Direct repression of MYB by ZEB1 suppresses proliferation and epithelial gene expression during epithelial-to-mesenchymal transition of breast cancer cells

Epithelial-to-mesenchymal transition (EMT), well described in development [1], enables carcinoma cells to invade local tissues and metastasize to distant sites [2]. EMT causes cell-cell detachment and basement membrane degradation, permitting cell migration aided by actin cytoskeletal rearrangements. EMT triggers myriad intracellular and extracellular signals, which combine to generate motile cells and provide protection against pro-death signals from the host and anticancer therapies, on the journey to secondary sites and while in the systemic circulation (reviewed in [3]).

ZEB1 (zinc-finger E-box-binding homeobox 1) is a dual zinc-finger, DNA-binding transcription factor, recognizing bipartite E-boxes (CACCTG, CAGGTG) and/or Z-boxes (CAGGTA) [4, 5]. ZEB1 as with ZEB2, Snail1 and 2, Twist1 and 2, TCF3 and 4, FoxC2, Goosecoid, KLF8 and Id1 orchestrate EMT transcriptional and morphologic changes (reviewed in [6]). In EMT, ZEB1 is a direct transcriptional repressor of E-cadherin [7] plakophilin3 [8], Crumbs3, HUGL2, and Pals1 [9, 10]. ZEB1 may also promote metastasis, as shown in a xenograft mouse model [10] and significantly higher ZEB1 expression is seen in human breast cancer cell lines of the more mesenchymal/invasive basal B subgroup [11‐13].

The transcription factor MYB is an oncogene in human leukemias, and in epithelial cancers of the colon and breast (reviewed in [14, 15]). MYB promotes proliferation and inhibits differentiation [14]. We have shown that MYB drives proliferation and suppresses apoptosis and differentiation in estrogen receptor (ER)-positive breast cancer cells in response to estrogen [16, 17], and that it is essential for mammary carcinogenesis in xenograft and transgenic models [18].

Mutual regulatory relations have been defined for MYB and ZEB1 in the hematopoietic system. MYB and Ets-1 synergize to overcome transcriptional repression of MYB by ZEB1 [19], and MYB has been shown to regulate ZEB1 expression in the developing inner ear [20]. Conversely, ZEB1 maintains tight regulatory control over MYB during T-cell differentiation [21]. However, the mechanism of this relation has not been defined, and it has not been reported in a solid tumor (cell) context.

A number of transcriptional repressors of CDH1 have been demonstrated to impede cell-cycle progression directly (reviewed in [22]). Colon cancer cells undergoing an EMT at the invasive front coincide with the region where ZEB1 is expressed [23] and display a downregulation of proliferation [24]. Conversely, miR-200 family members, which target ZEB mRNA for degradation [4], have been shown to have a pro-proliferative role [25, 26], thus promoting the growth of breast cancer cell metastases [27]. However, a pro-proliferative role has also been described for ZEB1, because in some contexts, it represses the cell-cycle inhibitors p21 and p73 [28, 29]. The current study sought to determine the ZEB1/MYB/proliferation interplay in the epidermal growth factor (EGF)-responsive PMC42 model of breast cancer EMT.

The PMC42 model system [6] comprises the parental cell line PMC42-ET (ET) and its more epithelial variant PMC42-LA (LA). Both lines exhibit EMT in response to EGF [30, 31], with marked differences in EMT-marker protein expression and arrangement [32]. Here we have identified an inverse relation between ZEB1 and MYB throughout these cell states, and also in the breast cancer cell lines MDA-MB-231 and MDA-MB-468. We showed that ZEB1 is a key player in promoting the mesenchymal phenotype and regulating the proliferative rate in ET cells through the direct transcriptional repression of MYB. Release of MYB repression promotes an epithelial phenotype, or MET, in which proliferation is restored. MYB and CDH1 gene expression were found to correlate in human breast cancer cell lines and in primary human breast tumors and metastases. Collectively our data link the transcriptional regulation of MYB by ZEB1 to the proliferative state of cells during EMT-MET in breast cancer, and also indicate a contribution of MYB to the epithelial phenotype.

