ZEB2 does not act through their canonical role as EMT inducing transcription factors
We sought to determine the ability of ZEB2 to induce a metastatic phenotype in our system. To accomplish this, we exogenously expressed ZEB2 in ER + MCF-7 and ZR-75-1 cell lines which are non-invasive and exhibit an epithelial phenotype. Expression was confirmed using qPCR analysis and nuclear localization of the GFP tagged protein (Supplementary Figure S1). As ZEB2 has been characterized as EMT-inducing transcription factors and are thought to function as transcriptional repressors [30, 31], we anticipated that exogenous ZEB2 expression would induce a transcriptional downregulation of CDH1 expression and of adhesion-related genes. We then evaluated how ZEB2 overexpression in the luminal cell lines affected genes that are associated with maintenance of an epithelial phenotype, tight junction and cell adhesion genes including claudins and desmosomes (CLDN1, CLDN3, CLDN4, CLDN7, DSP, PKP2, PKP3, GJB2, GJB3, ZO-3, CDH3) (Fig. 1A,B). The non-significant p-values for CLDN1, CLDN3, CLDN4, CLDN7, DSP, PKP3, GJB2, GJB3, ZO-3, CDH3 in MCF-7-ZEB2 and CLDN3, CLDN4, CLDN7, DSP, PKP3, GJB2, GJB3, ZO-3, CDH3 in ZR-75-1-ZEB2 cells compared to vector controls were 0.7154, 0.2433, 0.2102, 0.8714, 0.9248, 0.9141, 0.7784, 0.4884, 0.8317, 0.8737 and 0.0634, 0.2539, 0.9460, 0.1903, 0.2608, 0.8286, 0.3919, 0.7946, 0.4855, respectively .While cells overexpressing ZEB2 significantly increased PKP2 in MCF-7 and ZR-75-1 cells (p = 0.0211 and p = 0.0145, respectively) and CLDN1 in ZR-75-1 cells (p = 0.0041), these findings did not appear to be meaningful especially since the other cell adhesion and tight junction genes were not significantly altered in response to ZEB2 overexpression. Next, we sought to assess ZEB2 overexpression effects on EMT-related genes; ZEB2 overexpression did not alter expression of EMT genes, and neither suppressed CDH1 expression nor increased mesenchymal gene expression (CDH2, VIM, SNAI1, TWIST, PLAUR, NME) (Fig. 1C,D). Due to prominent roles of the miR200 family in regulating EMT plasticity, we then sought to determine how ZEB2 transcriptionally regulates this family. We found transcription of miR200 family members, miR200a, miR200b, miR200c and miR141 remained unchanged in ZEB2-overexpressing cells compared to vector controls. The non-significant p-values for miR200a, miR200b, miR200c and miR141 in MCF-7-ZEB2 and ZR-75-1-ZEB2 cells compared to vector controls were 0.6697, 0.2837, 0.5040, 0.4313 and 0.4321, 0.7498 0.2039 and 0.1617, respectively (Fig. 2A,B). Furthermore, analysis of cell morphology with crystal violet staining and brightfield imaging revealed that ZEB2 expressing cells maintained an epithelial like phenotype, as characterized by growth in colonies, cobblestone like shape, and maintenance of cell-to-cell contact (Fig. 2C). Follow-up morphology evaluation with phalloidin staining and fluorescence imaging confirmed these observations. As these results directly contradict our original hypothesis that ZEB2 induced the mesenchymal phenotype through suppression of epithelial genes and upregulation of mesenchymal genes, we further investigated potential mechanisms through which ZEB2 drives metastasis in luminal A breast cancers.
ZEB2 promotes cell motility and metastasis in luminal A breast cancer cell lines.
ZEB family member proteins have been characterized primarily in the context of induction of EMT [19–21], However, recent work demonstrated that the loss of expression of cell-cell junction proteins is not a necessary step in metastasis [32, 33], exemplified by observations of collective cell migration and invasion [34, 35]. Therefore, we sought to examine the ability of ZEB2 to potentiate a metastatic phenotype independent of their function in EMT in a murine model, and we compared these data to mice inoculated with ZEB1-overexpressing cells. MCF-7-vector, ZEB1, or ZEB2 expressing cells were injected into the mammary fat pad SCID/beige mice (5 mice per group) and monitored for tumor growth over 40 days. When the tumors reached 1000 mm3 in size, they were removed and the mice were monitored until they began to show signs of morbidity around 40 days later. Lungs were harvested, sectioned and H&E stained to examine metastasis. We found that tumor growth was not affected by ZEB1 nor ZEB2 expression (Fig. 3A). Furthermore, quantification of one section per mouse in each group revealed that mice with ZEB1 or ZEB2 overexpressing tumor cells had significantly greater number and overall area of metastatic lung lesions when compared to the control group (Fig. 3B-D).
Additionally, we performed assays measuring alterations in cell migration in ZEB1 or ZEB2 overexpressing cells in vitro. Contrary to what was observed in vivo, MCF-7 cell migration was increased by both ZEB1 and ZEB2 (Fig. 3E). ZR-75-1 cells overexpressing ZEB1 or ZEB2 were also significantly more migratory than vector cells (Fig. 3F). As these luminal A cell lines have little to no detectable endogenous ZEB2 expression and present with low motility, these data show that forced ZEB2 expression in these cell systems drove a motile and metastatic phenotype. These data are consistent with previous studies which have characterized ZEB2 as a driver of cell motility and metastasis in numerous cancers [21, 36].
