Cleavage and Cell Adhesion Properties of Human Epithelial Cell Adhesion Molecule (HEPCAM)*

Background: EPCAM was described as a cell adhesion molecule involved in regulation of proliferation through regulated intramembrane proteolysis. Results: Proteolysis was characterized in-depth, but cleavage or knock-out of HEPCAM did not affect adhesion. Conclusion: Direct adhesion through HEPCAM is questionable. Significance: Unraveling cleavage sites of EPCAM is crucial for developing inhibitors; however, its cell adhesion function may reveal context dependence. Human epithelial cell adhesion molecule (HEPCAM) is a tumor-associated antigen frequently expressed in carcinomas, which promotes proliferation after regulated intramembrane proteolysis. Here, we describe extracellular shedding of HEPCAM at two α-sites through a disintegrin and metalloprotease (ADAM) and at one β-site through BACE1. Transmembrane cleavage by γ-secretase occurs at three γ-sites to generate extracellular Aβ-like fragments and at two ϵ-sites to release human EPCAM intracellular domain HEPICD, which is efficiently degraded by the proteasome. Mapping of cleavage sites onto three-dimensional structures of HEPEX cis-dimer predicted conditional availability of α- and β-sites. Endocytosis of HEPCAM warrants acidification in cytoplasmic vesicles to dissociate protein cis-dimers required for cleavage by BACE1 at low pH values. Intramembrane cleavage sites are accessible and not part of the structurally important transmembrane helix dimer crossing region. Surprisingly, neither chemical inhibition of cleavage nor cellular knock-out of HEPCAM using CRISPR-Cas9 technology impacted the adhesion of carcinoma cell lines. Hence, a direct function of HEPCAM as an adhesion molecule in carcinoma cells is not supported and appears to be questionable.


Human epithelial cell adhesion molecule (HEPCAM) is a tumor-associated antigen frequently expressed in carcinomas,
which promotes proliferation after regulated intramembrane proteolysis. Here, we describe extracellular shedding of HEPCAM at two ␣-sites through a disintegrin and metalloprotease (ADAM) and at one ␤-site through BACE1. Transmembrane cleavage by ␥-secretase occurs at three ␥-sites to generate extracellular A␤-like fragments and at two ⑀-sites to release human EPCAM intracellular domain HEPICD, which is efficiently degraded by the proteasome. Mapping of cleavage sites onto three-dimensional structures of HEPEX cis-dimer predicted conditional availability of ␣and ␤-sites. Endocytosis of HEP-CAM warrants acidification in cytoplasmic vesicles to dissociate protein cis-dimers required for cleavage by BACE1 at low pH values. Intramembrane cleavage sites are accessible and not part of the structurally important transmembrane helix dimer crossing region. Surprisingly, neither chemical inhibition of cleavage nor cellular knock-out of HEPCAM using CRISPR-Cas9 technology impacted the adhesion of carcinoma cell lines. Hence, a direct function of HEPCAM as an adhesion molecule in carcinoma cells is not supported and appears to be questionable.
Human epithelial cell adhesion molecule (HEPCAM) 2 was first described in 1979 as a tumor antigen based on its ability to induce a humoral response in mice (1). HEPCAM is composed of a large extracellular domain (265 amino acids) linked to a short intracellular domain (26 amino acids) by a single transmembrane domain. In parallel to activities toward the development of EPCAM as a therapeutic target for adjuvant therapy of metastatic colon cancer (2), the first reports on the function as a homophilic, calcium-independent cell-cell adhesion molecule emerged in the mid-1990s (3,4). Ectopic expression of HEPCAM in murine fibroblasts and in the breast cancer cell line L153S resulted in enhanced formation of cell aggregates in suspension (4). The homophilic nature of adhesion was deduced from an enhanced segregation of HEPCAM-negative cells from aggregates of HEPCAM-positive cells in mixed culture assays (4). However, the effects of HEPCAM on pace of aggregation and morphology of transfected cells revealed weak. Initial characterization of HEPCAM functions was most extensively performed in fibroblasts and in poorly differentiated carcinoma cells with mesenchymal phenotype. Therefore, expression and function of HEPCAM were studied in its natural environment in normal epithelia and in carcinoma cells. HEP-CAM preferentially localized at cell-cell boundaries, and its aggregation capacity on cells in suspension was inhibited by an anti-HEPCAM antibody (3). In contradiction to EPCAM-mediated cell adhesion, expression of HEPCAM in cells relying on classical cadherins for cell-cell adhesion resulted in reduced membrane localization of E-and N-cadherins, ␣and ␤-catenins, impaired anchoring of cadherins to the cytoskeleton, and decreased adhesion (5).
The function of Epcam in normal epithelia and development was later analyzed in the mouse, claw frog, and zebrafish. Results of the knock-out of murine Epcam (mEpcam) by three different groups were not completely accordant. In the first published knock-out report, deletion of mEpcam resulted in embryonic lethality at day 12.5 of gestation due to placental defects in the differentiation or survival of parietal giant trophoblast cells (6). Later, Guerra et al. (7) and Lei et al. (8) independently published a perinatally lethal phenotype of mEpcam knock-out mice, due to severe intestinal problems, resembling a human lethal disorder termed congenital tufting enteropathy, which is associated with mutations of the epcam gene (9). Although Lei et al. (8) reported a certain degree of embryonic lethality, the reasons for these obvious discrepancies in phenotypes remain unknown. Furthermore, molecular mechanisms responsible for the observed congenital tufting enteropathy phenotypes were deviating. Guerra et al. (7) proposed a role for adherens junctions with a mislocalization of E-cadherin and ␤-catenin in the developing intestine (7), whereas Lei et al. (8) excluded the involvement of E-cadherin and ␤-catenin, which were properly located, and they claimed a function for mEpcam in the recruitment of claudins to tight junctions. A role for Epcam in the formation of functional adherens junctions via E-cadherin was further described during epiboly processes in the developing zebrafish embryo and in embryonic development of Xenopus laevis (10,11). Similar to reports by Nagao et al. (6), depletion of Epcam in Xenopus was lethal, suggesting an essential role for Epcam in embryonic development (11). Work by Zöller and co-workers (12) further revealed a physical interaction of Epcam with Claudin 7 and a regulatory role in the formation of metastases from rat carcinoma cells. A comparable beneficial effect of Epcam on invasion and migration was observed in Xenopus (11,13) and human breast cancer cell lines (14,15). In contrast, loss of Epcam during epithelial-to-mesenchymal transition (EMT) in circulating and disseminating tumor cells (16 -18) and in zebrafish was reported (19). Knockdown of EPCAM in esophageal carcinoma cells induced a mesenchymal phenotype along with increased migration and invasion (16) and conformed with a dynamic expression of EPCAM during tumor progression (20).
