The Voltage-Gated Sodium Channel Beta4 Subunit Maintains Epithelial Phenotype in Mammary Cells

The SCN4B gene, coding for the NaVβ4 subunit of voltage-gated sodium channels, was recently found to be expressed in normal epithelial cells and down-regulated in several cancers. However, its function in normal epithelial cells has not been characterized. In this study, we demonstrated that reducing NaVβ4 expression in MCF10A non-cancer mammary epithelial cells generated important morphological changes observed both in two-dimensional cultures and in three-dimensional cysts. Most notably, the loss of NaVβ4 induced a complete loss of epithelial organisation in cysts and increased proteolytic activity towards the extracellular matrix. Loss of epithelial morphology was associated with an increased degradation of β-catenin, reduced E-cadherin expression and induction of mesenchymal markers N-cadherin, vimentin, and α-SMA expression. Overall, our results suggest that Navβ4 may participate in the maintenance of the epithelial phenotype in mammary cells and that its downregulation might be a determining step in early carcinogenesis.


Introduction
Voltage-gated sodium channels β (Na V β) proteins, encoded by SCNxB genes, define a family of four transmembrane proteins with a short C-terminal intracellular domain and a large N-terminal immunoglobulin (Ig)-like extracellular domain [1]. These proteins were initially characterized as auxiliary subunits of sodium channels [2]. Indeed, they were isolated along with pore-forming voltage-gated sodium channel (Na V α) isoforms, which they interact with through covalent or non-covalent associations [3][4][5][6]. Hence, Na V β proteins regulate Na V α membrane trafficking as well as their biophysical [7] and pharmacological properties [8][9][10]. Along with these roles as sodium channel activity modulators, Na V β proteins have also been proven to conduct other specific cellular functions in excitable cells [11], such as an important role as cell adhesion molecules (CAMs), allowing for both trans-homophilic and trans-heterophilic cell-cell and cell-matrix adhesions in cells expressing Na V α isoforms [12]. Therefore, the expression and roles of Na V β are best known in excitable cells, such as neurons, in which they participate in the regulation of Na + influx but also control neurite outgrowth, axonal fasciculation and interaction with glial cells [13].
The most recently identified Na V β4 subunit, encoded by the SCN4B gene, is the least characterized member of the family. It was initially demonstrated to be expressed in the nervous system (dorsal root ganglia, spinal cord, and restricted areas or nuclei in the brain), in skeletal and cardiac muscle cells [14]. It shares sequence similarity with the Na V β2 subunit and engages in covalent interactions with Na V α in the extracellular Ig domain [6,14]. Na V β4 was demonstrated to control Na V α activity and particularly the generation of resurgent [15][16][17] or persistent [18,19] sodium currents in neurons. Na V β4 was also identified to participate in cell-cell adhesion [20] and neurite extension [21]. Evidence is shown that Na V β4 dysregulation is involved in epilepsy [22], in Rett syndrome [23], and mutations in the SCN4B gene have been linked to cardiac arrythmia [24][25][26][27] and sudden death syndromes [28].
Recently, we identified a critical role for Na V β4 in cancer progression [29]. Specifically, we have shown that Na V β4 was strongly expressed in normal epithelial cells and tissues from breast, colon, rectum, lung, and prostate but consistently downregulated in cancer samples, to be almost absent in high-grade primary and metastatic tumours [29]. In mammary cancer cells, reducing Na V β4 expression potentiated cell migration and invasiveness through the acquisition of a hybrid mesenchymal-amoeboid aggressive phenotype, which resulted in an increase in mammary tumour growth and a higher metastatic colonisation. This effect was independent of the Na V α channel activity [29]. Later, similar results were obtained in cervical cancer cells [30], and the preserved expression of SCN4B in papillary thyroid cancer was proposed to be a favourable indicator of a recurrence-free survival [31]. It was therefore suggested that the SCN4B gene might be considered as a metastasis suppressor gene [29]. Nevertheless, the function of Na V β4 in normal epithelial cells is not known. We therefore hypothesized that Na V β4 is important for epithelial phenotype. To test this, we explored the consequences of reducing Na V β4 expression in non-cancer MCF10A mammary cells. We showed that knocking-down Na V β4 induced a loss of epithelial phenotype that was reminiscent of early stages of carcinogenesis.

