Na_V1.5 Na⁺ channels allosterically regulate the NHE-1 exchanger and promote the activity of breast cancer cell invadopodia

Breast cancer is the primary cause of death by cancer in women worldwide, and patients mostly die because of metastases. At the cellular level, a determining step in the development of metastases depends on the capacity of cancer cells to degrade and migrate through the extracellular matrix (ECM). This is performed by invadopodia, which are actin-rich organelles that protrude into the ECM and are responsible for its proteolysis through the recruitment of both membrane-associated and extracellularly released soluble proteases (Linder et al., 2011). The proteolytic activity of invadopodia is highly dependent on the acidification of the peri-invadopodial extracellular compartment (Brisson et al., 2012) through the localised activity of Na⁺/H⁺ exchanger type 1 (NHE-1) (Busco et al., 2010; Magalhaes et al., 2011). Voltage-gated Na⁺ channels (Na_V) are abnormally expressed in cancer cells from different epithelial origins (e.g. breast, lung, prostate, colon and cervix) and, in melanoma, their function might be associated with cancer progression (Diaz et al., 2007; Hernandez-Plata et al., 2012; House et al., 2010; Roger et al., 2006; Roger et al., 2007; Yildirim et al., 2012). The Na_V1.5 isoform (also known as SCN5A) is overexpressed in breast cancer tissues, is associated with the development of metastases and is functional in highly invasive breast cancer cells (Fraser et al., 2005; Roger et al., 2003; Yang et al., 2012). In the highly aggressive human MDA-MB-231 breast cancer cell line, the activity of Na_V1.5 enhances ECM degradation by increasing the activity of acidic cysteine cathepsins through the acidification of the pericellular microenvironment (Brisson et al., 2011; Gillet et al., 2009). In the present study, we demonstrate that Na_V1.5 interacts with NHE-1 in focal ECM degradation sites corresponding to caveolin-1-containing invadopodia. We also found that Na_V1.5 activity controls Src kinase activity, cortactin phosphorylation, actin cytoskeletal dynamics and is responsible for the allosteric modulation of NHE-1, thus enhancing invadopodial proteolytic activity.

Invadopodia are key structures involved in matrix remodelling by cancer cells (Linder, 2007). Therefore, we analysed the ability of MDA-MB-231 cells to degrade ECM by growing them on a planar matrix of Matrigel™ containing DQ-gelatin, a fluorogenic substrate for proteases. Multiple subcellular areas of degradation were observed, which we designate ‘focal areas’; these colocalised with F-actin condensations (Fig. 1A). High-resolution structured-illumination microscopy (SIM) imaging of these areas showed intense spots of matrix degradation. Approximately 10% of the studied spots were surrounded by F-actin rings, forming structures of ∼2 µm in diameter in axial (xy) sections (Fig. 1B,C). These structures can also be seen on sagittal (xz) and coronal (yz) projections. They were located below the cell body and showed a substantial penetration into the ECM (>5 µm depth) (Fig. 1B; supplementary material Fig. S1A). This F-actin organisation might correlate with that observed in electron microscopy studies in the same cancer cells (Schoumacher et al., 2010) and with the protrusive structure of podosomes observed in dendritic cells (Gawden-Bone et al., 2010). One hypothesis that is consistent with these observations is that some invadopodia show a particular structure, namely that the protrusions into the ECM have a hollow tubular shape, with the centre showing the strongest proteolytic activity. Such a structure would allow a tight control of pH and protease activity inside the invadopodium. Alternatively, these images could be interpreted as being the endocytosed degradation products inside the invadopodium.