Besides this complex and intricate role in cell adhesion and tissue integrity, HEPCAM was associated early on with a proliferative state of epithelia, especially in carcinomas (21,22). This involvement in the regulation of proliferation and progression through the cell cycle was analyzed in-depth over the last decade. HEPCAM regulated proliferation of breast cancer cell lines (14), fibroblasts, and human embryonic kidney cells, in which it induced the transcription of the proto-oncogene c-MYC (23). To induce cell cycle progression, HEPCAM undergoes regulated intramembrane proteolysis (RIP), which includes a series of consecutive proteolytic cleavages of receptors within lipid bilayers (24,25). The regulated feature is conducted by sheddases within the extracellular domain of substrates, generating a C-terminal fragment (CTF), which is a substrate for ␥-secretase. Commonly, ␥-secretase cleaves CTFs at two distinct ⑀and ␥-sites to produce A␤-like and intracellular fragments (ICD). To date, numerous membrane proteins have been identified as targets of RIP, including prominent molecules such as amyloid precursor protein (APP) and NOTCH receptors (26,27). RIP of substrates has two major functions in that it can initiate signaling through ICDs of receptors and, additionally, result in degradation of substrates (28). Pathologic conditions, such as Alzheimer disease, result from abnormal processing of APP with formation of the disease promoting the A␤ fragment known to induce neurodegenerative plaques (27). RIP of EPCAM results in shedding of the extracellular domain HEPEX and in ␥-secretase-dependent release of the intracellular signaling domain HEPICD (29). Through interactions with the scaffolding protein FHL2 and ␤-catenin, HEPICD can translocate into the nucleus and bind to regulatory element of target genes, including cyclin D1 (29,30). Exact amino acid sequences involved in cleavage have been mapped for murine Epcam (31), but they remain unidentified for the therapeutic target HEPCAM.
In this work, we have investigated regulated cleavage of HEPCAM at the single amino acid level and then addressed its implication in cell adhesion. We demonstrate a broad cleavage pattern of EpCAM with numerous extra-and intracellular products. However, inhibition of cleavage did not affect adhesion of HEPCAM-expressing cells. Through the use of knockout and knockdown cell lines, we demonstrate that HEPCAM has no detectable effect on cell-matrix or cell-cell adhesion in the context of the carcinoma cells used herein. Thus, a general role of HEPCAM as an active cell adhesion molecule in carcinoma cells appears either lacking or context-and cell type-dependent.
Expression Vectors and Transfections-Human EPCAM at full length (314 amino acids) was cloned in fusion with yellow fluorescence protein (YFP) to generate HEPCAM-YFP. EPCAM-TF construct was cloned by the introduction of a TEV cleavage site (ENLYFQG) followed by a FLAG tag (DYKD-DDDK) between amino acid 223 and 224 of HEPCAM. The CTF mimic of human EPCAM termed Myc-CTF-FT-YFP consists of signal peptide sequence MPRLLTPLLCLTLLPALAAR-GLR, a Myc tag (EQKLISEEDL), CTF sequence of human EpCAM(251-315), a FLAG tag (DYKDDDDK), and the TEV recognition site (ENLYFQG) followed by YFP. All constructs were cloned into the 141 pCAG-3SIP expression vector by using NheI restriction enzyme sites. Transfections were performed with the MATra reagent (Iba, Goettingen, Germany) following the manufacturer's recommendations. Stable selec-tion of transfectants was performed with puromycin (4 ng/ml) starting at 1 day after transfection.
Adhesion Assays-For cell-matrix adhesion assays, 96-well flat bottom culture plates were coated with 30 -50 l of fibronectin (6 g/ml, Sigma, Munich, Germany), gelatin (0.2%, Sigma), poly-L-lysine (1:10, Sigma), laminin (15 g/ml, Sigma), Matrigel (40 l/ml, Pharmingen, Heidelberg, Germany) or kept uncoated. Cells were labeled with 1 mM calcein AM (Life Technologies, Inc., Schwerte, Germany) for 1 h at 37°C and washed three times in PBS. Calcein AM is a fluorescent acetomethoxy derivative of calcein, which stains live cells and can be used in rapid adhesion assays (32). Typically, 10,000 labeled cells were seeded in 50 l per well and kept for 2 h of adhesion time. Thereafter, actual measurements were washed twice with PBS (input control was not washed). Before measurement of calcein AM fluorescence in a Victor Wallac instrument, cells were lysed in 2% Triton X-100 in distilled water. For cell-cell adhesion assays, 50,000 of the indicated cells were plated in 96-well culture plates in 50 -200 l of medium and grown overnight to confluence, before the calcein AM-labeled cells were plated on this confluent layer, and measurements were performed as mentioned above.
