Monoclonal Antibodies Directed against Cadherin RGD Exhibit Therapeutic Activity against Melanoma and Colorectal Cancer Metastasis

There is a necessity to find new therapeutic targets to control metastatic spread in cancer. Therapeutic monoclonal antibodies (mAb) against EGFR or VEGF(R) are in clinical use for the treatment of advanced metastatic colorectal cancer. However, their impact on the final outcome of patients with metastasis is still limited (1), probably due to the difficulty of assessing the EGFR status of the patients (2, 3), adverse effects related to the wide distribution of these molecules in healthy tissues, and the lack of response in patients with KRAS mutations (4).

Cadherin 17 (CDH17), also known as liver–intestine (LI)-cadherin, is present in fetal liver and the gastrointestinal tract, exhibiting elevated expression during embryogenesis (5). CDH17 localizes to the basolateral domain of hepatocytes and enterocytes, where it mediates intercellular adhesion in a Ca⁽²⁺⁾-dependent manner to maintain tissue integrity in epithelia (6). In colon cancer, CDH17 is expressed at low levels in primary tumors or in regional lymph node metastases, as well as in poorly differentiated colon cancer tumors (7). However, CDH17 is overexpressed in advanced colorectal cancer liver metastasis (8), where it correlates with poor prognosis (9). It is also highly expressed in gastric cancer, esophagus carcinoma, pancreatic cancer (10), and hepatocarcinoma (11). CDH17 facilitates liver colonization and metastasis in orthotopic mouse colorectal cancer models after intrasplenic injection (9, 12). CDH17 binds α2β1 integrin through an RGD motif and induces β1 integrin activation, which leads to increased cell adhesion to Matrigel and type IV collagen, and increased proliferation (12). The CDH17 RAD mutant does not induce integrin signaling but rather leads to a reduction in tumor growth and liver colonization (12). In colorectal cancer metastatic cells, we have described a nonconventional situation in which α2β1 integrin binds to cadherins in a RGD-dependent manner, making it conformationally activatable and allowing it to bind to type IV collagen. This finding modifies the classical notion that α2β1 integrin binds type I collagen using the GFOGER motifs in a RGD- and conformation-independent way (13, 14). Previous data support a preferential trans activation model for cadherin/integrin interaction, although cis interaction cannot be ruled out (12). No activation effects were observed for the RGD motif on αv or α6β4 integrins.

RGD motifs are also present in vascular-endothelial (VE)-cadherin, CDH6 [fetal kidney (K)-cadherin], and CDH20. VE-cadherin is expressed in aggressive human melanoma cell lines and cutaneous melanomas (15), but not in poorly aggressive cell lines isolated from the same tumors (16). VE-cadherin has been involved in vasculogenic mimicry (the ability to form novel blood vessel–like structures) in uveal melanomas (16). It is also expressed in Ewing sarcoma (17) and promotes breast cancer progression (18). Recently, we have demonstrated that knocking down VE-cadherin suppresses the lung colonization capacity of melanoma or breast cancer cells inoculated in mice, whereas preincubation with VE-cadherin RGD peptides promotes lung metastasis for both cancer types (19). Like CDH17 RGD peptides, VE-cadherin RGD peptides cause β1 integrin activation, suggesting that the mechanisms of action for both cadherins are similar (12).

We hypothesized that blocking the cadherin RGD motifs would provoke an inhibition of liver and lung metastasis through α2β1 integrin inhibition. Here, we investigated the use of 9-mer peptides containing the CDH17 RGD motif and their flanking sequences to retrieve highly selective mAbs with antimetastatic activity in different cancer cell types expressing CDH17 and VE-cadherin. We developed a panel of CDH17 RGD-specific mAbs that inhibited β1 integrin activation, cell adhesion, and proliferation in colorectal and pancreatic cancer cells. This blocking effect was also effective in VE-cadherin–mediated β1 integrin activation of melanoma and breast cancer cells. RGD-specific mAbs were able to induce a significant increase in mouse survival after intravenous and intrasplenic injection of highly metastatic cells from melanoma and colorectal cancer causing lung and liver metastasis, respectively. Consequently, blocking the interaction between RGD cadherins and α2β1 integrin represents a promising therapy for distinct cancer metastases.

