ZEB1 Collaborates with ELK3 to Repress E-Cadherin Expression in Triple-Negative Breast Cancer Cells Free

Epithelial-to-mesenchymal transition (EMT) results in the transition of epithelial cells to a mesenchymal phenotype that is defined by the change in the expression of prototypical markers such as E-cadherin and vimentin. EMT was first described in embryonic development and has been demonstrated to play a crucial role in cancer cells by inducing aberrant motility, dissemination, and metastasis (1, 2).

EMT is executed by EMT-inducible transcription factors, mainly of the SNAIL, SLUG, TWIST, and ZEB families (3). Regarding EMT-inducible transcription factors, ZEB1 has been demonstrated to be a particularly potent factor that is associated with aggressive behavior, metastasis, and poor clinical prognosis in various tumor types, including pancreatic, lung, and breast cancer (4–6).

ZEB1 contains 2 zinc finger clusters that are responsible for binding to specific DNA sequences called E-box elements (7). In addition, ZEB1 contains several protein binding domains, including the CtBP interaction domain (CID), Smad interaction domain (SID), and p300-P/CAF binding domain (CBD; refs. 7–9). By recruiting corepressors or coactivators through these domains, ZEB1 can either suppress or activate the expression of its target genes (9, 10). For instance, ZEB1 functions as a transcriptional repressor of E-cadherin (10) by directly binding to the E-box located in the promoter of E-cadherin and recruiting the CtBP transcriptional cosuppressors (11) and/or the SWI/SNF chromatin-remodeling protein BRG1 (12). On the other hand, the direct binding of ZEB1 to YAP enhances the transcriptional expression of a subset of YAP target genes, converting ZEB1 from a transcription inhibitor into a transcription activator; this conversion is a strong predictor of poor clinical outcome in hormone receptor-negative breast cancer (13). Although there is accumulating evidence that ZEB1 is a pivotal factor in inducing EMT and metastasis of cancer in vitro and in vivo, a recent report revealed that ZEB1 knockout resulted in an incomplete inhibition of the metastatic ability of colorectal cancers, suggesting that there are additional factors that collaborate with ZEB1 to complete the phenotypic changes related to metastasis (14).

ELK3, a ternary complex factor belonging to the ETS family of transcription factors, is associated with various cellular phenomena such as wound healing, lymphangiogenesis, and vasculogenesis both in vitro and in vivo (15–17). In addition, the findings of multiple studies support the link between ELK3 and cancer metastasis (18–20). Suppression of ELK3 decreases migration and invasion ability of basal like breast cancer cells in vitro (21). In vivo studies show that a depletion of ELK3 leads to the formation of smaller tumors due to the inability of the tumor to become vascularized and oxygenated (17). In aggressive triple-negative (ER⁻, PR⁻, HER2⁻) breast cancer, ELK3 orchestrates metastasis through several distinct mechanisms, including the regulation of GATA3 expression, the production of VEGFC and the expression of cell-to-cell adhesion-related genes (18, 22, 23). The regulation of ELK3 activity is determined by a comprehensive regulatory signaling network, which is described as follows. First, having 2 transcriptional repressor domains at the N- and C-termini, ELK3 usually functions as a strong transcriptional repressor (24); however, ELK3 is transformed into a transcriptional activator when it is phosphorylated by Ras/ERK signaling (25). Second, ELK3 can be shuttled between the nucleus and cytoplasm depending on intracellular signaling, and nuclear ELK3 can be exported into the cytoplasm in response to stress-activated kinases, in particular JNK (26). For these reasons, despite the pleiotropic activity of ELK3 in both developmental and pathological progression, the underlying mechanism of ELK3 in regulating the expression of downstream target genes is poorly understood.

Recent studies suggest that ELK3 promotes liver fibrosis by regulating the EMT process (27). This result makes it feasible to speculate that the regulatory axis of ELK3 is in line with that of ZEB1 during the EMT process. If so, the linkage between ELK3 and ZEB1 activity could provide useful information to understand tumor metastasis as well as EMT.