Material and Methods
Bioinformatic analyses-Gene expression data were obtained from The Cancer Genome Atlas (TCGA) and Genotype-Tissue Expression (GTEx) databases using the UCSC Xena Browser (https://xenabrowser.net accessed on 19 December 2020). The IlluminaHiSeq (log2-normalized_count+1) files were downloaded from the "TCGA Breast Cancer (TCGA-BRCA)" cohort in order to compare expressions between adjacent non-tumoral tissues and primary tumour. From the "TCGA TARGET GTEx" cohort, the RSEM norm_count (log2-normalized_count+1) files were downloaded in order to compare expressions between adjacent non-tumoral tissues, primary tumour and metastases.
Cells and cell culture-The human breast cancer cell line MDA-MB-231 and noncancer mammary epithelial MCF10A cells were acquired from the American Type Culture Collection (ATCC Manassas, Virginia), through LGC Standards (Molsheim, France), and were cultivated at 37 • C in a humidified 5% CO 2 incubator. MDA-MB-231 cells were cultivated in Dulbecco's modified Eagle's medium (DMEM) supplemented with 5% foetal calf serum (FCS). The immortalized non-cancer mammary epithelial MCF10A cells were cultured in DMEM/Ham's F-12, 1:1 mix containing 5% horse serum (Dutscher, Bernolsheim, France), 10 µg/mL insulin, 20 ng/mL epidermal growth factor, 0.5 µg/mL hydrocortisone, and 100 ng/mL cholera toxin. A stable MCF10A cell line knocked-down for the expression of the SCN4B gene, coding for Na V β4, was generated using the CRISPR/Cas9 technique, as previously described [32] by transfection with the SCN4B Double Nickase Plasmid (sc-411001, Santa Cruz, France). Transfection was performed using Lipofectamine 2000 (Invitrogen, Villebon sur Yvette, France). Selection of a stable cell line called "MCF10A Crβ4" was performed using 10 µg/mL puromycin and was compared with a control cell line, thereafter called "MCF10A CTL", which received all treatments similarly to MCF10A Crβ4 but was not transfected with the guide RNA sequence specific for excision of the SCN4B gene. Efficiency of the CRISPR-mediated knock-down was assessed by western blotting, and the stability of cells was followed for a minimal duration of 6 weeks. Mycoplasma contamination tests were performed routinely (Lonza, MycoAlert™ Mycoplasma Detection Kit, Levallois-Perret, France).
Small interfering RNA transfection-MCF10A mammary epithelial cells were transfected with siRNA directed against human SCN4B mRNA (siβ4) or scramble siRNA as a control (siCTL), both of which were purchased from ON-TARGETplus siRNA (Horizon Discovery, Cambridge, UK). Cells were transfected using Pepmute™ siRNA Transfection Reagent (SignaGen laboratories, Rockville, MD, USA). Experiments were performed 24-48 h after transfection and efficacy of silencing was assessed by western blotting.
Cell viability-Cell viability was evaluated by the tetrazolium salt assay (MTT) as already described [33]. Briefly, cells were seeded in different densities indicated in the figure and grew for 4 days at 37 • C and 5% CO 2 in their normal culture medium. Cell viability was measured after incubation with 0.5 mg/mL MTT for 60 min at 37 • C by measuring the absorbance at 540 nm.
Cell proliferation-Cells were seeded at a density of 2 × 10 5 in wells of a 6-well plate. Then, 72 h after seeding, medium was removed, and cells were incubated in a fresh culture medium containing 10 µM 5-ethynyl-2 deoxyuridine (EdU). Cells were then washed in phosphate-buffered saline (PBS), trypsinized, and the incorporation of EdU was monitored by flow cytometry (BD FACS Canto, Becton Dickinson, Le pont de Claix, France) using the Click-iT™ Plus EdU Alexa Fluor488 kit (Invitrogen, Villebon sur Yvette, France).