Metastatic KM12SM human colon cancer cells were obtained directly from Dr. I. Fidler (The University of Texas MD Anderson Cancer Center, Houston, TX), whereas MDA-MB-468 triple-negative human breast cancer cells and metastatic BLM human melanoma cells were kindly provided by Dr. J. Teixidó (CIB-CSIC, Madrid, Spain). KM12SM cells were authenticated by short tandem repeat analysis. SKBR3 breast cancer cells were obtained from Dr. A Villalobos (IIB-CSIC Madrid, Spain). RKO, Colo320, and HT29 human colon cancer cells; A375 human melanoma cells; and the pancreatic cancer cell line PANC1 were purchased from the ATCC and passaged for all of the experiments fewer than 6 months after purchase. BLM, SKBR3, and MDA-MB-468 were not authenticated in our laboratory. All cell lines were tested regularly for Mycoplasma contamination and cultured in DMEM (Invitrogen) containing 10% FCS (Invitrogen) and antibiotics at 37°C in a 5% CO₂ humidified atmosphere.

CDH17 polyclonal antibodies (H-167 and C-17), RhoGDIα (G-2), VE-cadherin (BV9), and FAK (A-17) were purchased from Santa Cruz Biotechnology. CDH17 antibody (#141713) and Src (AF3389) were from R&D Systems. Blocking anti-β1 (P5D2), α4 (ALC 1/1), and α5 (P1D6) integrin antibodies were from Abcam. β1 integrin antibody specific for high-affinity conformation (HUTS-21) and pY397-FAK (#611722) were from BD Transduction Laboratories. Antibodies against phospho-Src (#2101), JNK (56G8), phospho-JNK (G9), ERK1/2 (L24F12), and phospho-ERK1/2 (#9101) were from Cell Signaling Technology. Anti-α3 integrin (AB1920) was from EMD Millipore. The antibody against α1 (TS 2/7) integrin was a kind gift from Dr. C. Bernabeu (CIB-CSIC).

CDH17-derived peptides (VSLRGDTRG, SLRGDTR, and LRGDT), VE-cadherin domain 2 (QGLRGDSGT), VE-cadherin domain 3 (SILRGDYQD), CDH6 (DQDRGDGSL), CDH16 (RAIRGDTEG), and CDH20 (DMDRGDGSI) peptides were synthesized using solid phase chemistry with a Focus XC instrument (AAPPTec). In the CDH17 9-mer VSLRGDTRG, Tyr at position 600 was replaced with a Val to facilitate synthesis and hydrophilicity. The cyclic RGD peptide Cilengitide was from Selleck Chemicals.

siRNAs against human α1 (SASI_Hs01_00067020), α3 (SASI_Hs01_00196571), and α2 integrin (19) subunits were purchased form Sigma-Aldrich and transfected using jetPRIME reagent (Polyplus Transfection).

All animal experiments in this study were conducted according to the national RD 53/2013 and European Union Directive 2010/63/EU. All animal protocols were approved by the ethics committee of the Instituto de Salud Carlos III (CBA22_2014-v2) and Community of Madrid (PROEX 278/14). Four female Balb/c mice were immunized 3× intraperitoneally using ovalbumin (OVA)-conjugated CDH17 peptide (VSLRGDTRG) as antigen—the first time together with 50 μg of peptide-OVA emulsified in Freund's complete adjuvant and the next two times with 25 μg of peptide-OVA emulsified in Freund's incomplete adjuvant. Then, mAbs were prepared according to standard procedures (20). The selection of clones was carried out using an indirect ELISA against the CDH17 protein expressed in E. coli and the peptide VSLRGDTRG coupled to BSA as described (20), flow cytometry against KM12SM cells, and β1 integrin activation assays. These antibodies were used at 10 μg/mL in the different experiments. mAbs were grown in HAT media, purified by Protein G, and dialyzed against PBS for final testing and characterization.

For flow cytometry, 2.5 × 10⁵ KM12SM cells in 100 μL were incubated with immune sera or control serum (diluted 1:50) or with the different antibodies (10 μg/mL) for 30 min at 4°C. After incubation, the cells were washed with 2 mL of PBS and incubated in the dark at 37°C for 30 minutes in the presence of Alexa Fluor 488 goat anti-mouse IgG antibodies (Thermo Fisher Scientific). Fluorescence was analyzed in a COULTER EPICS XL cytofluorometer. At least 10,000 events per sample were acquired, and cells were identified on the basis of their specific forward (FSC) and side (SSC) light–scattering properties. Mean fluorescence intensities (MFI) for the indicated antibodies are shown inside each panel.

