Somatic EpCAM mutations are present in a significant number of human cancers. To analyze somatic EpCAM mutations in human cancers, we queried the COSMIC database as well as 178 non-redundant datasets including 47,005 samples in the cBioPortal for Cancer Genomics [27, 28]. We identified 115 unique somatic/missense coding EpCAM mutations (Fig. 1). Depending on the dataset and cancer type, EpCAM mutations are present at a frequency between 0 to 5.13% in human cancers (Fig. S2). We identified multiple cancer-associated mutations which may affect the overall structure and function of EpCAM. C66Y EpCAM is an exemplary TY-1 domain EpCAM cancer-associated mutation identified in a liver cancer specimen. The cysteine residue is part of a critical disulfide bond, and the C66Y mutation is likely to extensively perturb EpCAM structure  and thereby disrupt function of the EpCAM TY-1 domain (Fig. 1). Analysis by PolyPhen-2  predicts the C66Y mutation to have a highly damaging effect (data not shown) on protein structure and function. We demonstrate below that the C66Y mutation also impacts EpCAM cellular localization as well.
WT, but not C66Y EpCAM, inhibits tumor cell invasion in vitro and in vivo.
EpCAM has been implicated in the regulation of cancer invasion. To investigate the role of cancer-associated EpCAM mutations, we initially focused on C66Y. We expressed WT, or C66Y EpCAM, in the human WHIM-3 breast and murine PyMT BO-1 mammary cancer cell lines. Both cell lines have minimal endogenous EpCAM expression and an invasive, mesenchymal phenotype with high CTSL activity. Cells were transduced with retroviruses expressing either GFP (control), WT EpCAM, or C66Y EpCAM. EpCAM expression was confirmed by flow cytometry (data not shown). Expression of WT EpCAM decreased invasion in vitro in both cell lines approximately 70% compared to cells expressing C66Y EpCAM or GFP (Fig. 2A-B). To extend these findings, we expressed WT or C66Y EpCAM in B16-F10 cells and performed both in vitro and in vivo studies. B16-F10 has minimal EpCAM expression, is known to be highly invasive, and is dependent on CTSL activity for migration and invasion . Specific ablation of CTSL decreased B16-F10 invasion in vitro, confirming that this cell line is dependent on CTSL for invasion (Fig. S3A-C). We transduced B16-F10 cells with WT or C66Y EpCAM and selected stable cell lines. We confirmed that expression of WT or C66Y EpCAM was comparable in these cell lines by protein immunoblot (Fig. 2B and S4, left panel). Expression of WT, but not C66Y EpCAM, significantly decreased B16-F10 tumor cell invasion in vitro (Fig. 2B, right panel), and the number of lung cancer metastases following tumor challenge in vivo (Fig. 2C, Fig. S3D). CTSL promotes tumor cell invasion and metastasis by degradation of the interstitial matrix and basement membranes. C66Y EpCAM failed to suppress CTSL-mediated lung metastasis in vivo (Fig. 2C). These results suggest that EpCAM expression has the potential to regulate cancer invasion in the B16-F10 cell line.
WT, but not C66Y EpCAM, inhibits CTSL activity.
Based on the known role of TY-1 domains in the regulation of CTSL (Fig. S1, Table S1), we tested the hypothesis that EpCAM can inhibit CTSL. First, we measured EpCAM expression and CTSL activity in a panel of cell lines. EpCAM expression was assessed by flow cytometry, and CTSL activity was assessed using a fluorescent substrate. We observed a striking inverse correlation between EpCAM expression and CTSL activity (Fig. 3A & 3B). As previously discussed, EpCAM [21, 22] and CTSL [23, 24] are both secreted into the extracellular space where they likely come in contact in the tumor microenvironment. To determine if soluble EpCAM can inhibit CTSL activity, we incubated recombinant EpCAM-Fc with SKOV3 cell lysates as a source of CTSL (SKOV3 cells have the highest CTSL activity [Fig. 3A] in the panel of tested cell lines). Recombinant EpCAM-Fc decreased CTSL activity in a dose dependent manner, and at 10 ng/mL, CTSL activity was suppressed approximately 60% (Fig. 3C). We also tested the ability of cancer cell lines transduced with EpCAM to inhibit CTSL. In cell line A549, which has minimal endogenous EpCAM expression (Fig. 3A), WT, but not C66Y EpCAM was able to significantly inhibit CTSL activity (Fig. 3D). To evaluate the potential role of the EpCAM TY-1 domain in the inhibition of CTSL, we generated EpCAM deletion mutants (Fig. 3E). Stably transduced A549 cell lines with EpCAM deletion mutants were cultured in serum-free media for 24 hours and assayed for CTSL activity. Only EpCAM deletion mutants with intact TY-1 domains were capable of inhibiting CTSL activity (Fig. 3F). As previously reported  and shown here, the cysteine at residue 66 forms a critical disulfide bond in the EpCAM TY-1 domain. The C66Y mutation likely impacts protein structure and CTSL inhibition. Taken together, these studies confirm that EpCAM is capable of inhibiting CTSL activity, presumably via the TY-1 domain.