We used a MagicRed® [MR-(RR)₂] substrate assay to determine the activity of the extracellular protease cathepsin B in the MDA-MB-231 cells growing on DQ-gelatin, and found that the focal areas of DQ-gelatin degradation in the Matrigel™ had an intense activity of cathepsin B (Pearson's coefficient, r = 0.65, Fig. 1D). This is consistent with our previous study suggesting that cathepsin B had a substantial role in the invasion of MDA-MB-231 cells (Gillet et al., 2009). Moreover, there was a strong colocalisation between Na_V1.5 channels and cathepsin B (r = 0.82, supplementary material Fig. S1B). Na_V1.5 also colocalised with MT1-MMP (also known as MMP14), which has been proposed to serve as a good invadopodia marker (Linder et al., 2011) (r = 0.69, Fig. 2A). We have previously shown that Na_V1.5 is located in caveolin-1-rich membrane fractions (caveolae) (Brisson et al., 2011). Using immunofluorescence imaging, we found a punctate labelling of Na_V1.5 that was highly correlated with caveolin-1 (r = 0.75, Fig. 2B). Caveolae are dynamic lipid raft platforms at the plasmalemma that allow protein clustering, and it is thought that they are important for the composition and functioning of invadopodia (Caldieri et al., 2009; Yamaguchi et al., 2009) given that they could recruit multiple components required for the focal degradation of the ECM (Caldieri and Buccione, 2010; Caldieri et al., 2012; Yamaguchi and Oikawa, 2010). Here, we observed that caveolin-1 strongly colocalised with F-actin condensation loci (r = 0.75, Fig. 2C). Proximity ligation assays indicated that there was a close association of Na_V1.5 and caveolin-1 in areas of DQ-gelatin degradation (Fig. 2D). Fig. 2E shows that invadopodial fractions from MDA-MB-231 cells, obtained as previously described (Busco et al., 2010) and characterised by the presence of the invadopodia markers cortactin, focal adhesion kinase (FAK) and MT1-MMP, contained NHE-1 as well as caveolin-1, Na_V1.5, and immature and mature forms of cathepsin B, which is the key protease in Na_V1.5-dependent ECM invasion (Gillet et al., 2009). Using SIM technology, we found that Na_V1.5 and NHE-1 were strongly colocalised (r = 0.92) in restricted domains protruding into the extracellular matrix (Fig. 3A; supplementary material Fig. S1C). Because the SIM apparatus is only equipped with two lasers, we assessed the colocalisation of Na_V1.5 and/or NHE-1 with ECM degradation spots in invadopodia by using confocal imaging of cells seeded on a layer of Matrigel™ containing BSA-BODIPY-FL. This showed that there were multiple protruding organelles involved in matrix proteolysis at the underside of the cell body, which strongly colocalised with NHE-1 and Na_V1.5 (Fig. 3B, supplementary material Fig. S1D). Proximity ligation assays also confirmed the close vicinity of NHE-1 and Na_V1.5 in focused matrix degradation spots (Fig. 3C). As can be seen in Fig. 3D, both Na_V1.5 and caveolin-1 coimmunoprecipitated with NHE-1, consistent with a (direct or indirect) interaction between these three membrane proteins. By contrast, Na_V1.5 and caveolin-1 were not immunodetected when a control IgG was used for precipitation.

NHE-1 is ubiquitously expressed at the plasma membrane of all mammalian cells, has a very low activity at the physiological intracellular pH (pHi) and activates rapidly when the cytoplasm becomes acidic, following a dimeric Monod–Wyman–Changeux cooperative mechanism (Lacroix et al., 2004). We next decided to investigate whether the presence of Na_V1.5 in breast cancer cells could modify the sensitivity of NHE-1 to internal protons. To do so, we constructed two stable cell lines derived from MDA-MB-231 cells: one expressing a shRNA targeting the expression of the SCN5A (shNa_V1.5) and another one expressing a shRNA that does not target any known gene (shCTL). ShNa_V1.5 reduced the mRNA expression of SCN5A gene by ∼80%, which resulted in the complete disappearance of Na⁺ currents in these cells (supplementary material Fig. S2A–C). Treating shCTL cells with 30 µM tetrodotoxin (TTX) to block Na⁺ currents reduced the relative invasion by ∼40% (Fig. 3E) without affecting cell viability (supplementary material Fig. S3C) or proliferation (Gillet et al., 2009). ShNa_V1.5 cells had a ∼40% reduced invasion compared to shCTL cells, and TTX had no supplementary effect. We then measured NHE-1 activity as a function of pHi, by measuring the fast kinetics of Li⁺ uptake, and showed that the presence of active Na_V1.5 channels in MDA-MB-231 cancer cells significantly (P<0.05) increased Li⁺ uptake in pHi range 6.4 to 7 (Fig. 3F; supplementary material Fig. S2D). The change in cooperativity strongly suggests that Na_V1.5 function in cancer cells allosterically increases NHE-1 activity in neutral ranges of pHi.

