Rock inhibition promotes NaV1.5 sodium channel-dependent SW620 colon cancer cell invasiveness

The acquisition of invasive capacities by carcinoma cells, i.e. their ability to migrate through and to remodel extracellular matrices, is a determinant process leading to their dissemination and to the development of metastases. these cancer cell properties have often been associated with an increased Rho-ROCK signalling, and ROCK inhibitors have been proposed for anticancer therapies. In this study we used the selective ROCK inhibitor, Y-27632, to address the participation of the Rho-ROCK signalling pathway in the invasive properties of SW620 human colon cancer cells. Contrarily to initial assumptions, Y-27632 induced the acquisition of a pro-migratory cell phenotype and increased cancer cell invasiveness in both 3- and 2-dimensions assays. This effect was also obtained using the other ROCK inhibitor Fasudil as well as with knocking down the expression of ROCK-1 or ROCK-2, but was prevented by the inhibition of NaV1.5 voltage-gated sodium channel activity. Indeed, ROCK inhibition enhanced the activity of the pro-invasive NaV1.5 channel through a pathway that was independent of gene expression regulation. In conclusions, our evidence identifies voltage-gated sodium channels as new targets of the ROCK signalling pathway, as well as responsible for possible deleterious effects of the use of ROCK inhibitors in the treatment of cancers.

www.nature.com/scientificreports/ which promote the formation of actin stress fibres, associate the cytoskeleton to the plasma membrane and generate contractile forces 13 . The expression of ROCK has been shown to be increased in several cancers, to correlate with a bad prognosis, and their activity has been demonstrated to substantially contribute to cancer progression [14][15][16] . Furthermore, several somatic mutations in genes encoding ROCK-1 or ROCK-2, leading to a gain-of-function, have been identified in several cancers and especially in colorectal cancer 17,18 . As such, ROCK are generally perceived as key players in cancer development and progression 8,19 , which led to consider the use of ROCK inhibitors in the treatment of cancers 7,20,21 . Indeed, the use of ROCK inhibitors such as Y-27632 or Fasudil (HA-1077) decreased the migration and invasion capacities of several cancer cell types 22,23 , among which colon cancer cells 19,24 . However, several other studies reported a pro-cancerous effect of ROCK inhibition by promoting the growth and migration of some cancer cells 25,26 , such as a gain in colon cancer cell proliferation and invasion [27][28][29] . Signalling pathways involved in such deleterious effects are still elusive, and certainly need to be better understood.
In colorectal cancer, voltage-gated sodium channels (Na V ), and notably the Na V 1.5 isoform, have been demonstrated to be abnormally upregulated and to participate to carcinoma cell invasiveness [30][31][32] . The functional link between Na V and ROCK signalling in colorectal cancer has not been investigated so far.
In this study, we explored the effect of ROCK inhibition, on the growth and invasiveness of SW620 human colon cancer cells, in both 2-and 3-dimensions models. We show that both the pharmacological and molecular inhibition of ROCK-1 and ROCK-2 induce SW620 cancer cell invasiveness, by promoting the activity of the pro-invasive voltage-gated sodium channel Na V 1.5 activity.

ROCK inhibitor Y-27632 promotes the acquisition of a migratory phenotype and invasive capacities in SW620 human colon cancer cells.