Here, we specifically test the hypothesis that ELK3 and ZEB1 collaborate with each other to endow the mesenchymal characteristics of MDA-MB-231; these characteristics constitute a critical phenotype of the cells for aggressive metastatic behavior. Using multiple databases for correlation analysis, we first explored the possibility that the expression of ELK3 and ZEB1 are closely related to breast cancer cells. We performed comprehensive molecular and cellular analyses and showed that ZEB1 is a direct transcriptional activator that contributes to the regulation of ELK3 expression. We also revealed that ELK3 and ZEB1 form transcriptional repressor complexes through direct protein-to-protein interactions. More importantly, we demonstrated that ELK3 is another partner of ZEB1 that represses the expression of E-cadherin, a representative protein that initiates EMT.

Information on the plasmids and sequences used for primers and siRNAs is indicated in the supplementary information section, as shown in Supplementary Tables S1 and S2.

The human embryonic kidney 293T cell line was used for the coimmunoprecipitation (Co-IP) assay and expression of various genes. The well-known triple-negative breast cancer cell line MDA-MB-231 and the human breast adenocarcinoma cell line MCF7 were cultured in DMEM (Gibco) with 10% FBS (Gibco) and 1% penicillin/streptomycin (Gibco) at 37°C and 5% CO₂. The cells were detached with 0.05% trypsin/EDTA (Gibco), and the cell numbers were estimated with a hemocytometer. Transient plasmid DNA transfection was performed with Lipofectamine 2000 (Invitrogen) and Lipofectamine 3000 (Invitrogen) according to the manufacturer's protocols.

Total RNA was extracted by manual methods using TRIzol (Invitrogen), and 1 μg of cDNA was synthesized using the LeGene Express 1st Strand cDNA Synthesis System (LeGene Biosciences Inc.) according to the manufacturer's instructions. qRT-PCR was performed using synthetic cDNAs and TOPreal qPCR 2× PreMIX (Enzynomics). The expression of the target genes was normalized to that of GAPDH. The PCR primers are listed in Supplementary Table S3.

For immunocytochemistry, each cell line was seeded on cover slips in 12-well plates. After 24 hours of incubation, the cells were fixed using 4% formaldehyde (Sigma) for 15 minutes at room temperature (RT), washed 3 times, and permeabilized with 0.5% Triton X-100 (Sigma)/PBS (Gibco). After the cells were washed 3 times, they were blocked with blocking buffer (1% BSA/0.1% Tween 20/PBS) for 1 hour. Then, primary antibodies (1:500) were added to each well and incubated overnight at 4°C. After washing out the primary antibody solution, the cells were incubated with Alexa Fluor 488-conjugated and Alexa Fluor 594-conjugated secondary antibodies (Santa Cruz Biotechnology) for 1 hour at RT, and then fluorescent images were observed under a fluorescence microscope Nikon ECLIPSE TE2000-U and analyzed with NIS-Elements software (Nikon).

For Western blots, cells were lysed with RIPA buffer (Cell Signaling Technology). Total cell lysates were separated by SDS-PAGE and transferred to polyvinylidene difluoride membranes (Bio-Rad). The membranes were blotted with antibodies.

For immunoprecipitation (IP), 293T cells were transfected with pcDNA3.1-V5-ZEB1 vectors (or domain-dependent mutants) or pLenti-ELK3-cMyc-DDDK vectors (or domain-dependent mutants) for 48 hours using Lipofectamine 2000 (Invitrogen). The cell lysates were extracted by 1× RIPA buffer (Cell Signaling Technology) and 100× protease/phosphatase inhibitor cocktail (Pierce Biotechnology, 78440). After quantification of protein concentration, cell lysates were immunoprecipitated by anti-Flag, anti-V5, or anti-ZEB1 antibodies overnight at 4°C. The protein–antibody solution was incubated with protein A/G beads (Santa Cruz Biotechnology) for 2 to 4 hours and washed twice with lysis buffer. The samples were boiled with 2× SDS buffer for 10 minutes at 95°C and were subjected to Western blotting as described previously. The antibodies used in this study are summarized in Supplementary Table S4.

