Breast cancer cell migration is regulated through junctional adhesion molecule-A-mediated activation of Rap1 GTPase

The MCF7 breast cancer cell line was obtained from the European Collection of Cell Cultures through Sigma-Aldrich (Poole, UK) and maintained in minimum essential medium (Sigma-Aldrich) supplemented with 10% fetal bovine serum (FBS) (Lonza, Basel, Switzerland), 2 mM L-glutamine, 50 U/mL penicillin, 50 μg/mL streptomycin (Invitrogen Corporation, Carlsbad, CA, USA), and 1% non-essential amino acids (Sigma-Aldrich). Cells were maintained at 37°C in humidified air with 5% CO₂.

Human breast tissue samples from lumpectomy or mastectomy patients were gathered with informed consent in protocols approved by the Beaumont Hospital medical ethics (research) committee and were used to generate mammary epithelial primary cell cultures with minor modifications from published methods [26]. In brief, tissue biopsies from within the tumor (T) and from histopathologically normal non-tumor margins (N) were incubated in penicillin/streptomycin/neomycin (Invitrogen Corporation); minced in Dulbecco's modified Eagle's medium/F12 (Sigma-Aldrich) containing penicillin/streptomycin/neomycin, 10% FBS, 10 μg/mL insulin, 5 μg/mL fungizone, 100 U/mL hyaluronidase 1-S, and 200 U/mL collagenase (Sigma-Aldrich); and agitated for 2 hours at 37°C. Cells were pelleted and washed before being cultured in mammary epithelial growth medium (Lonza) at 37°C with 5% CO₂. Trypsin/ethylenediaminetetraacetic acid and soybean trypsin inhibitor were used to subculture confluent flasks. Cells were harvested after one passage to generate enough material for IP assays and one or two passages for protein isolation. Only patient-matched primary cultures were used (Table 1).

JAM-A siRNAs were transfected into MCF7 breast cancer cells using the N-TER nanoparticle siRNA transfection system (Sigma-Aldrich). In brief, cells (1.5 × 10⁶ cells/10-cm dish) were seeded 16 hours prior to transfection. Cells were washed in phosphate-buffered saline (PBS) and transfected with nanoparticles containing 75 nM final concentration of JAM-A siRNAs, negative control siRNA, or mock (N-TER reagent only) in minimum essential medium. Assays were performed after transfection for 48 hours at 37°C with 5% CO₂. Two alternative JAM-A siRNAs were used for all experiments. Representative results from a single siRNA are shown for clarity.

All antibodies used for Western blot and immunofluorescence analyses were obtained from commercial sources: rabbit polyclonal anti-JAM-A (Zymed Laboratories Inc., now part of Invitrogen Corporation), MAb13 rat anti-β1 integrin (BD Biosciences, San Jose, CA, USA), rabbit anti-β2, rabbit anti-β3, rabbit anti-β4, rabbit anti-β5, rabbit anti-αV, rabbit anti-α4, rabbit anti-α5 integrins (Cell Signaling Technology, Danvers, MA, USA), rabbit anti-actin (Abcam, Cambridge, UK), rabbit anti-Rap1 (Millipore Corporation, Billerica, MA, USA), rabbit anti-AF-6 (Invitrogen Corporation), mouse anti-PDZGEF2 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA), and mouse J.10.4 inhibitory anti-JAM-A (Santa Cruz Biotechnology, Inc.). Horseradish peroxidase (HRP)-conjugated secondary antibodies were anti-mouse and anti-rat (Sigma-Aldrich) or anti-rabbit (Cell Signaling Technology). IgG isotype control antibodies were mouse, rabbit, and rat (Sigma-Aldrich). The Rap1 pharmacological inhibitor was GGTI-298 (Calbiochem/Merck, Darmstadt, Germany).

