SWI/SNF chromatin-remodeling factor Smarcd3/Baf60c controls epithelial-mesenchymal transition by inducing Wnt5a signaling

Abstract

We previously identified a gene signature predicted to regulate the epithelial-mesenchymal transition (EMT) in both epithelial tissue stem cells and breast cancer cells. A phenotypic RNA interference (RNAi) screen identified the genes within this 140-gene signature that promoted the conversion of mesenchymal epithelial cell adhesion molecule-negative (EpCAM-) breast cancer cells to an epithelial EpCAM+/high phenotype. The screen identified 10 of the 140 genes whose individual knockdown was sufficient to promote EpCAM and E-cadherin expression. Among these 10 genes, RNAi silencing of the SWI/SNF chromatin-remodeling factor Smarcd3/Baf60c in EpCAM- breast cancer cells gave the most robust transition from the mesenchymal to epithelial phenotype. Conversely, expression of Smarcd3/Baf60c in immortalized human mammary epithelial cells induced an EMT. The mesenchymal-like phenotype promoted by Smarcd3/Baf60c expression resulted in gene expression changes in human mammary epithelial cells similar to that of claudin-low triple-negative breast cancer cells. These mammary epithelial cells expressing Smarcd3/Baf60c had upregulated Wnt5a expression. Inhibition of Wnt5a by either RNAi knockdown or blocking antibody reversed Smarcd3/Baf60c-induced EMT. Thus, Smarcd3/Baf60c epigenetically regulates EMT by activating WNT signaling pathways.

Fig 1

SUM149 and SUM229 breast cancer cells maintain epithelial and mesenchymal populations. (A and B) FACS analysis of SUM149 and SUM229 cells showing EpCAM⁻ and EpCAM^+/high populations. (C) Epithelial and mesenchymal properties of EpCAM^+/high and EpCAM⁻ populations shown by phase microscopy and immunostaining with nuclear DAPI (blue) stain and anti-E-cadherin (green) and anti-EpCAM (red) antibodies. Arrowheads indicate the EpCAM⁻ cells in the parental SUM229 and SUM149 cell lines. Bars, 100 μm. (D and E) Reduced EpCAM and E-cadherin protein expression with elevated fibronectin, N-cadherin, and vimentin protein in EpCAM⁻ cells. The presence and level of various proteins in the parental (P), EpCAM^+/high (+), and EpCAM⁻ (−) SUM149 (E) and SUM229 (F) breast cancer cells are shown. The data in panels A to E are representative of the data from at least two independent experiments. (F and G) Reduced EpCAM and E-cadherin mRNA expression and elevated N-cadherin, vimentin, and TGFβ expression in EpCAM⁻ cells than in EpCAM^+/high cells. (H and I) Elevated mRNA expression of EMT-inducing transcription factors Snail, Slug, and Twist in EpCAM⁻ cells compared to EpCAM^+/high cells measured by qRT-PCR. The values in panels F to I are means plus standard errors of the means (SEMs) (error bars) of three independent experiments performed in triplicate. (J) Increased invasiveness of mesenchymal EpCAM⁻ compared to epithelial EpCAM^+/high cells through growth factor-reduced Matrigel-coated Transwell chambers. Statistical significance was evaluated by an unpaired Student's t test and indicated as follows: ∗∗∗, P value < 0.001; ∗∗, P value < 0.01. Values are means plus SEMs of two independent experiments performed in triplicate.

Fig 2

RNAi screening strategy for EMT regulatory genes. (A) Schematic depicting the RNAi screening strategy. (B) Separation of EpCAM^+/high and EpCAM⁻ SUM149 cells by FACS shows less than 0.5% EpCAM^+/high contamination in the EpCAM⁻ cells. Data depict FACS analysis of three individual cell sorts used for the primary RNAi screen. Q2, quadrant 2. (C) Acquisition of the epithelial phenotype by mesenchymal EpCAM⁻ cells following siRNA-mediated knockdown of Snail, Slug, or both Snail/Slug genes as shown by immunostaining with nuclear DAPI (blue) stain, anti-E-cadherin (green), and anti-EpCAM (red) antibodies. Bar, 500 μm. (D) Elevated EpCAM and E-cadherin percent cellular fluorescence quantitated on a single-cell basis following siRNA-mediated knockdown of the genes indicated. Values are means plus SEMs (error bars) of 8 independent wells. (E) Elevated mRNA expression of epithelial markers EpCAM and E-cadherin following siRNA-mediated knockdown of Snail and Slug measured by qRT-PCR. Values are means plus SEMs of three independent experiments performed in triplicate. (F) Representative plate image from the RNAi screen shows expression of epithelial markers E-cadherin (green) and EpCAM (red) by immunofluorescence in EpCAM⁻ SUM149 cells following siRNA-mediated knockdown of target genes in duplicate wells shown in columns 2 to 11 in rows A to H; siGAPDH in columns 1 and 12, rows B to D and E to H, respectively; siSnail/Slug in columns 1 and 12, rows E to H and A to D, respectively; and siUBB in wells A1 and H12. Positive hits are indicated by white numerals as follows: 1, Fhl1; 2, EphA4; 3, Rnf130. Data are representative of duplicate RNAi screens for 140 genes performed in triplicate.