Cysts production and analyses-A 40 µL layer of Geltrex ® (Sigma-Aldrich, St. Quentin Fallavier, France) was added into the wells of Nunc™ Lab-Tek™ II Chamber Slide™ (Thermo Scientific, Illkirch, France), mimicking the extracellular matrix. The chamber slide was incubated at 37 • C for 20 min. MCF10A CTL or MCF10A Crβ4 cell suspensions were prepared in medium containing 25 µL/mL Geltrex ® and seeded at 10,000 cells per 400 µL in the chamber slide wells. The culture medium of cells was replaced every 2 days by 400 µL containing 25 µL/mL Geltrex ® . Pictures of cysts were taken in bright field before changing medium (Invitrogen EVOS M7000, Thermofisher, Illkirch, France) in order to monitor growth and circularity of cysts. After 3 weeks of culture, cysts were washed twice in PBS, then fixed in 4% paraformaldehyde (PFA, Invitrogen, Villebon sur Yvette, France) for 30 min at room temperature. Cysts were washed three times for 10 min in 100 mM glycine solution, then permeabilised for 5 min in a 0.5% Triton X-100 solution (Sigma-Aldrich, St. Quentin Fallavier, France) and washed three times for 10 min with PBS. Unspecific site blocking was be realized by incubating for 1h in 5% BSA solution. Primary antibodies in 150 µL of 5% BSA solution were added to the wells and incubated at 4 • C overnight. Wells were washed twice for 10 min in PBS, then 150 µL of fluorescent secondary antibodies were added for 1h at room temperature. Hoechst 33342 (1/1000, Invitrogen, Villebon sur Yvette, France) was used to visualize cell nuclei and Phalloidin-AF594 (Invitrogen, Villebon sur Yvette, France) to visualize the F-actin network. Wells were finally washed four times with PBS, and micrographs were obtained using confocal microscopy using a 20× objective (LEICA SP8 STED, Nanterre, France). Primary antibodies used were dedicated to identifying β-catenin (Cell Signalling Technology D10A8, 1/200, Ozyme, Saint-Cyr-L'Ecole, France), E-cadherin (Invitrogen 13-1700, 1/1000, Invitrogen, Villebon sur Yvette, France), Vimentin (Abcam ab92547, 1/100, Paris, France). Growth and circularity of cysts were analysed using the ImageJ software (https://imagej.nih.gov/ij/ accessed on 19 December 2020).
Data presentation and statistical analysis-Data are displayed as mean ± standard error of the mean (sem) when following a normal distribution, or as individual points centred by a diagram showing the median when not following a normal distribution.
One-way ANOVA followed by a Dunn's Multiple Comparison Tests, two-way ANOVA, Mann-Whitney rank sum tests, and paired Student t-tests were used to compare different conditions as indicated in the figure legends. Statistical significance is indicated as: *, p < 0.05; **, p < 0.01, and ***, p < 0.001, while "ns" stands for not statistically different.
Data availability-The authors declare that all other data supporting the findings of this study are available within the paper, and its supplementary information files or available from the authors upon request.

Results
Bioinformatic gene expression analyses, using The Cancer Genome Atlas (TCGA) and the UCSC Xena browser (https://xenabrowser.net accessed on 19 December 2020), confirmed initial studies [29] indicating that the SCN4B gene, coding for Na V β4, is significantly down-regulated in all breast cancer stages compared with adjacent non-tumoral breast tissues (Figure 1a). A lower expression of SCN4B was even identified in stage IIA compared with stage I (Figure 1a). In line with these initial results, Na V β4 protein expression was almost 10 times lower in the human mammary cancer MDA-MB-231 compared with the non-cancer MCF10A cells (Figure 1b). Importantly, this non-cancer, and non-tumorigenic, cell line was demonstrated not to express any functional pore-forming Na V subunit [34,35], while human breast cancer cells express the Na V 1.5 isoform which enhances invasive capacities [36,37]. In order to assess the potential role of Na V β4 in non-cancer mammary cells, we developed, using the CRISPR/Cas9 technique, a cell line derived from the parental MCF10A with a permanent knockdown of Na V β4, named MCF10A Crβ4, as well as a control cell line MCF10A CTL. The efficiency of the Na V β4 knock-down was verified by western blotting indicating