SCN4B acts as a metastasis-suppressor gene preventing hyperactivation of cell migration in breast cancer

Abstract

The development of metastases largely relies on the capacity of cancer cells to invade extracellular matrices (ECM) using two invasion modes termed 'mesenchymal' and 'amoeboid', with possible transitions between these modes. Here we show that the SCN4B gene, encoding for the β4 protein, initially characterized as an auxiliary subunit of voltage-gated sodium channels (Na_V) in excitable tissues, is expressed in normal epithelial cells and that reduced β4 protein levels in breast cancer biopsies correlate with high-grade primary and metastatic tumours. In cancer cells, reducing β4 expression increases RhoA activity, potentiates cell migration and invasiveness, primary tumour growth and metastatic spreading, by promoting the acquisition of an amoeboid-mesenchymal hybrid phenotype. This hyperactivated migration is independent of Na_V and is prevented by overexpression of the intracellular C-terminus of β4. Conversely, SCN4B overexpression reduces cancer cell invasiveness and tumour progression, indicating that SCN4B/β4 represents a metastasis-suppressor gene.

Figure 1. SCN4B/β4 protein is expressed in normal epithelial cells of human breast tissues and is downregulated in cancer cells.

(a,b) β4 protein (expression of the SCN4B gene) was analysed by immunohistochemistry on human breast tissue samples. (a) The expression of β4 protein was strong in epithelial cells of mammary acini (some examples are indicated by the black arrows), and not in non-epithelial cells of normal breast tissues. (b) In breast cancer tissue, the expression of β4 protein was strong in normal epithelial cells of mammary acini (black arrows), but significantly reduced in cancer cells (tumour area indicated by the red arrow, ‘T'). Scale bars, 50 μm.

Figure 2. SCN4B down regulation in human breast cancer tissues associates with poor prognosis.

(a–d) SCN4B/β4 protein expression was analysed by immunohistochemistry on breast tissue microarrays. Samples were stratified in ‘no staining', ‘weak staining' or ‘strong staining' groups. (a) Proportion of samples showing no (white), weak (gray) or strong (black) β4 staining in normal breast, compared with mammary hyperplasia/dysplasia and cancer (mixed grades) samples. The number of samples per condition is indicated in brackets. SCN4B/β4 protein staining was stronger in normal compared with cancer samples (χ², P<0.001), and in hyperplasia/dysplasia compared with cancer samples (χ², P<0.001). (b) SCN4B/β4 staining from indicated samples. Scale bars, 50 μm. (c) Proportion of samples showing no, weak or strong SCN4B/β4 staining in cancer samples, from grade I to III, and in lymph node metastases (LNM) samples. The number of samples per condition is indicated in brackets. β4 staining was stronger in grade I cancer samples compared with more advanced cancer samples (grades II, III and LNM) (χ², P<0.001). (d) Representative pictures of β4 staining from grade I, II and III ductal carcinoma samples. Scale bars, 50 μm. (e) Expression of the SCN4B gene in non-cancer (n=29) and in invasive breast carcinoma tissues (n=145) was analysed from The Cancer Genome Atlas (TCGA). RNA level is expressed as reads per kilobase per million (RPKM). Box plots indicate the first quartile, the median and the third quartile; whiskers indicate minimum and maximum values; squares show the means. SCN4B gene was significantly downregulated in cancer tissues (Mann–Whitney rank sum test, MW, P<0.001). (f,g) Prognostic analyses of gene expression in breast cancers, performed using the Breast Cancer Gene-Expression Miner. (f) Kaplan–Meier Any Event (AE)-free survival analyses, performed on data pooled from cohorts for the expression of SCN4B gene (n=1,024 patients). AE is defined as being metastatic relapse (MR) or patient death. A weak expression of SCN4B gene (≤ median of the pooled cohorts) was associated with a decrease in the AE-free survival (P=0.0005). (g) Kaplan–Meier MR-free survival analyses were performed for the expression of SCN4B (n=661 patients). A weak expression of SCN4B (≤ median of the pooled cohorts) was associated with a decrease in the MR-free survival (P=0.0013). Cox results are displayed on the graph.