Scratch Assay, Immunofluorescence, and Laser Scanning Confocal Microscopy-For fluorescence staining, cells were plated on glass slides and grown to confluence before medium was changed to 0% FCS. After 16 h, a scratch was performed, and after 24 h the cells were washed with PBS, fixed in 3% paraformaldehyde, and permeabilized with ice-cold methanol. Cells were treated with horse serum and incubated with anti-HEPCAM and anti-vimentin antibodies (HEPCAM, VU1D9, Cell Signaling Technology; vimentin, 3B4 Dako, Hamburg, Germany, 1:500, 1 h). Cells were incubated with a biotin antimouse antibody and stained with Alexa-linked anti-biotin antibodies. Cells were covered with VectaShield containing DAPI to stain nuclei. HEPCAM and vimentin staining were analyzed using a TCS-SP2 scanning system, a DM-IRB inverted microscope, and LAS AF software (Leica).
CRISPR-Cas9-mediated Knock-out-Cells were transfected with two different CRISPR-Cas9 constructs targeting exon 2 and exon 3 of the human epcam gene. Transfected cells were sorted for the expression of green fluorescence protein (GFP) and separated in GFP high and GFP low cells in a FACSAria device (BD Biosciences, Heidelberg, Germany) before single cell cloning into 96-well plates at 0.5 cells/well. Knock-out clones were identified upon flow cytometry using HEPCAM-specific antibodies.
Mapping of Cleavage Sites-Crystal structure of HEPEX cisdimer (Protein Data Bank code 4MZV) and a coarse-grained molecular dynamics model of EPCAM transmembrane helix dimer (34) were used to map identified cleavage sites to structural data. Model of HEPEX cis-dimer with TEV cleavage site and FLAG epitope inserted immediately after glutamate 223 was prepared using the Robetta modeling server (Washington, D. C.) (35) running Rosetta software (36). All structure figures were prepared in PyMOL (Schrödinger).
Inhibitors-Inhibition of ␣-, ␤-, and ␥-secretase was performed as described in detail in Hachmeister et al. (31) with the exception of both ADAM protease inhibitors GI254023X and GW280264X (Sigma, Munich, Germany, and kind gift by Prof. Stefan Rose-John), which were used at concentrations of 3 M.
RNA Isolation, cDNA Synthesis, and Quantitative Real Time-Polymerase Chain Reaction-Total RNA was extracted from cells using the RNeasy Plus universal kit (Qiagen, Hilden, Germany) and reverse-transcribed to cDNA with the QuantiTect Reverse Transcription Kit (Qiagen). cDNA was amplified using SYBR Green PCR MasterMix (Qiagen, Germany) and genespecific primers in a LightCycler 480 device (Roche Diagnostics). Normalizations across samples were performed using geometric average of constitutive gene expression of gapdh. Gene expression levels were calculated as ⌬CT (CT gene of interest Ϫ CT endogenous control ). The following primers were used: epcam forward, GCAGCTCAGGAAGAATGTG, and reverse, CAGCCAGCTTTGAGCAAATGAC; gapdh forward, TGC-ACCACCAACTGCTTAGC, and reverse, GGCATGGAC-TGTGGTCATGAG.
Purification of Recombinant Proteins and Atomic Force Microscopy (AFM)-Recombinant HEPEX-Fc and DSG3-Fc, both stably expressed in HEK and CHO cells, respectively, were purified from cell culture supernatant as described before (37). hepex was PCR-amplified using 5Ј-GAT CAA GCT TGC CGC CAC CAT GGC GCC CCC GCA GGT CC-3Ј as a forward and 5Ј-GAT CGT CGA CTT TTA GAC CCT GCA TTG AGA-3Ј as a reverse primer. The PCR product was cloned into ps521 vector (kind gift from Pascal Schneider, ISREC, Switzerland). AFM measurements were performed at 37°C in Hanks' buffered saline solution buffer using the force spectroscopy mode of a Nanowizard III (JPK Instruments, Berlin, Germany) AFM setup to detect possible single molecule interactions as detailed before (38). Briefly, Si3N4 tips of the cantilever (MLCT, Bruker, Mannheim, Germany) and freshly cleaved mica sheets (SPI Supplies, West Chester, PA) were functionalized with the same amounts of recombinant proteins or BSA using polyethylene glycol spacers (39). To measure single molecule interactions, the AFM tip was repetitively lowered to and retracted from the mica sheet. Binding events were detected by continuously measuring and analyzing the cantilever deflection. In all measurements, a set point of 0.2 nanonewtons and a z-range of 300 nm were applied. For HEPEX measurements, the pulling speed was modulated between 1 and 10 m/s with resting times on the mica of either 0.1 or 0.5 s. Measurements with DSG3-Fc (positive control) and BSA (negative control) were operated with a pulling speed of 1 m/s and a resting time of 0.1s. At least 500 force-distance cycles at different positions on the mica were recorded for each cantilever/mica combination. Antibodies were incubated for 60 min at a concentration of 40 g/ml for anti-HEPEX-Ab and 75 g/ml for the anti-Dsg3 mAb (AK23, Biozol, Eching, Germany).