a median reduction of protein expression of 85% (Figure 1c). Reduction of Na V β4 had important effects on the MCF10A cell morphology, as observed in conventional 2-dimensional cultures (Figure 1d). Particularly, MCF10A Crβ4 showed an elongated morphology characterised by an increased cell length as compared with MCF10A CTL (Figure 1e, 89.9 ± 5.1 vs. 33.1 ± 1.7 µm, respectively). Furthermore, while the MCF10A CTL cells showed a tendency for being closely associated in clusters, as epithelial cells do, the MCF10A Crβ4 cells were more scattered and established less intercellular interactions (Figure 1f). These morphological changes were also identified by epifluorescence imaging in fixed cells labelled to visualize cell nuclei and the F-actin network (Figure 1g). Image analysis demonstrated that MCF10A Crβ4 showed a greater cell surface than the MCF10A CTL cells (Figure 1h, 2318 ± 123 vs. 1062 ± 52 µm 2 , respectively). Interestingly, MCF10A Crβ4 also appeared to have a stronger intensity of F-actin labelling (Figure 1i), characterised by the presence of actin stress fibres (Figure 1g). These initial data suggested that the loss of Na V β4 expression affected epithelial cell phenotype and intercellular organization. Importantly, the loss of Na V β4 expression did not induce the expression of the SCN5A gene coding for Na V 1.5 (Supplementary Figure S1a). We next investigated how the loss of Na V β4 affected cell polarisation in three dimensions, using a cyst forming assay in a 3D extracellular matrix [38]. We found that, in 3D cultures, the MCF10A CTL cells formed regular cysts with a general spheroid morphology (Figure 2a), characterized by a circularity index (calculated from pictures) approaching 0.9, maintained during the 3 weeks of the experiment (Figure 2b). The MCF10A CTL cysts demonstrated a regular growth that was slightly slowed after 14 days of culture (Figure 2c). After this time, the number of remaining individualized cysts tended to decrease, mostly due to the dispersion of some cells into the extracellular matrix (Figure 2d). The MCF10A Crβ4 cysts also exhibited a generally spheroid shape, albeit demonstrating a lower circularity index (Figure 2a,b), but their growth was slower compared with the MCF10A CTL cysts (Figure 2c). This could be partly due to a reduced cell proliferation rate. Indeed, loss of Na V β4 expression slightly reduced MCF10A cell viability after 5 days of culture (Supplementary Figure S1b), which appeared to be due to a slower DNA synthesis (Supplementary Figure S1c,d). However, it also appeared that cells evaded from the MCF10A Crβ4 cysts to invade the 3D extracellular matrix (Figure 2a, arrows), thus resulting in a complete disorganisation of cysts and in a reduction of the number of individualized cysts with time (Figure 2a,d). We thus characterized the cyst ultrastructure. Cysts were fixed and stained in order to visualise cell nuclei, vimentin expression, and F-actin network by fluorescence imaging. The MCF10A CTL cysts demonstrated a typical and regular epithelial organisation with compacted cell nuclei uniformly distributed within the structure, expression of vimentin at the external periphery of the cyst (equivalent to the basal side of the epithelial layer), contiguous cells displaying a cortical submembrane F-actin network, and the appearance of a central lumen (Figure 2e). By contrast, the MCF10A Crβ4 cysts were completely disorganized, lacked a central lumen, and did not show any apico-basal polarisation (Figure 2e). Cell nuclei appeared to be less compacted, the vimentin expression was more intense and diffuse, and the cortical F-actin network was disrupted to be more diffuse inside the cytosol. Furthermore, multiple cells detached from the cyst to invade the extracellular matrix. Because this characteristic might require proteolytic activities, we explored the capacity of the MCF10A cysts to degrade the fluorogenic substrate DQ-Gelatin when incorporated into the extracellular matrix, after 7 days of growth (Figure 2f). The MCF10A CTL cysts demonstrated an almost undetectable proteolytic activity. By contrast, the MCF10A Crβ4 cysts showed a strong proteolytic activity towards the extracellular matrix (green fluorescence) and a strong