Cells were detached with 2 mmol/L EDTA in PBS, washed with PBS, resuspended in DMEM, and incubated with a CDH17 nine-amino acid peptide (VSLRGDTRG) containing the RGD motif for 25 minutes at 37°C in the presence of immune or control serum (diluted 1:50) or the indicated antibodies (10 μg/mL). Peptides were used at 1 μg/mL in the different experiments (at approximately 1 μmol/L). After incubation, cells were subjected to flow cytometry assays using antibodies specific for β1 integrin in high-affinity conformation and Alexa Fluor 488 goat anti-mouse IgG antibodies. Fluorescence was analyzed in a COULTER EPICS XL cytofluorometer. This experiment was repeated for the other CDH17 RGD peptides (seven and five amino acids), VE-cadherin domains 2 and 3 [VE-cad (D2) and VE-cad (D3)], and CDH6.

KM12SM cells were cultured on Matrigel-coated cover slides. Then, cells were fixed with 4% paraformaldehyde in PBS and permeabilized with 1% Triton X-100. After washing, cells were incubated for 40 minutes with primary antibodies (H167 and 25.4.1 at 10 μg/mL; 6.5.2, 6.6.1, 12.4.1, and control antibody at 30 μg/mL) in PBS with human gamma-globulin (40 μg/mL) at room temperature. Cells were then incubated 25 minutes with secondary antibodies coupled with Alexa Fluor 488 and 4,6-diamidino-2-phenylindole (DAPI). Samples were mounted with Fluorescence Mounting Medium (Dako), and images were captured using a TCS-SP5-AOBS confocal microscope with 63× oil immersion objective.

For internalization assays, KM12SM cells were cultured on Matrigel-coated cover slides and treated with the indicated mouse CDH17 antibodies (10 μg/mL) for 1 hour. Then, cells were fixed, permeabilized, and stained with rabbit anti-LAMP1 antibody (MyBioSource), followed by incubation with secondary antibodies (anti-mouse IgG coupled with Alexa Fluor 488 and anti-rabbit IgG coupled with Alexa Fluor 568). Zoom images were taken with 100× oil immersion objective.

Cancer cells (5 × 10³) were resuspended in serum-free DMEM and added upon 96-well plates previously coated with Matrigel (50 μL). After 24-hour incubation, images of the wells were taken under a microscope with 10× phase contrast and tubules from five different wells were counted by two different observers (21). Tubules were defined as tubes when formed by multiple single cells arranged in a row (one-cell thickness) or a thick bundle (thickness of several cells) connecting two cell “islands.”

Cells were incubated 3 hours in serum-free DMEM, detached with 2 mmol/L EDTA, washed, and treated with anti-RGD mAbs (10 μg/mL) for 40 minutes. Then, cells were added to plates previously coated with Matrigel (0.4 μg/mm²) for 30 minutes in the presence of either control IgG or the mAbs. Finally, cells were detached as before and lysed with lysis buffer [1% Igepal, 100 mmol/L NaCl, 2 mmol/L MgCl₂, 10% glycerol, protease inhibitors (cOmplete Mini, Roche), and phosphatase inhibitor cocktails 2 and 3 (Sigma-Aldrich) in 50 mmol/L Tris-HCl]. Protein extracts were separated in SDS-PAGE and transferred to nitrocellulose membranes, incubated with primary antibodies, washed, and incubated with HRP-conjugated secondary antibodies (Sigma-Aldrich). Membranes were visualized using SuperSignal West Pico Chemiluminescent Substrate (Thermo Fisher Scientific).

Cancer cells were seeded at 1 × 10⁴ cells/well on 96-well plates and incubated for 48 hours at 37°C in DMEM with 0.5% serum in the presence of the indicated antibodies, followed by 1-hour incubation with thiazolyl blue tetrazolium bromide (MTT; 0.6 mg/mL; Sigma-Aldrich). Cell viability was determined by absorbance at 560 nm and comparison with control cells collected ad initium. MTT assays were used as a surrogate of cell proliferation, as no increase in cell detachment or cell death was observed after mAb treatments.

