Effects of EpCAM overexpression on human breast cancer cell lines

EpCAM is frequently overexpressed in human invasive breast cancer [1]. We reported EpCAM overexpression to be an independent prognostic marker for poor overall survival in node-positive breast cancer [2, 3]. An independent group confirmed this finding in node-negative patients too [4]. Moreover, the magnitude of EpCAM antigen expression semiquantitatively assessed by immunohistochemistry showed a dose-dependent relationship with survival. In this retrospective analysis the patient subgroup with breast carcinomas overexpressing both EpCAM and Her-2/neu had the worst prognosis [5]. Targeting EpCAM with a humanized monoclonal antibody (Adecatumumab) in a randomized multi-centre phase II trial for the treatment of metastatic breast cancer yielded an expression- and dose-dependent reduction in formation of new metastatic lesions [6]. Recently, a trifunctional anti-EpCAM antibody (Catumaxomab) has received approval by the European Commission for the treatment of patients with EpCAM positive tumours [7].

EpCAM has initially been identified as a cell adhesion molecule located within intercellular adherens junctions, where it modulates cadherin-mediated cell adhesion and promotes epithelial cell migration and proliferation. EpCAM expression is not only involved in epithelium formation and epithelial-mesenchymal transition during organ development and tissue repair but also contributes to epithelial cell transformation [8, 9]. Regarding EpCAM target genes, overexpression of EpCAM was found to be associated with enhanced transcription and translation of the proto-oncogene c-myc and the cell cycle proteins cyclin A and E in human epithelial 293 cells as well as in murine NIH3T3 fibroblasts [10]. Furthermore, proteome analysis revealed the epidermal fatty acid binding protein E-FABP, a major target of c-myc, to be upregulated upon EpCAM expression in HEK293 cells. Enhanced E-FABP expression correlated with EpCAM expression levels in squamous cell carcinoma lines and in primary head and neck carcinomas [11]. Very recently, the proteolytic shedding of the intracellular domain of EpCAM (EpICD) was shown to confer a mitogenic signal, participating in a multimeric nuclear complex together with FHL2, β-catenin and Lef-1 for the induction of target gene transcription in FaDu hypopharynx and HCT-8 colon carcinoma cells [12, 13]. Furthermore, our group described that DNA methylation is a potential mechanism for the regulation of EpCAM expression [14].

Knowledge on the role of EpCAM in the process of carcinogenesis, tumour progression and metastasis needs further elucidation. Presumably, consequences of EpCAM overexpression and signaling may strongly depend on the tumour type, stage and the tumour microenvironment. This assumption is corroborated by the simple clinical observation that the prognostic impact of EpCAM expression depends on tumour type, disease stage and host antitumour immunity [12, 15]. Contradictory findings from various cell culture systems support the view that EpCAM expression can modulate cell proliferation, differentiation and migration, but the outcome of modulation is strongly dependent on cell type and origin [16‐18].

So far little data exist on EpCAM signaling in breast cancer. The impact of EpCAM expression in human breast cancer cell lines was investigated in loss-of-function studies by silencing EpCAM expression in EpCAM-positive breast cancer cell lines, which resulted in a decrease in cell proliferation, migration and invasiveness, with a concurrent increase of the detergent-insoluble protein fractions of E-cadherin and α- and β-catenin. Importantly, those observations could be confirmed only partially with the weakly EpCAM-positive non tumourigenic breast cancer cell line MCF-10A [17]. Since EpCAM signaling and function has been studied primarily in EpCAM-positive breast cancer cell lines with siRNA-based knockdown, we aimed to generate overexpression breast cancer cell lines and characterize these cell line models by analysing migration, proliferation and transcriptional changes.

Only few commercially available immortalized human breast cancer cell lines lack EpCAM expression. Hs578T and MDA-MB-231 cells were found to express none or only very little EpCAM mRNA and protein in comparison to established and well characterized breast cancer cell lines such as MCF-7 or SK-BR-3 [25].

