Paget's seed and soil hypothesis has long postulated that cancer cells, or the "seeds", will only grow in a specific microenvironment, or "soil" [24, 29–31]. Indeed, despite the fact that tumors are continually shedding cells, very few circulating tumor cells actually establish metastases, suggesting that post-extravasation survival is a crucial rate-limiting step . The clinical observations that breast cancer displays a characteristic pattern of metastasis, specifically to the lung, liver, bone, and brain, indicate that these organs provide the most conducive microenvironment for metastatic growth. In addition, cancer cells themselves may exhibit an inherent gene signature predisposing them to homing to a particular organ site [14, 18]. The precise environmental factors that enable the organotropism of metastases are yet to be fully discovered, but even less well known is why only a tiny fraction of circulating carcinoma cells form metastases.
Prior to extravasation, cancer cells must survive through invasion and emigration, anchorage-independent dissemination, and extravasation into the ectopic organ. These behaviors are thought to be conferred by molecular changes as a result of EMT. However, post-extravasation, cancer cells encounter a new set of challenges, notably integration within organ parenchyma and establishment of blood supply, which mesenchymal-like cells appear poorly equipped to handle. Despite the importance of EMT in promoting metastatic progression, there is mounting evidence that EMT is not an irreversible switch in cancer cell phenotype. Analysis of primary tumors and their corresponding metastases reveal that even though an EMT may have occurred to engender metastases, the phenotypes of the two can be strikingly similar. E-cadherin expression has been detected in lymph node and non-nodal metastases in carcinomas not limited to breast . Re-expression of adhesion molecules could therefore be one way in which the secondary organ microenvironment promotes survival of metastatic cells as cadherin-cadherin engagement promotes activation of cell survival signaling pathways .
To ascertain whether these earlier reports of E-cadherin-expressing metastases held for intraductal breast carcinomas, we surveyed a small set of matched primary and metastatic tumors. Some 2/3 of metastases to the lung, liver, or brain expressed increased E-cadherin compared to the primary tumors, which largely exhibited aberrantly low to negative E-cadherin expression. Not all metastases exhibited high levels of E-cadherin expression, which is not surprising as metastases are known to evolve and give rise to further disseminations, suggesting that a second EMT may occur within more aggressive nodules.
Interestingly, E-cadherin expression even within metastases was heterogeneous, with increased E-cadherin expression seemingly correlated with proximity to normal parenchymal cells. This heterogeneity suggests that constant interaction with hepatocytes in liver may be necessary. Still, despite these observations, it was possible that these E-cadherin-positive tumor cells disseminated from the primary tumor as epithelioid cells and formed secondary metastatic lesions. Thus, we sought to provide proof-of-principle that cancer cells could be engineered to approach a mesenchymal-to-epithelial reverting transition by altering E-cadherin expression, either exogenously or via the microenvironment. We first hypothesized that we could engineer a MErT in MDA-MB-231 cells by expressing wild-type E-cadherin or by sequestering the E-cadherin-associated catenins with a non-binding E-cadherin construct. After transfecting the MDA-MB-231 cells with the cytosolic domain of E-cadherin linked to the MHC external domain, we saw that the dominant negative protein sequestered α-, β- and p120-catenins. The advantage of using this dominant negative is that the catenin signaling could be parsed from other activities of the extracellular domain of E-cadherin including cell adhesion through trans-ligation and EGFR cis-modulation [23, 34, 35]. While neither construct could completely revert MDA-MB-231 cells to an epithelial phenotype, expression of either construct resulted in morphological transformations and behavioral changes noted as suppression of migration and invasion. Our results also corroborate the findings of other studies focusing on the role of E-cadherin as a tumor or invasion suppressor [36–38].
When cultured in a hepatic microenvironment, MDA-MB-231 exhibited a similar reversion to an epithelial phenotype, both in morphology and E-cadherin re-expression. The nature of the signals that drive the reversion back to an epithelioid phenotype are not known and likely to be complex. Initial studies found that neither conditioned media nor hepatocyte-derived matrix could trigger E-cadherin re-expression in this breast carcinoma line, though the combination of the two was noted to lead to a weak re-expression of E-cadherin (data not shown). Re-expression secondary to loss of methylation of the E-cadherin promoter was also observed in the cell line MDA-MB-435 (Additional file 3), which is now considered to be a melanoma derivative, but is nonetheless useful as this neurectodermal lineage expresses E-cadherin as melanocytes but loses expression during melanoma progression . Furthermore, this reversion is not likely unique to the liver microenvironement, based on the findings in human metastases and in our in vivo mouse model. Recently, we have found that lung parenchymal cells can drive E-cadherin expression in prostate tumor cells . A recent study suggests that laminin-1 may be one component of the extracellular matrix that contributes to E-cadherin re-expression . One key difference between our studies is the microenvironment used to induce E-cadherin re-expression in MDA-MB-231 cells. While Benton et al used a three-dimensional laminin-1 hydrogel, we chose to simulate a secondary organ microenvironment by culturing breast cancer cells with hepatocytes, thereby exposing them to hepatocyte-derived soluble factors and extracellular matrix. Their finding of DNMT1 downregulation as the mechanism for E-cadherin expression was not observed in our system (data not shown), suggesting that tissue architecture may induce MErT by alternative mechanisms. Thus, the search for this signaling 'cocktail' is likely to be complex and lies beyond the scope of the present communication.