Results
Extracellular Cleavage of HEPCAM-Extracellular shedding is a prerequisite for subsequent cleavage of various cell surface receptors by ␥-secretase, including HEPCAM (29). HEK293 cells were stably transfected with an expression plasmid for a fusion of HEPCAM with yellow fluorescence protein (HEPCAM-YFP). Cleavage of HEPCAM-YFP was analyzed in membrane assays in dependence of inhibitors of ADAM proteases (GI254023X and GW280264X) and ␥-secretase (DAPT), respectively. After 22 h of incubation at 37°C in the absence of inhibitors, HEPCAM-YFP was cleaved to generate a membrane-associated C-terminal fragment (HEPCAM-CTF-YFP) and an intracellular fragment HEPICD-YFP (Fig. 1A). Both ADAM protease inhibitors GI254023X and GW280264X strongly decreased cleavage to HEPCAM-CTF-YFP and HEPICD-YFP (Fig. 1A). DAPT treatment inhibited formation of HEPICD-YFP and resulted in accumulation of HEPCAM-CTF-YFP, whereas combinations of ADAM and ␥-secretase inhibitors abrogated HEPICD-YFP and strongly reduced HEPCAM-CTF-YFP formation (Fig. 1A). Sizes of naturally generated HEPICD-YFP and recombinant HEPICD-YFP comprised of the predicted 26 intracellular amino acids of HEPCAM slightly differed. Naturally generated HEPICD-YFP had a higher apparent molecular weight, suggesting cleavage within the transmembrane domain (Fig. 1A). This is in line with earlier findings of an apparent molecular mass of 5 kDa for endogeneous HEPICD, despite a calculated molecular mass of 3 kDa (29).
To evaluate the impact of cleavage on levels of full-length EPCAM, HEK293-EPCAM-YFP cells were treated for 24 h with cycloheximide to inhibit protein de novo synthesis. Additionally, cells were treated with GI254023X and C3 or DMSO. Inhibition of EpCAM cleavage resulted in accumulation of fulllength EpCAM (Fig. 1A, right panel). This accumulation was quantified upon flow cytometry analysis and represented a 15-20% stabilization of EpCAM at the cell membrane (data not shown).
Further proof of a quantitative processing of HEPCAM-CTF-YFP to hEpICD-YFP and of its anticipated localization at the membrane was obtained upon fluorescence confocal microscopy imaging. HEK293 cells stably expressing HEP-CAM-CTF-YFP were treated with DMSO or DAPT before imaging of YFP fluorescence. In the absence of DAPT, YFP fluorescence was low and only detectable in intracellular speckles, although it located to the plasma membrane after inhibition of cleavage through ␥-secretase (Fig. 1B).
Exact amino acids involved in cleavage of HEPCAM are not yet known. Therefore, a variant of HEPCAM incorporating a TEV protease cleavage site and a FLAG epitope located 42 amino acids N-terminally of the predicted transmembrane domain (HEPCAM-TF; Fig. 1C) was expressed in HEPCAMnegative HEK293 cells. Correct cell surface expression was confirmed using HEPEX-and FLAG-specific antibodies in flow cytometry and immunoblot analysis (Fig. 1D). Cell culture supernatants from HEK293-HEPCAM-TF stable transfectants were used to immunoprecipitate HEPEX ectodomains. After TEV protease treatment, the remaining fragments composed of FLAG epitope and extracellular cleavage sites were analyzed upon mass spectrometry (schematic representation in Fig. 1C). Calculated and determined masses of fragments reproducibly identified in three independent experiments, including mass errors and deduced charges, are compiled in Fig. 1E. Three major cleavage peaks were detected in HEPCAM-TF-transfected cells compared with control cells carrying empty vector only, i.e. ⌬pCAG HEK293 cells (Fig. 1F), and were confirmed in FaDu hypopharynx and HCT8 colon carcinoma cell lines, however with reduced ␣-peak intensities in HCT8 cells (Fig. 1G). Peaks ␣1 and ␣2 corresponded to cleavages at amino acids aspartate 243/proline 244 and proline 244/glycine 245. Both peaks were reduced in comparison with the ␤-peak in cells treated with the ADAM protease inhibitor GI254023X and became the predominant cleavage peaks after treatment of cells with the phorbol ester PMA, which is a reported strong inducer of ADAM proteases (40) and thus served as a positive control (Fig. 1H). Peak ␤, corresponding to cleavage at amino acid tyrosine 250/tyrosine 251, was strongly reduced in cells treated with the ␤-secretase 1 (BACE1) inhibitor C3 (Fig. 1I). Peak amplitudes were calculated from representative spectra to determine specificity of inhibitors. Here, maximal peak amplitudes were divided by baseline values and are given as ratios in Fig. 1J. Smaller peaks most probably represented single amino acid trimming by endopeptidases, starting from tyrosine 250 toward the N terminus (asterisks in Fig. 1F). One additional peak (marked with # in Fig. 1), with an m/z difference of 91 Da, was not assignable to the actual amino acid sequence, i.e. a glycine (75 Da) and was not induced by PMA.
Thus, extracellular shedding of HEPCAM primarily occurs at two distinct ADAM protease sites and one BACE1 site to generate three major ectodomains. Alignments of murine and human Epcam amino acid sequences disclosed an identical BACE1 cleavage site but differing ADAM protease cleavage sites. ␣and ␤-cleavage sites were in closer proximity in HEPCAM than in mEPCAM (Fig. 1K).