Transcriptomic analysis reveals alterations in estrogen signaling pathways in ZEB2 expressing cells
To further explore the effects of ZEB2 on the phenotype of luminal breast cancer cells, we performed RNA seq analysis of MCF-7 cells expressing vector or ZEB2. We performed upstream regulator analysis using Ingenuity Pathway Analysis (IPA) using a z-score cutoff of 1.5. Several compounds regulating estrogen signaling were predicted to be inactivated in ZEB2 expressing cells. β-estradiol and estrogen were predicted to be inhibited in ZEB2, but this difference was only significant in ZEB2 overexpressing cells (Table I). We then analyzed the data using GSEA for altered hallmark pathways. Notably, downregulation of estrogen response pathways was observed in ZEB2 overexpressing cells (Fig. 4A, Supplementary Figure S2), demonstrating divergence in function between these two homologous family members. Furthermore, examination of breast cancer patient data based on DNA microarray analysis using bc-GenExMiner v4.5  revealed significantly higher expression of ZEB2 in ER negative (ER-) tumors (Fig. 4B). We next interrogated the correlation in gene expression of these factors with a subset of estrogen signaling genes ESR1, FOXA1, GATA3, PGR, TFF1 and BCL2 based on our RNA sequencing data, using the IPA program. This analysis revealed a significant negative trend and inverse association ZEB2 expression with ESR1, FOXA1, GATA3 and BCL2 (Table II). PGR and TFF1 associations were not significant. These findings were similar to correlation plots that demonstrated a negative correlation between ZEB2 and the estrogen signaling genes based on DNA microarrays in a cohort of breast cancer patients with ER + tumors, as obtained using the publicly available bc-GenExMiner v3.0 database  (Supplementary Figure S3).
Estrogen response is abrogated by ZEB2 in ER + breast cancer
The ERα signaling pathway plays a critical role in cell proliferation, motility, and survival in response to estrogen stimulation [37, 38]. We sought to confirm these predicted alterations in ERα signaling in vitro. Baseline gene expression of ERα complex members was assessed by qPCR. ESR1, FOXA1, and GATA3 expression was repressed by ZEB2 compared to vector control in both MCF-7 and ZR-75-1 cells (Fig. 5A). The members of this complex have been shown to be essential to ERα genomic function in breast cancer cells . Furthermore, FOXA1 is necessary for transcription of ERα [39, 40]. We then sought to interrogate the role of ZEB2, specifically, in estrogen response as characterized by transcriptional alterations in ERα signaling target genes. MCF-7 and ZR-75-1 vector and ZEB2 expressing cells were treated with estrogen for 24 hours and analyzed for expression of the estrogen responsive genes BCL2, PGR and TFF1 using qPCR. ZEB2 expressing cells had a diminished estrogen response when compared to estrogen treated vector control cells, though the response was not eliminated entirely (Fig. 5B, C). Additionally, MCF-7 cells were analyzed for estrogen response element promoter activity using luciferase assay. We found that ZEB2 cells had diminished activity of the ERE promoter upon estrogen treatment (Supplementary Figure S4). Notably, baseline ERE activity should not change dramatically because it is theoretically very lowly activated in the DMSO condition so it cannot be decreased in any meaningful way. In this line of thought, it should only be activated after estrogen treatment. We observed ERE to be activated in both vector and ZEB2-overexpressing cells, but to a decreased extent in the ZEB2 group. Taken together, these results demonstrate a novel role for ZEB2 in the regulation of estrogen response in luminal A breast cancer cells.
We then investigated the effects of ERα signaling on MCF7 and ZR75 proliferation in vector and ZEB2 expressing cells. E2 treatment increased cell proliferation in vector control cells (Fig. 5C, D); conversely, ZEB2 expressing cells showed not alterations in proliferation upon E2 stimulation. These data reflect the pathway activation patterns predicted in our RNA seq data, as well as the patterns of gene expression seen in breast cancer patient tumors in vivo. It is of note that in cells expressing ZEB2, baseline growth did not change compared to vector control indicating that the cells are not only insensitive to estrogen growth signals, but perhaps more importantly, that they are capable of functioning without ER signaling. This estrogen independent proliferation led us to led us to investigate the response of ZEB2 specifically to estrogen targeted therapies, and whether this decrease in estrogen response induced by ZEB2 conferred resistance to those therapies.
To address this, we investigated the effects of ERα inhibition in ZEB2 expressing cells compared to cells expressing the vector control in our ER + cell lines in a colony assay. We found that ZEB2 expressing cells had a significantly higher proportion of clonogenic survival in the presence of ICI 182,780 when compared to the vector control (Fig. 6A), indicating that ZEB2 confers a distinct survival advantage in ER + cells when estrogen signaling is abrogated. Next, we found that expression of ZEB2 protected cells from ICI 182,780 induced suppression of proliferation in both MCF-7 and ZR-75-1 cell lines (Fig. 6B, C). Additionally, migration of cells expressing ZEB2 was not affected by ICI 182,780 or estrogen treatment when measured by transwell migration assay (Supplementary Figure S5), demonstrating that ZEB2 mediated changes in migration occur independent of estrogen signaling.