Figure 3. Expression of the SCN4B/β4gene in human breast cancer cell lines and contribution to cancer cell invasiveness.

(a) The expression of SCN4B gene was studied by RT–qPCR in human mammary epithelial non-cancer MCF-10A and cancer MCF7, MDA-MB-468, MDA-MB-435s and MDA-MB-231 cell lines. Results are expressed as relative to that of HPRT-1 gene (n=7–12) and presented as mean values±s.e.m. *, significantly different from MCF-10A at P<0.05 (MW). (b) The expression of SCN4B/β4 protein was studied by densitometric analysis of western blot experiments in same cells as in a. Results are given as the amount of SCN4B/β4 protein relative to that of HSC70 (n=5) and presented as mean values±s.e.m. *, significantly different from MCF-10A at P<0.05 (MW). The image on top shows a representative western blotting experiment. (c) The expression of SCNxB genes was analysed in MDA-MB-231-Luc cells by reverse transcription–PCR. Plasmids encoding human SCNxB genes were used as positive controls for PCR primers. (d) Representative western blotting experiments showing protein expression for β1 (SCN1B), β2 (SCN2B) and β4 (SCN4B) in MDA-MB-231-Luc cells. (e) Cells were transfected with scrambled siRNA (siCTL) or with siRNA directed against the expression of the SCN1B gene (siSCN1B), the SCN2B gene (siSCN2B) or the SCN4B gene (siSCN4B). The efficacy of siRNA transfection was assessed by western blotting experiments 48 h after transfection. HSC70 was used as a control for sample loading. (f) Representative images of fixed and haematoxylin-stained MDA-MB-231-Luc cells on invasion inserts. Cancer cells were transfected with scrambled siCTL or with specific siRNA. Scale bars, 50 μm. (g) Summary of cancer cell invasiveness results (n=8) for MDA-MB-231-Luc cells transfected with siCTL or siSCNxB. Results were expressed relative to siCTL and presented as mean values±s.e.m. ***, statistically different from siCTL at P<0.001 (Student's t-test). (h) Representative image of a zebrafish embryo injected in the yolk sac with MDA-MB-231-Luc cells stained with CM-Dil and showing sites of colonization. Scale bars, 500 μm. Below is a magnification of the highlighted region containing human cancer cells (see arrows) colonizing organs of the embryo. (i) Zebrafish colonization index of siCTL or siSCN4B cells. Numbers in brackets indicate the number of embryos examined for each condition, from three different experiments. Results are presented as mean values±s.e.m. **, statistically different from siCTL at P<0.01 (Student's t-test).

Figure 4. Loss of SCN4B/β4 expression promotes human cancer cell invasiveness independently of the pore-forming Na_V subunit.

(a) Cancer cell invasiveness was assessed, using Matrigel-invasion chambers, from MDA-MB-231-Luc cells stably transfected with null-target shRNA (shCTL), SCN5A-targeting shRNA (shSCN5A) or SCN4B-targeting shRNA (shSCN4B), in the absence (−) or presence (+) of 30 μM TTX. The results from 8 to 16 independent experiments were expressed relative to control cells transfected with shCTL in the absence of TTX. ***, statistically different from shCTL at P<0.001 and #, statistically different from shSCN4B in the absence of TTX at P<0.05. (b) Cancer cell invasiveness was likewise assessed in shCTL or shSCN4B cells, transiently transfected with null-target siRNA (siCTL) or SCN5A-targeting siRNA (siSCN5A). The results from 12 independent experiments were expressed relative to shCTL cells transfected with siCTL. ***, statistically different from the shCTL/siCTL condition at P<0.001 and #, statistically different from shSCN4B/siCTL at P<0.05. (c) Cancer cell invasiveness was assessed in MDA-MB-231-Luc cells stably expressing the SCN5A-targeting shRNA (shSCN5A), not expressing the Na_V1.5 protein, and transiently transfected with null-target siRNA (siCTL) or SCN4B-targeting siRNA (siSCN4B). This effect was assessed in the absence (−) or presence (+) of two TTX concentrations (3 or 30 μM), or 30 nM of the Na_V1.8 inhibitor A803467. The results from six independent experiments were expressed relative to shSCN5A cells transfected with siCTL, in the absence of any Na_V inhibitor. NS stands for no statistical difference and *** denotes a statistical difference from shSCN5A/siCTL at P<0.001. (d) Cancer cell invasiveness was assessed using Matrigel-invasion chambers for MDA-MB-468 breast, H460 and A549 non-small-cell lung, and PC3 prostate cancer cells transfected with null-target siRNA (siCTL, black bar) or SCN4B-targeting siRNA (siSCN4B, red bars). Cancer cell lines known to express or not functional Na_V channels are indicated as Na_V⁺ and Na_V⁻, respectively. The results from 3 to 12 independent experiments were presented and are expressed relative to the results obtained with the same cells transfected with siCTL. *, different from siCTL at P<0.05 and *** at P<0.001. Statistics presented in this figure were performed using ANOVA for multiple group comparison (a–c) or Student's t-test (d). All results presented in this figure are mean values±s.e.m.

