The discovery that tumor growth and metastasis are codependent on the formation of neovascularization1 revealed many potential protein targets for cancer treatment.2, 3, 4 The angiogenic process depends on proliferation of vascular endothelial cells, migration of tumor cells to the vascular endothelium followed by cell adhesion, and finally invasion of the endothelium.5 A family of highly conserved adhesion molecules, known as integrins, is central to the regulation of these processes. Integrins are heterodimeric transmembrane receptor complexes composed of noncovalently associated α and β chains, and recognize the arginine–glycine–aspartic acid (RGD) sequence present in their extracellular matrix (ECM) ligands.6 At present, 18 α and 8 β subunits are known; these form 24 different αβ heterodimers with different specificity for various ECM cell-adhesive proteins.7 Ligands for various integrins include fibronectin, collagen, laminin, von Willebrand factor, osteopontin, thrombospondin and vitronectin, all components of the ECM. Certain integrins can also bind to soluble ligands, such as fibrinogen or to other adhesion molecules on adjacent cells.7
Integrin αvβ3 is one of the most well-characterized integrin heterodimers and is one of the several heterodimers that have been implicated in tumor-induced angiogenesis.8, 9 While sparingly expressed in mature blood vessels, αvβ3 is significantly upregulated during angiogenesis in vivo.10 The expression of αvβ3 correlates with the aggressiveness of the disease in breast and cervical cancer as well as in malignant melanoma,11, 12, 13 and recent studies have suggested that αvβ3 may be useful as a diagnostic or prognostic indicator for some tumors.14 Integrin αvβ3 is expressed on some invasive tumors, including metastatic melanoma15, 16 and late-stage glioblastoma,17 thus contributing to their malignant phenotype. Integrin αvβ3 is particularly attractive as a therapeutic target because of its relatively limited cellular distribution. It is not generally expressed on epithelial cells, and minimally expressed on other cell types.18, 19, 20 Furthermore, αvβ3 antagonists, including both cyclic RGD peptides and monoclonal antibodies (mAbs), significantly inhibit cytokine-induced angiogenesis and the growth of solid tumor on the chick chorioallantoic membrane.21
Another αv integrin heterodimer, αvβ5, is more widely expressed on malignant tumor cells and is quite likely involved in VEGF-mediated angiogenesis.8, 22 It has been shown that αvβ3 and αvβ5 promote angiogenesis via distinct pathways: αvβ3 through bFGF and TNF-α, and αvβ5 through VEGF and TGF-α.8 Another study has shown that inhibition of Src kinase can block VEGF-induced, but not bFGF-induced, angiogenesis.23 These results strongly imply that bFGF and VEGF activate different angiogenic pathways that require αvβ3 and αvβ5, respectively. Therefore, αvβ3 and αvβ5 are attractive targets for antitumor therapies, and may provide therapeutic benefit in the treatment of solid tumors.
A number of integrin αvβ3 and αvβ5 antagonists are being developed for use as angiogenesis inhibitors.24 These include Vitaxin, a humanized form of mouse anti-human αvβ3 mAb LM609,25, 26, 27 a fully human anti-human αv mAb CNTO95,28 cyclic RGD peptides,29 and synthetic small molecule RGD mimetics.30, 31 We have developed a chemically programmed antibody approach that is unique, wherein small synthetic molecules and catalytic mAb form a reversible covalent bond that results in the reprogramming of the specificity of the antibody both in vitro and in vivo.32 The chemically programmed antibody (cp38C2) is prepared by covalently linking a small synthetic molecule, RGD mimetic SCS-873, which binds specifically to integrins αvβ3 and αvβ5, to aldolase mAb 38C2 (Fig. 1). The complex of the SCS-873 mimetic and mAb 38C2 was spontaneously formed in vivo, the half-life of the SCS-873 mimetic in the cp38C2 complex was increased in circulation by more than 2 orders of magnitude relative to the SCS-873 alone, and cp38C2 effectively reduced tumor growth in mouse models.
In the current study, we evaluated the effect of the preformed cp38C2 complex on the growth and metastases of M21 human melanoma in mouse models. Further, advancing our earlier study, the current study employs the preformed complex of small molecule and antibody which is, as we discuss, a more relevant format for the anticipated clinical studies. In addition, this mode of administration allows for a direct comparison of the therapeutic effect of small molecule antibody, small molecule complex, and conventional antibody, all of which is addressed in extensive detail in the current study.
Material and methods
Human umbilical vein-derived endothelial cells (HUVEC) and mouse endothelial cell line MAEC were cultured as described previously.33 SV40 transformed, mouse endothelial pancreatic islet cell line MS1 (MILE SVEN 1) was purchased from American Type Culture Collection (Manassas, VA), and was maintained in DMEM, supplemented with 4 mM L-glutamine, 1.5 g/L sodium bicarbonate, 4.5 g/L glucose, 1 mM sodium pyruvate, 10% FCS, and antibiotics. Human melanoma cell lines M21 and M21L were obtained from Dr. D. A. Cheresh (The Scripps Research Institute, La Jolla, CA), and were maintained in RPMI medium 1640 containing 10% FCS and antibiotics.