MDA-MB-231 cancer cells display a ‘fibroblast-like’ shape when grown on a planar layer of Matrigel™, exhibiting a rich F-actin labelling in the barbed-end of the lamellipodial leading edge, as well as at the rear of the cell. This shape, characteristic of highly invasive cells, was lost when cells did not express Na_V1.5 channels or were treated with TTX (Fig. 4A,B). In these two cases the F-actin cytoskeleton was disturbed, showed uniform submembrane localisation and was associated with a change in cell morphology characterised by the increase of the circularity index of cancer cells (Fig. 4C). The actin-nucleation-promoting factor cortactin (encoded by the gene CTTN) is a crucial regulator of the actin cytoskeletal dynamics and, consequently, of the formation and function of invadopodia. Cortactin is known to bind (through a SH3 domain) and to activate the Arp2/3 complex, which is responsible for the formation of branched F-actin and the enhancement of cell migration (MacGrath and Koleske, 2012). Cortactin also binds and inhibits the actin-severing protein cofilin (encoded by the gene CFL1). The phosphorylation of cortactin releases cofilin and leads to the increase of polymerizing actin barbed ends (Oser et al., 2009). By performing prognostic analyses of gene expression in cohorts of breast cancer patients (bc-GenExMiner v3.0; http://bcgenex.centregauducheau.fr; Jézéquel et al., 2012) we found that the overexpression of both CTTN [P<0.0001; hazard ratio (a measure of survival in patients with the condition compared with that of those without), 1.49, 95% confidence interval, 1.26–1.70] and CFL1 (P<0.0001; hazard ratio 1.40, 95% confidence interval, 1.19–1.64) genes were associated with a reduced survival without metastatic relapse (supplementary material Fig. S3A,B) indicating the crucial role of these proteins in cancer cell invasion and the development of metastases. Here, we found that inhibiting NHE-1 by the use of ethyl-isopropyl amiloride (EIPA) also increased cell circularity in both shCTL and shNa_V1.5 cells (Fig. 4C). This could be because the activity of cofilin is pH dependent, and, hence, the intracellular alkalinisation induced by NHE-1 activity might disrupt the interaction between cofilin and cortactin thus enhancing the initiation of actin polymerisation (Frantz et al., 2008; Pope et al., 2004). However, there are partially additive effects on cell circularity when combining both TTX and EIPA as compared with the use of one of these two inhibitors separately. We hypothesise that, independently of its regulation of NHE-1 activity, Na_V1.5 could modulate F-actin polymerisation and invadopodia formation. Cortactin can also be released from cofilin when phosphorylated on tyrosine residues (Y421 and Y466) by Src kinase, which regulates its ability to promote branched actin nucleation by modulating its ability to recruit binding partners (Oser et al., 2010; Oser et al., 2009). We found that inhibiting the activity of Na_V1.5 channels with TTX, which had no effect on cell viability (supplementary material Fig. S3C), rapidly reduced the level of pY421-cortactin [by 26.4±8.7% after 5 minutes, 36.0±9.8% after 10 minutes and by 48.5±16.2% after 24 hours] (Fig. 4D). This was not due to a general dephosphorylation mechanism (supplementary material Fig. S3D–F) and was not observed in cells that did not express Na_V1.5 (shNa_V1.5). An analysis of the levels of pY421-cortactin at various time points of the TTX treatment, revealed that there were reductions in the amount of pY421-cortactin at all time points analyzed (5, 10, 20 and 30 minutes, and 1, 2 and 24 hours of treatment) compared to the control conditions (data not shown). By contrast, blocking Na_V1.5 had no effect on the phosphorylation of cofilin on S3 (Fig. 4E), which has been shown to be critical for binding to actin (Arber et al., 1998). This suggests that Na_V1.5 might regulate Src kinase and not LIM kinase. Indeed, we found that TTX induced a reduction in Src kinase activity (Fig. 4F). As a result, a reduction of both invadopodia formation and activity was expected upon inhibition Na_V1.5. The number of pixels corresponding to the colocalisation of F-actin condensation areas and focal spots of DQ-gelatin proteolysis were quantified per cell, giving a ‘matrix-focalised-degradation index’ (Fig. 4G; supplementary material Fig. S4). Inhibiting Na_V1.5 with TTX, and NHE-1 with EIPA, resulted in 89% and 93% inhibition of the matrix focal degradation, respectively. Combining TTX and EIPA reduced matrix degradation by 96% (Fig. 4G).