Figure 5. Loss of SCN4B/β4 expression maintains Na_V1.5-mediated persistent current and dependent extracellular matrix degradation.

(a) Sodium current (I_Na)–voltage relationships in shCTL (black squares, n=18) and in shSCN4B (red circles, n=22) cells. There was a significant difference at P<0.001 between the two conditions in the voltage range between −40 and +40 mV. (b) Activation (filled circles)– and availability (filled squares )–voltage relationships in shCTL (black symbols) and shSCN4B (red symbols) cells. (c) I_Na peak and I_Na persistent currents obtained from shCTL (black trace) and shSCN4B (red trace) cells for a membrane depolarization from −100 to −30 mV. (d) Mean values±s.e.m. of I_Na persistent currents obtained for a membrane depolarization from −100 to −30 mV from 18 shCTL and 21 shSCN4B cells. NS, not statistically different. (e) Mean values±s.e.m. of I_Na persistent/I_Na peak currents ratios obtained from 18 shCTL and 21 shSCN4B cells. **, statistically different from shCTL at P<0.01. (f) Dose–response effect of TTX on the inhibition of I_Na peak elicited by a membrane depolarization from −100 to −5 mV in shCTL (black squares, n=8–12) and in shSCN4B (red circles, n=7–12) cells. Data were fitted with the Hill equation and IC₅₀ values were 2.02±0.10 and 2.24±0.11 μM for shCTL and shSCN4B cells, respectively. (g) Intracellular pH measurements using the BCECF-AM probe, in NH₄Cl-acidified shCTL (black trace) and shSCN4B (red trace) cells in the absence of NaCl. NaCl (130 mM) was added at the time indicated (arrow). (h) H⁺ efflux measurements after the addition of NaCl in conditions similar to g (n=20). Results are expressed as mean values±s.e.m. NS, no statistical difference. (i) MDA-MB-231 shCTL or shSCN4B cells were cultured on a Matrigel matrix containing DQ-Gelatin. A ‘Matrix-Focalized-degradation index' was calculated as being F-actin foci (red labelling, phalloidin-Alexa594) co-localized with focused proteolytic activities (green) (n=442 cells for shCTL and 448 cells and shSCN4B). Results are expressed as mean values±s.e.m. NS, no statistical difference. (j) Representative pictures showing matrix degradation areas (green spots) and F-actin foci (red spots). Merging points (coloc), which appear as white pixels, were counted. Numbers of white pixels per cell were normalized to the mean value obtained in shCTL cells. Statistics were performed using Student's t-test.

Figure 6. The loss of SCN4B/β4 expression promotes human cancer cell migration and invasiveness in two and three dimensions.