To characterize the participation of the ROCKdependent signalling pathway in the invasive capacity of human colon cancer cells, SW620 cells were treated with the potent ROCK-1 and ROCK-2 inhibitor Y-27632 at a concentration of 10 µM. This concentration is conventionally used in both in vitro and in vivo experiments to inhibit ROCK activity 33,34 . We first assessed the effect of such a treatment over time on colon cancer cell morphology, and identified that it was responsible for the acquisition of a more elongated shape, characterized by a significant reduction of the cell circularity index (Fig. 1A). While there might be some differences according to the cell type and mode of migration 35 , an elongated morphology of cancer cells is generally associated with a mesenchymal invasive phenotype 36 . Voltagegated sodium channels (Na V ) have been demonstrated to be critical inductors of carcinoma cell invasiveness 37,38 . In colon [30][31][32] as in breast cancer 36,[39][40][41][42][43] , the Na V 1.5 isoform has been identified to be abnormally upregulated and associated with invasive and metastatic potencies, and to control cell morphology. Its inhibition leads to an increased circularity in breast cancer cells 36,40 , while its expression and activity promotes the acquisition of an elongated mesenchymal phenotype 43 . Therefore we also tested the effect of the Na V inhibitor tetrodotoxin (TTX) used at a concentration of 30 µM that blocks > 90% Na V 1.5 currents (Suppl. Figure 1), but identified no significant effect in the morphology of these cells harbouring a roundish morphology (Fig. 1A). Then, we compared the capacity of cancer cells to invade extracellular matrices in control conditions or in presence of Y-27632. For this purpose, SW620 spheroids were grown in a Matrigel-composed 3-dimensional matrix (Fig. 1B). Morphology and growth of spheroids were analysed over a total time duration of 96 h by time-lapse microscopy, along with the capacity of cells to disseminate from spheroids and to invade the matrix. We also tested the effect of Figure 1. The ROCK inhibitor Y-27632 enhances 3D colon cancer cell invasiveness. (A) Left, representative micrographs of SW620 human colon cancer cells in control condition (vehicle 0.1% DMSO, CTL) or treated for 24 h, 48 h or 72 h in presence of 30 µM TTX or 10 µM Y-27632. Scale bar, 30 µm. Right, a cell circularity index was calculated from micrographs taken in absence or presence of Y-27632 at the three different times (n = 100 cells for each condition). This was performed using the Fiji software after having manually delineated the shape of cells. ***Statistical difference at P < 0.001 (Mann-Whitney rank sum test) versus CTL condition of corresponding time. NS stands for not statistically different. (B) Representative phase contrast micrographs (× 10 amplification objective) taken from SW620 colon cancer cells grown as spheroids in a 3-dimension matrix composed of Matrigel™. Pictures of spheroids were taken at different incubation times (0 h, 24 h, 48 h, 72 h, 96 h ) in control condition (CTL, vehicle) or treated with Y-27632 (10 µM), NaV channel inhibitor tetrodotoxin (TTX, 30 µM), or both (Y-27632 + TTX). Scale bar, 100 µm. (C) A spheroid circularity index was calculated over time from time-lapse micrographs in the four experimental conditions indicated above, CTL (vehicle, black line), Y-27632 (10 µM, red line), TTX (30 µM, green line) and the combination Y-27632 + TTX (30 µM, blue line) (n = 12, 8 and 5 spheroids per condition, respectively). Only the TTX group showed as significant reduction of circularity compared to the CTL condition, starting at time 36 h (**P < 0.01, Mann-Whitney rank sum test). There was no other statistical difference between groups. (D) Spherical volumes were calculated over time from time-lapse micrographs in the three experimental conditions indicated in B, CTL (vehicle), Y-27632 (10 µM), TTX (30 µM) and Y-27632 + TTX (30 µM) (n = 12, 8 and 5 spheroids per condition, respectively). The co-treatment with Y-27632 and TTX significantly reduced the volume of spheroids as compared to the treatment with Y-27632 alone, or with TTX alone, at times indicated on the figure (*P < 0.05, Mann-Whitney rank sum test). There was no other statistical difference. (E) The surface of extracellular matrix (ECM) invasion by SW620 cancer cells, at distance from spheroids, was calculated over time in the four experimental conditions indicated in C. The treatment with Y-27632 significantly increased the invasion area as compared to the CTL condition ( # P < 0.001, Mann-Whitney rank sum test). The co-treatment with Y-27632 and TTX significantly reduced the surface of invasion as compared with the treatment with Y-27632 alone (**P < 0.01; ***PP< 0.001, Mann-Whitney rank sum test), but did not differ from the CTL condition. www.nature.com/scientificreports/ TTX alone, and of the combination