Both Hs578T and MDA-MB-231 cells were stably transfected with the EpCAM cDNA containing construct pIRESpuro3_EpCAM. The resulting cell lines were designated Hs578T^EpCAM and MDA-MB-231^EpCAM; the cell lines transfected with the empty control vector pIRESpuro3 were named Hs578T^control and MDA-MB-231^control, respectively. Parental MDA-MB-231 cells were originally derived from a malignant pleural effusion. MDA-MB-231 cells are highly proliferative, invasive in vitro and tumourigenic in nude mice [26]. Hs578T cells were derived from an invasive ductal carcinoma, show low invasive potential and are non-tumourigenic [27, 28]. Of note, both cell lines lack E-cadherin expression (due to promoter methylation), which is a common feature among carcinoma cell lines with low or absent EpCAM expression.

In the transfected, constitutively EpCAM-expressing human breast cancer cell lines described here, we observed a reduced expression of Wnt signaling associated components after EpCAM overexpression. We found the downregulation of mRNA levels for SFRP1 in both Hs578T^EpCAM and MDA-MB-231^EpCAM cells, and TCF7L2 in MDA-MB-231^EpCAM only. These are two important negative regulators of Wnt signaling [32]. The results on proliferation and Wnt signaling activation between cell lines were not consistent. As such, Wnt signaling was activated significantly in MDA-MB-231^EpCAM but not in Hs578T^EpCAM. In contrast to these data, proliferation and chemosensitivity was increased in Hs578T^EpCAM but not in MDA-MB-231^EpCAM.

Canonical Wnt signaling hinges on four complexes: the Wnt-frizzled receptor complex in the membrane, the β-catenin destruction complex, the nuclear TCF/LEF/β-catenin transcriptional activator complex and the nuclear TCF/LEF transcriptional repressor complex [33]. SFRP1 acts as a repressor, it binds directly to Wnt ligands and competes with their ability to activate the frizzled receptors [34]. SFRP1 siRNA treatment in the immortalized, non-transformed human mammary epithelial cell line MCF-10A caused a modest but reproducible increase in luciferase activity in a TCF/LEF activation assay. This activity was further enhanced by addition of Wnt-3A conditioned media [34]. As expression of SFRP1 in breast cancer counteracts Wnt signaling, we proposed that its reduction leads to enhanced Wnt signaling in breast cancer cells.

In MDA-MB-231^EpCAM cells mRNAs for both SFRP1 and TCF7L2 were down-regulated compared to the corresponding empty vector line and Wnt signaling was about 20% more active. Of note, Low SFRP1 expression in human breast cancer has been reported by several groups independently [35, 36]. Both low SFRP1 and TCF7L2 levels have been described in a subset of breast carcinomas derived from infiltrating ductal carcinoma [32]. However, reduction of SFRP1 expression only, was not sufficient to enhance Wnt signaling in Hs578T^EpCAM cells compared to Hs578T^control cells. Probably, the reduction of both SFRP1 and TCF7L2 is required to more potently enhance Wnt signaling in breast cancer cell lines, and a contribution of additional factors cannot be excluded.

The transcription factor TCF7L2 belongs to the TCF/LEF (T-cell factor/lymphoid enhancer factor) protein family and is expressed in many isoformal variants which can coexist in one cell. The short isoforms lack the C-terminal protein binding sites (CtBP) which convert the long variants of TCF7L2 into a repressor [37]. It is therefore conceivable that downregulation of a repressive variant of the TCF7L2 protein in MDA-MB-231^EpCAM cells might favour Wnt signaling mediated through other activating TCF/LEF proteins.