Natural killer (NK) cells were isolated from the spleen of nude and severe combined immuno-deficient (SCID) mice, using the MACS system, according to the manufacturer's recommendations (Miltenyi Biotech, Auburn, CA). Non-NK cells (i.e., B cells, T cells, dendritic cells, macrophages, granulocytes and erythroid cells) were depleted with a cocktail of biotin-conjugated antibodies against CD19, CD4 (L3T4), CD8a (Ly-2), CD5 (Ly-1), Ly-6G (Gr-1) and Ter-119, and anti-biotin MicroBeads. Purity of NK fractions was >95%, as determined by flow cytometry.
Antibodies and other reagents
The generation and purification of mouse mAb 38C2 (IgG2a isotype) has been described,34 and it is commercially available from Sigma–Aldrich (St. Louis, MO). Mouse anti-human integrin αvβ3-specific mAb LM609 (IgG1 isotype) was obtained from Dr. D. A. Cheresh (The Scripps Research Institute), and is also commercially available from Chemicon International (Temecula, CA). Mouse anti-human integrin αvβ5-specific mAb P1F6 (IgG1 isotype) and purified human integrins αvβ3 and αvβ5 were purchased from Chemicon. Integrin αIIbβ3 was purchased from Enzyme Research Laboratories (South Bend, IN). FITC-conjugated donkey anti-mouse IgG polyclonal antibodies and horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG polyclonal antibodies were from Jackson ImmunoResearch Laboratories (West Grove, PA).
Human vitronectin was purchased from Promega (Madison, WI). Human plasminogen-free fibrinogen (VWF/Fn depleted) was purchased from Enzyme Research Labs (South Bend, IN). Human bFGF and human VEGF were obtained from R&D Systems (Minneapolis, MN). Recombinant HIV-1 IIIB Tat protein was purchased from ImmunoDiagnostics (Woburn, MA). The calcein AM was purchased from Molecular Probes (Eugene, OR). The synthesis of β-diketone (JW) conjugated to BSA (JW-BSA), SCS-397 and SCS-873 were previously described.32, 34
A stock solution of 10 mg/mL (66.6 μM) 38C2 IgG in PBS (pH 7.4), stored at 4°C, was used. Stock solution of 8.73 mg/mL (10 mM) SCS-873 in 100% ethanol, stored at −80°C, was used. 38C2/SCS-873 complex (cp38C2) was formed by incubating 3.3 μM 38C2 with 6.6 μM SCS-873 for 60 min at room temperature in 50 μL of either PBS (in vivo studies), metal buffer (ELISA and FACS assay) or culture medium, appropriate for the assay condition.
Analysis of integrin binding by ELISA
Costar 96-well ELISA plates (Corning, Acton, MA) were coated with 75 ng of antigen (human αvβ3, αvβ5, αIIbβ3 or BSA) in 25 μL of metal buffer (25 mM Tris-HCl, pH 7.4, 137 mM NaCl, 1.25 mM KCl, 1 mM MgCl2, 1 mM CaCl2) and incubated overnight at 4°C. After blocking with 150 μL of TBS/3% BSA for 1 hr at 37°C, 50 μL of 1 μg/mL mAb 38C2, cp38C2, LM609 or P1F6 solution was added into each well, and the plates were incubated for 2 hr at 37°C. Washing and detection were performed essentially, as described,35 using HRP-conjugated goat anti-mouse IgG antibody (diluted 1:3,000 in TBS/1% BSA).
Analysis of integrin binding by flow cytometry
M21, M21L, HUVEC, MAEC or MS1 cells were detached by mild trypsinization with 0.025% trypsin, 0.01% EDTA in HEPES-buffered saline solution (HBSS, BioWhittaker), washed and resuspended at 106 cells/mL in FACS metal buffer (1% (w/v) BSA, 25 mM HEPES, 0.03% (w/v) NaN3 in metal buffer, pH 7.4). Aliquots (100 μL) containing 105 cells were distributed into wells of a V-bottom 96-well plate (Corning), and incubated for 40 min at RT with 100 μL of cp38C2 at 25 μg/mL, or with control mAbs LM609 and P1F6 at 5 μg/mL in FACS metal buffer, supplemented with 1 mM MnCl2. mAb 38C2 at 25 μg/mL was used as a negative control. Cells were washed once with 200 μL of FACS metal buffer and incubated for 40 min at RT with 100 μL of FITC-conjugated donkey anti-mouse IgG antibody, diluted to 1:100 in FACS metal buffer. Cells were washed twice, resuspended in 200 μL of FACS metal buffer, and transferred to FACS-tube for analysis in a FACS Scan II flow cytometer (Becton Dickinson, Mountain View, CA).