Adhesion and invasion assays were performed as previously described (22). Briefly, cancer cells were labeled with BCECF-AM (Sigma-Aldrich), detached with EDTA/PBS, and incubated in serum-free DMEM in the presence of the indicated antibodies (10 μg/mL) for 10 minutes at 37°C. Then, 6 × 10⁴ cells in 100 μL were added to 96-well plates previously coated with Matrigel (0.4 μg/mm²) or type I collagen (Millipore; 0.3 μg/mm²), and the plates were incubated for 25 or 30 minutes at 37 °C, respectively. Subsequently, nonadhered cells were removed by three washes with DMEM. Bound cells were lysed with 1% SDS in PBS, and the extent of the adhesion was quantified using a fluorescence analyzer, POLARstar Galaxy (BMG LABTECH).

For invasion assays, 6 × 10⁴ cells were loaded onto 8-mm pore-size filters coated with 35 μL of Matrigel (1:3 dilution; BD Biosciences) in Transwell plates (Sigma-Aldrich) in the presence of antibodies. The lower compartment of the invasion chamber was filled with 5% serum DMEM. After 24 hours, noninvading cells were removed from the upper surface of the filter, and cells that migrated through the filter were fixed with 4% paraformaldehyde, stained with crystal violet, and counted under a microscope.

The ethics committees of the Consejo Superior de Investigaciones Científicas and Community of Madrid approved all protocols used. Swiss nude mice (Charles River; n = 4–6 per condition) were inoculated in the spleen with 1.5 × 10⁶ KM12SM cells in 0.1 mL PBS. A day after inoculation, mice were subjected to removal of the spleen to avoid local growth of the tumors. Then, these mice were inoculated intravenously with control antibodies or with anti-RGD antibodies 25.4.1 and 6.6.1 (50 mg/kg of weight, divided in seven doses) starting 2 days after inoculation and for 2 weeks. Mice were inspected daily for signs of disease, such as abdominal distension, locomotive deficit, or tumor detectable by palpation. When signs were visible, mice were euthanized, subjected to necropsy, and inspected for metastasis in liver.

For liver and lung colonization assessment, mice were inoculated in the spleen or tail vein respectively, with 1 × 10⁶ KM12SM, BLM, or MDA-MB-468 cells, and euthanized 96 hours after inoculation. RNA was isolated from the liver using Trizol (Thermo Fisher Scientific) and retrotranscribed, and 0.3 μg cDNA was subjected to PCR with Taq DNA polymerase (Thermo Fisher Scientific) to amplify human GAPDH as previously described (19). As a control, a 20-cycle amplification of murine β-actin was performed.

Data were analysed by Student t test (two conditions) or by one-way ANOVA followed by Tukey–Kramer multiple comparison test (more than two conditions). Histograms showed the average of the assessed values, and the error bars showed the SD. Each experiment was carried out at least three times. Survival curves were plotted with Kaplan–Meier technique and compared with the log-rank test. The minimum acceptable level of significance in all tests was P < 0.05.

Different commercial CDH17-specific antibodies were first tested for their capacity to inhibit β1 integrin activation and cell proliferation in metastatic colon cancer cells. CDH17 polyclonal antibodies (H167, C-17, or 141713) raised against the ectodomain or the carboxy-terminal domain of CDH17 failed to inhibit β1 integrin activation induced by CDH17 RGD peptides (Fig. 1A). We therefore prepared mouse polyclonal antibodies and mAbs against the domain six polypeptide of CDH17, which contains the RGD domain. These antibodies inhibited β1 integrin activation to a limited extent (<50%; Fig. 1A). In contrast, mouse polyclonal antibodies from mice immunized with the CDH17 peptide VSLRGDTRG inhibited β1 integrin activation at about 80% to 90%. These data confirmed the hypothesis that RGD-specific antibodies inhibited CDH17 RGD peptide-induced activation of β1 integrin. In addition, RGD peptide-specific antibodies significantly inhibited cell adhesion (Fig. 1B) and proliferation (Fig. 1C), whereas full-length CDH17 or domain six antibodies did not affect cell adhesion or proliferation of CDH17⁺ colon cancer cells. The adhesion to Matrigel was β1 integrin dependent, as confirmed using β1 integrin blocking antibodies. In summary, a selective immune response against the CDH17 RGD motif displayed β1 integrin blocking activity together with a clear reduction in cell adhesion and proliferation.