of both Y-27632 and TTX. Y-27632 induced no significant change in spheroid morphology, assessed by calculating a circularity index (Fig. 1C), or growth ( Fig. 1D) as compared to the control condition. However, cancer cells treated with Y-27632 demonstrated a significant 3-time increase in the capacity to disseminate from the spheroid and to invade the extracellular matrix (ECM) (Fig. 1E). Importantly, the use of the Na V inhibitor tetrodotoxin (TTX) used at a concentration of 30 µM prevented the induction of ECM invasion mediated by Y-27632 (Fig. 1E), while having no or mild effect on spheroid circularity and growth ( Fig. 1C, D). At the concentrations used, neither Y-27632 nor TTX interfered with SW620 cell viability (Suppl. Figure 2A). These results showed that the ROCK inhibitor Y-27632 unexpectedly induces 3D invasion of SW620, at least partially in dependence on Na V activity. We then performed 2D-invasion assays using transwell inserts containing an 8-µm-pore-sized filter that we covered by either Matrigel (300 µg/mL) or collagen I (500 µg/mL). These contain characteristic components of ECM that are generally invaded by epithelial cancer cells during the metastatic process, Matrigel mimicking the basement membrane while collagen I mimics ECM of conjunctive tissues. In both cases, the inhibition of Na V activity using TTX reduced cancer cell invasiveness by about 40-50% (median factor of 0.51 on Matrigel, and of 0.58 on collagen I) compared to the control condition, while Y-27632 induced a significant increase of cancer cell invasiveness (by a median factor of 6.6 on Matrigel, and of 7.0 on collagen I). Again, the co-administration of TTX partially prevented (by approximately 45%) the effect of Y-27632 ( Fig. 2A). To decipher whether this effect could be due to non-specific effect of Y-27632, or rather due to the inhibition of ROCK, we tested the effect of another selective and potent ROCK inhibitor, Fasudil, on the 2D-invasion of Matrigel-coated inserts. Fasudil (20 µM) also induced an increase of SW620 invasiveness by a median factor of 2.1, albeit responsible for a reduction of cell viability (Suppl. Figure 1B). The effect of Fasudil on cell invasiveness was prevented by approximately 40% in presence of TTX (Fig. 2B). Then, we reduced the expression of either ROCK-1 or ROCK-2, using a mix of three specific sequences of silencing RNA for each target (siROCK-1 or siROCK-2) and compared to the transfection of a null-target siRNA (siCTL). These experiments resulted in a decrease of ROCK-1 or ROCK-2 mRNA expression, as assessed by RT-qPCR, by 67.3 ± 2.3% (n = 3) and by 50.1 ± 2.3 (n = 3), respectively. These reduced expressions mediated by siRNA were confirmed at the protein level by western blotting experiments (Fig. 2D) and we recorded a reduced protein expression compared to the siCTL condition of 70.7 ± 14.8% (n = 4) and 54.5 ± 18.4% (n = 4) for ROCK-1 and ROCK-2, respectively. In siCTL cells, TTX reduced the invasive capacity by about 50% (median 0.52), indicating the participation of Na V in basal invasiveness (Fig. 2C). Knocking down the expression of ROCK-1 increased the invasive capacity of SW620 cancer cells by a median factor of 1.42, and this effect was prevented by TTX. Knocking down the expression of ROCK-2 also increased the invasive capacity of SW620 cancer cells, by a median factor of 1.99. Again, this effect was prevented by the use of TTX. By contrast to these results, knocking down the expression of RhoA, by 75.7 ± 0.3% (n = 3) as assessed by qPCR, using a mix of three specific sequences of silencing RNA (siRhoA), had no effect on SW620 cancer cell invasiveness Figure 2. ROCK inhibitors enhance 2D colon cancer cell invasiveness dependently on Na V 1.5 channels. (A) Summary of SW620 colon cancer cell invasiveness studies performed either on Matrigel-coated (300 µg/mL) or Collagen I-coated (500 µg/mL) invasion inserts in control condition (vehicle, 0.1% DMSO), in presence of TTX (30 µM), Y-27632 (10 µM) or the combination Y-27632 + TTX. Results, from 7-8 independent experiments, are expressed relative to the CTL condition which appears as a dashed line. ***P < 0.001 compared to CTL condition (Mann-Whitney rank sum test); ## P < 0.01 compared to the condition performed in presence of TTX (Mann-Whitney rank sum test). (B) Summary of SW620 colon cancer cell invasiveness studies performed on Matrigel-coated (300 µg/mL) invasion inserts in control condition, in presence of TTX (30 µM), Fasudil (20 µM) or the combination Fasudil + TTX. Results, from 8 independent experiments, are expressed relative to the CTL condition (vehicle, 0.1% DMSO) which appears as a dashed line. ***P < 0.001 compared to CTL condition (Mann-Whitney rank sum test); ## P < 0.01 compared to the condition in presence of TTX (Mann-Whitney rank sum test). (C) Effect of silencing the expression of ROCK-1, or ROCK-2 or RhoA using specific siRNA (siROCK1, siROCK2 or siRhoA, respectively), compared to the transfection of irrelevant siRNA (siCTL), on SW620 colon cancer cell invasiveness. These experiments were performed in absence (CTL) or presence of TTX (30 µM). Results are expressed as ratios of mean results obtained with siCTL cells in CTL condition (vehicle). The dashed line indicates a ratio of 1. Results are from 12 to 15 independent experiments and were analysed using Mann-Whitney rank sum tests. ***P < 0.001 compared to CTL condition in siCTL cells; *P < 0.05 compared to CTL condition in siCTL cells; +++ P < 0.001 when comparing TTX to the CTL condition in siROCK1 cells or in siROCK2 cells. NS stands for not statistically different.  