), using the BIAevaluation program. Each goat anti-rat Fc surface was regenerated at the end of each cycle by a quick injection of 30 mM HCl. Double-referencing was applied to eliminate responses from the reference surface and buffer-only control. Affinity constant (K_D) was obtained by simultaneously fitting the association and dissociation phases of the sensorgram from the analyte concentration series using the 1:1 Langmuir model from the BIAevaluate software.

Primary mouse endothelial cells were incubated with anti-mouse α5β1 antibodies at 10 μg/ml for 16 hr, after which cell membrane phosphatidylserine was detected using Oregon Green 488 conjugated Annexin V. Harvested cells were washed with Annexin binding buffer [(ABB; 10 mM HEPES, 140 mM NaCl, 2.5 mM CaCl₂ (pH 7.4]) and incubated in ABB (100 μl) containing Annexin V (10 μl). Following washing, cells were suspended in ABB containing 0.5 μg/mL propidium iodide (200 μl) and assessed by flow cytometry.

HUVEC or murine lung EC's (5 × 10⁵ cells/ml) were seeded in Hank's balanced salt solution (HBSS) containing 3 mg/ml fibrinogen and 200 μg/ml aprotinin (Roche) at 37°C, as described previously 21. For each condition, EC/fibrinogen/aprotinin mixture (500 μL) was quickly transferred to a 24-well tissue culture plate containing α-thrombin (1 U) and gently mixed. The resulting fibrin matrices were polymerized at 37°C for 20 min. Antibody was added in Medium 200 [1 ml, without LSGS and containing 20% human or mouse serum (Rockland Immunochemicals), and rhTGF-α (0.01 μg/ml), rhHGF (0.1 μg/ml) and rhVEGF and/or rmVEGF (0.1 μg/ml)]. Tubes formed over 6 days, with one medium change (Medium 200 plus serum only) at day 3. For visualization of tubes, matrices were fixed with 4% formaldehyde for 4 hr and stained overnight with 1 ml phalloidin-Alexa 488 (1.75 U/ml) in 50% fetal calf serum (FCS) containing 0.25% saponin. Vessel formation was visualized using the Discovery-1 High Content Screening System (Molecular Devices) and quantified using the accompanying software.

Xenograft tumors derived from cells inoculated subcutaneously in ICR-SCID or SCID-Beige mice (Taconic Farms) were frozen in OCT compound and stored at -70°C. Cryostat tissue sections (4-5 μm) were fixed in acetone for 10 min, air dried and incubated in 0.03% H2O2 for 10 min. After successive incubations with Avidin block, Biotin block and Protein block solutions for 15 min each, samples were treated with the indicated antibodies or biotin-conjugated rat anti-mouse CD31 mAb MEC 13.3 or biotin-conjugated IgG2a isotype control (2.5 μg/ml; BD Pharmingen) for 60 min. Sections were developed using the Vectastain Elite ABC kit (Vector Laboratories) and stable diaminobenzidine (Dako). All staining procedures were performed using a Dako Autostainer at room temperature.

Six- to eight-week-old ICR SCID or SCID-Beige female mice, obtained from Taconic Farms and maintained in micro-isolator cages, received subcutaneous injections on the right flank of 5 × 10⁶ A673 human rhabdomyosarcoma or 1 × 10⁷ SVR murine angiosarcoma cells. Tumors were allowed to establish for 7–18 days, reaching an average of 50–100 mm³, as determined by caliper measurement. Animals were distributed into groups of ten and received vehicle (PBS) or rat anti-mouse integrin α5β1 antibody (200 μl at 1.0 mg/ml). Reagents were delivered by intraperatoneal injection twice or thrice weekly for the duration of the studies. Tumor volume was measured twice weekly, and clinical and mortality observations were performed daily according to Institutional Animal Care and Use Committee regulations.

To generate antibodies directed against murine integrin α5β1, Sprague-Dawley rats were immunized with mouse α5β1-Fc fusion protein or with affinity-purified integrin from mouse placenta. Supernatants from the resulting hybridomas were subjected to a number of assays designed to identify clones that produced antibodies encompassing characteristics that define volociximab, including heterodimer specificity, blocking activity, high affinity binding to the Fc fusion protein and binding to endothelial cells 21. Over sixty clones were screened for these functional properties.

Selected antibodies represented a range of binding characteristics. These included antibodies specific for α5, β1 or α5β1 heterodimer, as determined by ELISA (Figure 1A), and spanned a range of relative affinities (Figure 1B) and abilities to block binding of α5β1 to fibronectin (Figure 1C). Interestingly, one antibody, 321.1, increased binding of integrin to fibronectin, suggesting that it may recognize a ligand-induced binding site (LIBS). In general, most α5-, β1- and α5β1 heterodimer-specific antibodies recognized integrin α5β1 on immortalized endothelial cells (Figure 1D). Taken together these assays identified a number of antibodies that resemble volociximab in vitro.

