The isolation and characterization of TICs from human tumors and cell lines have been limited because these cells represent a rare population of cells within the tumor and also because of our lack of understanding of their molecular signatures. In this paper, we have described the isolation of TIC-like cells by exogenous expression of the OCT4 TF in primary breast cell preparations. We have also shown that OTBCs exhibit an overlapping gene signature with claudin-low carcinomas.
The relatively low (0.1% to 0.01%) frequency of mesenchymal colonies in the transduced samples suggests that a subpopulation of cells is the target of OCT4. It is possible that, in addition to inducing an expansion of a relatively undifferentiated and rare subpopulation of cells in the mammary gland (possibly a self-renewal stem or an early progenitor cell or both), OCT4 induces global epigenetic reprogramming in an epithelial target cell type of the breast. It is well documented that OCT4 is an essential reprogramming factor [42
] and is sufficient to reprogram neural stem cells toward an induced pluripotent state [43
]. In epithelial and other tissues, it is generally accepted that stem/progenitor cells reprogram at a higher frequency than more differentiated somatic cells, and this also suggests that the target cells mediating the OCT4 phenotype are not fully differentiated. However, to study whether OCT4 induces genome-wide epigenetic remodeling, global changes in DNA and histone methylation need to be evaluated in OTBCs.
To confirm the epithelial origin of OTBCs, we evaluated their differentiation potential by placing the OTBCs in differentiation conditions and performing a detection of specific CKs, which are a hallmark of epithelial cells. In 3D culture conditions, OTBCs formed TDLUs, which were morphologically very similar to those reported for breast stem and cancer stem cells [24
]. When OTBCs were placed in 2D cultures, small populations of cells stained positive for myoepithelial markers (CK14, SMA, and Maspin) or luminal CKs (CK19) or both. These experiments demonstrated that OTBCs had an epithelial origin, and the cell target of OCT4 was possibly a primitive stem/progenitor cell. In self-renewal conditions, OTBCs exhibited antigenic signatures characteristic of prospective stem cells of the breast, such as low levels of CD133, high CD49f, and an absence of EpCAM expression. Given the current understanding of prospective signatures in the mammary gland hierarchy [8
], these antigenic signatures are consistent with a putative breast stem/early progenitor cell identity. The finding that all OTBCs analyzed were EpCAM-
suggested that OTBCs do not originate from prospective luminal-restricted progenitor cells, which are presumably EpCAM+
. However, it is also possible that OTBCs originate from myoepithelial CD10+
restricted progenitors. In addition to being enriched in prospective stem cell signatures, OTBCs were enriched in the tumorigenic, cancer stem cell CD44+
signature. Consistently, we found that as few as 50 cells derived from our OTBC lines was sufficient to generate tumors with metastatic colonization abilities in nude mice.
Histopathological analysis of the tumors in nude mice confirmed the epithelial origin of OTBCs. All OTBCs analyzed generated poorly differentiated carcinomas of the breast and revealed an epithelial morphology with a relatively high nuclear-to-cytoplasmic ratio and brisk mitotic activity. These tumors were negative for ER, PR, and HER2 and were positive for both OCT4 and the mesenchymal marker VIM. Examination of CK staining also revealed that a subset of tumor cells was immunoreactive for CK/pan-keratin, which further supports their classification as a poorly differentiated carcinoma.
To gain a mechanistic understanding of how OCT4 immortalized and transformed the target cells, we performed gene expression microarray experiments. The comparison of genome-wide transcriptional profiles of OTBCs with their parental lines revealed a gene signature that was over-represented in the newly discovered claudin-low intrinsic subtype of breast cancer. Claudin-low carcinomas were recently identified by Herschkowitz and colleagues [44
] and further characterized by using a large database of human breast tumors [6
] and cell lines [6
]. Although claudin-low tumors are relatively rare (representing up to 11% of all breast cancers), they are associated with poor patient survival [6
]. Claudin-low carcinomas uniquely express low levels of tight and adherent junction genes, including claudins and E-cadherin [6
]. Hallmarks of these tumors include enrichment in EMT markers (VIM and Twist) and putative TIC markers (CD44+
Recent genome-wide analysis suggests that this newly discovered intrinsic subtype of breast cancer is closely related to putative EpCAM-
mammary stem cells [8
]. Basal-like breast cancer, which is associated with mutations in the tumor suppressor gene BRCA1
, appears to be more closely related to an EpCAM+
luminal-restricted progenitor cell population [8
]. Further support for the hypothesis that claudin-low carcinomas may arise from primitive stem/progenitor cells is provided by clinical data, which show that TICs are enriched in patients with breast cancer after neo-adjuvant therapy [45
]. Recent gene expression microarray analyses of these TICs revealed enrichment in EMT gene signatures [11
]. Similarly, OTBCs exhibited enrichment in mesenchymal markers and TIC features. Compared with their parental lines, OTBCs upregulated the EMT TFs SNAIL
, and ZEB1/2
as well as microRNAs associated with EMT, such as miR-200s family members and miR-205. EMT has been associated with stemness. The forced expression of EMT TFs in immortalized breast epithelial cells led to stem cell-like characteristics and induction of TIC surface antigens [12
Recently, ectopic expression of OCT4
was shown to enhance malignancy and induce EMT in lung adenocarcinoma cell lines [46
]. This finding confirms our results that link OCT4
as potential oncogenes, which drive EMT processes in the mammary tissue. OCT4 expression was recently demonstrated in the MMTV-Wnt1 mouse models of breast cancer [47
]. Recent work on epithelial ovarian cancer has shown that pluripotency TFs, such as OCT4 and NANOG, are overexpressed in poorly differentiated epithelial ovarian cancers. Furthermore, the RNAi knockdown of OCT4 in these cells prevented or blocked their ability to generate spheroids [16
]. Likewise, a similar report in the MCF-7 breast cancer cell line demonstrated that the knockdown of OCT4 induced tumor cell death [17
]. Our loss-of-function studies also outlined the crucial role of OCT4 and its downstream targets in maintaining self-renewal and EMT in our OTBC lines. We found that the hESC NOS target ZIC1 was upregulated in all OTBCs. Recent reports have suggested that ZIC1 is overexpressed in brain [49
] and lung [50
] tumors. Analysis of transcriptional profiles of large cohorts of human tumors revealed that ZIC1
mRNA is overexpressed in poorly differentiated carcinomas, including breast cancers [13
]. We found that siRNA-mediated knockdown of ZIC1 suppressed the ability of OTBCs to form spheroids in vitro
, outlining an important role of ZIC1 as a potential oncogene in claudin-low carcinomas. These data suggest that OTBCs can be used as model systems to identify oncogenic targets in claudin-low carcinomas.