Although downregulation of tumour suppressors is a common mechanism during cancer progression, activation of Wnt signaling via the suppression of repressors observed in our breast cancer cells differs from the mechanism discovered for colon cancer, where Wnt pathway activation occurs through loss-of-function mutations of negative pathway components e.g. adenomatous polyposis coli (APC) or gain-of-function mutations in genes activating Wnt signaling (i.e. β-catenin) [38]. Our findings in human breast cancer lines underscore the complexity of Wnt pathway in human cancer [33].

EpCAM and β-catenin proteins were detectable in nuclear lysates of both MDA-MB-231^EpCAM and Hs578T^EpCAM, but the nuclear staining for EpCAM was negative in confocal microscopy. As the antibody used for microscopical analysis was directed against the extracellular domain, a detection of the shedded cytoplasmic fragment EpICD was not possible. Regarding the results otained for full-length EpCAM in the fractionation experiment, we suppose that EpCAM has a perinuclear location and cofractionates with nuclear proteins. However, further experiments are required to establish a relationship to the perinuclear space and the endoplasmatic reticulum.

The activation of TCF/LEF by β-catenin is a crucial step in the direct activation of Wnt target genes. In many studies the expression of key target genes such as cyclin D and c-myc is used as a read-out for Wnt signaling activity [10, 39]. Upregulation of cyclin D and c-myc has been found to be strongly associated with breast cancer progression [38‐40]. However, although expressed at moderate levels in parental cell lines, no significant upregulation of both genes could be detected in Hs578T^EpCAM and MDA-MB-231^EpCAM cells (data not shown). In line with our observation, it has been reported that in the absence of distinct TCF7L2 variants the regulation of common TCF/LEF target genes such as cyclin D1, c-myc, MMP7 and c-jun is not detectable in renal cell carcinoma and colorectal cancer cell lines [41]. As each cell type is able to express multiple isoforms of TCF7L2 the combination of long repressory and short activating variants of TCF7L2 adds even more complexity to the signaling events, thus target gene expression can be affected in many different ways [37, 42]. Furthermore, the type of trigger and feedback loops within the Wnt pathway have been shown to shape the gene expression pattern and thus the cellular response in normal and transformed epithelial cells [43]. Of note, Western blot analysis of MDA-MB-231^EpCAM cells showed reduced levels of both short and long TCF7L2 variants.

The role of ITF-2, which is strongly down-regulated in MDA-MB-231^EpCAM cells, remains unclear. ITF-2 is a basic-helix-loop-helix (bHLH) transcription factor. Its interaction with the inhibitors of DNA binding Id-1 and Id-2 proteins has been described. In complex with Id proteins, ITF-2 can modulate the mammary epithelial cell phenotype and malignant transformation [44]. Zhai et al showed an interference between ITF-2 and Wnt signaling [45]. Furthermore, the appearance of two isoformal variants, ITF-2A and ITF-2B, differing substantially in domain architecture and function, has been reported [30].

Finally, an increase in invasion and migration after EpCAM overexpression was not observed. Such complex processes are governed by multiple molecular changes. The addition of a single gene might not be sufficient and essential interaction partners might be absent in our analysed cell lines. As such, EpCAM has been described to be a transcriptional target of p53 [46]. A loss of wild-type p53 during breast cancer progression does probably not result in a sole increase of EpCAM expression but also in the regulation of other important target genes.

The effects of constitutive expression of the EpCAM oncogene are dependent on the properties of parental cell lines. Further, the functional alterations observed in vitro were marginal, potentially due to the lack of accessory molecules which might be necessary for EpCAM to exert its full oncogenic potential. In MDA-MB-231^EpCAM cells a nuclear accumulation of ß-catenin and a significant upregulation of Wnt reporter assay activity were observed. In Hs578T^EpCAM cells, overexpression was accompanied by a considerable decreased expression of the Wnt pathway inhibitor SFRP1, by an increased proliferation rate and enhanced sensitivity to Docetaxel. Certainly, future investigations are warranted to define the functional relationship between increase of EpCAM expression, Wnt signaling and oncogenic features.