Escherichia coli BL21(DE3) transformed with pGEX4T-1-ELK3 was cultured in LB broth (plus 1000× ampicillin), and ELK3 protein expression was induced by 0.1 mmol/L isopropyl β-D-1-thiogalactopyranoside (IPTG) at 30°C for 5 hours. Escherichia coli were cultured in 100 mL of LB broth, and the cells were sonicated under suitable conditions (45 amp, 10 seconds on and 20 seconds off, 2 times). Cell lysates were incubated with 300 μL of GST bead resin (ElpisBio) at 4°C overnight. The bound proteins were washed with PBS twice and 1% Triton X 100/PBS was added. 293T cell lysates transfected with pcDNA3.1-V5-ZEB1 were divided in equal quantities and mixed with purified GST-ELK3. The mixture of cell lysates and GST-ELK3 was incubated at 4°C overnight. After washing with PBS 3 times, the samples were boiled in 2× SDS buffer for 10 minutes at 95°C. Samples were separated by SDS-PAGE and analyzed by Coomassie blue staining and Western blotting.

A total of 3.0 × 10⁵ cells were seeded in 60 mm dishes with culture media. The cells were transfected with plasmids using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocols. Cells were harvested 48 hours after transfection, and luciferase activity was measured using the Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer's protocols. The values of firefly luciferase were normalized to the respective values of Renilla luciferase.

Cells were harvested with Dulbecco's Phosphate Buffered Saline (DPBS; Gibco), and 37% formaldehyde was added to a final concentration of 1% and incubated for 15 minutes at RT. Glycine was added to a final concentration of 125 mmol/L for 5 minutes at RT, and the cells were washed with cold PBS by centrifugation of the samples at 200 × g for 10 minutes at 4°C 3 times. The cell pellet was lysed in 400 μL of 1× cell lysis buffer (Cell Signaling Technology) containing protease/phosphatase inhibitor cocktail (Pierce Biotechnology). The cell lysates were sonicated in 8 rounds of 8 pulses for 10 seconds on ice at 50% amplitude and incubated for 10 minutes on ice. After centrifugation at 13,000 rpm for 15 minutes at 4°C, the supernatants were normalized to bicinchoninic acid (BCA) using a BCA solution (Thermo Fisher Scientific) according to the manufacturer's protocols. The supernatants were divided equally and mixed with 40 μL of Dynabead protein G and 2 μg of primary antibodies for 2 hours at RT or overnight at 4°C. The complexes were washed sequentially with 1× RIPA buffer, 1× RIPA buffer (500 mmol/L NaCl), LiCl buffer, and TE buffer twice for 10 minutes. Then, 5 μL of 20 mg/mL proteinase K and 3 μL of 10% SDS were added to separate the DNA–protein complex. The DNA was purified by phenol/chloroform extraction and ethanol precipitation. PCR was performed with the extracted DNA.

We reported the SEM for the results. qRT-PCR, Western blotting, and IP were performed 3 times to ensure repetitive results. For patient survival analysis, statistical calculations were performed using the statistical package SPSS 13 (SPSS). The Kaplan–Meier method was used in the survival analysis, and survival curves were constructed. We classified the gene expression as low or high according to the median expression of each gene and then analyzed the survival rate according to the gene expression (i.e., Low/Low, Low/High, High/Low, and High/High).

ZEB1 is often highly expressed in aggressive basal subtypes or triple-negative breast cancer cells, and its expression is correlated with a high risk of breast cancer metastasis that results in the poor survival of patients (13, 28). Because ELK3 is also highly expressed in the basal-like and normal-like/claudin-low breast cancer cell lines and plays a role in orchestrating the metastasis of breast cancers (18), we investigated whether the expression pattern of ELK3 is correlated with that of ZEB1 in breast cancer.

We first examined 49 breast cancer cell lines that model the luminal, basal, and claudin-low subtypes of primary tumors at the transcript level (29). In silico analysis suggested that the expression of ELK3 is positively correlated with that of ZEB1 in the analyzed breast cancer cell lines and ZEB1/ELK3 expression is relatively high in basal type (Fig. 1A; Supplementary Fig. S1). Then, we analyzed the expression of ELK3 and ZEB1 in 3 luminal type (MCF7, SKBR3, T47D) and 3 basal type (MDA-MB231, Hs578T, BT20) breast cancer cell lines. The expression of ELK3 and ZEB1 was relatively high in the basal type compared with the luminal type at both the transcript and protein levels (Fig. 1B and C; Supplementary Fig. S2). Immunocytochemical staining revealed that both ELK3 and ZEB1 were colocalized in the nuclei of MDA-MB231, BT20, and HS578T cells (data not shown). To examine the correlation of ELK3 and ZEB1 expression in clinical samples, we next analyzed the correlation of ELK3 and ZEB1 expression in patients with breast cancer using The Cancer Genome Atlas (TCGA) database (30). The analysis revealed that the expression of ELK3 and ZEB1 had a strong positive correlation in the group of patients (r = 0.799; Fig. 1D). Finally, we assessed the expression pattern of ELK3 and ZEB1 in 4 different molecular subtypes (luminal-A, luminal-B, HER2-E, basal-like) of breast cancer tumors using the Breast Cancer Gene-Expression Miner database (bc-GenExMiner v4.1) based on 36 published genomic datasets. The expression of ELK3 and ZEB1 was positively correlated in all 4 molecular subtypes of clinical samples, and the correlation was the highest in basal-like breast cancer tumors, which are characterized by an aggressive metastatic phenotype with poor prognosis (Fig. 1E). Based on these results, we concluded that the expression of ELK3 is highly associated with that of ZEB1 in breast cancers, especially in basal-like subtypes, both in clinical samples and cell lines.