invasion by cells originating from the cyst (Figure 2f,g). Because cell-cell junctions appeared to be disrupted in the MCF10A Crβ4 cysts, we then analysed the expression of adherens junction proteins β-catenin and E-cadherin. While the MCF10A CTL cysts were characterized by a strong colocalization of β-catenin and E-cadherin at the cell plasma membrane, the MCF10A Crβ4 cysts demonstrated a weaker and diffuse expression of both proteins with no colocalization (Figure 3a). These observations prompted us to investigate the expression of β-catenin depending on that of Na V β4. Consistent with the more mesenchymal phenotype, the mRNA level of gene coding for β-catenin, CTNNB1, was significantly higher in the MCF10A Crβ4 compared with MCF10A CTL cells (40% median increase, Figure 3b). In contrast, β-catenin protein level was down-regulated in the MCF10A Crβ4 compared with MCF10A CTL cells (52% median decrease, Figure 3c). A possible explanation for this apparent discrepancy could be an increased recycling of β-catenin in the MCF10A Crβ4 cells. Therefore, we incubated the MCF10A CTL and MCF10A Crβ4 cells with the proteasome inhibitor MG132 for different time durations and assessed levels of β-catenin (Figure 3d). These led to an increased immunodetection of β-catenin in both cell lines. Nevertheless, the increase in β-catenin protein expression was significantly higher in the MCF10A Crβ4 than in MCF10A CTL cells (Figure 3e). Similarly, the MCF10A cells transfected with specific silencing RNA targeting Na V β4 (siβ4, Figure 3f) also displayed a reduced expression of β-catenin compared with scramble siRNA (siCTL, Figure 3g), and this expression was restored upon treatment with MG132 ( Figure 3h). These results suggested that the loss of the plasma membrane Na V β4 led to the disruption of intercellular junctions and to a reduced half-life of β-catenin. Together with the loss of the epithelial morphology in individual cells, the disorganisation and the increased extracellular invasiveness of cysts, the disruption of adherens junctions pointed in favour of a transition towards a mesenchymal-like phenotype. Therefore, we next examined the expression of genes associated with either epithelial phenotype (CDH1, coding for E-cadherin) or with mesenchymal phenotype (CDH2, coding for N-cadherin; SNAI1, SNAI2, TWIST, ZEB1 coding for transcription factors promoting epithelial-tomesenchymal transition, VIM, coding for vimentin, and ACTA2, coding for α-SMA). The loss of Na V β4 expression was associated with a significant reduction in the epithelial marker CDH1 expression, and an increase in the expression of mesenchymal markers CDH2, SNAI2, TWIST, ZEB1, VIM, and ACTA2 (Figure 4a). Only the SNAI1 expression was not modified. These changes observed at the transcriptional level were confirmed at the protein level for E-cadherin, N-cadherin, vimentin, and α-SMA (Figure 4b). Finally, we explored the possibility to rescue the epithelial phenotype in MCF10A Crβ4 by overexpressing Na V β4 (Figure 4c). This partially restored the CDH2, VIM, and ACTA2 expression (Figure 4d). relatively to the initial number at day 1 (n = 50 and 83 MCF10A CTL and MCF10A Crβ4 cysts, respectively, from 6 independent experiments). Statistically different from MCF10A CTL: *, p < 0.05; **, p < 0.01 (ANOVA test). (e) The MCF10A CTL and MCF10A Crβ4 cysts were then fixed and stained with Hoechst 33342 to visualize cell nuclei (blue staining), as well as with phalloidin-594 to visualize the F-actin network (red staining). Cysts were also immunostained with primary rabbit anti-vimentin antibodies and secondary AF488-coupled anti-rabbit antibodies (green staining). The white asterisk indicates the presence of a lumen inside the MCF10A CTL cysts. Representative images from 6 independent experiments. Scale bar, 25 µm. (f) Proteolytic activity of the cysts was assessed by including DQ-gelatin, which emits green fluorescence when degraded, in the 3D matrix. Representative images from 8 independent experiments. Scale bar, 150 µm. (g) An index of DQ-gelatin degradation was calculated from images in (f), from 16 individualized MCF10A CTL and 21 MCF10A Crβ4 cysts at day 7. Statistically different from the MCF10A CTL cysts: **, p < 0.01 (Student's t-test).