Mouse tumor models
Confluent cultured human melanoma tumor cells M21 were harvested by incubation with 5 mL of trypsin solution (0.25%) and washed twice with PBS. Viable cells were counted by trypan blue exclusion, and cell fractions containing over 95% of viable cells were used in this study. M21 cells were suspended in PBS, and 0.2 mL (2 × 106 cells/mouse) of the suspension was inoculated s.c. into right flanks of 20 female nude mice (8 weeks of age) on day 0. Four groups of 5 mice were formed, and the animals were treated between days 1 and 20 after tumor induction. Treatment involved 200 μL i.v. injections of either PBS alone (groups 1 and 2) or 2.5 mg/mL mAb 38C2 in PBS (groups 3 and 4) once a week, on days 1, 8 and 15. In addition, 50-μL i.p. injections of 10 mg/mL SCS-873 in solvent (50% PBS, 25% DMSO and 25% ethanol) (groups 2 and 4) or solvent alone (groups 1 and 3) were given on days 2, 5, 8, 11, 14, 17 and 20. Tumor volumes of treated animals were measured over the skin in 2 dimensions, using a slide caliper, every third day starting on day 9, and the tumor volume was calculated according to the following formula, 1/2(width)2 × length. Toxicity was monitored by determining the body weight of the mice once a week. All animals were killed on day 42; the tumors were completely dissected and then weighed. Results are reported as means ± SD for each group. Simultaneously, an experiment was performed with 2 groups of 5 nude mice, except that the treatment involved 200 μL i.v. injections of either cp38C2 at 0.5 mg/mL (3.3 μM 38C2 plus 6.6 μM SCS-873) or SCS-873 alone at 6 μg/mL (6.6 μM) in PBS once a week on days 1, 8 and 15.
Female SCID mice (9 weeks of age) were used for the experimental metastases studies. In one study, 4 groups of 8 mice were inoculated on day 0 with 200 μL of human melanoma M21 cells (2 × 106 cells/mouse) i.v. into the tail vein. Two hundred microliters of cp38C2, 38C2, LM609 (all at 1.25 mg/mL) or SCS-873 (15 μg/mL) was administered i.p. in 0.2 mL of PBS on days 1, 2, 5, 8, 11 and 14. After the treatment, the animals were checked daily, body weights were recorded weekly, and moribund mice were killed. The study was terminated after 200 days, and the surviving mice were killed. The lungs were recovered from all mice, fixed in 10% buffered formalin, embedded in paraffin, sectioned and stained with H&E for routine histological examination by light microscopy. In another study, 4 groups of 8 SCID mice were inoculated with M21 melanoma cells, and treated as described earlier with different cp38C2 antibody concentrations: 1.25 mg/mL, 0.5 mg/mL or 0.125 mg/mL. The control group of mice was treated with mAb 38C2 at 1.25 mg/mL.
In vitro cell proliferation assay
Cell proliferation was evaluated as previously reported.32 Briefly, 96-well plates were precoated with 50 μL of 2 μg/mL vitronectin, fibrinogen or HIV-1 tat protein overnight at 4°C, washed twice with 100 μL HBSS and blocked with 100 μL cell culture media for 30 min at 37°C. To examine monolayer cell growth, a total of 2.5 × 103 (HUVEC), 2.5 × 103 (MAEC), 1 × 103 (MS1), 1 × 103 (M21) or 1 × 103 (M21L) cells were plated, treated with the indicated concentrations of SCS-873 or cp38C2, ranging from 0.01 to 100 μM (from 0.005 to 5 μM for cp38C2), and incubated at 37°C for 64 hr in a humidified CO2 incubator. [3H]thymidine (ICN Biomedicals, Irvine, CA) was added to 0.5 μCi per well (1 Ci = 37 GBq) during the last 16 hr of incubation. The cells were frozen at −80°C overnight, and subsequently processed on a multichannel automated cell harvester (Cambridge Technology, Cambridge, MA), and the [3H]thymidine incorporated into DNA was determined by using a liquid scintillation beta counter (Beckman Coulter). The background was defined by running the same assay in the absence of SCS-873 or cp38C2. The inhibition was calculated according to the following formula: (background − counts with treatment)/background × 100%. All experiments were performed in triplicate.
Complement-dependent cytotoxicity and antibody-dependent cellular cytotoxicity assays
For complement-dependent cytotoxicity (CDC), 106 M21, HUVEC or MS1 cells were labeled with 100 μCi of Na251CrO4 (ICN Biomedicals) for 1 hr at 37°C, and washed 4 times with HBSS. Fifty microliters of 51Cr-labeled cells (2 × 105 cells/mL) and 50 μL of cp38C2, at concentrations ranging from 0.1 to 100 nM, were distributed in triplicates into U-bottomed 96-well plates (Corning). Then, 100 μL (1:2.5 dilution) of Low-Tox-H (MS1 cells) or MA (M21 and HUVEC) rabbit complement was added to each well, and the plates were incubated for 1 hr at 37°C in a tissue culture incubator. Plates were centrifuged, and 175 μL of culture supernatants were collected, and radioactivity was measured in a Beckman 7000 gamma counter. The cell lysis was calculated using the following formula: (counts with treatment − spontaneous release)/(maximum release − spontaneous release) × 100%, in which spontaneous release (8–12%) was measured by incubating radiolabeled cells in the absence of complement, and maximum release was obtained by adding 1% Triton X-100 to the radiolabeled cells. The mouse anti-human HLA-ABC class I mAb (IgG2a) and LM609 mAb (IgG1) were used as positive and negative controls, respectively.
For antibody-dependent cellular cytotoxicity (ADCC), 50 μL of 51Cr-labeled M21 or MS1 cells (2 × 105 cells/mL) and 50 μL of cp38C2 at a concentration of 100 nM were distributed into U-bottomed 96-well plates and incubated for 15 min at RT. Then, 100 μL of murine NK cells, freshly isolated as described earlier, were added in triplicates to each well at the indicated effector/target ratio. After the plates were incubated at 37°C in a tissue culture incubator for 8 hr, and cell lysis was calculated as described earlier for the CDC assay, except that the spontaneous release (10–15%) was measured by incubating radiolabeled cells in the absence of NK cells. The cp38C2 alone was used as the negative control mAb.