To investigate the biological mechanisms of the correlation between ZEB1 and ELK3 expression, we observed the effects of ZEB1 suppression on ELK3 expression in MDA-MB231 cells. When siRNA targeting ZEB1 was transfected into MDA-MB231 cells, ELK3 expression was suppressed at both the transcript and protein levels (Fig. 2A). As ZEB1 and various miRNAs function in a double-negative feedback loop to coordinate tumor metastasis (31), we assessed whether the suppression of ELK3 expression was caused by the siZEB1-mediated activation of ELK3 targeting miRNA by analyzing the mRNA stability of ELK3. When actinomycin D was treated with siNS- or siZEB1-transfected MDA-MB231 cells for the indicated times, the suppression of ZEB1 did not have a significant effect on the degradation rate of ELK3 mRNA, indicating that the ELK3 suppression with siZEB1 transfection was not caused by the activation of ELK3-targeting miRNAs (Fig. 2B). Although ZEB1 mainly functions as a transcriptional repressor, it can also form a transcriptional activator complex with cofactors such as YAP1 (13). Therefore, we performed a luciferase reporter assay to analyze whether ZEB1 functions as a transcriptional activator of the ELK3 promoter. When the luciferase reporter plasmid of the ELK3 promoter was transfected into MDA-MB231 or MCF7, it was activated by the cotransfection of the ZEB1-expressing plasmid (Fig. 2C). Furthermore, activation of the luciferase reporter of the ELK3 promoter showed a dose-dependent response to the ZEB1-expressing plasmid in MDA-MB231 cells (Fig. 2D). To determine whether ZEB1 directly functions as a transcriptional activator of ELK3, we performed chromatin immunoprecipitation analyses with an anti-ZEB1 antibody against the ELK3 promoter region between −81 bp and +160 bp, which has a 9 potential ZEB1 binding sites. E-cadherin, which is a well-known direct downstream target of ZEB1, was used as a positive control for the experiment, and the analysis demonstrated that ZEB1 directly binds to the ELK3 promoters (Fig. 2E). Overall, we concluded that ZEB1 functions as a transcriptional activator of ELK3 in breast cancer cells.

It has been reported that murine ELK3 mediates transcriptional repressor activity through interaction with CtBP1 (24), and CtBP is a representative corepressor of ZEB1 in various types of cells, including T cells and cancer cells (11). Indeed, Cytoscape network analysis revealed that ELK3 and ZEB1 share CtBP as a binding partner (Fig. 3A). Therefore, we explored whether ZEB1 and ELK3 directly bind each other. Notably, Co-IP of ectopically expressed V5-tagged ZEB1 or Flag-tagged ELK3 from HEK293T extracts suggested that there is a protein-to-protein interaction between ELK3 and ZEB1 (Fig. 3B). Endogenous binding of ELK3 to ZEB1 was confirmed by Co-IP with cell extracts of MDA-MB231 (Fig. 3C). GST pull-down assays with GST-fused ELK3 further demonstrated that ELK3 directly binds to ZEB1 protein (Fig. 3D). To identify the binding domain of ELK3 that interacts with ZEB1, we constructed 3 mutants of ELK3 by deleting the functional domain of ELK3, such as the transcriptional activation domain, C-terminal inhibitory domain (CID), and N-terminal inhibitory domain (NID; Fig. 3E). Figure 3F shows that Flag-ELK3 deletion mutants 1 or 2 that were deleted with the C domain or CID domain, respectively, could be immunoprecipitated with V5-ZEB1, whereas deletion mutant 3 of ELK3 loses binding affinity to ZEB1.