www.nature.com/scientificreports/ (Fig. 2C). Taken together, these results suggested that the promotion of SW620 cancer cell invasiveness was dependent on Na V activity and indeed due to the inhibition of both ROCK-1 and ROCK-2. Interestingly, both Y-27632 and Fasudil reduced MDA-MB-231 human breast cancer cell circularity (Suppl. Figure 2C), and Y-27632 increased the invasive capacity of these cells, in which Na V 1.5 has been demonstrated to promote pro-metastatic capacities 40,44 , by a median factor of 1.51 (Suppl. Figure 1D). To address the specific regulation of the Na V 1.5 channel, encoded by the SCN5A gene, which has been previously identified as an important enhancer of SW620 cancer cell invasiveness 30,31 , we developed two cell lines derived from SW620, one stably expressing a small hairpin RNA specific for targeting SCN5A gene expression (shNa V 1.5) and the other stably expressing a null-target small hairpin RNA (shCTL). As shown in Fig. 2E (top panel), a fast inward sodium current could be recorded in shCTL but not in shNa V .1.5 cells. These two cell lines were treated with Y-27632 (10 µM) or its vehicle (CTL) and cancer cell invasiveness through Matrigel-coated inserts was assessed. As anticipated, in CTL condition, shNa V 1.5 cells demonstrated a 65%-lower invasion capacity compared to shCTL cells. Furthermore, the Y-27632-mediated induction of invasion was 2.5-fold lower in shNa V 1.5 cells compared to shCTL cells (Fig. 2E, lower panel). The reduced expression level of Na V 1.5 proteins in shNa V 1.5 cells was also confirmed by western blotting (Fig. 2F).

RocK inhibitors increase na V 1.5 protein expression and activity in SW620 human colon cancer cells.
To further explore the possible regulation of SCN5A expression by the ROCK signalling pathway, we measured its transcription level, by RT-qPCR, over a time range from 4 to 24 h treatment, with either Y-27632 or Fasudil treatments. Results obtained indicated no significant regulation of SCN5A expression by ROCK inhibitors at the mRNA level, during this time-scale (Fig. 3A). However, an increased level of Na V 1.5 proteins was observed after 48 h treatment with Y-27632 ( Fig. 3B-E). This appeared to be statistically increased by a median factor of 1.28, as compared to the CTL (vehicle) condition when assessed by western blotting experiments (Fig. 3C), and a significant increase in the mean fluorescence intensity (MFI) value by 1.52 times was recorded found under Y-27632 treatment by flow cytometry in non-permeabilized cells (Fig. 3E). This increased level of Na V 1.5 proteins was also observed after 48 h treatment with Fasidul (Suppl. Figure 3A,B).
This increased protein level at the cell surface, yet difficult to observe in immunocytochemistry experiments (Suppl. Figure 3C), was responsible for an increased Na V activity at the plasma membrane of SW620 cells. Indeed, a 48 h-long treatment with Y-27632 was responsible for an increased amplitude of Na V -mediated transient inward currents (I Na ) (Fig. 4A). There was no shift in the I Na -voltage relationship with a threshold of activation between − 60 and − 55 mV and a reversal potential between + 55 and + 60 mV, but a significant increase (P < 0.05) of the maximal amplitude (obtained for a depolarizing step from − 100 to − 10 mV), from − 9.68 ± 3.78 (n = 23) to − 19.61 ± 4.00 (n = 33) pA/pF in CTL and Y-27632 conditions, respectively (Fig. 4B). There was no significant change in both activation-voltage and inactivation-voltage relationships (Fig. 4C). There was a significant increase (P < 0.05) in the peak I Na current density recorded from a holding potential of -100 to a depolarizing step of − 5 mV, from − 8.73 ± 2.75 (n = 22) to − 20.44 ± 4.29 (n = 33) pA/pF in Y-27632-treated cells (Fig. 4D). There was also a significant increase (P < 0.01) in the persistent I Na current density recorded at the end of this same 30 mslong depolarizing step, from − 0.86 ± 0.08 (n = 22) to − 1.42 ± 0.14 (n = 33) pA/pF (Fig. 4E), for CTL and Y-27632 conditions, respectively. These results therefore suggested that the inhibition of ROCK promotes Na V 1.5 activity, by increasing the stability of the channel at the plasma membrane of cancer cells.