Intramembrane Cleavage of HEPCAM-To investigate intramembrane cleavage of EpCAM by ␥-secretase, a shortened EpCAM construct bypassing the need for ectodomain shedding was used. To this end, HEK293 cells were stably transfected with a mimic of HEPCAM-CTF-YFP composed of an N-terminal c-MYC epitope, 15 amino acids of the extracellular domain, and transmembrane and intracellular domains in FIGURE 1. Extracellular shedding of HEPCAM by ADAM and BACE1 proteases. A, cleavage of HEPCAM-YFP was assessed in stable HEK293 transfectants using membrane assays. Cells were treated with DMSO, GI254023X, and GW280264X in combination with DAPT as indicated. The presence of HEPCAM-CTF-YFP and HEPICD-YFP is indicated. Lysates from stable transfectants of HEPICD-YFP in HEK293 cells were separated as a size control. Proteins were detected with a GFP-specific antibody. Shown is one representative experiment from three independent experiments. B, plasma membrane localization of HEPCAM-CTF-YFP after inhibition of ␥-secretase activity. HEK293 cells stably transfected with HEPCAM-CTF-YFP were either treated with DMSO (left) or with the ␥-secretase inhibitor DAPT (right). After 10 h, HEPCAM-CTF-YFP expression and localization were assessed using YFP fluorescence in laser scanning microscopy. Nucleic DNA was visualized with DAPI. C, schematic representation of HEPCAM-TF used for proteomics and mass spectrometry approaches. Shown are TEV cleavage sites and FLAG epitopes along with fragments generated during cleavage. D, immunoblot and flow cytometry detection of EpCAM-TF in stable HEK293 transfectants. Lysates of wild-type HEK293 (HEK WT), control transfectants (HEK⌬pCAG), and HEPCAM-TF transfectants (HEK TF) were separated in a SDS-PAGE and proteins detected with HEPCAM-and FLAG-specific antibodies. Stable HEK TF cells were analyzed by flow cytometry with isotope (dark gray), HEPCAM-, and FLAG-specific antibodies (light gray). Shown are the representative results from three independent experiments. E, tabular representation of peptides identified upon mass spectrometry, including calculated and determined mass in daltons, errors, and charges. F and G, representative mass spectrum of processed supernatants from HEK⌬pCAG and HEK TF (F) or HCT8 and FaDu cells (G), including two ␣-, one ␤-cleavage, and trimming sites within HEPEX. H and I, representative mass spectrum of processed supernatants from HEK TF pretreated with DMSO, GI254023X, or PMA (H) or C3 (I). J, calculated relative peak intensities (peak max /baseline). K, alignment of ␣and ␤-cleavage sites in human and murine EpCAM.
fusion with YFP (MYC-CTF-FLAG-TEV-YFP). The length of 15 amino acids was chosen because it corresponds to the shortest CTF generated after BACE1 cleavage (see Fig. 1I). Addi-tional FLAG tag and TEV sites were cloned C-terminally between HEPICD and YFP, to process cleavage fragments (see scheme in Fig. 2A). Amounts of HEPCAM-CTF and HEPICD in DMSO-treated MYC-CTF-FLAG-Tev-YFP expressing HEK293 cells were very low and associated with degradation bands, whereas ⌬pCAG control cells expectedly showed no expression (Fig. 2B). Inhibition of ␥-secretase with DAPT resulted in accumulation of two CTF fragments with substantially differing molecular weights, suggesting additional trimming of CTFs and robust processing of CTF fragments to hEpICD by ␥-secretase, which, however, could not be visualized under these conditions (Fig. 2B). As such, low levels of HEPICD in DMSO-treated cells implied efficient degradation of fragments after their release from HEPCAM-CTF, and MYC-CTF-FLAG-TEV-YFP HEK293 cells were treated with the proteasome inhibitor clasto-lactacystin-␤-lactone. This led to accumulation of HEPICD with a molecular weight slightly higher than recombinant HEPICD (Fig. 2B), thus confirming previous data (Fig. 1A).
Supernatants of MYC-CTF-FLAG-TEV-YFP HEK293 cells were collected after 48 h of culture and subjected to anti-c-MYC immunoprecipitation before analysis of cleavage sites using mass spectrometry. Calculated and determined masses of fragments reproducibly identified in three independent experiments, including mass errors and deduced charges, are compiled in Fig. 2C. ␥-Cleavage of HEPCAM-CTF was reproducibly observed at three major sites, ␥1, ␥2, and ␥3, corresponding to cleavage at amino acids valine 273/valine 274, valine 274/ valine 275, and valine 275/valine 276. Corresponding peaks were lacking in control cell line ⌬pCAG or after DAPT treatment (Fig. 2, D and E). One additional very minor peak could be assigned to a fragment lacking valine 273 (labeled # in Fig. 2, D and E). All major ␥-cleavage sites were confirmed with carcinoma cell lines HCT8 and FaDu expressing MYC-CTF-FLAG-TEV-YFP (Fig. 2D). Calculated molecular weights corresponded to an N-terminally trimmed signal peptide with two remaining amino acids (leucine and arginine). Further peaks (labeled * in Fig. 2, D and E) corresponded to the same ␥-secretase cleavage sites ␥1-3, although with a differently trimmed signal peptide comprised of two additional N-terminal amino acids (arginine and glycine).
Subsequently, lysates of MYC-CTF-FLAG-TEV-YFP HEK293 cells were subjected to YFP-specific immunoprecipitation using GFP-trap beads, TEV digest, and FLAG-specific immunoprecipitation. Enriched fragments were analyzed using mass spectrometry. Calculated and determined masses of fragments reproducibly identified in three independent experiments, including mass errors and deduced charges, are compiled in Fig. 2C. Cleavage of HEPCAM-CTF was observed at two distinct ⑀1 and ⑀2 sites, representing cleavage at amino acids valine 284/valine 285 and valine 285/leucine 286, and corresponding peaks were lacking in control cell line ⌬pCAG or after DAPT treatment (Fig. 2, F and G). Both ⑀-cleavage sites were confirmed with carcinoma cell lines HCT8 and FaDu expressing MYC-CTF-FLAG-TEV-YFP (Fig. 2F). Hence, HEP-CAM-CTF is processed by ␥-secretase to generate two distinct A␤-like fragments, which are released in the extracellular space, and two cytoplasmic HEPICD fragments. A comparison of cleavage sites of murine and human Epcam disclosed high similarities with one identical ␥-cleavage (␥1) and two identical ⑀-cleavages (⑀1 and 2) (Fig. 2H).