In hESCs, OCT4 acts as a gatekeeper of self-renewal and master regulator of a TF network [33
]. Indeed, knockdown of OCT4 in hESCs or epigenetic silencing of its promoter irreversibly blocks self-renewal and pluripotency and triggers differentiation gene programs. Consistent with the ability of OTBCs to maintain self-renewal, we found that these lines also activated the endogenous hESC TF network. We speculate that overexpression of OCT4 in a subpopulation of cells in the mammary gland was able to maintain these cells in a 'locked in' and undifferentiated state, limiting them from undergoing downstream lineage specification gene programs. This explanation is consistent with a mouse model of OCT4
cDNA overexpression, which demonstrates that OCT4 generates hyperplasia of the skin and colon by possibly targeting progenitor cells [15
Although the exact mechanism by which OCT4 triggers the TIC-like phenotype needs further investigation, we speculate that gain of self-renewal ability is a complex genome-wide phenomenon that requires endogenous reactivation of a TIC self-renewal TF network. This model is consistent with our microarray data, which show that direct targets of NANOG, OCT4, and SOX2 (NOS targets), which are reasonably well characterized in hESCs, are also differentially regulated in OTBCs relative to their parental lines. Thus, OTBCs could mimic or even corrupt a basic hESC self-renewal TF network, which involves protein-protein associations acting in a combinatorial manner at specific promoter sites [50
]. The characterization of this TIC-like TF network and specifically how this protein network differs from hESCs will require further study. In a TIC, this network may similarly involve associations between TFs, such as OCT4 and NANOG, and co-activator or co-repressor complexes as well as chromatin remodelers. This combinatorial occupancy of factors at specific promoters could result in the activation of potential oncogenes and self-renewal gene programs as well as the repression of selected tumor suppressor genes.
Importantly, our data suggest that NOS targets are regulated differently in TICs relative to hESCs. DKK1, an antagonist of the Wnt signaling pathway, is abundantly expressed in hESCs. In contrast, this target was found downregulated in OTBCs. Indeed, DKK1 is a secreted tumor suppressor in breast cancer [35
]. Wnt signaling in breast cancer has been linked to EMT through stabilization of Snail [55
] and upregulation of the EMT TFs SLUG and TWIST [56
]. Overexpression of DKK1
in a breast cancer cell line resulted in an inhibition of self-renewal ability. Thus, downregulation of the NOS target DKK1
in OTBCs is consistent with gain of self-renewal ability and mesenchymal characteristics via the upregulation of EMT TFs.
In hESCs, OCT4 represses TFs involved in pattern specification, such as homeobox-containing proteins. In contrast, homeobox-containing TFs were highly enriched in our OTBC upregulated gene signature. The homeobox TF SIX1
was found upregulated in the OTBCs relative to parental lines. Overexpression of SIX1
in the mouse mammary gland promoted an expansion of stem/progenitor cells and subsequent tumor development [58
]. In a parallel study, Micalizzi and colleagues [59
] reported that overexpression of SIX1
also facilitated breast cancer metastasis by induction of EMT.
In conclusion, our data support a mechanism by which differential regulation of downstream targets of OCT4
led to activation of oncogenes and downregulation of tumor suppressors. Since OTBCs sustained aberrant self-renewal, it is possible that these cells gained TIC features by selective amplification of spheroids. A compromised tumor suppressor repertoire could result in subsequent selection of clones possessing tumorigenic ability. The molecular events leading to upregulation of TF genes and downregulation of multiple tumor suppressor genes are unknown at present; however, genetic and epigenetic events might be involved. In hESCs, activation of TF networks is associated with promoter de-methylation and gene reactivation. Loss of tumor suppressor functions in OTBCs could also involve genetic and epigenetic silencing mechanisms. Downstream epigenetic regulators of OCT4, such as DNMT3a/b, could be involved in this concerted silencing of tumor suppressor genes [60
]. Alternatively, OCT4 and other self-renewal TFs could be associated with large silencing complexes involving HDACs, such as NODE [61
] and NuRD [62
], which have been well described in hESCs. Consistent with the idea of epigenetic modulation triggered by OCT4, we found that methyltransferase inhibitors and HDACis were able to partially reactivate tumor suppressor genes in OTBCs. In summary, our data describe the generation of novel claudin-low cell lines that could be used by breast cancer investigators to analyze genetic and epigenetic determinants of tumor initiation.