Cell adhesion strips coated with fibronectin (which binds αVβ1 and α5β1 integrins) or bovine serum albumin (BSA)-control substrate (Millipore Corporation) were rehydrated in 96-well plates with PBS for 30 minutes at room temperature. Cells (1 × 10⁵/well) were added and incubated for 2 hours at 37°C and 5% CO₂, washed in PBS, stained with 0.2% crystal violet (Sigma-Aldrich) in 10% ethanol for 5 minutes at room temperature, and rinsed in PBS. Cell-bound stain was solubilized by gently shaking with solubilization buffer (1:1, 0.1 M NaH₂PO₄:50% ethanol) for 5 minutes. Absorbance was measured at 560 nm on a microplate reader. Results from three independent experiments with three replicates per experiment were pooled.

Transwell migration assays were conducted on MCF7 cells following transient JAM-A gene-expression knockdown. Transwell chambers (8 μM pore) were coated with 10 μg/mL fibronectin (Sigma-Aldrich) overnight at 4°C, washed in PBS, and rehydrated with serum-free media for 30 minutes at 37°C. Media were removed and cells (1 × 10⁵/chamber) were added to upper chambers with 15% FBS in lower chambers as a chemoattractant. Chambers were incubated for 3 hours at 37°C and 5% CO₂. Migrated cells on the underside of the filter were fixed in 10% ethanol for 20 minutes prior to staining with 4,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich) for 10 minutes at room temperature. Membranes were mounted on glass slides and cells were enumerated using Cell B software (Olympus, Tokyo, Japan) to analyze multiple fluorescent micrographs. Results from three independent experiments with three replicates per experiment were pooled.

Scratch-wound migration assays were conducted on wild-type MCF7 cells after treatment with inhibitors. Confluent cell monolayers in 24-well plates were preincubated for 2 hours with 5 μg/mL mouse anti-JAM-A J.10.4 antibody, 5 μg/mL MAb13 anti-β1 integrin antibody, 10 μM Rap1 inhibitor GGTI-298, appropriate isotype-control IgG, or dimethyl sulphoxide (Sigma-Aldrich) vehicle control. A scratch wound was made using a pipette tip. Media were replaced and wounds were photographed at 0, 2, and 6 hours. Scion Image software (Scion Corporation Ltd., Frederick, MD, USA) was used to measure closure of the wound over time by averaging six individual measurements of wound size for each wound at each timepoint. Results from three independent experiments with three replicates per experiment were pooled.

Adherent cells were washed in Tris-buffered saline and then scraped and dounced in 1 mL of Rap1 activation lysis buffer containing 50 mM Tris-HCl, 0.5 M NaCl, 1% NP40, 2.5 mM MgCl₂, and 15% glycerol (Millipore Corporation) and 1% protease and phosphatase inhibitors (Sigma-Aldrich). Lysates were incubated with 30 μg of Ral GDS-RBD (Rap-binding domain) agarose slurry (Millipore Corporation) for 45 minutes at 4°C. Beads were pelleted and washed in lysis buffer. Bound proteins were recovered by boiling at 95°C for 5 minutes in 40 μL of 2X sample buffer. Each immunoblot depicted is representative of three independent experiments, and densitometry was conducted on triplicate experiments. Densitometric data were used to determine the average active-to-total Rap1 protein expression over triplicate experiments.

Cultured cells were washed in 10 mL of PBS, scraped, and dounced in Relax lysis buffer containing 100 mM KCl, 3 mM NaCl, 3.5 mM MgCl₂, 10 mM HEPES pH 7.4, and 1% Triton-X100 as well as protease and phosphatase inhibitor cocktails (Sigma-Aldrich). Protein samples were separated by SDS-PAGE under reducing conditions by using Tris-glycine running buffer. After electrophoresis, proteins were transferred to nitrocellulose membranes (Optitran; Sigma-Aldrich). Membranes were blocked in 5% milk for 1 hour. Protein expression was detected by using primary antibodies incubated overnight at 4°C. Membranes were washed and incubated for 1 hour with HRP-conjugated secondary antibodies. Antigen-antibody complexes were detected by using Western Lightning Enhanced Chemiluminescence reagent (PerkinElmer, Waltham, MA, USA). Each immunoblot depicted is representative of three independent experiments, and densitometry was conducted on triplicate experiments.