Cell adhesion assay
Costar 96-well black wall clear bottom plates (Corning) were coated at 4°C overnight, with 50 μL vitronectin in PBS (2 μg/mL). After blocking with 100 μL of PBS/1% BSA for 1 hr at RT, a calcein AM-based assay was used to quantify cell adhesion, as previously described in Trikha et al.36 Briefly, calcein AM-labeled M21, HUVEC or MS1 cells were resuspended at 5 × 105 cells/mL in an assay medium (M199/1% BSA/10 mM HEPES), and 100 μL of cells' suspension were incubated for 30 min at RT in the dark in the presence of SCS-873 or cp38C2, ranging from 0.005 to 10,000 nM. Incubation with LM609 mAb served as a positive control. The cells were then transferred onto 96-well plates coated with vintronectin, and cell adhesion was allowed to occur for 2.5 hr at 37°C. Adhesion was measured in a fluorescence plate reader (SpectraMax Gemini, Molecular Devices, Sunnyvale, CA) at 494/514 nm. Cell adhesion to BSA-coated wells served as a negative control. Extent of cell adhesion in the presence of various concentrations of inhibitor was plotted as a percent of cell adhesion in the absence of inhibitor. Each data point is the mean of triplicate determinations (± SD).
Cell migration assay
Migration of HUVEC toward angiogenic factors, such as VEGF and bFGF, (10 ng/mL each) and human M21 tumor and mouse MS1 endothelial cells toward 2.5% FCS was assayed in 8-μm pore size Costar transwells (Corning). The underside of the membrane was coated with vitronectin (2 μg/mL), overnight at 4°C, and then blocked with a solution of 1% BSA/PBS at room temperature for 60 min. Next, membranes were washed with PBS and dried. M21, HUVEC or MS1 cells were starved overnight in low serum medium (0.5% FBS), trypsinized, washed, resuspended in migration buffer (M199 medium, 0.1% BSA) at 2.5 × 105 cells/mL, and 200 μL of 5 × 104 cells was added to the upper chambers in the presence of cp38C2, at concentrations ranging from 10 to 500 nM. LM609 (50 nM), P1F6 (50 nM) or 38C2 (500 nM) mAbs were used as controls. The chambers were placed in a tissue culture incubator, and migration was allowed to proceed for 6 hr (4 hr for HUVEC) at 37°C. Migration was terminated by rinsing the filters in PBS, and fixing and staining with crystal violet (Sigma–Aldrich). Cells that did not migrate were removed from the filter tops by gentle wiping with wet cotton swabs. The extent of cell migration was determined by light microscopy, and images were analyzed using the Adobe Photoshop CS software (Adobe Systems, San Jose, CA). The software analyzes the total area occupied by the stained cells on the bottom side of the filter, and this is directly proportional to the extent of cell migration.36
The invasion assay was performed using essentially the same protocol used for the migration assay, described earlier, except that the cells were allowed to invade through the Matrigel-coated invasion chambers (Becton Dickinson) in response to 10 ng/mL angiogenic factors VEGF and bFGF for HUVEC and 5% FBS for M21 and MS1 cells. After 16 hr for HUVEC, 24 hr for MS1 and 48 hr for M21 invasion, the invasion assay was stopped, and the nonmigrated cells were removed, and the migrated cells were quantified as described earlier.
Unpaired two-tailed Student's t test and Mantel–Cox log-rank test were performed using SAS software. Differences were considered statistically significant at p < 0.05.
Characterization of cp38C2 binding to integrins αvβ3 and αvβ5
ELISA was used to determine the binding specificities of various antibodies. As expected, LM609 mAb bound only to integrin αvβ3, whereas P1F6 mAb bound only to integrin αvβ5. cp38C2 bound to both human integrins αvβ3 and αvβ5, although the binding to αvβ5 integrin was weaker than to αvβ3 (Fig. 2a). None of the antibodies tested crossreacted with integrin αIIbβ3. Covalent attachment of the RGD mimetic to the antibody was required for integrin recognition; mAb 38C2 alone did not bind to any of the integrins or cell lines tested (Fig. 2a and data not shown).
Flow cytometry analysis showed that cp38C2, bound to M21 melanoma cells, expresses both integrins αvβ3 and αvβ5 on their surface (Fig. 2b). In contrast, cp38C2 was unable to bind αvβ3 and αvβ5 integrin-negative M21L cells, thus confirming previous ELISA results, demonstrating selective αvβ3 and αvβ5 integrin recognition. cp38C2 was also bound to human (HUVEC) and mouse (MAEC and MS1) endothelial cell lines, confirming recognition of both human and mouse integrins by the chemically programmed antibody (Fig. 2c).