WT EpCAM physically interacts with CTSL
The ability of TY-1 domain proteins to inhibit CTSL activity typically depends on a physical interaction between the TY-1 domain protein and CTSL. To determine if EpCAM can physically interact with CTSL, we performed immunoprecipitation and protein immunoblot assays. We used the MDA-MB-468 breast cancer cell line, which expresses moderate amounts of both EpCAM and CTSL. MDA-MB-468 cell lysates were incubated with IgG, anti-EpCAM, or anti-CTSL antibodies, and immunoprecipitated proteins were immunoblotted for EpCAM or CTSL. EpCAM immunoprecipitation pulls down CTSL (Fig. 4A and S4), and CTSL immunoprecipitation pulls down EpCAM (Fig. 4B and S4), demonstrating a physical interaction between these proteins.
Many cancer-associated mutations prevent EpCAM cell surface expression and secretion and abrogate the ability to inhibit CTSL
EpCAM is a type-I transmembrane glycoprotein, which is predominantly localized on the cell surface. The extracellular domain of EpCAM is secreted (and/or cleaved) and is detectable in cell culture media and the serum and ascites of cancer patients. Soluble/secreted EpCAM is a 242 amino acid (aa) fragment lacking the signal peptide (23 aa), transmembrane domain (23 aa), and cytoplasmic tail (26 aa) . CTSL, while typically expressed in the endosome, has also been shown to be secreted from tumor cells and potentiates invasion in multiple cancer types [33, 34]. A study of congenital tufting enteropathy patients demonstrated that EpCAM mutations can alter cellular trafficking and localization of EpCAM protein to the cell surface . In that study, multiple germline EpCAM mutations prevented EpCAM expression at the cell surface.
To investigate whether cancer-associated EpCAM mutations affect its cellular localization, and/or its potential to inhibit CTSL, we cloned multiple cancer-associated EpCAM mutations in expression vectors alone, or fused to GFP at the C-terminal, and tracked EpCAM localization in vivo. As shown in Fig. 5A, flow cytometry and confocal microscopy demonstrate that WT EpCAM and some mutants (EpCAM-M115T) localize to the cell surface of epithelial MDCK cells, whereas other mutants (EpCAM-C66Y and EpCAM-L240A) localize in the cytosolic compartments. To verify that soluble EpCAM can inhibit CTSL activity, HEK-293T cells were then transfected with GFP-tagged EpCAM mutants and after 48 hours, conditioned media was collected to measure soluble EpCAM levels by ELISA and/or to test the ability of conditioned media to inhibit CTSL activity. As expected, EpCAM mutants that are not expressed at the cell surface were not detected in culture media by ELISA (Fig. 5B, right), and these conditioned medias could not inhibit CTSL activity from SKOV3 media (Fig. 5C, right). In contrast, WT EpCAM and EpCAM mutants expressed at the cell surface were readily detected in culture media (Fig. 5B, left), and conditioned media robustly inhibited CTSL activity (Fig. 5C, left). Together, these results demonstrate that secreted EpCAM inhibits secreted CTSL activity, while cancer-associated EpCAM mutations that prevent cell surface expression also prevent the ability to inhibit CTSL activity.
To confirm and extend these findings, we tested A549 lung cancer cells under multiple conditions. CTSL activity is required for A549 invasion  and at baseline A549 cells are less invasive compared to other CTSL-secreting cells. Addition of CTSL-rich SKOV3 conditioned media enhanced A549 invasion 3-fold, and this was abrogated by the CTSL inhibitor E64 (Fig. 5D, last 2 bar graphs). This confirms previous reports that CTSL contributes to A549 invasion. To assess the effect of cancer-associated EpCAM mutations on A549 invasion, A549 cells were plated on Matrigel chambers for 2 hours followed by the addition of conditioned media from transfected HEK-293T cells expressing WT or mutant EpCAM. After 30 minutes, CTSL-rich SKOV3 conditioned media was added and invasion was monitored over 24 hours. As expected, WT and cancer-associated EpCAM mutants expressed at the cell surface and secreted suppressed invasion, while cancer-associated EpCAM mutants with no expression at the cell surface did not affect invasion. Together, our findings suggest that soluble/secreted EpCAM can suppress extracellular CTSL protease activity via its TY-1 domain, while cancer-associated EpCAM mutants that are not expressed on the cell surface do not retain this function.