Breast cancer metastases in soft tissue and bone have been reported to reexpress CDH1 [55, 56]. We have proposed that the restoration of proliferation required for a secondary tumor to form may be linked to the reversion of EMT, or MET, and we suggest that MYB plays a central role here in promoting tumor growth and the epithelial phenotype. Furthermore, we have established a functional and reciprocal relation between ZEB1 and MYB that is prevalent in breast cancer systems. Direct repression of MYB by the EMT driver ZEB1 builds onto the paradigm through which tumor growth, invasion, and metastasis are integrated (Figure 7). Resistance to apoptosis [57], reduced proliferation [5, 58], and tumor dormancy [59, 60] have all been associated with EMT in the past, but the potential role of MYB in this process is novel.

Examining the consequences of MYB knockdown in LA cells (Figure 1A), and the overexpression of MYB expression in MDA-MB-231 cells (Figure 1B) and MCF-7 cells (see Additional file 8B) has shown (a) that the manipulation of MYB expression has a direct consequence on epithelial versus mesenchymal state, and (b) MYB expression is concordant with the epithelial phenotype in terms of marker expression, culture and colony morphology, and random migration in the monolayer. We have also found increased MYB expression on CDH1 knockdown in MDA-MB-468 cells (not shown).

How could MYB expression determine epithelial state? One possibility that we have considered is through promotion of KLF4 expression. KLF4 was noted as a novel gene upregulated by MYB via occupation of the KLF4 promoter in MCF-7 cells treated with estradiol [61]. KLF4 is a metastasis-suppressor gene that drives CDH1 expression, preventing EMT [62]. Thus in an actively growing, ER-positive primary tumor, MYB may transduce estrogen signaling into both proliferation [16] and promotion of the epithelial phenotype. Further studies will be needed to determine whether this MYB-KLF4 partnership occurs in secondary tumors, whether this may be a means through which MYB could mediate the restoration of CDH1 at these sites, and the functional importance of the epithelial aspect regulated by MYB.

We present evidence to suggest that MYB may repress ZEB1 and hence relieve ZEB1-mediated CDH1 transcriptional repression, restoring the epithelial phenotype. This is suggested by the upregulation of ZEB1 on the knockdown of MYB in LA cells (Figure 6(i)) and the repression of ZEB1 in MYB-overexpressing MDA-MB-231 cells (Figure 6(ii)), and also by previous work in hematopoietic cells [19]. A predicted MYB binding site was identified in intron 1 of ZEB1, the site which is mutated in mice with multiple developmental malformations, some of which were suggested to be related to dysfunctional EMT or MET [20]. Thus, similar to the ZEB1-miR200 reciprocal regulatory relation [4], our data suggest a mutually antagonistic negative-feedback loop between ZEB1 and MYB. ZEB1 and MYB may therefore act as biologic switches, translating molecular changes in the tumor microenvironment into alterations in cell state. These may be epithelial or mesenchymal, as illustrated in Figure 7.

We have shown that ZEB1 binds to the MYB gene in a region encompassing the promoter and intron 1 (Figure 2B). Furthermore, CAT-reporter experiments suggest that the overexpression of ZEB1 represses MYB promoter-intron 1 activity (Figure 2C). This finding is supported by ET-LA cell comparative studies (Figure 3), the characterization of ZEB1sh-ET cells (Figure 4), and a potential role in the EMT of MDA-MB-468 (hypoxia) and LA cells (EGF) (Figure 5A), leading us to conclude that ZEB1 directly targets and represses MYB. Our deletion-mapping studies suggest that ZEB1 can operate through alternative regions, such as the MYB promoter. Although we did not identify exact consensus binding sites for ZEB1 within the promoter, this suggests that ZEB1 may work through divergent binding sites or indirectly by impinging on the activity of a transactivator, such as MYB. Given that MYB is a gene that positively autoregulates itself via tandem MYB binding sites located in the proximal promoter region [54], and that ZEB1 has been shown to inhibit the transcriptional activity of MYB [19], we propose that ZEB1 represses MYB gene expression by intercepting the positive-feedback cycle of MYB. ZEB1 downregulates miR34a in a similar manner by directly repressing the miR34a-positive transactivating protein, ΔNp63, resulting in the acquisition of invasive tumor cell capabilities [63].

Likewise, our studies outline the consequences of ZEB1-mediated gene repression, that is, the suppression of epithelial gene expression and cellular proliferation in breast cancer cells.