Fig 3

RNAi screen identifies genes important for EMT. (A and B) Elevated EpCAM and E-cadherin percent cellular fluorescence quantitated on a single-cell basis following siRNA-mediated knockdown of the genes indicated. Values are means plus SEMs of 6 independent wells from the primary screen. Statistical significance was evaluated by an unpaired Student's t test and indicated as follows: ∗∗∗, P value < 0.001; ∗∗, P value < 0.01. (C and D) Acquisition of the epithelial phenotype by mesenchymal EpCAM⁻ cells following siRNA-mediated knockdown of the genes indicated is shown by immunostaining of SUM149 (C) and SUM229 (D) EpCAM⁻ cells with nuclear DAPI (blue) stain, anti-E-cadherin (green), and anti-EpCAM (red) antibodies. Bars, 100 μm.

Fig 4

EMT regulatory genes affect cellular invasiveness and expression of EMT-inducing transcription factors. (A) Expression levels of EMT regulatory genes in EpCAM^+/high cells compared to EpCAM⁻ cells measured by qRT-PCR. N-cad, N-cadherin. (B) Deconvolution of siRNA smart pools demonstrates knockdown of EMT regulatory genes with at least two individual siRNA oligonucleotides per gene in SUM149 EpCAM⁻ cells as shown by qRT-PCR. (C) Cellular invasiveness through Matrigel-coated Transwell chambers of SUM149 EpCAM⁻ cells following siRNA-mediated knockdown of the indicated EMT regulatory genes. Statistical significance was evaluated by an unpaired Student's t test and indicated as follows: ∗∗, P value < 0.01; ∗, P value < 0.05. Values are means plus SEMs of two independent experiments performed in triplicate. (D and E) Expression changes of EMT-inducing transcription factors in EpCAM⁻ SUM149 cells following siRNA-mediated knockdown of the indicated EMT regulatory genes measured by qRT-PCR. The values in panels A, B, D, and E are means plus SEMs of at least three independent experiments performed in triplicate.

Fig 5

Smarcd3/Baf60c is an epigenetic regulator of EMT. (A) Western blot comparing levels of expression of Smarcd3/Baf60c in EpCAM⁻ SUM149 cells before (−) and after siRNA-mediated knockdown of Smarcd3 (siD3) and following the rescue of Smarcd3 expression. (B) Mesenchymal EpCAM⁻ cells do not acquire the epithelial phenotype following the rescue of Smarcd3 expression as shown by immunostaining with nuclear DAPI (blue) stain, anti-E-cadherin (green) and anti-EpCAM (red) antibodies. Representative images of two independent experiments performed in quadruplicate are shown. Bar, 100 μm. (C) Expression of Smarcd3/Baf60c in HMECs at levels comparable to those in EpCAM⁻ SUM149 cells. EV, empty vector; D3₁ and D3₂, two independent HMEC clones expressing Baf60c. (D) Smarcd3/Baf60c-expressing HMECs (D3-HMECs) gain a mesenchymal phenotype with loss of epithelial properties as shown by phase microscopy and immunostaining of EV- and D3-HMECs with nuclear DAPI (blue) stain, anti-E-cadherin (green), anti-EpCAM (red), or antivimentin (green) antibodies. Bars, 500 μm. (E) Loss of epithelial markers EpCAM and E-cadherin and gain of mesenchymal markers fibronectin and vimentin in D3-HMECs shown by Western blotting. (F) Elevated mRNA expression of EMT-inducing transcription factors Lef1 and Zeb2 and reduced expression of epithelial and cell adhesion markers in D3-HMECs measured by qRT-PCR. Values are means plus SEMs of three independent experiments performed in triplicate. (G) Different antigenic profiles in EV- versus D3-HMECs demonstrated by FACS analysis with epithelial differentiation markers anti-EpCAM (FITC) and anti-Cd49f (PE-Cy5) or cancer stem-like markers anti-Cd44 (APC) and anti-Cd24 (FITC). (H) Loss of epithelial cell polarity in D3-HMECs shown by confocal microscopy of EV- and D3-HMECs stained with nuclear DAPI (blue) stain and anti-ZO1 (green) antibody. Images are representative of at least two independent experiments. Bars, 50 μm.