That E-cadherin re-expression is caused by loss of methylation suggests a functional mechanism by which the microenvironment modulates the mesenchymal to epithelial phenotypic switch. E-cadherin is predominantly downregulated in carcinomas at the post-translational and/or transcriptional levels. Regulation of E-cadherin is therefore unique among tumor suppressors in which loss or mutation appears to be the rule, but this epigenetic regulation of E-cadherin allows for increased phenotypic plasticity. We have previously reported that prostate cancer cells cultured with hepatocytes also re-express E-cadherin, but as a result of inhibition of the EGF receptor signaling [24, 35, 41, 42]. However, in breast cancers E-cadherin is silenced directly at the transcriptional level by promoter hypermethylation or indirectly through its transcriptional suppressors Snail, Slug, and Twist . No differences in expression of these transcriptional suppressors were observed following hepatocyte coculture (data not shown). In MDA-MB-231 cells, representative of the basal subtype of infiltrating ductal carcinomas, the CpG islands in the promoter region most proximal to the E-cadherin initiation site are fully methylated, which exerts a profound effect on mesenchymal nature. Demethylation of these islands by the chemical agent 5-aza-deoxycytidine causes re-expression of E-cadherin and loss of invasive ability [44–47]. Coculturing of MDA-MB-231 cells with primary hepatocytes resulted in loss of methylation of the E-cadherin promoter and expression of E-cadherin mRNA and protein. We observed that the loss of methylation was dependent on the proliferation of the cancer cells. This finding was not unique to the breast carcinoma cells, as the MDA-MB-435 line also demonstrated loss of promoter hypermethylation upon coculturing with hepatocytes. Importantly, this loss of methylation was at least semi-specific and not global as the imprinted H19 gene remained methylated. The ubiquitous transcription factor Sp1 has been implicated in the regulation of methylation status by binding loci of hemimethylated DNA, protecting sequences from de novo methylation, preferential demethylation, or passive demethylation mechanisms . Sp1 was necessary for loss of methylation in coculture (data not shown), strongly suggesting active signaling from the microenvironment.
The foundation of our findings rest on the epigenetic reversion observed when breast cancer cells are cocultured with primary hepatocytes. The epigenetic status of the primary tumor and disseminated metastases is most likely important, since primary tumors that have high E-cadherin levels have very little systemic disease [31, 49], suggesting that the epigenetic reversion at distant secondary sites is also relevant. The xenograft model in which E-cadherin negative MDA-MB-231 cells formed E-cadherin-negative primary tumors in the mammary fat pads but E-cadherin-positive micrometastases and the finding that at least some E-cadherin-negative primary breast carcinoma cells can re-express this molecule support the idea that this reversion is possible. Furthermore, the xenograft experiment demonstrates that the molecular changes can occur in the secondary site. However, these experiments do not mean that all E-cadherin-positive metastases necessarily arise from the reversion of E-cadherin-negative cancer cells. Further molecular dissections and a much larger breast tumor survey, challenging due to the paucity of matched primary and non-nodal metastases, are needed to determine the extent of this MErT in early metastatic seeding.
The potential implications of E-cadherin re-expression and MErT are many. There are several possible outcomes or combinations of outcomes after a cell extravasates into a metastatic target tissue: apoptosis, dormancy, or sustained proliferation, with the latter appearing the rarest . While E-cadherin typically mediates homotypic cell-cell adhesions, heterophilic ligation between different cell types has been documented [51–53]. Cancer cell adhesion has been shown to facilitate extravasation and colonization of distant organs [54, 55]. Phenotypic reversion to epitheliality in vivo may therefore enhance the integration and survival of cancer cells at the metastatic site by cloaking the cancer cell with epithelioid-like characteristics, or may act to transmit mitogenic signals. E-cadherin expression has also been shown to suppress cell growth, which may account for the dormancy period between clinical presentation of metastases . However, preliminary results in a parallel study reveal that one important survival advantage conferred by E-cadherin expression is increased resistance to cell death induced by chemotherapeutic agents such as camptothecin, doxorubicin, and taxol (data not shown). Cellular adhesion has long been implicated in intrinsic or acquired resistance of solid tumors to multiple anticancer therapeutics not restricted to chemotherapy [57, 58]. The addition of E-cadherin function blocking antibodies sensitizes multicellular spheroids to treatment with various chemotherapeutic agents and E-cadherin-positive cells are more resistant to staurosporine-induced cell death than E-cadherin-negative breast cancer cells . A similar survival advantage may be conferred when disseminated cells face apoptotic cytokines, thus providing a selective pressure that then confounds adjuvant therapies. The finding that E-cadherin re-expression and catenin sequestration can contribute to a MErT suggests that they may be appropriate therapeutic targets for preventing the establishment of metastases in breast cancer.