Inhibition of tumor growth in vivo by cp38C2
Human melanoma M21 cells express very high levels of integrin αvβ3 and a moderate level of integrin αvβ5. In one study, 2 × 106 M21 cells were inoculated into nude mice s.c. on day 0. Antibody 38C2 was administered weekly through i.v. injections of 200 μL of 2.5 mg/mL antibody, on days 1, 8 and 15. This same group of mice received SCS-873 i.p. injections of 50 μL of 10 mg/mL on days 2, 5, 8, 11, 14, 17 and 20. This treatment resulted in the formation of cp38C2 in vivo. Control groups of animals received 38C2 alone, SCS-873 alone or PBS. Beginning at day 12, a reduced tumor growth rate was observed in nude mice treated with cp38C2 formed in vivo, when compared with mice treated with PBS (Fig. 3a). On an average, 81% reduction in tumor growth was observed by day 42 in the cp38C2-treated group relative to the PBS-treated group of mice. In addition, 2 of 5 animals in cp38C2-treated group completely rejected the M21 tumor by day 27 and 33. These 2 mice remained tumor-free in excess of 160 days after challenge, in the absence of further cp38C2 treatment, until they were killed. Repeated inoculation of SCS-873 alone or mAb 38C2 alone into mice bearing M21 tumors did not result in lower tumor sizes when compared with the PBS-treated control group (Fig. 3a).
In the second study, antibody cp38C2 was preformed in vitro and administered through i.v. injection on days 1, 8 and 15. A group of control animals received SCS-873 alone on the same days. When cp38C2 was preformed, tumor growth was inhibited by 68% when compared with SCS-873-treated mice. Treatment with cp38C2 did not result in weight loss or behavioral changes during the course of therapy. Tumor weights, determined after killing on day 42, confirmed the reduction in tumor growth during the course of cp38C2 treatment (Fig. 3b).
Effect of cp38C2 on experimental metastases
We next studied the effect of cp38C2 on the survival of mice with lung metastases. M21 cells were injected i.v. into SCID mice, resulting in the formation of metastatic foci in the lung. Animals were then treated with 6 injections of cp38C2, 38C2 or SCS-873 given 3 times per week, starting on day 1, at a dose of 12.5 mg/kg mAb or 0.15 mg/kg of SCS-873. The antimetastatic effect of mAb LM609 was tested in parallel using the same treatment protocol. Survival curves show that animals treated with mAb 38C2 or SCS-873 alone began to die on day 50, apparently due to the difficulty in breathing caused by lung metastases. All 8 mAb 38C2-treated mice had died by day 97, and all 8 mice treated with SCS-873 had died by day 134 (Fig. 4a). Mean survival time for the group of mice treated with 38C2 or with SCS-873 was 75 ± 6 and 96 ± 11 days, respectively. All 8 animals treated with cp38C2 and 4 out of 8 mice treated with LM609 mAb survived the experimental period (survivors were killed at day 200). Compared with 38C2, treatment with cp38C2 led to a significant improvement in overall survival (p < 0.01), thus confirming the antitumor activity of cp38C2 in a lung metastasis melanoma model. Histological examination of the lungs of all animals showed diffuse metastases in the case of deceased animals and no metastatic foci in any of the long-term survivors protected by treatment with either cp38C2 or with mAb LM609 (Fig. 4b).
To confirm the effect of cp38C2 treatment against M21 tumor metastasis, 3 different doses of cp38C2 were tested. Similar to previous observations, treatment with mAb 38C2 was unable to protect the mice against M21 metastases, and all animals had died at day 91 (Fig. 4c). Mean survival time for the group of mice treated with 38C2 was 70 ± 5 days. In contrast, cp38C2 confirmed its strong antitumor effect against M21 metastases. Maximal protection was observed with the 12.5 mg/kg dose of cp38C2. Seven out of 8 mice were alive on day 200 (Fig. 4c). Two other groups of mice treated with 5 mg/kg and 1.25 mg/kg of cp38C2 had mean survival times of 135 ± 18 and 118 ± 14 days, respectively (p < 0.01). Photomicrographs, demonstrating histological features of lungs of deceased animals and of survivors protected by treatment with high doses of cp38C2, are shown in Figure 4d.
These data indicate that cp38C2 has very potent antitumor effects in mice bearing lung metastases of M21 melanoma. Since the tumor colonies in the lung tissue can grow without neo-angiogenesis,37, 38 this model suggests that blockade of human tumor-expressed integrins alone can inhibit tumor lung colony growth in mice independently of the antiangiogenic effects of cp38C2.
Effect of cp38C2 on cell proliferation
To better understand the mechanism of cp38C2, we evaluated its effect on proliferation, adhesion, migration, and invasion of M21 tumor and human (HUVEC) and mouse (MAEC and MS1) endothelial cells in vitro. We also evaluated the effect of SCS-873, and used M21L cells that do not express αvβ3/αvβ5 integrin as a control cell line. Cells were incubated in the presence of increasing concentrations of cp38C2 or SCS-873, and their proliferation was measured by [3H]thymidine incorporation after 64 hr. Dose-dependent inhibition of HUVEC, MAEC and MS1 cell proliferation was observed with SCS-873 and cp38C2, when endothelial cells were incubated on vitronectin-coated plates (Figs. 5a and 5b). In contrast, treatment with cp38C2 or SCS-873 had no effect on human M21 melanoma cell growth, suggesting that integrins αvβ3 and αvβ5 are important for endothelial cell proliferation but not for proliferation of M21 tumor cells. In addition, a similar pattern of inhibition of proliferation of endothelial cells, but not human tumor cells by SCS-873 was observed on fibrinogen, which has minimal interaction with αvβ5,39 and on HIV-1 tat protein, which contains an RGD-independent binding domain for αvβ5,40 suggesting that αvβ3 is the predominant integrin at work in this system (Figs. 5c and 5d).