We next constructed 3 deletion mutants of ZEB1, of which major functional domains were deleted (Fig. 3G). We found that all 3 mutants of ZEB1 that had a deletion of the zinc finger domain (deletion mutant 1), CtBP interaction domain (deletion mutant 2), or Smad binding domain (deletion mutant 3) could bind to ELK3 (Fig. 3H). These results suggest that the C-terminal domain of ELK3, downstream from ETS DNA binding domain, participates in binding to the zinc finger domain of ZEB1.

ELK3 functions as a transcriptional repressor and is transformed into a transcriptional activator form when it is phosphorylated by Ras-ERK signaling (25). Phosphorylation residues are localized at the C domain of ELK3, and our results suggest that the C domain binds to the ZEB1 protein. Therefore, we examined whether the direct interaction of ELK3 with ZEB1 hinders the phosphorylation of ELK3. As expected, ELK3, which is present in a phosphorylated form in 293T cells, became dephosphorylated when it was coexpressed with ZEB1 (Fig. 4A). Additionally, phospho-ELK3 (Ser357) was gradually dephosphorylated depending on ZEB1 concentration (Fig. 4B). These results suggest that an interaction with ZEB1 inhibits the phosphorylation of ELK3 and that the ZEB1–ELK3 complex may function as a transcriptional repressor.

ZEB1 has been previously demonstrated to form a transcriptional repressor complex to downregulate E-cadherin expression (32). Thus, we examined whether ELK3 functions as a transcriptional repressor of E-cadherin as a component of the ZEB complex. To test this possibility, we transfected siRNA targeting ELK3 into MDA-MB231 cells and examined the effect of ELK3 suppression on E-cadherin expression. Notably, E-cadherin expression was increased more than 3-fold 48 hours after transfection at both the transcript and protein levels (Fig. 4C). Contrary to E-cadherin, the expression of other EMT markers such as Snail, Slug, and Vimentin was not affected by the suppression of ELK3 (Supplementary Fig. S3).

We also analyzed the effect of siELK3 transfection in HS578T cells, another triple-negative breast cancer cell line, and confirmed that ELK3 suppression results in the activation of E-cadherin expression (Fig. 4D). Furthermore, ectopic expression of ELK3 in MCF7 cells resulted in a more significant suppression of E-cadherin expression than the ectopic expression of ZEB1 (Fig. 4E). These results suggest that ELK3 functions as a transcriptional repressor of E-cadherin, at least in breast cancer cells. To elucidate whether the association of ELK3 with ZEB1 is linked to the transcriptional repressor activity of ELK3 on E-cadherin expression, we next examined potential ELK3 binding sites in the early promoter region (−86 bp to +60 bp) of E-cadherin with the Eukaryotic Promoter Database (EPD; ref. 33). The 5′ proximal minimal promoter regions of E-cadherin (−86 bp to +60 bp) contain 3 E-box elements (5′-CACCTG-3′) that are a binding site of ZEB1. Interestingly, there are 2 ELK3 binding sites: one is located 14 bp upstream (+38), and the other is 13 bp downstream (+11) of the ZEB1 binding site at +24 of the E-cadherin promoter (Fig. 4F).

Considering that the DNA double helix is approximately 10.5 base pairs per turn, these sites are expected to be occupied by the ELK3 and ZEB1 complexes. To test whether ELK3 regulates E-cadherin promoter activity by binding to these sites, we constructed a luciferase reporter plasmid with the 5′ proximal minimal promoter regions of E-cadherin. Then, we analyzed the luciferase assay in the presence of ELK3- and/or ZEB1-expressing plasmids in MCF7 cells. Consistent with the previous result (Fig. 4E), ectopic expression of ELK3 or ZEB1 resulted in the suppression of E-cadherin reporter promoter activity in MCF7 cells, and the coexpression of ELK3 and ZEB1 more effectively suppressed E-cadherin promoter activity (Fig. 4G). To further confirm that ELK3 functions as a transcriptional repressor by binding to the +38 and +11 sites of the E-cadherin promoter, we generated 3 mutants of the E-cadherin promoter reporter that have a single mutation at the +38 binding site (M1) or at the +11 binding site (M2) and that have double mutations at the +38 and +11 binding sites (M3) (Fig. 4H). As shown in Fig. 4I, the repressor activity of ELK3 on M1, M2, and M3 was significantly weaker than on the wild-type E-cadherin reporter promoter. These results suggest that ELK3 binding to the E-cadherin promoter was hampered by the mutation of the 5′-GGAA-3′ sequence at the +11 and +38 sites into 5′-GGAG-3′. Overall, we concluded that ELK3 functions as a transcriptional repressor of E-cadherin, possibly by being a component of the ZEB1 complex at the proximal minimal promoter region of E-cadherin.