Discussion
Voltage-gated sodium channels (Na V ) are membrane spanning heteromeric complexes, composed of one large pore-forming α subunit (9 isoforms, Na V 1.1-1.9) associated with one or two smaller auxiliary β subunits (4 transmembrane β1-4, and one soluble β1B, isoforms), traditionally considered as features of excitable cells, because of their well-characterized participation in the generation of action potentials. However, it has been recognized that these channels are also expressed, and fully functional, in several carcinoma cells 30,[44][45][46][47][48][49][50] , where they are not associated with cellular excitability but rather to dedifferentiation of epithelial cells 43 , invasive properties and metastatic potencies 37,38,51 . Several Na V isoforms, mostly Na V 1.5, Na V 1.6 and Na V 1.7 depending on the cancer type 52 , have been shown to be abnormally expressed, but the origin of this dysregulated expression as well as the reasons for the association with a specific cancer tissue have not been identified. Whether these channels are regulated by intracellular signalling pathways in cancer cells is still unclear, and has not been fully characterized.
The Na V 1.5 isoform, which is the product of the SCN5A gene, was found to be highly overexpressed at both mRNA and protein levels in colon and breast tumours, compared to normal tissues, and was correlated with cancer recurrence, metastases development and reduced patients survival 30,42,53,54 . In tumours, Na V 1.5 was expressed and functional at the plasma membrane of cancer cells, thus giving rise to sodium currents, but the isoform expressed was a neonatal splice variants 32,54 , and not the adult splice variant isoform. This splice variant only differs from the adult one by a few amino acid residues. In colon and breast cancers, the activity of Na V 1.5 results in a small but persistent entry of Na + at the basal membrane potential, that was demonstrated to promote extracellular matrix degradation, cancer cell invasiveness in vitro 30,31,36,39,44,55 , primary tumour growth and metastases development in animal models 40,41 . This paved the way for the development of new small inhibitors of Na V 56 or to the repurposing of clinically used inhibitors for anticancer treatments 32,40,41,57,58 . On the opposite, drugs promoting Na V activity in cancer cells, such as the alkaloid veratridine, importantly increase invasive behaviours in in vitro experiments 31,32,44 .
In this study, we show that the inhibition of ROCK, using conventional ROCK inhibitors at classical concentrations used both in vitro and in vivo, increases the invasive capacities of SW620 human colon cancer cells, and also those of MDA-MB-231 human breast cancer cells. While ROCK inhibitors are generally used to inhibit cell migration and invasion 7,8,19 , we are not the first to demonstrate pro-invasive effects of ROCK inhibition 25,29 . In  www.nature.com/scientificreports/ B16 melanoma cells, Y-27632 induced invasion via enhanced AKT and ERK signalling pathways 25 . In SW620 colon cancer cells, Y-27632 was also shown to increase invasiveness through 3D matrices composed of collagen I, but the mechanisms involved were not identified 29 . www.nature.com/scientificreports/ In this study, we show that the two well-known and commonly used pharmacological inhibitors of ROCK, Y-27632 and Fasudil, both promoted 2D and 3D-invasive capacities, through both Matrigel-or collagen-Icomposed extracellular matrices. Importantly, an increased invasion was also confirmed when silencing either ROCK-1 or ROCK-2, using specific siRNA, thus ruling out non-specific effects of the pharmacological compounds. Pro-invasive effects of ROCK inhibition were importantly abrogated by the concomitant inhibition of Na V 1.5 channel, identifying this channel as a potential target of the ROCK signalling. Indeed, the activity of Na V 1.5 channel, i.e. peak and persistent sodium currents, was demonstrated to be increased with no change in voltage-dependencies for the activation or for the inactivation. This suggests an increased quantity of Na V 1.5 proteins at the plasma membrane. Interestingly, the inhibition of ROCK did not interfere with SCN5A gene expression, and there was no effect at the mRNA level. However, an increased level of Na V 1.5 proteins under ROCK inhibition treatment could be identified, thus favouring the hypothesis of an increased stability and halflife of Na V 1.5 proteins, maybe by reducing their recycling.