Complement-dependent and cp38C2 ADCC
Two potential effector mechanisms that might contribute to the efficacy of cp38C2 in vivo were investigated in vitro. We first examined the ability of cp38C2 to kill M21 tumor cells by CDC. In the presence of rabbit complement, cp38C2 (mouse IgG2a isotype) induced substantial cytotoxicity against M21 cells (Fig. 6a). CDC correlated directly with the mAb concentration and was undetectable at concentrations below 5 nM. These data suggest that a minimal number of antibody molecules had to be bound to the cell surface to mediate CDC. As expected, LM609 (mouse IgG1 isotype) failed to show any CDC activities on M21 tumor cells (Fig. 6a). There was no direct cytotoxicity of cp38C2 against M21 cells, at concentrations of 100 nM and below, in the absence of complement. Contrary to the proliferation results, only minimal lysis of mouse MS1 endothelial cells was observed in the presence of cp38C2 at concentrations >5 nM (Fig. 6b). Moreover, HUVEC proved to be complement-resistant cell line (data not shown). Thus, chemically programmed antibodies can functionally direct complement activities.
Another mechanism through which antibodies can destroy tumor cells is ADCC. Experiments with splenic NK cells from nude and SCID mice were performed with cp38C2 antibody-coated M21 and MS1 cells, at different effector/target ratios. As shown in Figure 6c, nude NK cells affected substantial (>25%) lysis of cp38C2-coated M21 target cells in a dose-dependent manner. SCID NK cells, however, revealed only marginal (<5%) lysis of M21 cells (Fig. 6d). Neither nude nor SCID NK cells affected any significant lysis of mouse MS1 endothelial cells (data not shown).
Effect of cp38C2 on cell attachment in vitro
Integrins αvβ3 and αvβ5 bind vitronectin41; therefore, we studied the ability of cp38C2 and SCS-873 to inhibit integrin-mediated adhesion of endothelial and human tumor cells to vitronectin. In this experiment, cells were seeded onto tissue culture plates coated with vitronectin and allowed to adhere for 2.5 hr, before quantitating the number of adherent cells. As shown in Figure 7, SCS-873 dose-dependently inhibited the adhesion of M21, HUVEC and MS1 cells to vitronectin (IC50's 11 nM, 30 nM and 3 μM, respectively). Antibody cp38C2 was as effective as SCS-873 in the inhibition of adhesion of M21 cells to vitronectin (IC50 7 nM) and 4–6 times more potent than SCS-873 in the inhibition of HUVEC and mouse MS1 endothelial cell adhesion to vitronectin (IC50's 7 and 500 nM, respectively). As seen in Figures 7a and 7b, LM609, which is specific for αvβ3, was 2.5 times less efficient than cp38C2 at inhibition of adhesion of human M21 tumor cells (IC50's 18 and 7 nM, respectively) and slightly better for HUVEC (IC50's 4.5 and 7 nM, respectively). As expected, LM609 had no effect on the adhesion of mouse endothelial cells, because it does not recognize mouse integrin αvβ3. Collectively, these data indicate that cp38C2 is a more potent inhibitor of endothelial cell attachment than SCS-873, and that the blocking of both integrins αvβ3 and αvβ5 is superior to blocking αvβ3 alone.
Migration of human melanoma and endothelial cells
We next evaluated the effect of cp38C2 on cell migration toward vitronectin. Figure 8 shows that the addition of cp38C2 significantly reduced the migration of M21, HUVEC and MS1 cells toward this matrix protein compared with controls (BSA and 38C2). Moreover, cp38C2 dose-dependently inhibited human tumor and human or mouse endothelial cell migration (Figs. 8b–8d). Similar results were obtained with LM609 (specific for αvβ3), P1F6 (specific for αvβ5) and the combination of both antibodies against migration of M21 and HUVEC, but not against mouse endothelial cells (Fig. 8d), which are not recognized by human integrin-specific antibodies, LM609 and P1F6. At doses higher than 25 nM, cp38C2 completely blocked M21 tumor cell migration, thus outperforming LM609, P1F6 and the combination of both (Fig. 8b). These findings suggest that endothelial and melanoma cells primarily use both integrins αvβ3 and αvβ5 to migrate toward vitronectin, and cp38C2 can inhibit both tumor and endothelial cell migration.
Human melanoma and endothelial cell invasion through matrigel
Since cp38C2 inhibited cell adhesion and migration, we investigated the effect of the chemically programmed antibody on invasion of tumor cells and endothelial cells, using Matrigel invasion chambers. Invasion is a multistep process that involves cell adhesion, degradation of the matrix and migration of cells through the degraded matrix. Invasion of M21, HUVEC and MS1 cells was inhibited by cp38C2 (Fig. 9), suggesting the involvement of integrins αvβ3 and αvβ5 in this process. The effect of cp38C2 on cell invasion was dose-dependent (Figs. 9b–9d). In contrast to the effect on migration, cp38C2, LM609 and P1F6 did not completely inhibit M21 tumor cell invasion (Fig. 9b). An enhanced blocking effect was observed using treatment with both LM609 and P1F6 mAbs, suggesting that both αvβ3 and αvβ5 are involved in tumor and endothelial cell invasion. As expected, only cp38C2 was effective in inhibiting the invasion of mouse endothelial cells (Fig. 9d). Collectively, these data suggest that blockade of integrins αvβ3 and αvβ5 by cp38C2 can effectively inhibit invasion of human melanoma and human or mouse endothelial cells.