Volociximab does not bind to murine α5β1 22, suggesting that differences exist in this critical functional epitope between mice and humans. A number of antibodies were cross-reactive for human integrin, including a majority of the α5 and heterodimer specific antibodies and all of those tested that bound β1 (Figure 2A). Immunohistochemical analysis of sections from human α5β1 positive (MDA-MB-231) and negative (C32) xenograft tumors further confirmed that antibodies that cross-reacted to human integrin by ELISA also did so on tissue sections (Figure 2B). For example, 517-2, which cross-reacts by ELISA, binds to tumor cells in the α5β1 positive MDA-MB-231 xenograft, and not to C32 xenograft tumor cells, but binds to murine α5β1 on the vasculature and stroma of both xenografts. Antibody 339.1, on the other hand, does not cross-react to human integrin by ELISA and does not bind tumor cells in either the MDA-MB-231 or the C32 xenograft, yet it recognizes murine α5β1 on vessels and stroma. The panel of antibodies, therefore, represents antibodies that bind human α5β1 and those that do not.

High affinity, function-blocking antibodies against integrin α5β1 have been reported to inhibit endothelial cell migration and tube formation and elicit cell death in vitro 151621. A subset of antibodies from the panel was assessed for these effects in murine cell-based assays. Inhibition of migration towards fibronectin was determined using the murine hemangioma line SVEC. Of the antibodies tested, 13E9-20, 517-2 and 339.1 were among the most potent (Figure 3A, B), each inhibiting migration by greater than 50% relative to an IgG control. The extent of inhibition for each antibody was statistically significant (p < 0.05) relative to control.

Volociximab has a very high affinity for human α5β1 21. We were therefore interested in identifying antibodies with similarly high affinity for mouse α5β1. Selected antibodies were purified from supernatants using Protein A affinity chromatography and assessed for binding to mouse and human α5β1 by BIAcore. Of the human cross-reactive set, 517-2 had the highest affinity (K_D = 0.21 nM) and of the non-cross-reactive set, antibody 339.1 bound tightest (K_D = 0.59 nM). The affinities of both antibodies were sub-nanomolar, comparing favorably with volociximab (K_D = 0.32 nM).

The highest affinity antibodies that inhibited α5β1 binding to fibronectin and cell migration, were human cross-reactive antibody 517-2 and mouse-specific antibody 339.1. These two antibodies, and a third, lower affinity antibody (K_D = 11.5 nM) that binds both mouse and human integrin, 13E9-20, were assessed for their ability to elicit cell death in mouse ECs in a manner analogous to that of volociximab on HUVEC. Primary murine endothelial cells were incubated with the antibodies for 16 hours and Annexin V binding was assessed by flow cytometry. 339.1 induced significant apoptosis in these cells whereas 517-2 and 13E9-20 did not, despite 517-2 and 339.1 having similar affinities for mouse integrin (Figure 4A). 339.1 also induced cell death in the T-antigen transformed murine endothelial cell line SVR, with an EC₅₀ of 5.3 nM (Figure 4B), compared to an EC₅₀ of 2.5 nM for volociximab on HUVEC. This finding suggests that although 339.1 and 517-2 have similar biological capabilities and similar affinities, initiation of the cell death program requires binding to a highly specific epitope, specifically recognized by 339.1.

To determine if 339.1 also inhibited the formation of capillary-like structures in an in vitro angiogenesis assay, endothelial cells isolated from mouse lungs were induced to form tubes in a fibrin matrix in the presence or absence of antibody. Antibody 339.1 significantly inhibited tube formation (Figure 5A, B; p = 0.018), although this activity was not as robust as the inhibitory effects exerted by the human cross-reactive antibodies or volociximab on HUVEC tube formation (data not shown; 21). This suggests that murine lung microvascular cells are less sensitive, in general, to anti-α5β1 activity, relative to human umbilical vein endothelial cells, in this assay. Nevertheless, these results demonstrate that antibodies in this panel can inhibit the ability of mouse and human endothelial cells to adopt a vessel-like structure in a three dimensional assay.

Antibodies against human α5β1, including volociximab, inhibit angiogenesis and impede tumor growth in vivo 2122. As volociximab does not cross-react with rodent integrin, these earlier studies were conducted in rabbit and monkey models. We therefore considered whether 339.1 affects tumor growth in syngeneic and xenograft tumor models. In the murine endothelial-derived angiosarcoma model, SVR, 339.1 targets both tumor and tumor vascular endothelial cells. 339.1, administered intraperitoneally to mice bearing established tumors, thrice weekly at 10 mg/kg, inhibited tumor growth by approximately 60% (p = 0.007; Figure 6A). In the A673 human xenograft model, 339.1 targets vascular endothelial cells, but not tumor cells. 339.1 inhibited tumor growth in these models by approximately 40% relative to tumors from control-treated mice (p = 0.022; Figure 6B). As in the C32 and MDA-MB-231 xenografts, vessels stained positive for integrin with 339.1 in both of these xenograft models (data not shown). Sections from A673 tumors treated with 339.1 were assessed for vessel density by immunohistochemistry using an antibody against anti-CD31 (Figure 6C). Vascular density was significantly decreased in tumors from treated animals compared to tumors that received vehicle (p = 0.011; Figure 6D); statistically significant results were obtained in a second experiment employing 25 mg/kg 339.1 three times weekly (15 animals per group; 47% inhibition, p < 0.001), whereas two additional experiments utilizing the 25 mg/kg dosing schedule did not show significant effects (5 animals per group each; 1%, p = 0.96 and 10%, p = 0.82). These results confirm that 339.1, like volociximab, impedes tumor growth, at least in part, by inhibiting angiogenesis.