The pivotal role of EMT in oncogenic progression from tumor initiation to metastasis has been a driving force in advancing our understanding of the transcriptional regulatory network of EMT. Among transcriptional factors such as Snail, Slug, Twist, Zeb1, and Zeb2, which play fundamental roles in EMT, Zeb1 is the most recognized factor that regulates cancer cell plasticity to promote metastasis (34). Depending on the binding protein that forms a transcriptional complex, Zeb1 functions as either a transcriptional repressor or activator to regulate the cell-to-cell interaction of proteins, including E-cadherin and cytoskeletal markers such as vimentin (12, 13).

Here, we identified ELK3 as a novel factor in the ZEB1 axis that regulates cancer cell plasticity during the EMT process. Two major conclusions can be derived from our findings (Fig. 4J). First, ZEB1 functions as a transcriptional activator of ELK3, at least in MDA-MB231 and MCF7 cells. Second, ELK3 forms a transcriptional repressor complex with ZEB1 to repress E-cadherin expression. Although accumulating evidence has supported the pivotal role of ELK3 in oncogenicity, little is known about the transcriptional regulatory mechanism of ELK3, except that ELK3 is regulated by KLF4 upon bacterial endotoxin exposure in macrophages (35). The finding that ELK3 expression is directly regulated by ZEB1 in cancer cells suggests that ELK3 plays a critical role in the regulation of EMT by being a member of the ZEB1 axis. In line with this evidence, ELK3 has been reported to contribute to the progression of liver fibrosis by regulating the expression of EMT-related markers (27). Although Zeb1 is highly expressed in metaplastic carcinomas of triple-negative breast cancer, whether the ZEB1 expression level is associated with the clinical outcomes of breast cancer patients remains inconclusive (28, 36). Because both ELK3 expression and transcriptional activity are closely associated with ZEB1 in MDA-MB231, we expected that ELK3 might be the component that clarifies the potential value of ZEB1 as a clinical marker to predict disease-free survival. Unexpectedly, Kaplan–Meier plots with data from TCGA breast cancer dataset (30) showed that the survival rate of the patient group with a high expression of ELK3 and ZEB1 was better than that of patients with a low expression of ELK3 and ZEB1 (Supplementary Fig. S4). The analysis also revealed that the patient group with a high ELK3 and a low ZEB1 expression had the worst clinical outcomes. Furthermore, high expression of ZEB1 and ELK3 did not predict overall survival or progression-free survival of TNBC patients (data not shown). One possible explanation for this puzzling result is that cellular context dependent additional factors are involved in the ELK3–ZEB1 complex-mediated EMT process. Notably, the effect of ZEB1 overexpression on the ELK3 promoter activity in MDA-MB231 is relatively stronger than in MCF7 (Fig. 2C), indicating that ZEB1/ELK3 axis may be controlled by cell type specific mechanism. In pancreatic cancer, histone deacetylases HDAC1 and HDAC2 form a complex with ZEB1 to repress E-cadherin expression (32). Therefore, it will be interesting to address whether other factors, including HDAC and CtBP, are involved in the activity of the ELK3–ZEB1 complex in different cancer cell types. We expect that a comprehensive expression analysis of the components of the ELK3–ZEB1 complex will provide a precise prediction about the patient's clinical outcome, such as disease recurrence and overall survival. Although CtBP is known as a traditional corepressor of ZEB1 (37, 38), the ZEB1 mutation that is unable to bind CtBP did not affect the transcriptional repressor activity of ZEB1 in several cancer cell lines, including MCF7, implying that there is an additional factor that supports the transcriptional repressor activity of ZEB1, at least in the analyzed cancer cell lines (12). Our data showed that mutant ELK3, of which the CtBP binding domain was deleted (Del2), completely lost its ability to repress E-cadherin promoter activity in MCF7 (Supplementary Fig. S5), despite Del2 being able to form a protein complex with ZEB1 (Fig. 3F). A combination of the findings of previous reports and our data suggest that CtBP forms a transcriptional repressor complex with ZEB1 and ELK3 by binding to the CID of ELK3 but not to the CID of ZEB1. This finding provides a plausible hypothesis that ELK3 functions as a coupler between CtBP and ZEB1 to form a transcriptional repressor complex for E-cadherin.