In the current study, we have characterized the efficacy of chemically programmed mAb cp38C2 against human melanoma and human or mouse endothelial cells in vivo and in vitro. In vivo, growth of human melanoma tumors in nude mice was significantly reduced by treatment with cp38C2, using 2 distinct therapeutic regimens. The first therapeutic regimen studied involved the separate administration of SCS-873 and mAb 38C2 to examine the prospect of in vivo self-assembly of the cp38C2 complex, since it may be possible in future studies to deliver chemical programming agents into a pre-existing reservoir of 38C2-like antibody, prepared through immunization or gene delivery. The second therapeutic regimen studied involved in vitro preparation or programming of the cp38C2 complex that was then administered as a conventional mAb therapeutic. Delivery of the preformed complex is a substantially more controlled treatment than in vivo self-assembly, and would likely be the administration route of first choice if this drug were to enter clinical studies.
Self-assembly of cp38C2 in vivo provided an 81% mean reduction in tumor growth, while the preformed complex provided a 68% mean reduction in tumor growth relative to the PBS-treated control groups. Although the mean difference in tumor growth inhibition seen in the 2 therapeutic regimens was not statistically significant, the overall effect was achieved by injecting 300 μg of total cp38C2 (3.6 μg of SCS-873) preformed in vitro, when compared with the injection of 2.5 μg total 38C2 and of 3.5 mg total SCS-873 to form the complex in vivo. In addition, treatment with cp38C2 significantly improved the survival of mice-bearing melanoma metastases in lungs. After 200 days, all the survivors had lungs free of tumor metastases, thus highlighting the potential importance of targeting more than one protein for antitumor efficacy. In all in vivo studies, the chemically programmed antibody demonstrated a therapeutic effect that was far superior to that observed following treatment with the SCS-873 programming agent itself, even when SCS-873 was administered at an overwhelming ˜1,000-fold molar excess when compared with cp38C2, 3.5 mg of SCS-873 vs. 3.6 μg of SCS-873 found in the cp38C2 complex.
In vitro results demonstrated that cp38C2 has potent antiangiogenic properties in a variety of assays. Through dual αvβ3 and αvβ5 integrin targeting, cp38C2 was effective in the inhibition of adhesion of M21 melanoma cells and endothelial cells to the integrin ligand, vitronectin. cp38C2 also inhibited proliferation of human and mouse endothelial cells, but not M21 tumor cells. Previous studies demonstrated that anti-integrin mAb or RGD peptides reduced cell growth by inhibiting integrin ligation, detaching cells from the ECM and inducing an anchorage-dependent apoptosis, termed as anoikis.21, 42, 43 Our proliferation and adhesion data show that anchorage loss did not trigger any pro-apoptotic signals in human M21 melanoma cells. In addition, cp38C2 also inhibited migration and blocked invasion of M21 melanoma tumor and human and mouse endothelial cells. These studies are in agreement with previous observations of the abilities of antibody- and RGD-based peptide antagonists of integrin αvβ3/αvβ5 to inhibit angiogenesis, thus leading to regression of tumors in animal models.9, 28, 31, 44, 45
During angiogenesis, αvβ3 is upregulated on the surface of activated endothelial cells, enhancing migration, proliferation and invasion.9, 46 There is also evidence that αvβ5, one of the most ubiquitous members of the integrin family,22 may play a distinct yet important role in angiogenesis. Using two different models of angiogenesis, it was shown that mAb LM609 that binds to αvβ3 selectively inhibited bFGF-stimulated angiogenesis. In contrast, mAb P1F6, an antibody that blocks αvβ5, selectively inhibited angiogenesis stimulated by VEGF.8 This suggests that the 2 integrins participate in angiogenesis through distinct pathways. Trikha et al.28 found that antibody blockage of either αvβ3 by LM609 or αvβ5 by P1F6 reduced endothelial cell adhesion and migration in vitro, but that the 2 agents had to be combined to reach the level of inhibition achieved by the mAb CNTO95 that binds to all αv integrins. The ability to block tumor growth through multiple pathways may prove critical to effectively inhibiting angiogenesis and tumor growth in vivo. In the same study, CNTO95 displayed potent antiangiogenic effects in both rodent model and novel nonhuman primate model in cynomolgus monkeys.28 In the study by Rader et al.,32 we showed that cp38C2 could inhibit tumor growth through an antiangiogenic effect. In that study, αvβ3/αvβ5-negative human colon carcinoma cells SW1222 were injected into nude mice. Administration of cp38C2 significantly inhibited the growth of αvβ3/αvβ5-negative tumors by blocking the growth of mouse blood vessels. The hypothesis that cp38C2 can inhibit tumor growth indirectly, presumably via inhibition of angiogenesis, is also supported by the observation that cp38C2 has no effect on M21 cell proliferation, but inhibits endothelial cell proliferation in vitro.