While the inhibition of ROCK increased Na V 1.5 activity and related invasive capacity in SW620 colon cancer cells, it could appear surprising that the inhibition of RhoA expression induced no effect on cell invasion. Indeed, ROCK-1 and ROCK-2 are well-known effectors of activated RhoA, but could also be activated by RhoC 59 . Our results may also appear to be in contradiction with results obtained in MDA-MB-231 breast cancer cells, in which RhoA silencing reduced cell invasiveness by reducing the mRNA expression of SCN5A and therefore Na V 1.5-mediated sodium current, thus indicating that RhoA could exerts a tonic effect on the expression of Nav1.5 in these cancer cells 60 . Furthermore, in the same study, the authors identified the existence of a positive feedback of Nav1.5 channel expression on that of RhoA 60 . In our study, performed in a different cell type, we did not measure the effect of RhoA silencing on Na V 1.5 currents, but identified no significant effect on cell invasiveness. This could possibly be explained by a different stoichiometry in Rho GTPase or by different levels of basal activation.
Our results also appear in apparent contradiction with those obtained by Vishnubhotla and collaborators who reported that ROCK-2 was highly expressed in SW620 colon cancer cells, mainly distributed in invadopodiallike structures and that its knock-down reduced the depth of invasion into a 3D-scaffold of type I collagen 61 . However, our results are in line with those recently reported by Libanje and collaborators who demonstrated that colorectal cancer cells mainly harbour a collective mode of invasion and that ROCK-2 inhibition triggers the initial induction of leader cell formation and induces collective invasion from cysts 62 . Therefore, these apparently contrasting results should probably be studied with regard to the different growing modes and environmental conditions (2D vs 3D, type and stiffness of the extracellular matrix, etc.) and the different modes of invasion that could be induced in these situations.
Taken together, our results clearly demonstrated that the inhibition of ROCK could induce procancerous effects, and most specifically promote cancer cell invasion. These results also identify Na V 1.5 channels as potential targets of ROCK. Therefore, the use of ROCK inhibitors for anticancer purposes should be tightly controlled and probably restricted to cancers in which Na V channels have not been identified, in order to avoid adverse effects.

Methods
Chemicals, antibodies and pharmacological compounds. Tetrodotoxin was purchased from LATOXAN (France) and used at the final concentration of 30 µM prepared in PBS, thus blocking the activity of Na V 1.5 channels 44 . Y-27632 was purchased from SELLECKCHEM (France) and used at the final concentration of 10 µM (prepared in 0.1% DMSO), a conventional concentration used both in vitro and in vivo experiments 33,34 . Fasudil (HA-1077) was purchased from SIGMA-ALDRICH (France) and used at the final concentration of 20 µM prepared in a saline solution (PBS) as previously reported 63 . Fluorescent probes and conjugated antibodies were purchased from THERMOFISHER SCIENTIFIC (France). All other drugs and chemicals were purchased from SIGMA-ALDRICH (France).
Cell lines and culture. SW620 human colon and MDA-MB-231 human breast cancer cell lines were purchased from the American Type Culture Collection (LGC PROMOCHEM, France). Cancer cells were grown at 37 °C in a 5% CO 2 incubator, in a humidified atmosphere. MDA-MB-231 breast cancer cells were cultured in DMEM supplemented with 5% foetal calf serum (FCS). SW620 colon cancer cells were cultured in DMEM supplemented with 10% FCS. SW620 shCTL and shNa V 1.5 stable colon cancer cells, respectively expressing a null-target and a SCN5A expression product-targeting small hairpin RNA were generated as previously described 40,64 using the Giga Viral vectors Plateform (University of Liège, Belgium). Briefly, these two cell lines were obtained by transduction with a lentiviral vector encoding a short hairpin RNA (shRNA) specifically targeting human SCN5A transcripts (shNa V 1.5 cell line) or a null-target shRNA (shCTL cell line). The sequence encoding shSCN5A, inhibiting the expression of Na V 1.5 protein, was 5′-GCT GGA CTT TAG TGT GAT TAT CTC GAG ATA ATC ACA CTA AAG TCC AGC -3′. We constructed a lentiviral vector expressing a null-target shRNA (pLenti-shCTL), with the following sequence: 5′-CCT AAG GTT AAG TCG CCC TCG CTC GAG CGA GGG CGA CTT AAC CTT AGG -3′.
