The findings of this study reveal that naturally occurring, genetically unperturbed non-spindle carcinoma includes a spectrum of carcinoma cells exhibiting phenotypes ranging from the E-cad+ Vim− epithelial end to the E-cad− Vim+ mesenchymal end. This result supports the concept of EMP, where carcinoma cells differentiate along the epithelial–mesenchymal spectrum in real tumors. EMT is characterized by a loss of proteins associated with the polarized epithelial phenotype and by synthesis of proteins associated with the mesenchymal morphology and involves a transition from polarized epithelial units to individual motile spindle cells [28]. Typically, the loss of E-cad and a switch in the usage of the intermediate filament from cytokeratin to Vim is observed in complete EMT [29]. Our study revealed that the majority of the cells in the M clusters, identified through scRNA-seq, exhibited a complete loss of CDH1 and EpCAM accompanied by a significant increase in VIM and CDH2 expression. These cells demonstrated enhanced migration, invasion, and anchorage-independent tumorigenicity, indicating that E-cad− Vim+ cells represent the complete EMT counterpart of E-cad+ Vim− cells. However, the HM cells maintained a single epithelioid morphology throughout the culture, deviating from the typical spindle morphology observed in the mesenchymal end of the EMT process in many experimental systems [29]. The current study’s discovery of an absence of spindle Vim+ tumor cells in xenografts and primary tumor sections challenges previous reports of an association of spindle morphology with the mesenchymal end of EMT. Our observation of high ZEB1 expression in the HM cells aligns with previous findings of elevated ZEB1 expression in cancer cells lacking morphological indicators of complete EMT [30, 31]. This finding demonstrates that EMT-TFs may be active, even in cells lacking overt spindle morphology [10], which potentially indicates that EMT plays a nuanced role in real-world genetically unperturbed tumors. Our results offer a plausible explanation for the discrepancy between current histopathological assessments and experimental systems, suggesting complete EMT in tumors does not always manifest as spindle morphology and that such morphology may be limited to rare tumor subtypes.
Although EMT and MET are widely recognized as crucial processes in tumor advancement [32], limited evidence has been provided of the dynamic changes in these processes in real-world, genetically unperturbed tumors. In this study, our primary culture system enabled sequential tracking of phenotypic changes in cancer cells; such a task is difficult to complete through static analyses of pathological tissue sections or tumor samples collected at specific points in tumor development [9]. Our findings indicate that all epithelial-predominant HE cells and their progenies exhibited spontaneous EMT during culture passages, suggesting that spontaneous EMT is a common event. By contrast, the consistent E-cad− Vim+ phenotype observed in the HM cells and their progenies throughout culture passages implied that spontaneous MET is a less common event. This observation was corroborated by scRNA-seq data revealing the frequent presence of cancer cells expressing VIM in the HE cells or E cluster. Our findings align with those reported by Sarrio et al. who indicated that EpCAM+ mammary epithelial cells can spontaneously transition to mesenchymal EpCAM− cells, whereas the reverse transition occurs less frequently [33]. Furthermore, the larger tumor size and the lack of evident EMT in the HE cell xenografts, along with the consistent presence of MET in the HM cell xenografts, suggest a distinct role of EMT and MET in tumor progression. Whereas EMT is associated with increased invasiveness during tumor progression, MET appears to be essential for secondary tumor outgrowth. These results are consistent with previous observations of a reversion from a mesenchymal to an epithelial state during metastatic outgrowth in a mouse tumor model, even in the absence of experimental modifications to EMT inducers [11].
Through scRNA-seq, we comprehensively identified TFs that exhibited sequential and significant upregulation as cancer cells progressed through EMT. Our study identified ZEB1 as the earliest TF to exhibit significant upregulation, with its upregulation coinciding with VIM expression. Notably, for the majority of these TFs, after they were activated, consistent upregulation was maintained throughout the entire EMT process. This finding indicates that EMT-upregulated TFs, once initiated, remain active in upregulating downstream TFs, which amplifies the cascade that underlies the complex event of EMT, instead of being sequentially activated at distinct EMT stages. The earliest upregulation of ZEB1 during the initiation of the EMT process, the precise correlation between ZEB1 and VIM expression identified through scRNA-seq and immunofluorescence staining, and the reversal of EMT upon ZEB1 knockdown collectively confirm the pivotal role of ZEB1 in initiating and sustaining the EMT phenotype in breast cancer. Although our results are based on a single tumor sample, the significant correlation between Vim and ZEB1 expression in the TNBC cohort and the consistent ZEB1 expression in all spindle metaplastic components of MpBC, which is pathologically associated with EMT, substantiate the biological importance of ZEB1 in acquiring Vim expression and the EMT phenotype in breast cancer. This finding provides compelling in vivo evidence supporting ZEB1 as a core EMT-TF [8, 10], aligning with previous research highlighting enriched EMT features in TNBC cancer cells expressing Vim [34, 35] and linking EMT features with ZEB1 expression in tumor budding [30]. The current study findings have potential clinical implications: ZEB1-expressing cancer cells demonstrate high resistance to therapy but sensitivity to ferroptosis-inducing drugs [10, 36]. Moreover, aggressive tumors with an EMT phenotype often exhibit high expression of PD-L1 indirectly induced by ZEB1, resulting in immunosuppression in breast cancer [10, 37]. Combining ferroptosis-inducing drugs with immune checkpoint inhibitors may offer therapeutic benefits in carcinomas exhibiting an EMT phenotype.