Other synthetic dual αvβ3/αvβ5 integrin inhibitors have been created and have demonstrated antiangiogenic and antitumor efficacy in vivo. For example, the peptidomimetic compound S247 demonstrated antiangiogenic and antimetastatic effects against colon tumor xenografts in mice dosed with 70 mg/kg daily.45 Similar results were reported for the RGD mimetic SCH 221153, which inhibited human melanoma tumor growth and angiogenesis in mice, when dosed twice daily at 20 and 50 mg/kg.31 A cyclic peptide EMD 121974 is currently in clinical trials in patients with advanced solid tumors.47 These observations point to the potential importance of dual integrin antagonism in inhibiting tumor growth and angiogenesis. Moreover, it should be noted that the chemically programmed antibody approach demonstrated here allows for a once a week dosing schedule, making compliance in chronic therapy more likely.
The ability to inhibit integrin function on tumor cells directly, in addition to blocking integrins on angiogenic endothelium, may prove critical in halting the growth of some tumors. Many αv integrins have been suggested to play critical roles in the biology of tumors. For example, changes in expression of αvβ3 and αvβ5 have been associated with the progression, growth and dissemination of melanomas.12, 48 In the present study, cp38C2 was equally effective against tumor and mouse endothelial cells in all in vitro assays, except proliferation. In our mouse lung metastasis model, however, where tumor growth is most likely angiogenesis-independent, treatment with cp38C2 significantly prolonged the overall survival of mice, suggesting the direct inhibition of αvβ3/αvβ5-positive melanoma tumor growth or lung engraftment. In the same model, we observed an effect of SCS-873 alone on survival of the mice, although the effect was much less pronounced than that of cp38C2. Taken together with the fact that SCS-873 did not affect primary tumor growth in vivo, we theorize that the major antitumor effect was its ability to inhibit tumor growth directly.
In the mouse model assay of solid tumor growth, cp38C2 was able to bind and block integrins on both tumor cells and angiogenic endothelium, and caused a marked reduction in tumor size by the end of the study. In this model, dosing with SCS-873 alone, at ˜1,000-times the molar dosage of the compound present in cp38C2, was ineffective. The lower amount of SCS-873 antagonist required in the chemically programmed approach may also be significant with some compounds, where excessive dosing leads to the generation of toxic metabolites. Both in vivo and in vitro data collectively suggest that through combined blockade of integrins αvβ3 and αvβ5 on tumor and endothelial cells, cp38C2 may have multiple mechanisms of action that contribute to its observed antitumor efficacy in animal models. In in vitro experiments, cp38C2 exhibited very high CDC activity against M21 tumor cells, and also showed a fairly high level of ADCC activity when tested in combination with nude NK cells. Our demonstration that chemically programmed antibodies can functionally direct CDC as well as ADCC indicates that the full range of antibody Fc functions should be accessible, and that these 2 mechanisms may contribute to the in vivo antitumor activity of cp38C2 in mouse models.
Chemically programmed antibodies may have a number of advantages over conventional antibody-based approaches. One of the most important features of this methodology is that the antibody can be chemically programmed to target proteins other than αvβ3 and αvβ5. For example, using appropriate compounds linked to β-diketone, cp38C2 can be programmed to bind other receptors. Multiple programming should also be available for 38C2 by using a mixture of different compounds to provide tumor and tumor-activated endothelial cell targeting through different receptors. Therefore, this approach may offer significant features related to multiple specificities not found in conventional antibodies.
Another important feature of 38C2 is that it has been already humanized using a rational design strategy in preparation for human studies.49 Two antibody-based integrin antagonists are currently being developed for cancer treatment. The first is Vitaxin, the humanized form of the murine anti-human αvβ3 antibody LM609.26, 27 A dose-escalating phase I study in cancer patients demonstrated that it was safe.25 The other antibody is CNTO95, a fully human mAb that recognizes the family of αv integrins, developed at Medarex (Princeton, NJ) by immunizing mice that were transgenic for part of the human immunoglobulin repertoire.28 A phase I clinical trial of CNTO95 is ongoing in patients with solid tumors. Cilengitide, a peptide antagonist of αvβ3 and αvβ5, has also proven safe in phase I trials.47 These preclinical and clinical observations demonstrate the importance of targeting multiple integrin pathways to inhibit angiogenesis.
Finally, human/mouse crossreactivity of SCS-873 and cp38C2 is also very important. The vast majority of anti-human integrin specific antibodies were raised in mice and do not react with mouse integrins. This makes their preclinical evaluation in mouse models difficult or impossible. The in vivo antitumor effect of LM609 was evaluated by using a SCID mouse model of human angiogenesis.21 In this system, αvβ3-negative human melanoma cells were injected into full thickness human skin, grafted onto SCID mice. Because LM609 does not crossreact with mouse integrins, its antiangiogenic activity was attributed to blockade of human αvβ3 receptors in the vasculature of the human skin.21 One limitation of this study is that xenografted tumors would grow even in the absence of human vasculature, because the mouse vasculature can sustain tumor growth. Like LM609, chimeric anti-human αvβ3 and αIIbβ3 mAb 7E3 (Abciximab) and fully human anti-human αv integrins mAb CNTO95 do not crossreact with mouse integrins.
In conclusion, chemically programmed mAb cp38C2 designed to target αvβ3 and αvβ5 integrins is a unique and promising immunotherapeutic. cp38C2 inhibited tumor growth and angiogenesis in vitro and in vivo. A combination of cp38C2 and additional programming for other known tumor receptors may be promising as a strategy to inhibit the growth of primary tumors and metastases in patients.