From our previous data, it seemed counterintuitive to use RGD peptides for cancer therapy, as they might activate α2β1 integrin in cancer epithelial cells. Cilengitide, a cyclic RGD pentapeptide targeting integrins αvβ3 and αvβ5, was unsuccessfully tested in clinical trials for glioblastoma (23). We did not observe any effect of Cilengitide on β1 integrin activation either as a ligand, probably due to the different flanking sequences and conformation, or as an inhibitory drug (Fig. 1D). Therefore, not all RGD sequences are appropriate for α2β1 integrin activation.

Mouse mAbs were prepared against the CDH17 RGD 9-mer peptide and functionally tested for their inhibitory capacity of β1 integrin activation. After initial anti-peptide ELISA, four clones, 6.5.2, 6.6.1, 12.4.1, and 25.4.1, were selected for characterization (Supplementary Table S1). Only mAbs 6.6.1 and 6.5.2 were positive for immunoprecipitation (Supplementary Fig. S1A). mAbs did not work in Western blot analysis (data not shown) and were only weakly positive for flow cytometry analysis of surface CDH17, except for antibodies 6.6.1 and 12.4.1 (Supplementary Fig. S1B). By confocal microscopy, the staining capacity of the RGD antibodies was good except for antibody 12.4.1 (Fig. 2A). We thus tested antibody internalization after CDH17 binding as a potential tool for toxin payload delivery to tumors. We observed a similar internalization in lysosomes (assessed by colocalization with the lysosomal marker LAMP1) for the anti-RGD mAb 25.4.1 and the control anti-CDH17 antibodies (Fig. 2B), suggesting an endocytosis of CDH17 after antibody binding.

The four purified mAbs were functionally active and inhibited the activation of β1 integrin (Fig. 2C), cell adhesion (55%–68%; Fig. 2D), and cell proliferation (Fig. 2E) in KM12SM colon cancer cells and in PANC1, a CDH17⁺ pancreatic cancer cell line (12). We obtained similar results with HT-29, a different CDH17⁺ colorectal cancer cell line (Supplementary Fig. S2A). In contrast, colorectal Colo320 cells that are negative for CDH17 expression were not affected by the RGD mAbs (Supplementary Fig. S2B). Collectively, these results support that the RGD mAbs preferentially recognize the conformational epitopes in the native CDH17 and underscore their value for in blocking cell adhesion and proliferation in colorectal and pancreatic cancer cell lines that express CDH17.

Next, we investigated the minimal length of the RGD motif necessary to induce β1 integrin activation in combination with the peptide-blocking capacity of the specific mAbs. In RKO colon cancer cells, we compared the 7-mer SLRGDTR and the 5-mer LRGDT with the positive control 9-mer peptide VSLRGDTRG. Peptides were used at 1 μg/mL in the different experiments. In the 9-mer peptide–treated cells, mAbs caused a strong inhibition of β1 integrin: antibody 25.4.1 caused a complete inhibition, followed by 6.6.1 (90% inhibition), and then 12.4.1 and 6.5.2 to a lesser degree (Fig. 3A). Shorter peptides induced integrin activation but to a lesser extent (50%). The 7-mer peptide was partially inhibited by mAbs 6.6.1 and 25.4.1 and was totally blocked by 12.4.1 (Fig. 3B and C). Regarding the 5-mer, mAbs had a minor inhibitory effect, likely due to the short length of the peptide supporting only a weak interaction with the binding site of the antibody (Fig. 3B and C).