(a) Cancer cell invasiveness was assessed using Matrigel-invasion chambers from shCTL or shSCN4B MDA-MB-231 cells, in the absence (−) or presence of the protease inhibitors GM6001 (10 μM), leupeptin (200 μM) or E64 (100 μM). Results from three to seven independent experiments are presented and were expressed relative to shCTL cells in the absence of inhibitors. Results are expressed as mean values±s.e.m. *** denotes a statistical difference from the shCTL at P<0.001, and # indicates a statistical difference from shSCN4B at P<0.05 (ANOVA). (b) Cancer cell migration of shCTL and shSCN4B cells measured by time-lapse microscopy to track the movement of cells over 180 min, 1 frame per min (n=20 representative cells in each condition). Distances are indicated in μm. (c) The speed of migration (in μm min⁻¹) was analysed in shCTL and shSCN4B cells from time-lapse experiments and results shown were obtained from 106 and 96 cells, respectively. *** denotes a statistical difference from the shCTL at P<0.001 (MW). (d) The track length of cell migration (in μm) was analysed over 180 min in shCTL and shSCN4B cells from time-lapse experiments and results shown were obtained from 106 and 96 cells, respectively. *** denotes a statistical difference from the shCTL at P<0.001 (MW). (e) Three-dimension (3D) invasiveness of shCTL and shSCN4B cells, embedded inside Matrigel, was measured by time-lapse microscopy to track the movement of cells over 48 h (1 frame per 30 min) in the absence (CTL) or presence of the MMP inhibitor GM6001 (10 μM) (n=13 representative cells in each condition). Distances are indicated in μm. (f) The track length of 3D cell invasiveness (in μm) was analysed over 48 h in shCTL and shSCN4B cells from time-lapse experiments and results shown were obtained from 30 cells in each condition. ** and *** denote statistical difference from the shCTL, CTL condition at P<0.01 and P<0.001, respectively. ## denotes a statistical difference from the shSCN4B, CTL condition at P=0.002. § denotes a statistical difference from the shCTL, GM6001 condition at P=0.038 (Dunn's test). (c,d and f) Box plots indicate the first quartile, the median and the third quartile; whiskers indicate minimum and maximum values; squares show the means. Error bars encompass 95% of data samples.

Figure 7. The loss of SCN4B/β4 expression promotes RhoA-dependent amoeboid cell transition and migration.

(a) F-actin was stained with phalloidin-AlexaFluor594 in shCTL and shSCN4B cells and a cell circularity index was calculated (n=88 cells per condition). Results are expressed as mean values±s.e.m. ***P<0.001 from shCTL (Student's t-test). (b) Representative SEM micrographs of shCTL and shSCN4B cells. Scale bars, 10 μm. (c) Number of filopodia-like structures per cell, counted from SEM pictures in shCTL and shSCN4B cells (n=60 and 66 cells, respectively). ***P<0.001 from shCTL (MW). (d) Number of blebs per cell, counted from SEM micrographs in shCTL and shSCN4B cells (n=82 cells per condition). ***P<0.001 from shCTL (MW). (e) Representative confocal micrographs of shCTL and shSCN4B cells for which F-actin was stained with phalloidin-AlexaFluor488 (green) and nuclei with DAPI (blue), scale bar 20 μm. For enlargements images, scale bars are 5 μm. (f) Number of filopodia per cell, counted from confocal micrographs in shCTL and shSCN4B cells (n=46 and 66 cells, respectively). ***P<0.001 from shCTL (MW). (g) Length of filopodia, measured from confocal micrographs, in shCTL and shSCN4B cells (n=1,070 and 593 filopodia, respectively). ***, statistically different from shCTL at P<0.001 (MW). (h) SEM observations of shSCN4B cell invasion 24 h after cells were seeded on a layer of Matrigel (4 mg ml⁻¹). The coloured structure is the tip of the cell still observable above the Matrigel layer, while penetrating inside the matrix. Scale bar, 10 μm. (i) Western blots showing total and active GTP-bound forms of RhoA, Rac1 and Cdc42, pulled down by GST-RBD in shCTL and shSCN4B cells. (j) Quantification of GTP-bound RhoGTPases (active), normalized to total protein level, and expressed relatively to that of shCTL (n=5). **, statistically different from shCTL at P<0.01 and * at P<0.05 (MW). (k) shCTL or shSCN4B cancer cell invasiveness, in the absence (−) or presence of blebbistatin (50 μM) (n=3). Results are expressed as mean values±s.e.m. ***P<0.001 from shCTL, and ^###P<0.001 from shSCN4B (ANOVA). (l) Left panel, in situ proximity ligation assays showing a strong proximity between SCN4B/β4proteins and RhoA in shCTL cells (red dots, left panel) and the absence of any proximity signal in shSCN4B cells (right panel). Scale bars, 50 μm. (c,d,f,g,j) Box plots indicate the first quartile, the median and the third quartile; whiskers indicate minimum and maximum values; squares show the means. Error bars encompass 95% of data samples.