Cleavage Site Localization in Three-dimensional Structures of HEPCAM-Crystal structure analysis and coarse-grained molecular dynamics simulation revealed a prevalent occurrence of HEPCAM as a cis-dimer at the cell surface (34). Extracellular cleavage sites were mapped onto the three-dimensional structure of HEPEX dimers to address their accessibility. Both ␣-sites are located in a groove in the membrane-distal part of dimers, on a loop connecting to the top region of dimers, termed the "ridge on C-terminal domain" (34). ␤-Site resides in a membrane-proximal and even more shielded part of the native protein (Fig. 3, A-F). Both localizations imply restricted accessibility of cleavage sites within cis-dimers and suggest a need for partial and/or temporal dissociation of cis-dimers for cleavage. Mapping of TEV-FLAG insertion site onto the monomer and cis-dimer revealed that the inserted segment is located near the dimer interface (Fig. 3, A-F). Insertion at this site could not have an effect on the overall structure of EpCAM subunits because it is part of a surface-exposed connecting region. The insertion of more than 10 amino acids might still influence cisdimer/monomer equilibrium; however, no significant differences in cleavage rates were observed between TEV-FLAGcontaining and wild-type EpCAM variants (data not shown). Also, it is possible that the TEV-FLAG insertion does not affect cis-dimerization at all, which is supported by a model of HEPEX-TF where the TEV-FLAG segment could easily adopt a conformation compatible with cis-dimerization (Fig. 3, B and  C). Comparably, addition of YFP did not impair HEPICD generation and functions, neither in vitro nor in vivo (29). Intramembrane ␥and ⑀-sites were mapped onto a model of HEPCAM transmembrane helix dimer calculated using coarsegrained molecular dynamics simulations. ␥-Sites and ⑀-sites are located N-and C-terminally of the helix dimer crossing region formed by amino acids valine 276 to valine 280 and hence do not interfere with major structural features of dimers and are readily accessible within cis-dimers (Fig. 3, G and H).
Endocytosis of HEPCAM-Because of its optimum at pH 4.5 for catalytic activity, BACE1 is functional in acidic intracellular compartments, including the trans-Golgi network and endosomes (41). Consequently, shedding through BACE1 was suggestive of an internalization of HEPCAM into acidic intracellular compartments through endocytosis. To address this hypothesis, HEPCAM-positive esophageal carcinoma cell line Kyse520 was used based on the reported gradual loss of HEP-CAM expression during processes of cell migration (16). Kyse520 cells were cultured to fully confluent monolayers, scratched with a sterile pipette tip, and HEPCAM expression and localization were assessed upon immunofluorescencebased confocal laser scanning microscopy. Compared with cells present in structured epithelial monolayers, migrating Kyse520 cells were characterized by loss of cell surface HEPCAM and increase of intracellular speckled HEPCAM staining (Fig. 4A). Double staining of HEPCAM and mesenchymal marker vimentin demonstrated the shift of migrating Kyse520 cells toward a more mesenchymal phenotype (Fig. 4B). Comparable cytoplasmic HEPCAM speckles were observed in head and neck carcinoma cell line FaDu (Fig. 4C).
Next, cells were grown to full confluence, incubated for 15 min with anti-HEPCAM antibody targeting the extracellular domain HEPEX, fixed, permeabilized, and stained with a FITClabeled secondary antibody. Anti-HEPCAM antibodies internalized via endocytosis were detected as strong cytoplasmic speckles (Fig. 5A). Overlay of HEPEX and hEpICD stainings displayed full-length HEPCAM (yellow speckles) and intracellular HEPICD (red speckles) (Fig. 5B), depicting endocytosed full-length HEPCAM and intracellular RIP product hEpICD. Short term inhibition of clathrin-independent endocytosis upon treatment of cells with nystatin for 45 min had no effect or a minor effect on the occurrence of intracellular hEpEX speckles, whereas inhibition of clathrin-dependent endocytosis by chlorpromazine resulted in nearly complete lack of intracellular hEpEX speckles (Fig. 5C). Thus, human EPCAM is subject to clathrin-dependent endocytosis and subsequent cleavage by BACE1 in carcinoma cell lines. Because of toxicity upon long term treatment of cells with endocytosis inhibitors, effects on the half-life of EPCAM, which was reported to be ϳ20 h (42), could not be assessed.
Cleavage of EpCAM Does Not Affect Cell Adhesion-One of the primary functions assigned to HEPCAM related to the formation of cell-cell adhesion via homophilic interactions of HEPCAM molecules on opposing cells (3, 4, 43). Cleavage of HEPCAM is required for RIP-dependent signaling, but additionally it has the potential to decrease cell surface expression of the protein and thus might impact on HEPCAM-dependent adhesion in a dual manner through signaling and protein availability. FaDu hypopharynx carcinoma cells were pretreated 24 h with either DMSO as a solvent control or a combination of ␣-, ␤-, and ␥-secretase inhibitors (GI254023X, C3, and DAPT). Thereafter, cells were stained with calcein AM, a fluorescent acetomethoxy derivative of calcein, which is cell-permeable and stains live cells, and cultured for a further 2 h in the presence of DMSO or inhibitors on Matrigel matrix-coated 6-well plates. After washing and lysis, calcein AM was measured as a surrogate for adherent cells and compared with calcein AM fluorescence of the input without washing steps, which was set to 100% (Fig. 6, A and B). Treatment of FaDu cells with GI254023X, C3, and DAPT did not influence cell-matrix adhesion to Matrigel, which represented 26.90 and 28.3% of input, respectively (Fig. 6D, left panel). Alternatively, FaDu cells were pretreated for 24 h with GI254023X, C3, and DAPT or DMSO as a control and plated on 6-well plates. Subsequently, inhibitor-and control-treated FaDu cells were stained with calcein AM and plated on top of the abovementioned layer of inhibitorand control-treated FaDu cells (Fig. 6C). Cell-cell adhesion was assessed after 2 h in the presence of DMSO or inhibitors as the percentage of calcein AM fluorescence of washed versus untreated samples. Again, treatment of FaDu cells with ␣-, ␤-, and ␥-secretase inhibitors had no measurable effect on cell-cell adhesion, which represented 20.81 and 20% of cell input, respectively (Fig. 6D, right panel). Because HEPCAM was described as a calcium-independent homophilic cell adhesion molecule, cell-matrix and cell-cell adhesion assays were repeated in the absence of any calcium in culture media. Overall adhesion levels were significantly reduced upon withdrawal of calcium from 20 -30 to 6 -9%, which was expected considering FIGURE 4. Internalization of HEPCAM upon migration and mesenchymal transition. A, confluent Kyse520 esophageal carcinoma cells were subjected to a scratch assay, and HEPCAM was visualized using specific antibodies in immunofluorescence staining and confocal laser scanning microscopy. B, HEPCAM and vimentin were simultaneously detected in migrating cells using specific antibodies in immunofluorescence staining and confocal laser scanning microscopy. C, HEPCAM was visualized in FaDu hypopharynx carcinoma cells using specific antibodies in immunofluorescence staining and confocal laser scanning microscopy. In each stainings, nuclear DNA was detected with DAPI. Shown are representative results from two independent experiments. the presence of E-cadherin in FaDu cells. However, neither cellmatrix nor cell-cell adhesion was affected by the treatment of FaDu cells with ␣-, ␤-, and ␥-secretase inhibitors in the absence of calcium (Fig. 6E). Hence, regulated intramembrane proteolysis of HEPCAM had no impact on cell adhesion properties.