Protein extraction was conducted as above. Equal protein concentrations from control siRNA, JAM-A siRNA, and mock-transfected cells were subjected to JAM-A IP. Preclear was conducted via rotation of protein lysates for 1 hour at 4°C with protein G-sepharose beads (GE Healthcare, Chalfont St.Giles, Bucks, UK). IP antibodies (J.10.4 anti-JAM-A antibody or isotype control mouse IgG; 4 μg/mL) were rotated for 1 hour at 4°C. Bound antibody was retrieved by rotation with protein G-sepharose beads for 3 hours at 4°C. Beads were washed in Relax lysis buffer, and bound proteins were recovered by boiling at 95°C for 5 minutes in 40 μL of 2X sample buffer. Western blotting was performed as above for JAM-A, AF-6, PDZ-GEF2, Rap1, and β1-integrin. Each IP and associated immunoblot was conducted three times for cell lines and once each per primary culture.

Our previous study reported that high JAM-A expression levels in breast tissues from patients with invasive breast cancer correlated with reduced metastasis-free survival [19]. We and others have demonstrated that, likely owing to reductions in β1-integrin protein, knockdown of JAM-A gene expression in epithelial cells results in decreased collective cell migration [19, 27, 28] (Supplementary Figure S1A in Additional file 1). Indeed, integrins perform key roles at several key steps required for cell migration: adhesion assembly, disassembly, and turnover [29, 30]. Taken together, our results to date suggest a role for JAM-A in promoting breast cancer cell migration through a β1-integrin-dependent pathway. To investigate this hypothesis, we used the MCF7 breast cancer cell line, which expresses high endogenous levels of JAM-A and β1-integrin. JAM-A protein levels were knocked down in these cells using a nanoparticle delivery system to transiently transfect cells with siRNA targeting JAM-A. Transiently-tranfected cells were analyzed after 48 hours for JAM-A knockdown at the protein level. JAM-A siRNA transfected cells displayed a 90% reduction in JAM-A protein expression (Figure 1a). Similar effects were observed with a second JAM-A siRNA construct (Supplementary Figure S1B in Additional file 1).

To investigate the link between JAM-A, β1-integrin, and cell migration, the ability of transfected cells to adhere to and migrate across a fibronectin substrate was analyzed (Figure 1b, c). Fibronectin binds multiple β1-integrin proteins through an RGD domain [31] and thus was used to facilitate analysis of the effects of JAM-A knockdown. First, transfected cells were allowed to adhere to fibronectin-coated transwell supports prior to cell staining and absorbance measurement. JAM-A knockdown cells showed an approximate 40% reduction in adhesion to fibronectin (Figure 1b) (P = 0.02) but not to control transwells coated with BSA (data not shown). Next, transfected cells were allowed to migrate across fibronectin-coated supports with 8-μm pores prior to fluorescent staining and enumeration. Here, in addition to showing a reduction in collective cell migration demonstrated in previous reports [19, 27, 28], JAM-A knockdown cells showed an approximate 50% reduction in individual cell motility in these matrix-specific transwell migration assays (Figure 1c) (P = 0.01), further underlining the important role of JAM-A in the regulation of breast cancer cell migration. However, JAM-A antagonism did not reduce invasion of MCF7 cells across Matrigel-coated filters in classic invasion assays (Supplementary Figure S2C in Additional file 2), owing to the fact that MCF7 cells are virtually non-invasive in this model (Supplementary Figure S2A in Additional file 2).