Figure 8. SCN4B/β4 protein overexpression inhibits cancer cell invasiveness.

(a) CTL, shSCN4B and oeSCN4B cancer cell invasiveness, in the absence (−) or presence of TTX (30 μM), expressed relative to oeCTL cells in the absence of TTX (n=6). Results are expressed as mean values±s.e.m. ***, different from CTL at P<0.001, ** at P<0.01. ^###, different from shSCN4B at P<0.001 (ANOVA). NS, no statistical difference. (b) Cancer cell invasiveness (n=6) from oeCTL and oeSCN4B cells, in the absence (−) or presence of TTX (30 μM), expressed relative to oeCTL cells in the absence of TTX. Results are expressed as mean values±s.e.m. **, different from oeCTL at P<0.01. NS, no statistical difference (ANOVA). (c) I_Na–voltage relationships in oeCTL (black squares, n=15) and oeSCN4B (green triangles, n=43) cells. There was a significant difference at P<0.05 between the two conditions in the voltage range between −45 and +45 mV. (d) Activation (filled circles)– and availability (filled squares)–voltage relationships obtained in the same oeCTL (black symbols) and oeSCN4B (green symbols) cells as in c. (e) Mean values±s.e.m. of I_Na persistent currents obtained for a membrane depolarization from −100 to −30 mV from 15 oeCTL and 43 oeSCN4B cells. *P<0.05 from oeCTL (Student's t-test). (f) Mean values±s.e.m. of I_Na persistent/I_Na peak current ratios in same conditions as in e. ***P=0.001 (Student's t-test). (g) oeCTL or oeSCN4B cells were cultured on a Matrigel-composed matrix containing DQ-Gelatin, and a ‘Matrix-Focalized-degradation index' was calculated (n=77 and 69 cells for oeCTL and oeSCN4B, respectively). ***, statistically different from oeCTL at P<0.001 (MW). (h) Cell circularity index was calculated from oeCTL and oeSCN4B cells (n=73 cells per condition). Results are expressed as mean values±s.e.m. ***, statistically different from oeCTL at P<0.001 (Student's t-test). (i) Speed of migration (μm min⁻¹) of oeCTL and oeSCN4B cells analysed from time-lapse experiments (n=47 per condition). ***, statistically different from oeCTL at P<0.001 (MW). (j) Western blots showing total and active GTP-bound forms of RhoA, Rac1, Cdc42, pulled down by GST- in oeCTL and oeSCN4B cells. (k) Quantification of GTP-bound RhoGTPases in oeSCN4B cells, normalized to its total protein level, and expressed relatively to that of oeCTL cells (n=4). *, statistically different from the oeCTL at P<0.05. NS, no statistical difference (MW). (g,i,k) Box plots indicate the first quartile, the median and the third quartile; whiskers indicate minimum and maximum values; squares show the means. Error bars encompass 95% of data samples.

Figure 9. SCN4B/β4 protein inhibits cancer cell invasiveness through its intracellular C-terminus but not through its extracellular Ig-like domain.