Effect of Cellular Knock-out of HEPCAM on Cell Adhesion-To further address the contribution of HEPCAM to adhesion of carcinoma cells, we generated CRISPR-Cas9-mediated knockout clones of FaDu hypopharynx cells. Fig. 7A depicts wild-type FaDu cells (FaDu-WT), one CRISPR-Cas9 clone with wild-type expression of HEPCAM (FaDu-CC-#L21ϩ), and three knockout clones (FaDu-CC#L13, #L20, and #H4). Statistical analysis of cell surface expression from three independent experiments demonstrated 100% expression of HEPCAM in clone FaDu-CC-#l21ϩ compared with untreated wild-type cells, whereas all knock-out clones lacked expression of HEPCAM (Fig. 7B). Expression of E-cadherin was assessed in parallel in all clones. Except for knock-out clone CC#L20, which dis- FIGURE 5. Endocytosis of HEPCAM in carcinoma cells. A, live Kyse520 esophageal carcinoma cells were incubated for 15 min with anti-HEPEX antibody, fixed, and permeabilized before staining with an Alexa488-linked secondary antibody (green fluorescence). Visualization of HEPCAM was conducted via laser scanning confocal microscopy. Besides membrane localization, HEPCAM appeared as speckles within the cytoplasm. B, live Kyse520 cells were treated as mentioned above, and HEPICD was stained with a specific antibody in combination with an Alexa648-linked secondary antibody (red fluorescence). Full-length HEPCAM was detected as yellow stained after overlay of HEPEX and HEPICD staining, and released HEPICD appeared as red speckles. C, live Kyse520 esophageal carcinoma cells were treated for 30 min with nicastrin (clathrin-independent) or chlorpromazine (clathrin-dependent) endocytosis inhibitors. Thereafter, HEPCAM was visualized as described in A. Shown are representative results from three independent experiments. played a 69% increase in E-cadherin expression, all clones were unchanged with respect to E-cadherin expression (Fig. 7,  A and B). Complete lack of HEPCAM protein was further confirmed in whole cell lysates (Fig. 7C). Analysis of epcam transcription using real time quantitative RT-PCR demonstrated slightly increased mRNA levels in FaDu-CC#L21ϩ cells, complete lack in clones FaDu-CC#L20 and #H4, and wild-type level of expression in clone FaDu-CC#L13 (Fig. 7D). Because HEP-CAM protein was lacking in clone #L13, point mutations within and/or mutations located 3Ј of the amplified mRNA segment are probable.
Cell-matrix adhesion of FaDu wild-type and EpCAM-positive and -negative CRISPR-Cas9 clones was assessed. Each cell clone was cultured on Matrigel matrix-coated 6-well plates. Alternatively, wild-type or knock-out FaDu cells were plated in 6-well plates and calcein AM-stained FaDu wild type and EpCAM-positive and -negative CRISPR-Cas9 clones were plated on top of these cell layers (Fig. 7E). Neither cell-matrix nor cell-cell adhesion was affected by cellular knock-out of HEPCAM in FaDu carcinoma cells (Fig. 7F). Withdrawal of calcium reduced overall adhesion of FaDu cells from 25-35% down to 5-13%, but cellular knock-out of HEPCAM also had no measurable impact on cell-matrix and cell-cell adhesion (Fig. 7G). Complementation of EpCAM knock-out upon increased activity of E-cadherin appears unlikely given the unchanged E-cadherin expression with the exception of clone  FaDu-CC#L20. Hence, adhesion was not affected by EpCAM knock-out under the assay conditions.