Voltage-gated Na⁺ channels (Na_V) have long been considered as being characteristic of excitable cells (Catterall, 2012). However, different Na_V isoforms have been found in non-excitable cancer cells and their function enhances cancer cell invasiveness (Roger et al., 2006). In highly invasive breast cancer cells and in high-grade breast cancer biopsies, the overexpression of the Na_V1.5 isoform has been associated with ECM remodelling and an increased probability of developing metastases (Gillet et al., 2009; Yang et al., 2012). Indeed, Na_V1.5 promotes ECM proteolysis through NHE-1-dependent acidic activation of the extracellular cathepsins B and S (Brisson et al., 2011; Gillet et al., 2009). NHE-1 is already known to have a substantial role in extracellular acidification and in the invasiveness of cancer cells (Bourguignon et al., 2004; Busco et al., 2010; Cardone et al., 2005; Stock and Schwab, 2009), and any signalling pathway increasing its activity could make cancer cells become more aggressive. In this study, we showed for the first time that Na_V1.5 interacts with and allosterically increases NHE-1 activity in a range of pHi between 6.4 and 7.0. We propose that these interactions occur in caveolae of the invadopodial compartment and are responsible for an increased ECM degradative activity. Indeed, caveolae are membrane domains that ‘build’ protein–protein signalling complexes (Harvey and Calaghan, 2012) that are required for both invadopodia formation and ECM degradative activity (Caldieri et al., 2012; Caldieri et al., 2009; Yamaguchi et al., 2009). As such, caveolae could be a signalling platform in invadopodia for Na_V- and NHE-1-dependent cancer cell invasiveness. Because of their invaginated structure, caveolae also provide a space with restricted diffusion for ionic fluxes that could be responsible for local changes in membrane potential and signalling cascades. We also found that Na_V1.5 function in cancer cells enhanced Src kinase activity and Y421 cortactin phosphorylation, which could contribute to the acquisition of an aggressive phenotype. Although the mechanism has not yet been elucidated, this could be in line with the observation that membrane depolarisation regulates the actin polymerisation:actin depolymerisation ratio and thus, the F-actin network directly under the plasmalemma (Callies et al., 2011). This effect on actin polymerisation could also be dependent on the interaction of actin with NHE-1 through the association with actin-binding proteins of the ERM (ezrin, radixin and moesin) family (Baumgartner et al., 2004; Cardone et al., 2005).

Na_V channels have also been identified in other invasive cells, such as macrophages (Carrithers et al., 2007) and microglial cells (Black and Waxman, 2012), and blocking their activity has been shown to reduce cell invasiveness. It has also been proposed that this effect depends on the regulation of formation of podosomes (Carrithers et al., 2009).

In conclusion, we have shown that Na_V1.5 channels promote the degradative invadopodial activity in breast cancer cells, both by modifying the F-actin cytoskeleton and enhancing the activity of NHE-1. The participation of Na_V channels in the constitution and activity of ECM degrading compartments could be a general feature shared by invasive cancer and non-cancer cells.

Tetrodotoxin was purchased from Latoxan (France). Fluorescent probes were purchased from Tebu-Bio (France). Other drugs and chemicals were purchased from Sigma-Aldrich (France). Secondary horseradish peroxidase (HRP)-conjugated antibodies were from Santa Cruz Biotechnology.

MDA-MB-231 human breast cancer cells were cultured in DMEM supplemented with 5% fetal calf serum (FCS), and grown at 37°C in a humidified 5% CO₂ incubator.