To study the applicability of our mAbs to other RGD cadherins in other tumors, we examined the capacity of VE-cadherin and the CDH6, CDH16, and CDH20 RGD peptides to activate β1 integrin in metastatic melanoma and breast cancer cell lines that did not express CDH17 (Supplementary Fig. S3). Incubation with 9-mer RGD peptides from VE-cad (D2), VE-cad (D3), CDH6, and CDH20 (but not CDH16) caused β1 integrin activation in all cancer cell lines (Fig. 3D). We therefore investigated whether mAbs could block the β1 integrin activation caused by VE-cadherin and CDH6 RGD peptides in RKO cells. VE-cad (D2) was partially blocked by the mAbs 6.6.1 and 12.4.1, whereas the VE-cad (D3) site was fully blocked by 6.6.1, 12.4.1, and 25.4.1 (Fig. 3E and F). For CDH6 RGD, we observed a complete β1 integrin activation inhibition by mAb 12.4.1, a partial blocking effect for mAbs 6.6.1 and 6.5.2, and no effect for 25.4.1 in RKO cells (Fig. 3G). Overall, the mAb blocking effects were more significant for the VE-cad (D3) motif, which displays a flanking sequence YSI/L (RGD) similar to CDH17.

To confirm the broad applicability of cadherin RGD-specific antibodies for other cancer metastatic cell lines, we tested two melanoma (BLM and A375) and two breast (MDA-MB-468, a triple negative cell line, and SKBR3) cancer cells expressing VE-cadherin, but not CDH17 (Supplementary Fig. S3; ref. 19). RGD mAbs inhibited the high-affinity conformation of β1 integrin (Fig. 4A and Supplementary Fig. S4A), causing a significant reduction in cell adhesion (Fig. 4B and Supplementary Fig. S4B), proliferation (Fig. 4C and Supplementary Fig. S4C), and invasion capacity through the extracellular matrix (ECM), particularly in breast cancer cells (Fig. 4D and Supplementary Fig. S4D). A blocking anti-β1 integrin antibody was used as a positive control, and antibodies against full-length VE-cadherin and control antibody were used as as negative controls (Fig. 4). Interestingly, the antibody effects on adhesion, proliferation, and invasion were not associated to the reduction of cadherin expression in the cell surface (Fig. 4E), interference with tubule formation (Fig. 4F and Supplementary Fig. S5A), or alterations in endothelial permeability (Supplementary Fig. S5B). These results suggest that the RGD antibodies inhibit β1 integrin activation in VE-cadherin⁺ cancer cell lines without interfering with cadherin homotypic cell–cell interactions, similar to RGD-independent cell aggregation mechanisms observed in CDH17⁺ cells (12).

To confirm that the blocking capacity of the RGD mAbs was specific for α2β1 integrin, we investigated for other α integrin subunits in the cancer cells. We detected the presence of a significant expression of α1, α2, and α3 on the surface of both KM12SM and BLM cells (Supplementary Fig. S6A). Then, we performed siRNA silencing experiments for α1, α2, and α3 integrins in both cell lines (Supplementary Fig. S6B and S6C). Although RGD antibodies caused a significant reduction of adhesion to Matrigel and type I collagen in KM12SM and BLM cells silenced for α1 and α3 integrins, they did not cause any alteration in α2-silenced cells (Supplementary Fig. S6D). Together, these results confirm the specific blocking capacity of the RGD mAbs on α2β1 integrin.

To investigate whether the integrin signaling pathway was affected by the blocking RGD mAbs, representative KM12SM, PANC1, BLM, and MDA-MB-468 cell lines were cultured on Matrigel in the presence of the four RGD mAbs or a control IgG. In KM12SM colon cancer cells, the four RGD mAbs diminished the activation of FAK, JNK, and ERK kinases, which correlate with a decrease in cell adhesion and proliferation, but they did not affect Src or AKT activation (Fig. 5). In PANC1 pancreatic cancer cells, the mAbs provoked a clear decrease in JNK, ERK, and Src activation but had no effect on FAK or AKT. The mAbs (except for 6.5.2) inhibited JNK, ERK, and Src activation in BLM (melanoma) and MDA-MB-468 (breast) cancer cell lines, without affecting FAK activation. AKT activation was only reduced in melanoma. Src and AKT activities have been correlated with cell invasion (19). In summary, RGD mAbs caused different type-specific effects on cell signaling after blocking the integrin pathway activation in the four tested metastatic cell lines. These results confirm cell type–specific differences observed in VE-cadherin cell signaling on migration and invasion with respect to CDH17 colon cancer cells on cell adhesion.