Discussion
Pore-forming Na V α and auxiliary Na V β proteins of voltage-gated sodium channels were initially characterized in excitable cells in which they are responsible for triggering and propagation of action potentials. However, over the past years, it has been shown that both Na V α and Na V β subunits are dysregulated in cancers in which they have nonexcitable roles. Most of the time, their overexpression in carcinoma cells has been associated with cancer progression, and their activity was shown to promote pro-cancerous properties [39]. Pore-forming Na V 1.5, Na V 1.6, and Na V 1.7 appear to be specifically upregulated in cancer cells and their activity, through an inward sodium current, have been shown to promote invasive properties [40]. It has also been found that some Na V β subunits are upregulated in cancer and that they bear important roles in cancer cell biology and in cancer progression [39]. In contrast, the Na V β4 subunit has recently been identified as being expressed at high levels in normal epithelial tissues, but consistently downregulated in cancer samples. This was identified in breast, lung, prostate, colorectal [29], cervical [30] and papillary thyroid [31] cancer. In these studies, the preserved expression of the SCN4B gene was considered as a favourable biomarker of metastasis-free survival, suggesting that it could be a metastasis-suppressor gene.
The expression of the SCN4B gene was shown to be lower in highly invasive cancer as compared with weakly invasive or non-cancer cells [29,30,41]. Both in vitro and in vivo experiments, performed with different human cancer cell lines, demonstrated that reducing the Na V β4 expression potentiated cell migration, invasiveness, and tumour progression, while overexpressing it had opposite effects [29,30,42]. In breast cancer cells, the loss of Na V β4 promoted RhoA activity and the acquisition of a hybrid mesenchymal-amoeboid phenotype associated with highly invasive capacities [29]. However, the role of Na V β4 in normal epithelial cells is not known, and whether its loss might participate in events of early carcinogenesis has never been elucidated.
In this study, we confirmed that SCN4B/Na V β4 is downregulated in cancer compared with non-cancer cells and tissues. Furthermore, we demonstrated that its repression in non-cancer mammary epithelial cells induced dramatic morphological and functional changes associated with a loss of epithelial polarisation, disruption of epithelial junctions, overexpression of genes associated with a mesenchymal phenotype, and acquisition of pro-invasive capacities. Consequently, the loss of Na V β4 completely abrogated the establishment of 3D epithelial structures (cysts). The loss of Na V β4 favoured degradation of β-catenin, thus reducing its half-life and leading to the disruption of adherens junctions. All these effects were completely independent on the expression of the pore-forming Na V 1.5 channel, which is a marker of breast cancer aggressiveness [39]. It can be speculated that the Na V β4 subunit, expressed at the plasma membrane of epithelial cells, stabilizes β-catenin in proximity to E-cadherin. The loss of Na V β4 might lead to cytosolic release of β-catenin part of which could translocate to the nucleus and induce the expression of EMT-related genes, and a second part could undergo proteasomal degradation. These cellular events might be critical during carcinogenesis.
However, what could trigger the repression of SCN4B during cancer transformation is still unknown. Interestingly, recent studies pointed out the involvement of several miRNAs that are dysregulated in some cancers as important regulators of the SCN4B expression. In colorectal cancer, the increased expression of miR-424-5p in tumour samples was associated with poor prognosis [43]. In this study, the authors demonstrated that SCN4B was directly inhibited by miR-424-5p, thus promoting colon cancer cell proliferation, migration, and invasion [43]. Another miRNA, miR-3175, has recently been shown to be overexpressed in prostate cancer and to participate in cancer cell growth and invasion. Knocking down miR-3175 in prostate cancer cells increased the SCN4B and E-cadherin expression, inhibited the N-cadherin expression, and importantly reduced cell proliferation, migration, and invasion [44]. Together, these recent studies brought new elements on the involvement of miRNA in the promotion of cancer progression through the regulation of the SCN4B expression in tumours. As such, exploring the expression of these miRNAs during carcinogenesis might be of importance.
In conclusion, this study demonstrated the critical role played by Na V β4 in the maintenance of an epithelial phenotype in normal cells, and how its repression might contribute to cellular dysplasia and early carcinogenesis. In this context, maintaining its expression level in epithelial cells would be determinant in order to prevent, or delay, tumour transformation.