We constructed a lentiviral vector encoding a short hairpin RNA (shRNA) specifically targeting human SCN5A transcripts using the same protocol as previously described (Jelassi et al., 2011). Briefly, the sequence encoding shNa_V1.5 was obtained by PCR elongation of two partially complementary primers: 5′-GGATCCCCAAGGCACAAGTGCGTGCGCAATTCAAGAGA-3′ and 5′-AAGCTTAAAAAAAGGCACAAGTGCGTGCGCAATCTCTTGAA-3′. We also constructed a lentiviral vector expressing an untargeted shRNA (pLenti-shCTL), by using the following primers: 5′-GGATCCCCGCCGACCAATTCACGGCCGTTCAAGAGACG-3′ and 5′-AAGCTTAAAAAGCCGACCAATTCACGGCCGTCTCTTGAACG-3′. These experiments resulted in shNa_V1.5 and shCTL cell lines. Cell viability was assessed as previously described (Brisson et al., 2011; Gillet et al., 2009).

Patch pipettes were pulled from borosilicate glass to a resistance of 4–6 MΩ. Na⁺ currents were recorded, in whole-cell configuration, under voltage-clamp mode using an Axopatch 200B amplifier (Axon Instrument, USA) as previously described (Gillet et al., 2009). The external Na⁺ solution was (in mM): NaCl 140, KCl 4, MgCl₂ 1, CaCl₂ 2, D-glucose 11.1, HEPES 10, adjusted to pH 7.4 with NaOH (1M). The external Li⁺ solution used LiCl 140 mM to replace NaCl. Intrapipette solution was (in mM): K-Glutamate 125, KCl 20, CaCl₂ 0.37, MgCl₂ 1, Mg-ATP 1, EGTA 1, HEPES 10, adjusted to pH 7.2.

Cell invasiveness was analysed as previously described (Brisson et al., 2011) using culture inserts with an 8-µm pore size filter covered with Matrigel™ (Becton Dickinson, France).

Cells seeded on 24-well plates were acidified in Li⁺- and Na⁺-free solutions containing 2.5 µM nigericin and 140 mM KCl, and were calibrated at various pH values ranging from 5.2 to 7.2 as previously described (Lacroix et al., 2004; Milosavljevic et al., 2010). Measurements were performed after a 3-minute incubation in the uptake medium supplemented with 3 mM LiCl, followed by four rapid rinses in ice-cold PBS. Cells were solubilised in 0.25M nitric acid, and Li⁺ was measured using atomic absorption spectrometry (Zeeman furnace system, Thermo Scientific). The initial rates of NHE-1 were calculated as the EIPA (10 µM)-sensitive Li⁺ accumulation per well divided by the protein quantity.

Invadopodia, embedded into a thick layer of gelatin, were isolated and separated from cytosolic and membrane fractions as previously described (Busco et al., 2010). Primary antibodies used were: mouse anti-NHE1 (Chemicon), mouse anti-HSC70 (Santa-Cruz Biotechnology), rabbit anti-caveolin-1 (Cell Signaling), goat anti-β-adaptin (BD Biosciences), rabbit anti-human Na_V1.5 (Sigma-Aldrich), rabbit anti-focal adhesion kinase (FAK, Santa-Cruz Biotechnology) and rabbit anti-human cathepsin B (Fitzgerald).

Cells were washed with PBS and lysed in presence of a lysis buffer (50 mM Tris-HCl, pH 7, 100 mM NaCl, 5 mM MgCl₂, 10% glycerol, 1 mM EDTA), containing 1% Triton-X-100 and protease inhibitors. Coimmunoprecipitations (co-IP) experiments were performed using Bio-Adembeads PAG magnetic beads (Ademtech, France) according to standard protocols. Two types of lysate–antibody mixes were prepared and incubated overnight at 4°C. Each mix contained 500 µg total cell lysate in 500 µl lysis buffer and 1 µg antibody (mouse anti-NHE1 or mouse control IgG). Cortactin and cofilin were immunodetected using mouse anti-cortactin, rabbit anti-pY421 cortactin, rabbit anti-cofilin and rabbit anti-pS3 cofilin antibodies, respectively (all from Millipore).