Tests assessing for mycoplasma contamination were performed once a week (LONZA, MycoAlert Mycoplasma Detection Kit). www.nature.com/scientificreports/ SCIENCES) were used. Samples were incubated at 37 °C for 60 min. Real time PCR experiments were performed as previously described 44 . Results obtained from cell lines are expressed as the relative gene expression using the comparative 2 -ΔΔCt method 65 with PPIA and HPRT1 as reference genes. Primers sequences can be found in Table 1. Analogue signals were filtered at 5 kHz, and sampled at 10 kHz using a 1440A Digidata converter. Cell capacitance and series resistance were electronically compensated by about 60%. The P/2 sub-pulse correction of cell leakage and capacitance was used to study Na + current (I Na ). Peak and persistent sodium currents were recorded by depolarizing the cells from a holding potential (HP) of − 100 mV to a test pulse of − 5 mV for 30 ms every 500 ms, and the amplitude of the persistent current was measured at the end on this 30 ms-long protocol. Sodium current-voltage (I Na -V) relationships were determined by stepwise depolarizing the membrane from a HP of − 100 mV to a maximal voltage of + 60 mV, with 5-mV increments, for 50 ms and at a frequency of 2 Hz. Inactivation-voltage relationships were obtained by applying 50 ms-long prepulses using the I Na -V curve procedure, followed by a depolarizing pulse to − 5 mV for 50 ms. Cell viability. SW620 cancer cells were seeded in a 24-well plate at the density of 15.10 3 cells per well. Media were changed every day, and after 5 days growing, the number of viable cells was assessed by the tetrazolium salt assay as previously described 44 and normalised to the appropriate control condition (vehicle, 0.1% DMSO). Acquisitions were taken at 570 nm using a spectrophotometer TECAN Nanoquant Infinite 200 Pro (France).

Transfection of small interfering
Three-dimensions (3D) invasion model. 3D-spheroids were generated from the SW620 human colon cancer cell line using 96-well round-bottom Ultra Low Attachment (ULA) plates (CORNING, New York, USA) inhibiting cell attachment. Briefly, 500 cells/ well were seeded in 200 µl DMEM + 10% FCS in ULA plates. Fortyeight hours later, when spheroid were formed and visible, Matrigel at a final concentration of 300 µg/mL was added to the culture medium. Four hours after the addition of Matrigel, treatments were performed, adding 50 μL of fresh culture medium with the indicated compounds (10 µM Y-27632 and/or 30 µM TTX) into each well. Plates were then centrifuged at 300×g for 3 min, and placed on the motorized stage of the NIKON TI-E microscope (NIKON, France) equipped with a time lapse system, in a controlled atmosphere chamber at 37 °C, 5% CO 2 and saturated with humidity. Images were taken every 30 min, for a total period of 96 h of acquisition at a 100× magnification, using the NIKON DS-Qi2 camera connected to the microscope. Spheroid growth and morphology were analysed using Fiji software (https ://image j.net/Fiji, National Institute of Health, USA). Several parameters were evaluated after having manually delineated the shape of spheroids: circularity index, spherical volume and matrix invasion area. Spherical volume (V) was estimated by measuring spheroid perimeter (P) and calculated with the formula V = P 3 /6π 2 . Matrix invasion area was evaluated by subtracting the invasion area at time t from the initial spheroid area at t = 0.