The role of MYB in EMT is controversial, and most likely is context dependent. MYB has been positively implicated in the induction of an EMT of avian embryonic neural crest cells [64] and recently in the induction of Snail2 expression in EMT of embryonic kidney, colon carcinoma, chronic myeloid leukemia-blast crisis, and neuroblastoma cells [65]. Knopfova et al.[66] recently showed increased migration and Matrigel invasion, but not collagen invasion, in MYB-transfected-MDA-MB-231 cells, which contrasts our live-cell imaging results for random monolayer migration. In further contrast to our results, it has been shown that TGF-β-induced EMT of ER-positive breast cancer cells is dependent on MYB upregulation, in which putative miR-200 family binding sites were identified in the MYB mRNA, and the downregulation of these miRs was pinpointed as the mechanism driving this EMT [67]. We have observed a significantly higher level of expression of miR-200 family in LA cells compared with ET (unpublished observation), but increased MYB expression in the LA cells that overexpress all miR-200 family members. In addition, in MYB-overexpressing MCF-7 or MDA-MB-231 cells, we did not observe an induction of Snail2 (Additional file 8B, C) nor did we find any association between MYB induction and EMT, but found the opposite (Figure 5A; Additional file 6A, B). We also found no association between MYB and Snail2 in Luminal versus Basal breast cancer subtypes (see Additional file 10). These dramatic differences could be due to cellular context, because much of our work was performed in the PMC42 model system, which clusters with Basal A/Basal breast cancer cell lines (T. Blick, E. Tomascovic-Crook, unpublished data). However, our results were supported by similar findings in luminal, ER⁺ breast cancer cells such as MCF-7 and T47D. It seems more likely that our findings are specific to ZEB1, which we find to be expressed later during the EMT, after SNAI1 and 2 induction in the EGF and EGF/staurosporine-induced EMT in our system [39]. Of the CDH1 repressor genes, ZEB1 has been found by others to be the more-dominant EMT driver, important in stabilizing and maintaining the mesenchymal phenotype in breast cancer systems [5, 9, 68]. It is possible that MYB contributes to early EMT responses by driving Snail 1 and 2 expression while repressing ZEB1, until stronger driving forces activate ZEB1, suppress MYB, and thus suppress SNAI1 and 2. Further work is required to delineate these potentially distinct EMT states and the capricious role that MYB may play.

Comparison of the EMT gene-expression profiles in control versus ZEB1 knockdown PMC42-ET cells (Figure 4A, part iii) may provide insights into specific gene-expression changes that may occur during MET at the secondary tumor site. For example, PAX6 was significantly upregulated in ZEB1sh-ET cells. PAX6 has been described as a tumor suppressor, suppressing invasiveness of glioblastoma cells and MMP2 expression [69], both of which are hallmarks of EMT [6]. PAX6 has also been implicated in promoting the proliferation of breast cancer cells [70], and we have reported that PAX6 is a direct MYB target gene in the neurogenic zones of the brain [71]. Thus the upregulation of PAX6 may also have contributed to the enhanced proliferative rate observed for ZEB1sh-ET. CDH5 expression was significantly downregulated in ZEB1sh-ET cells (Figure 4A, iii). CDH5 has been shown to be induced in an EMT of mammary tumor cells, influencing the levels of Smad2 phosphorylation and upregulating the expression of TGF-β target genes [72]. CD24 was found to be significantly increased in the ZEB1sh-ET cells, correlating with increased CD24 receptor expression, as determined by FACS (data not shown), consistent with our previous findings that EMT traits in human breast cancer cell lines correlate with the CD44(hi/+)/CD24(lo/-) stem-cell signature [12, 73]. MMP1 expression was significantly reduced in the ZEB1sh-ET cells. Therefore, the ZEB1sh-ET cells reexpress markers of the epithelial phenotype.

We have shown that ZEB1 transcriptionally represses MYB to achieve two outcomes: to downregulate cell proliferation and to stabilize the mesenchymal phenotype. Both outcomes are likely to be beneficial to the survival of the cancer cell in the systemic circulation and at sites of dissemination. EMT has been linked to the acquisition of cell dormancy and reduced chemotherapeutic efficiency (reviewed in [74]). In addition, a sustained mesenchymal state may enable escape from immune-related apoptotic signals via expression of factors such as SNAI1, which may act to suppress these signals and the immune system itself [75]; reviewed in [76]. An ongoing debate in relation to the treatment of metastatic breast cancer is whether to customize therapies that reactivate proliferation in dormant, circulating tumor cells and hence improve chemotherapeutic efficiency, or to maintain dormancy in these rogue cells [77, 78]. Each approach is equally problematic, but gene-targeted modulation along the ZEB1-MYB axis may provide an avenue to tackle this foremost clinical dilemma.