In this study, we classified ZEB2, MXD4, ID1, ID3, SMARCA1, FOXO1, ZBTB16, and HIPK2 as early upregulated TFs in the EMT process because of their upregulation (VIM 3–4) immediately following ZEB1 upregulation (VIM 2–3). Thus, these TFs were considered more likely than late upregulated TFs to play a causal role in EMT. ZEB2 is well-recognized as a core EMT-TF [8, 10]. Although experimental evidence indicates potential context-dependent interactions among these core EMT-TFs [10, 38–40], such interactions in real-world genetically unperturbed cancers remain to be explored. In this study, we identified and validated both ZEB1 and ZEB2 as spontaneous EMT regulators on the basis of their correlations with VIM expression, upregulation of ZEB2 occurring promptly after upregulation of ZEB1 in the EMT process, and an observed reversal of EMT upon their knockdown. Our findings provide evidence of the cooperative role of ZEB1 and ZEB2 in the acquisition of the EMT phenotype in primary breast cancer.
The members of the ID gene family heterodimerize with basic helix-loop-helix (bHLH) TFs, which inhibits their DNA binding and affects development [41, 42]. Among these members, ID1 has been implicated in EMT in breast cancer [43]; the involvement of ID3 in EMT remains unclear. MXD4, part of the MAD gene family encoding bHLH TFs that heterodimerize with MAX protein, forms a transcriptional repression complex [44]. Although the role of MXD4 in EMT is not yet fully understood, it was reported to be upregulated in MCF10A cells undergoing EMT induced by TGF-β1 stimulation [12]. The roles of the other TFs that our study revealed to undergo early upregulation in EMT, namely SMARCA1, FOXO1, ZBTB16, and HIPK2, have yet to be defined. However, unlike the observations related to ZEB1 and ZEB2, which indicate a precise correlation with VIM expression, the observation that most other early upregulated TFs exhibited some level of expression, even at the epithelial end (VIM 0–1), does not provide support for the TFs playing a direct causal role in VIM upregulation. However, their downregulation upon ZEB1-knockdown-mediated EMT reversion substantiates their involvement in acquiring the mesenchymal phenotype throughout the EMT process and suggests a causal role of ZEB1 in their upregulation. The underlying molecular mechanisms of this process warrant further investigation.
In contrast to ZEB1 and ZEB2, SNAI2 and TWIST1 exhibited significant upregulation during the mesenchymal end stage of the EMT course. However, the observation of sustained minimal expression of SNAI2 and consistently high levels of TWIST1 throughout the early and middle stages of EMT did not provide support for these TFs playing the roles of EMT regulators in this context. In addition to consistent minimal expression of SNAI1 throughout the EMT process and the absence of enriched expression of those EMT-related TFs in the mesenchymal cluster [8], our findings revealed the context-dependent nature of these EMT-TFs in both EMT-related and non-EMT-related regulation in genetically unperturbed breast cancer in real-world scenarios. Furthermore, our finding indicates that upregulation of core EMT-TFs may not initiate EMT but may be caused by TFs in the EMT cascade under specific conditions. Our observation of CREB3L1 being enriched in mesenchymal-predominant HM cells through bulk RNA-seq, its significant upregulation at the mesenchymal end of EMT through scRNA-seq, and its lack of EMT reversion upon CREB3L1 knockdown aligns with this notion. This finding indicates that such TF serves as a marker for the mesenchymal end stage but not as a regulator in the EMT process.
Although we demonstrated the traits and TFs associated with spontaneous EMT, in our culture system, we did not analyze the effect of the tumor microenvironment. However, through in situ RNA-seq, we identified EMT as a prominent gene signature distinguishing Vim+ tumor cells from Vim− tumor cells and observed significantly increased expression of ZEB1, ZEB2, and CREB3L1 in Vim+ tumor cells relative to that in paired Vim− tumor cells within the sampled ROIs in the corresponding primary tumor section. The in vivo data validate our culture system findings, indicating that upregulation of these TFs is intrinsic to carcinoma cells undergoing EMT rather than initiated by the tumor microenvironment [45, 46]. Our discovery, coupled with the finding of a correlation of VIM expression with increased NESs of EMT pathways, confirms Vim as a marker for the spontaneous EMT phenotype in breast cancer. This aligns with the poor prognostic effect of Vim expression in breast cancer [30, 47]. In addition, our identification of HIF1A upregulation during the EMT process (VIM 4–5) indicates that HIF1A upregulation is intrinsic to carcinoma cells undergoing EMT rather than being induced by a hypoxic microenvironment. Although we relied on a single primary carcinoma sample, our finding that gene signatures upregulated in the EMT process in this system correlate with poor patient outcomes validates the biological significance of this model.
In summary, our results demonstrate the presence of EMP and indicate that real-world carcinoma cells undergoing spontaneous EMT may not have a spindle morphology, even in the presence of complete EMT. We illustrated that EMT is a preferred process, whereas MET plays a crucial role in secondary tumor outgrowth. We discovered that TFs that were sequentially and significantly upregulated as carcinoma cells progressed through the EMT process and that ZEB1 plays a pivotal role in initiating and sustaining the EMT phenotype. Our findings confirm the context-dependent role of core EMT-TFs in EMT. Our results offer insights into the characteristics and TFs of spontaneous EMT in genetically unperturbed non-spindle carcinoma in real-world scenarios.