For “in vivo” metastasis experiments, we first examined the effect of the antibodies on liver and lung homing capacity caused by colon and melanoma cancer cells, respectively. Tumor cells were inoculated in the spleen (for liver metastasis) or tail vein (for lung metastasis) of nude mice together with the RGD mAbs. Mice were euthanized at 96-hours postinoculation to avoid unspecific binding to target organs. mRNA was then extracted from liver and lung. In melanoma and breast cancer cells treated with mAb 12.4.1, no human GAPDH was detected in lung or liver (Fig. 6A). Furthermore, in colon cancer cells, liver homing was completely inhibited after tumor cells were treated with the RGD antibodies (Fig. 6A).

We next investigated the effects of RGD-specific mAbs on mouse survival after liver (colon) or lung metastasis (melanoma). For colon cancer, highly metastatic KM12SM cells were inoculated in the spleen to induce liver colonization through the hepatic portal vein. After 48 hours, mice were treated on alternate days for 2 weeks, with a total dose of 50 mg of antibody per kg of mouse weight. When signs of disease were evident, animals were euthanized and livers were removed for visual inspection of metastatic nodules. Otherwise, surviving animals were sacrificed at days 80 and 90 for lung and liver metastasis, respectively. Kaplan–Meier survival curves showed that all control mice died by day 50, although those inoculated with mAb 6.6.1 were still alive at this time point; further, only half of them developed liver metastasis by 90 days, and the rest survived to the experimental endpoint (Fig. 6B). mAb 25.4.1 caused an intermediate survival (50 days), whereas the half-life of mice inoculated with control antibody was 27 days. The number of metastatic nodules observed in liver was substantially reduced or abolished in half of the treated mice (Fig. 6C). For lung metastasis in melanoma, mice inoculated with mAbs 6.6.1 and 25.4.1 presented a prolonged survival in most of the treated animals. Lung inspection revealed a large number of metastases in control mice but very few or none in treated mice at the experimental endpoint (Fig. 6C). In summary, blocking the cadherin RGD-induced activation of α2β1 integrin constitutes a promising strategy for treating lung and liver metastasis in melanoma and colorectal cancer.

After many unsuccessful attempts in the last decades to target integrins with RGD peptides for cancer therapy, we propose here a novel strategy. Rather than using ECM RGD motifs, we have demonstrated that cadherin RGD motifs are efficient and selective targets for α2β1 integrin inhibition in metastatic cells of four different types of tumors. We have shown the protective effects of cadherin RGD-specific antibodies using two different cancer cell types (colorectal and melanoma) and two metastatic settings (lung and liver metastasis). In addition, RGD mAbs were effective in blocking the integrin pathway activation in pancreatic and breast cancer cell lines. The capacity to inhibit the metastatic capacity of different cancer cell lines enhances the value and promise of these mAbs as potential therapeutic agents to control metastatic spread in different solid tumors.

We proved that (i) peptide sequences containing the CDH17 RGD motif elicited blocking antibodies for β1 integrin activation in metastatic cells, (ii) the RGD peptide length required for integrin activation was as short as seven amino acids, (iii) mAbs raised against the CDH17 RGD motif were equally effective against β1 integrin activation with VE-cadherin and CDH6 RGD peptides, (iv) RGD mAbs blocked the integrin signaling pathway cascades in a cancer-dependent manner, and (v) RGD-specific mAbs significantly delayed and decreased liver and/or lung metastasis in colorectal and melanoma cancers. Moreover, because these mAbs did not affect the cell–cell homotypic interactions of cadherins (12, 19), their side effects on endothelial cells and vascular integrity might be avoided. The selectivity conferred by the cadherin RGD flanking sequences might avoid unspecific binding of the mAbs to the RGD motifs of the ECM proteins (fibronectin, collagen, etc.).

Integrins are key molecules in multiple cellular processes required for cancer progression and metastasis (see ref. 24 for a review). Integrin-targeted treatments for cancer therapy have been the focus of intense research efforts in the last decades, and integrin-targeted antibodies are under clinical evaluation for different diseases (24, 25). Historically, αvβ3, α4β1, and α5β1 integrins have been preferentially associated with cancer metastasis therapy (26). However, growing evidence underscores the critical relevance of α2β1 integrin as a key regulator of cancer metastasis (12, 19, 27, 28) despite some conflicting results with a mouse mammary tumor model (29, 30). Note that mouse CDH17 and VE-cadherins do not have RGD motifs (12), suggesting that alternative signaling pathways are used for metastasis progression in mice. A key role for α2β1 integrin has been well established in melanoma; rhabdomyosarcoma; and gastric, prostate, and colon cancer metastasis (9, 31–35). Therefore, we speculate that RGD mAbs might theoretically be applicable for all of these human tumors.