Because we analyzed only the proximal minimal promoter region of E-cadherin, it is possible that there are other regulatory sites bound by the ELK3–ZEB1 complex or ELK3 on the extended region of the E-cadherin promoter. Indeed, ectopic expression of ELK3 could suppress E-cadherin under siRNA-mediated ZEB1 depletion conditions in MDA-MB231 cells, indicating that ELK3 itself, independent of ZEB1, is able to function as a transcriptional repressor of E-cadherin (Supplementary Fig. S6). TCGA database analysis shows that ELK3 expression is correlated with other EMT transcription factors such as ZEB2 and SNAI2 (Supplementary Fig. S7), suggesting that ELK3 might have ZEB1-independent activity to regulate EMT process in cancer cells.

To examine the correlation of ELK3 and ZEB1 expression in clinical samples, we next analyzed the correlation of ELK3 and ZEB1 expression in breast cancer patients using TCGA database (30). The analysis revealed that the expression of ELK3 and ZEB1 had a strong positive correlation in the group of patients (r = 0.799; Fig. 1D). Finally, we assessed the expression pattern of ELK3 and ZEB1 in 4 different molecular subtypes (luminal-A, luminal-B, HER2-E, basal-like) of breast cancer tumors using the Breast Cancer Gene-Expression Miner database (bc-GenExMiner v4.1) based on 36 published genomic datasets. The expression of ELK3 and ZEB1 was positively correlated in all 4 molecular subtypes of clinical samples, and the correlation was the highest in basal-like breast cancer tumors, which are characterized by an aggressive metastatic phenotype with poor prognosis (Fig. 1E). Based on these results, we concluded that the expression of ELK3 is highly associated with that of ZEB1 in breast cancers, especially in basal-like subtypes, both in clinical samples and cell lines.

In addition to repressing E-cadherin expression, ZEB1 regulates other target genes, such as Crumbs3, HUGL2, and Pals1-associated tight junction (PATJ), and leads to the reduced adhesion and increased invasiveness of breast cancer cells (39). Furthermore, ZEB1 represses the expression of stemness-inhibiting microRNAs such as miR-200 (40, 41), implying that cancer cell plasticity and stemness are linked by ZEB1. Therefore, it is important to determine the extent to which ELK3 is involved in the pleiotropic role of ZEB1 in the EMT process. The identification of other targets of the ELK3–ZEB1 complex will greatly extend our understanding of the intrinsic activity of ELK3 that is predicted to take part in the progression of EMT.

Another important point is that ZEB1 regulates the phosphorylation of ELK3. As shown in Fig. 4B, the ELK3 phosphorylation level decreased depending on the concentration of ZEB1. Because phospho ELK3 is known to function as a transcriptional activator and the ELK3–ZEB1 complex functions as a transcriptional repressor, it would be important to further investigate whether ZEB1 influences the nature of ELK3 as a transcriptional repressor or transcriptional activator.

No potential conflicts of interest were disclosed.

Conception and design: H.-J. Cho, N. Oh, J.-H. Park, K.-S. Park

Development of methodology: H.-J. Cho, N. Oh, J.-H. Park, K.-S. Park

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.-J. Cho, N. Oh, K.-S. Kim, H.-K. Kim, E. Lee

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.-J. Cho, N. Oh, J.-H. Park, S. Hwang, S.-J. Kim, K.-S. Park

Writing, review, and/or revision of the manuscript: H.-J. Cho, J.-H. Park, K.-S. Park

Study supervision: S. Hwang, K.-S. Park

This research was supported by the Ministry of Education, Science, and Technology (NRF-2019R1A2C1003581) and by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2019R1A6A1A03032888).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.