We and others have previously reported that JAM-A knockdown induces a concomitant reduction in β1-integrin in epithelial cells [10, 19, 27]. To probe whether this was a specific effect of JAM-A knockdown on β1-integrin in breast cancer cells, the expression of several integrin subunit proteins was analyzed following JAM-A knockdown in MCF7 breast cancer cells. Total protein isolates were prepared from MCF7 cells transfected with either JAM-A siRNA or control siRNA, and the protein expression levels of a panel of alpha- and beta-subunit integrins were assessed by Western blot analysis (Figure 2a, b). As previously stated, JAM-A siRNA transfectants achieved a greater than 90% reduction in JAM-A protein expression in comparison with control siRNA transfection. No change in the expression of β3-, β4-, or β5-integrins was observed upon JAM-A knockdown, as evidenced by densitometric analysis of triplicate immunoblots (Figure 2b). Expression of α4-integrin was not detected in MCF7 cells. Similar to previous studies [10, 19, 27], a 50% reduction in the expression of β1-integrin was observed upon transient JAM-A knockdown. Furthermore, reductions of approximately 30% and 60%, respectively, were observed for protein expression levels of αV- and α5-integrin subunits, both of which have been widely reported to complex with β1-integrin to form functioning fibronectin receptors [32]. These results further suggest that JAM-A exerts specific effects on β1-integrin heterodimers.

We next sought to investigate the signalling events connecting JAM-A engagement at the cell-cell interface to β1-integrin function at the cell-matrix interface. Given the spatial separation of these proteins, we hypothesized that their physical association was unlikely. To test this, JAM-A IPs were conducted on protein isolates from MCF7 cells transfected with JAM-A or control siRNA (Figure 2c). A reduction of β1-integrin total protein expression was confirmed in input samples of equal protein concentration. However, no bands were observed in JAM-A immunoprecipitated lanes immunoblotted for β1-integrin, confirming an absence of physical association between JAM-A and β1-integrin. We therefore concluded that other signalling proteins must be responsible for linking β1-integrin function to JAM-A.

One candidate protein, the integrin activator Rap1 GTPase [33], has been shown to regulate tissue polarity, lumen formation, and invasive potential in human breast epithelial cells [25]. We therefore investigated whether JAM-A knockdown using siRNA or functional inhibition using JAM-A inhibitory antibody altered Rap1 expression and activity (Figure 2d). Western blot analyses of MCF7 breast cancer cells showed a marginal decrease in total protein expression of Rap1 following JAM-A knockdown. However, a near abolition of active (GTP-bound) Rap1 was observed in JAM-A knockdown cells. In an effort to further confirm the regulation of Rap1 activation status by JAM-A, MCF7 cells were treated with an inhibitory antibody to JAM-A (J10.4). Western blot analysis demonstrated negligible changes in total Rap1 expression but dramatic reductions in active Rap1 following JAM-A inhibition. Indeed, reductions of 50% (P <0.01) and 80% (P <0.02) were observed for expression ratios of active Rap1 to total Rap1 following JAM-A knockdown and inhibition, respectively (Figure 2e), strongly indicating that JAM-A regulates downstream activation of Rap1. However, it is notable that the ratios of active-to-total Rap1 were not equivalent between knockdown and inhibition conditions, because JAM-A knockdown (but not inhibition) induced a reduction in the total levels of Rap1 protein.

Our results thus far suggested the existence of a JAM-A signalling pathway in MCF7 breast cancer cells, and that this pathway is initiated by JAM-A signalling via Rap1 and β1-integrin and culminates in cancer cell migration. We next reasoned that, if this pathway were linear, inhibition of any step should elicit inhibitory effects on cell migration similar to those induced upon inhibition of JAM-A alone. To investigate this hypothesis, we first treated MCF7 cells with the Rap1 pharmacological inhibitor GGTI-298 and verified that active Rap1 is reduced following treatment (Figure 3a). GGTI-298 also significantly reduced MCF7 cancer cell migration in scratch-wound assays (Figure 3b) (P = 0.015). Next, we demonstrated that treatment of MCF7 cells with an inhibitory antibody targeting β1-integrin elicited a similar reduction in MCF7 cell migration, from approximately 32% wound closure to 18% wound closure at 6 hours (Figure 3c) (P = 0.001). This was mirrored upon JAM-A inhibition, in which cell migration was observed to decrease from approximately 32% to 15% wound closure after 6 hours (Figure 3d) (P <0.0001). Combined treatment of MCF7 cells with inhibitors of Rap1, β1-integrin, and JAM-A resulted in a decrease in cancer cell migration from approximately 35% to 18% wound closure after 6 hours (Figure 3e) (P <0.0001). When the migratory differences between treatments and controls were quantitatively compared, no additive effects in response to inhibitor combinations versus single inhibitors alone were observed (Figure 3f). This indicated that JAM-A, Rap1, and β1-integrin are likely to function together in a linear signalling pathway in breast cancer cells.