α2β1 integrin has been therapeutically targeted using blocking antibodies against α2 integrin subunit (GRB-500), which is currently in clinical trials for multiple sclerosis and ulcerative colitis (27). Another α2 integrin blocking mAb demonstrated high value in the inhibition of breast carcinoma cell growth (36). Regarding β1 integrin, volociximab (M-200), a monoclonal targeting α5β1 integrin is currently in clinical trials for solid tumors (37). In general, α2β1 integrin targeting is considered safe and tolerable (27). The RGD cadherin selectivity of our antibodies avoids the adverse effects caused by indiscriminate targeting of integrins using other therapeutic approaches, such as cyclic RGD molecules (38). A cyclic RGD peptide (Cilengitide) targeting the ligand binding site of αvβ3 and αvβ5 integrins reached clinical trials for glioblastoma but did not reach significant disease outcome improvement (39). In our hands, Cilengitide was completely ineffective in inducing or blocking α2β1 integrin activation (Fig. 1D).

Metastasis development has been associated with enriched cancer stem cell populations (40). β1 integrin is critical to preserve stem cell populations (41). Indeed, β1 and α6 integrins are enriched in cancer stem cells (24, 40). Interestingly, these two integrins are abundantly present in KM12SM colorectal metastatic cancer cells (8). Other stem cell markers present in KM12SM cells are ALDH1, ALCAM, and CD44 (9). These data suggest that the KM12SM metastatic cells have some stemness features.

Cancer cell types activate different integrin pathways for metastasis; therefore, it was not surprising to observe differences in the blocking effects on downstream signaling caused by the mAbs in the different cancer types. Cadherin RGD-specific antibodies blocked the integrin signaling pathways in a cancer cell–dependent manner. As a general rule, proliferation pathways were always inhibited, whereas adhesion and invasion were cell type dependent. These data support that β1 integrin is a critical molecule for the activation of multiple pathways required for metastatic colonization.

In summary, our data support that antibodies against the cadherin RGD motifs reduced the proliferation, adhesion, and invasion capacity of metastatic cancer cell lines by inhibiting the activation of α2β1 integrin. The extended mice survival demonstrates the potential therapeutic effect in cancer metastasis, specifically in colorectal cancer and melanoma. Moreover, it is likely that this capacity might be extended to other neoplasias such as pancreatic and breast cancer. Based on the different biochemical properties (antigen recognition by ELISA, immunoprecipitation, confocal microscopy, and yield and metastatic inhibition capacity), mAb 6.6.1 seems to be the most convenient antibody for further therapeutic developments. The next steps should include the humanization process of the most effective mAbs, particularly 6.6.1, for potential clinical application. Finally, the crucial role of α2β1 integrin in fibrosis, platelet-mediated thrombosis, and angiogenesis (29) might increase considerably the application range of cadherin RGD mAbs to other diseases.

J.I. Casal holds ownership interest (including patents) in Protein Alternatives SL. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R.A. Bartolomé, J.I. Casal

Development of methodology: R.A. Bartolomé, J.I. Imbaud

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.A. Bartolomé, C. Aizpurua, M. Jaén, S. Torres, J.I. Imbaud

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.A. Bartolomé, C. Aizpurua, M. Jaén, S. Torres, J.I. Casal

Writing, review, and/or revision of the manuscript: R.A. Bartolomé, J.I. Imbaud, J.I. Casal

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Calviño

Study supervision: J.I. Imbaud, J.I. Casal

The authors appreciate the technical assistance of Mercedes Dominguez (Centro Nacional de Microbiología, ISCIII, Majadahonda, Madrid) for the preparation of the monoclonal antibodies.

This research was supported by grant BIO2015-66489-R from the MINECO, Foundation Ramón Areces, and PRB2 (IPT13/0001-ISCIII-SGEFI/FEDER).