Two-dimensions cancer cell invasiveness. Cancer cell invasiveness was assessed using culture inserts with 8-µm pore size migration filters (BECTON DICKINSON, France), that we covered with a film of Matrigel (500 µg/mL) or collagen I (300 µg/mL). The upper chamber of the insert was then seeded with 1 × 10 5 cells in  TGC TGT CTT TGG GAC CTT  GTC TGC  129   SCN5A  Na V 1.5  TTC CTG GGG TCC TTC TAC CT TTT CCT TCT CCT CGG TCT CA 103   ROCK1  ROCK1  AGC GGT TGG AAC ACC TGA TT AAC CGA CCA CCA GTC ACA TT 94   RHOA  RhoA  CGC TTT TGG GTA CAT GGA GT GAG CAG CTC TCG TAG CCA TT  Western blotting experiments. Cells were washed twice with PBS and lysed in presence of a lysis buffer (50 mM Tris, pH 7, 100 mM NaCl, 5 mM MgCl 2 , 10% glycerol, 1 mM EDTA), containing 1% Triton-X-100 and protease inhibitors (SIGMA-ALDRICH, France). Cell lysates were cleared by centrifugation at 16,000×g for 10 min. Total protein concentrations were determined using the Pierce® BCA Protein Assay Kit (THERMO FISHER SCIENTIFIC, France). Protein sample buffer was added and the samples were boiled at 100 °C for 3 min. Total protein samples were electrophoretically separated by sodium dodecyl sulphate-polyacrylamide gel electrophoresis in 10% gels, and then transferred to polyvinylidene fluoride membranes (MILLIPORE, USA). Na V 1.5 proteins were detected using anti-human Na V 1.5 rabbit polyclonal primary antibodies (1/1000, Ref S0818, SIGMA-ALDRICH) and horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG secondary antibody at 1:2000 (TEBU-BIO, France). ROCK-1 and ROCK-2 proteins were detected using HRP-conjugated primary antibodies (SANTA CRUZ references G-6 sc-17794 and D-11 sc-398519, respectively) used at the working dilution of 1/1000. HSC70 protein was detected as a sample loading control using anti-HSC70 mouse primary antibody at 1:30,000 (TEBU-BIO) and HRP-conjugated anti-mouse-IgG secondary antibodies at 1:2000 (TEBU-BIO). In some other conditions β-actin was used as a sample loading control using anti-β-actin-HRP primary antibody at 1:1000 (C4, SANTA CRUZ ref sc-47778). Proteins were revealed using electrochemiluminescence-plus kit (PIERCE ECL Western Blotting Substrate, THERMO FISHER SCIENTIFIC, France) and captured on a PXi acquisitions system (SYNGENE, UK). Densitometry analysis of protein bands was performed using the Gel Tool from Fiji software (Scientific image analysis software available at https ://fiji.sc). All original blots are provided in Supplementary Figs. 4-7. Epifluorescence imaging. SW620 colon cancer cells were cultured for 24-48 h on glass coverslips, before receiving ROCK inhibitor (or vehicle treatment) for a duration of 48 h. Cells were then washed twice in PBS, before being fixed with 3.7% ice-cold paraformaldehyde prepared in PBS. Cell permeabilization was obtained using a solution containing 50 mM NH 4 Cl, 1% BSA and 0.02% saponin. Saturation of epitopes was achieved by incubating for 2 h with a solution containing 3% BSA and 3% Normal Goat Serum (NGS). Epifluorescence microscopy was performed with a NIKON TI-S microscope, and images were analysed using the NIS-BR software (NIKON, France). Fluorescent probes and conjugated antibodies were purchased from THERMO FISHER SCIENTIFIC (France).
Flow cytometry. SW620 cancer cells were seeded in T-25 flasks at the density of 1 × 10 6 cells per flask.
Twenty-four hours after cell seeding, treatments were performed, adding 0.1% DMSO (vehicle, control) or 10 µM Y-27632 into each flask. Cells were incubated for 48 h and detached using accutase solution. Cells were re-suspended in PBS supplemented with EDTA and 0.1% FBS. For each condition, 4 × 10 5 cells were incubated with 1.6-μg anti-Na V 1.5 antibody (ASC-005, ALOMONE LABS, Israel) for 30 min at 4 °C followed by incubation with secondary antibody Alexa-Fluor-488 (1:500, A-11008, THERMO FISHER SCIENTIFIC, France) for 45 min at 4 °C. Background controls were obtained after incubation with secondary antibody alone. All data were performed using a BD FACSMelody cell sorter (BECTON DICKINSON, San Jose, USA) and analyzed using FlowJo software (TREE STAR, USA).

Statistical analyses.
Statistical analyses were performed using SigmaStat 3.0 software (SYSTAT SOFT-WARE INC.). Normality of sample distribution was tested prior to conduct any comparison between groups. When normality failed, and/or equal variance test failed, non-parametric statistical tests were performed (Mann-Whitney rank sum test) and data were displayed as box plots indicating the first quartile, the median, and the fourth quartile, whiskers indicating the minimal and maximal values, and square dots indicating the means. When normality and equal variance were obtained, parametric tests (Student's t test) were performed. In these cases, results were presented as mean ± SEM. P values are indicated on figures. NS stands for "not statistically different".