(a) Cartoon showing the transmembrane structure of the β4 protein, encoded by the SCN4B gene. The extracellular domain contains an Ig-like structure. After introduction of synonymous nucleotide substitutions in the SCN4B sequence, we have generated a sequence that is not recognized by the small hairpin RNA targeting SCN4B expression. This sequence has been inserted into a pSec expression vector in order to overexpress the full-length β4 protein (called ‘Full-length') in shSCN4B cells. Alternatively, we have also created truncated versions of the β4 protein: one containing a deletion of its intracellular C-terminus, from residue K185, and called ‘ΔC-ter', and one containing a deletion of its extracellular N-terminus up to residue T161, and called ‘ΔN-ter'. The nucleotide sequences were inserted into the pSec mammalian expression vector. (b) Cancer cell invasiveness was assessed using Matrigel-invasion chambers from shCTL and shSCN4B, transfected with an empty expression vector (pSec), or transfected with ‘ΔN-ter', ‘ΔC-ter' or ‘Full-length' encoding sequences. Results are expressed as mean values±s.e.m. ^###, statistically different from shSCN4B/pSec at P<0.001 and ^## at P<0.01. NS, not statistically different (ANOVA). (c) The speed of migration (in μm min⁻¹) was analysed from time-lapse experiments with shCTL and shSCN4B cells, transfected with an empty expression vector (pSec), or transfected with ‘ΔN-ter', ‘ΔC-ter' or ‘Full-length' encoding sequences, and results shown were obtained from 30 cells in each condition. ***, statistically different from shCTL at P<0.001. ^###, statistically different from shSCN4B/pSec at P<0.001. NS, not statistically different (Dunn's test). (d) Quantification of GTP-bound RhoA in shSCN4B cells, transfected with empty vector (pSec), with ‘ΔN-ter', ‘ΔC-ter' or ‘Full-length' encoding sequences. The activity of GTP-bound (active) RhoAGTPase was normalized to its total protein level, and was expressed relatively to that of shSCN4B/pSec cells (n=3). *, statistically different from shSCN4B/pSec at P<0.05 (MW). (c,d) Box plots indicate the first quartile, the median and the third quartile; whiskers indicate minimum and maximum values; squares show the means. Error bars encompass 95% of data samples.

Figure 10. SCN4B expression inversely correlates with primary tumour growth and metastatic development.

(a) Top cartoon, transendothelial migration experiment. Bottom box plot, quantification of the number of shCTL, shSCN4B or oeSCN4B cancer cells migrating through the endothelium (HUVEC monolayer) and the 8-μm pore-sized filter of the migration transwell, expressed as a relative number to shCTL (4 independent experiments). (b) Top cartoon, transendothelial invasion experiment. Bottom box plot, quantification of the number of shCTL, shSCN4B or oeSCN4B cancer cells migrating through the extracellular matrix (matrigel) coating the 8-μm pore-sized filter of the invasion transwell, then endothelium (HUVEC monolayer), expressed as a relative number to shCTL (three independent experiments). (c) Bioluminescent imaging (BLI) performed in NMRI nude mice tail vein-injected with MDA-MB-231-Luc cells that do not express (shSCN4B), or which overexpress, the SCN4B protein (oeSCN4B). Representative ex vivo lung BLI, after organ isolation, at completion of the study (ninth week after cell injection). (d) BLI quantification of excised lungs from mice injected with shSCN4B cells (n=7) and mice injected with oeSCN4B cells (n=8) (MW). (e) Mean±s.e.m. in vivo BLI value of tumours (expressed in c.p.m.) as a function of time recorded in the whole body of mice. * denotes a statistical difference from the shSCN4B group at P<0.05 (Student's t-test). (f) Representative bioluminescent images of mammary tumours in shSCN4B and oeSCN4B experimental groups (23rd week after cell implantation). (g) Immunohistochemical analyses of primary mammary tumours obtained from same mice as in Fig. 8e,f implanted with shSCN4B cells (top image) or with oeSCN4B cells (bottom image). Slides were counterstained with haematoxylin (blue labelling), and incubated with anti-mouse SCN4B/β4 antibodies and immunohistochemistry was performed using the streptavidin-biotin-peroxidase method with diaminobenzidine as the chromogen (brown labelling). Scale bars, 100 μm. (h) Immunohistochemical analyses of lungs obtained from the same mice as in Fig. 6e,f, implanted with human shSCN4B cells (top image) and oeSCN4B cancer cells (bottom image). Slides were counterstained with haematoxylin (blue labelling), and human breast cancer cells were identified using anti-human cytokeratin 7 immunohistochemical (brown) labelling. Scale bars, 100 μm. (a,b,d) Box plots indicate the first quartile, the median and the third quartile; whiskers indicate minimum and maximum values; squares show the means. Error bars encompass 95% of data samples.