Cells were cultured on Matrigel™ (4 mg/ml)-coated coverslips for 24 hours then washed in PBS, fixed with 3.7% ice-cold paraformaldehyde in PBS. For confocal imaging, cells were permeabilised with 0.1% Triton X-100 and saturated with 0.1% gelatin in PBS. For epifluorescence and SIM imaging, they were permeabilised with a solution containing 50 mM NH₄Cl, 1% BSA and 0.02% saponin, and saturated in 3% BSA and 3% normal goat serum (NGS). Coverslips were incubated with primary antibodies then washed with 0.1% BSA in PBS and incubated with secondary fluorescent-conjugated antibodies. F-actin was stained with phalloidin–Alexa594. Primary antibodies used were anti-human Na_V1.5 (Sigma-Aldrich), anti-NHE1 (Abcam 4E9, or Santa-Cruz 4E9), anti-caveolin 1 (Cell Signaling). Fluorescent-conjugated antibodies used were anti-rabbit-Ig–TexasRed, anti-rabbit-Ig–AlexaFluor488, anti-goat-Ig–DyLight488, anti-rabbit-Ig–AlexaFluor350, anti-mouse-Ig–AlexaFluor568 and anti-mouse-Ig–AlexaFluor488. MagicRed substrate [MR-(RR)₂,, 1:1000, Immunochemistry Technologies] was used to assess cathepsin B activity. In some experiments, cells were transfected with LifeAct-TagGFP2 plasmids (Ibidi) allowing the visualisation of F-actin dynamics in living cells.

Gelatinolytic activity was assessed by culturing cells for 24 hours on a planar Matrigel™ matrix containing 25 µg/ml DQ-Gelatin®. Epifluorescence microscopy was performed with a Nikon TI-S microscope and analysed using both NIS-BR software (Nikon, France) and ImageJ^© software 1.38I (http://rsbweb.nih.gov/ij). A circularity index was calculated as ⁠. A value approaching 0 indicates an increasingly elongated shape, whereas a value of 1.0 indicates a perfect circle. Pearson's coefficients were calculated with the ImageJ colocalisation plugin.

SIM experiments were performed at the Nikon Imaging Center at the Institut Curie-CNRS (Paris), using a Nikon Eclipse Ti inverted microscope equipped with a MCL Piezo stage, two lasers (488 nm, 100 mW and 561 nm 100 mW), a 100× CFI Plan Apo TIRF (NA 1.49) oil immersion objective and an EM-CCD Camera iXon DU897 (Andor).

Cells were seeded on a layer of Matrigel™ (4 mg/ml) containing a substrate for digestion localisation (BSA-BODIPY-FL, 30 µg/mL) and observed at 600× magnification using a oil immersion objective with a Leica TCS SP5 II AOBS laser scanning confocal microscope equipped with He/Ne-633 and Argon-488 lasers with 495–519 (B2-A) and 642–660 (Cy5) nm excitation filters. Proximity ligation assays were performed according to standard protocols using the Duolink-‘In-cell Co-IP’ kit (OLink Biosciences) (Söderberg et al., 2006).

Src kinase activity was assessed using the ProFluor® Src family kinase assay (Promega, France) according to the manufacturer's instructions. shCTL MDA-MB-231 cells were treated for 10 minutes with 30 µM TTX or with a control solution (CTL, vehicle), then were lysed. Src kinase activities were assessed from total cell lysate samples containing 3 µg total proteins.

Data are displayed as mean±s.e.m. (n = sample size). Student's t-tests were used to compare groups showing Gaussian distributions and homogenous variances, alternatively a Mann–Whitney rank sum test was performed. Statistical significance is indicated as: *P<0.05; **P<0.01 and ***P<0.001.

We thank I. Domingo and C. Le Roy for technical and administrative assistance, J.-C. Pagès, C. Collin, M.-L. Jourdan and V. Joulin for the development of shRNA and the Nikon Imaging Center at the Institute Curie-CNRS Paris.