Fig 6

Smarcd3/Baf60c expression induces a claudin-low gene signature. (A) Gene expression profiles of D3-HMECs cluster with the claudin-low gene signature. Heat map compares gene expression of D3-HMECs, Slug-HMECs, and Snail-HMECs to the 9-cell line claudin-low predictor. Upregulated genes (red) and downregulated genes (green) are indicated. (B and C) Venn diagrams depict shared genes between D3-, Slug-, and Snail-HMECs compared to the claudin-low predictor or compared to each other. (D) Claudin-low human tumors show the highest expression of D3-HMEC genes among breast cancer subtypes, as demonstrated by the mean expression of D3-, Slug-, and Snail-HMEC genes across the subtypes of breast cancer in the UNC337 data set. P values were calculated by comparing gene expression means across all subtypes using an analysis of variance (ANOVA) test. Each plus symbol represents a distinct tumor sample within the data set. (E) Differentiation scores compare D3-, Slug-, and Snail-HMECs relative to EV-HMECs showing the lowest differentiation propensity of D3-HMECs. The P value was calculated by comparing gene expression means across all breast epithelial cell lineages.

Fig 7

Phenotypic comparison of Snail-, Slug-, and Smarcd3-induced EMT. (A) Loss of the epithelial phenotype with gain of mesenchymal properties in D3-, Slug-, and/or Snail-HMECs as shown by phase microscopy and immunostaining with nuclear DAPI (blue) stain, anti-E-cadherin (green), and anti-EpCAM (red) antibodies. (B) Western blotting shows reduced protein expression for epithelial markers EpCAM or E-cadherin and elevated protein expression of mesenchymal markers fibronectin and vimentin in Snail-, Slug-, and D3-HMECs. (C) Elevated gene expression of mesenchymal markers N-cadherin, vimentin, and Smarcd3/Baf60c in Snail-, Slug-, and D3-HMECs measured by qRT-PCR. (D and E) Elevated gene expression of EMT-inducing transcription factors and reduced expression of epithelial cell-cell adhesion markers in Snail- and Slug-HMECs measured by qRT-PCR. Values are means plus SEMs of three independent experiments performed in triplicate. The images in panel A and data in panels B and C are representative of the results of at least two independent experiments.

Fig 8

SWI/SNF complex regulates noncanonical WNT signaling through induction of Wnt5a expression. (A) Elevated expression of Wnt5a and TGFβ in D3-HMECs shown by qRT-PCR. Values are means plus SEMs of three independent experiments performed in triplicate. (B) Activation of noncanonical WNT signaling with inhibition of canonical WNT signaling shown by Western blotting for elevated phosphorylation of PKCβ and reduced nuclear (nuc) but not whole-cell (wc) β-catenin protein expression. (C) Induced Brg1 protein expression with the expression of Smarcd3/Baf60c in HMECs shown by Western blotting with the antibodies indicated. (D) Brg1 coimmunoprecipitates with FLAG-Smarcd3/Baf60c in HMECs. IP, immunoprecipitation; α-FLAG, anti-FLAG antibody. The results shown in panels B to D are representative of at least two independent experiments. (E to G) Elevated Smarcd3/Baf60c, Brg1, and H3K4me3 on the promoters of the indicated genes in D3-HMECs compared to EV-HMECs as measured by ChIP–qRT-PCR. Values are means plus SEMs of three independent experiments performed in duplicate.

Fig 9

Inhibition of Wnt5a restores epithelial adherens junctions. (A) Wnt5a expression correlates with the mesenchymal phenotype of EpCAM⁻ cells. Elevated gene expression of Wnt5a in Slug-HMECs is similar to D3-HMECs measured by qRT-PCR. (B) Elevated gene expression of Wnt5a in EpCAM⁻ SUM149 and SUM229 cells compared to EpCAM^+/high SUM149 and SUM229 cells measured by qRT-PCR from two independent FACS sorts. Values in panels A and B are means plus SEMs of two independent experiments performed in triplicate. (C) siRNA-mediated knockdown of Wnt5a partially restores the epithelial phenotype of D3-HMECs as shown by phase microscopy. (D) siRNA-mediated knockdown of Wnt5a increases mRNA expression of epithelial and cell-cell adhesion markers and decreases expression of the mesenchymal marker vimentin and the EMT-inducing transcription factor Lef1. Data are measured by qRT-PCR, and values are means plus SEMs of three independent experiments performed in triplicate. (E) Western blotting showing increased protein expression of EpCAM and E-cadherin with loss of PKCβ phosphorylation in D3-HMECs transfected with siWnt5a. (F) Reduced invasiveness through growth factor-reduced Matrigel-coated Transwell chambers of D3-HMECs treated with siWnt5a or two different concentrations of the Wnt5a blocking antibody (α-Wnt5a) compared to untreated D3-HMECs. Statistical significance was evaluated by an unpaired Student's t test and indicated as follows: ∗∗∗, P value < 0.001; ∗∗, P value < 0.01. Values are means plus SEMs of two independent experiments performed in triplicate.