Using IP strategies, we next sought to determine the JAM-A associations and signalling events that may affect downstream Rap1 activation. Total protein was extracted from MCF7 cells transfected with control siRNA or JAM-A siRNA, and JAM-A protein-binding partners were co-immunoprecipitated using an anti-JAM-A antibody (Figure 4). We first tested for the presence of Rap1 in JAM-A immunoprecipitates but did not detect any co-association. We therefore reasoned that intermediate signalling proteins must link JAM-A to the downstream activation of Rap1. Knockdown of the Rap1-associating protein [34] AF-6 (afadin) had been shown to result in reductions in active Rap1 and colorectal cancer cell migration [27]. We therefore investigated whether JAM-A co-associated with AF-6 in MC7F breast cancer cells. JAM-A immunoprecipitates from MCF7 protein lysates were immunoblotted with AF-6, and co-association between a pool of JAM-A and AF-6 was detected (Figure 4). Similar effects were observed using a second, independent JAM-A siRNA construct (Supplementary Figure S1B in Additional file 1). In addition to demonstrating a role for AF-6, previous studies had demonstrated a role for the Rap1 activator PDZ-GEF2 in both lung cancer cell adhesion [35] and colorectal carcinoma cell migration [27]. Similarly, JAM-A immunoprecipitates immunoblotted for PDZ-GEF2 confirmed co-association between a pool of JAM-A and PDZ-GEF2 in MCF7 breast cancer cells (Figure 4).

Our accumulating results suggested a breast cancer signalling pathway in which JAM-A associates with AF-6 and PDZ-GEF2 to activate Rap1 and regulate β1-integrin-mediated cell migration. To determine the ex vivo relevance of this putative JAM-A signalling pathway, we sought to verify our results in primary cultures isolated from multiple tissues of patients with breast cancer (Table 1). Primary cultures were generated from three patient-matched tumor (1T, 2T, and 3T) and non-tumor (1N, 2N, and 3N) tissues, and total protein was extracted for investigation of JAM-A co-associations. JAM-A IP was conducted on all samples followed by immunoblotting for JAM-A, AF-6, PDZ-GEF2, total Rap1, and β1-integrin (Figure 5a). Analysis of total protein levels demonstrated a small increase in JAM-A expression in patient tumor cultures relative to non-tumor cultures (Figure 5a input and Supplementary Figure S3 in Additional file 3). Negligible differences in protein expression between tumor and non-tumor cultures were observed for AF-6, Rap1, and β1-integrin (Figure 5a input and Supplementary Figure S3 in Additional file 3). Notably, however, after densitometric pooling, tumor cultures displayed a reduced total protein expression of PDZ-GEF2 in comparison with non-tumor cultures (Supplementary Figure S3 in Additional file 3). Analysis of JAM-A immunoprecipitates demonstrated co-association between JAM-A and a pool of AF-6 and PDZ-GEF2 but not Rap1 or β1-integrin (Figure 5a), which mirrors data acquired from MCF7 breast cancer cells (Figure 4). Pooled densitometric analysis of tumor-to-non-tumor ratios following IP experiments revealed a trend toward increased association of both AF-6 and PDZ-GEF2 with JAM-A, suggesting enhanced formation of a JAM-A/AF-6/PDZ-GEF2 signalling complex in tumor cells (Supplementary Figure S3 in Additional file 3). Together, our findings provide compelling evidence of a novel role for the cell-cell adhesion protein, JAM-A, in influencing breast cancer cell migration at the cell-matrix interface via regulation of Rap1 and β1-integrin.