To exclude potential cell type-specific and CRISPR-Cas9related effects, HEPCAM-positive Kyse30 esophageal carcinoma cell lines were stably transfected with control and HEPCAM short hairpin RNA (shRNA) expression vectors. HEPCAM mRNA expression was reduced by 90% in EpCAM-shRNA versus control-shRNA Kyse30 clones (Fig. 8A). Calcein AM-stained control and knockdown clones were plated on untreated (plastic), gelatin-, fibronectin-, and Matrigel matrix-coated 6-well plates. Cell-matrix adhesion was evaluated as calcein AM fluorescence after 2 h. Knockdown of HEPCAM had no effect on adhesion of Kyse30 cells to plastic, gelatin, fibronectin, and Matrigel (Fig. 8B). Control and knockdown clones were then cultured as monolayers in 6-well plates and calcein AM-stained control and knockdown cells plated on these cell layers. Cell-cell adhesion was unaffected by expression levels of HEPCAM (Fig. 8C). Because shRNA-mediated knockdown includes stable selection of cells, effects of short term knockdown of HEPCAM via transient siRNA transfections were addressed. Kyse30 esophageal carcinoma cells were transiently transfected with control and HEPCAM-specific siRNA. After 24 h, cell surface levels of HEPCAM were analyzed upon flow cytometry and revealed a 70% reduction in average (Fig. 8D). Reduction of HEPCAM resulted in a subtle up-regulation of adhesion to plastic, gelatin, fibronectin, and Matrigel but did not affect cell-cell adhesion (Fig. 8, E and F), despite increased migration of Kyse30 cells upon knockdown of HEPCAM (16). In line with the observed lack of impact of HEPCAM reduction on cell adhesion, Kyse30 cells overexpressing HEPCAM-YFP were not distinguishable from control YFP transfectants with respect to cell-matrix and cell-cell adhesion (Fig. 8, G and H). Finally, naturally occurring variants of Kyse520 cells, which differ in HEPCAM expression and migration capacity (16), were analyzed. Kyse520 high cells displayed increased HEPCAM expression by 1 order of magnitude as compared with Kyse520 low cells (Fig. 8I). A slight reduction in adhesion of Kyse520 low cells to plastic and fibronectin was measured; however, differences were not significant and did not apply to adhesion to gelatin (Fig. 8J). Adhesion of both subclones to Kyse520 low cells was enhanced as compared with Kyse520 high cells, again speaking against a role for HEPCAM as a major active cell adhesion molecule (Fig. 8K).
Adhesive Properties of Recombinant EpEX-Finally, we studied possible HEPEX interactions on the single molecule level using AFM. A flexible AFM cantilever with a sharp tip (inner diameter ϳ20 nm) and mica sheets were functionalized with recombinant HEPEX-Fc via a flexible polyethylene glycol spacer (Fig. 9A). The tip was repeatedly brought in contact with the mica and retracted again. Deflections of the cantilever were monitored by the use of a laser beam and were converted into forces acting on the cantilever. The resulting force curves of each approach/retract cycle can be screened for specific binding events between the molecules attached to tip and mica (44). Testing HEPEX-Fc-functionalized tips and mica sheets with a pulling speed of 0.1 m/s and a contact time of 0.1s resulted in binding events detected in less than 5% of all force curves, which was only slightly higher than the negative control in which BSA was coupled to the tip and mica (Fig. 9B). Importantly, an antibody targeting the far N terminus of HEPEX did not further reduce the number of interaction events. Similar results were observed with longer contact times and higher pulling speeds (data not shown). As a positive control, we tested a cadherin-type adhesion molecule, DSG3, which is crucial for keratinocyte cohesion (45). Similar as shown previously (46), binding events occurred in roughly 15% of all force curves and were specifically reduced by an antibody targeting the adhesive part of the extracellular domain to levels we observe when mea-suringHEPEXinteractions.Althoughthesesinglemoleculemeasurements do not entirely rule out an adhesive function of EpCAM, they support our cell-based experiments and demonstrate that EpCAM does not show the behavior of a typical cell-cell adhesion molecule.

Discussion
Human EPCAM is a transmembrane protein that is frequently and highly expressed in numerous carcinoma entities (47). Based on the availability of the large extracellular domain for therapeutic antibodies, HEPCAM is a long-known target for cancer therapies (48 -50). More recently, Epcam was recognized as a marker for pluripotent stem cells in mice and humans (51,52) and for tissue stem cells (53,54). Frequent overexpression of HEPCAM was explained by its ability to foster proliferation and oncogenic behavior (14,23), which is based on regulated intramembrane proteolysis and formation of a signaling intracellular domain termed hEpICD (29). Besides regulating expression of central molecular switches of proliferation such as cyclin D1 (30), HEPICD is involved in embryonic stem cell pluripotency and differentiation (55). Knowledge of the exact amino acids involved in cleavage of HEPCAM was lacking, and thus, the spectrum and precise composition of HEPCAM fragments remained unknown.
Studies of cleavage of murine Epcam demonstrated variable cleavages and degradation of the protein (31). To delineate exact cleavage sites and study their availability, we applied biochemical and mass spectrometry analysis on recombinant HEPCAM variants in combination with structure-based approaches. Three major extracellular and five major intramembrane cleavage sites, generating numerous different cleavage products, have been identified. Extracellular cleavage primarily occurs at two ␣-sites by ADAM proteases and at one ␤-site by BACE1. This is reminiscent of APP, which represents the central pathogenic agent in Alzheimer disease (27,56,57). APP can either be cleaved in a nonamyloidogenic pathway by proteases of the ADAM family or in an amyloidogenic pathway by BACE1 (27). It must be noted that both cleavage processes involve different subcellular compartments. ␣-Sheddases are functional at plasma membranes and neutral pH, whereas BACE1 is located intracellularly in membranes of the trans-Golgi network and endosomes, because of its acidic pH optimum for enzymatic activity (41). These spatial restrictions are circumvented by clathrin-dependent endocytosis of HEPCAM in cancer cells. Endocytosis of HEPCAM was assumed on the basis of efficacious delivery of toxins into cancer cells via conjugation to EPCAM-specific antibodies (58), but formal proof was lacking. In support of endocytosis, SILAC-based search for interactors of murine EpCAM yielded FLOTILIN, clathrin, RAB proteins, and the clathrin-dependent cargo adaptor DAB2 as potential intracellular ligands (data not shown). Balzar et al. (59) claimed the presence of an internalization NPXY motif within the cytoplasmic domain of EpCAM. This motif, which is also known as coated-pit internalization signal necessary for interaction with clathrin, is however not present in published sequences of human EPCAM. Nonetheless, endocytosis in combination with proteolytic cleavage by BACE1 and degradation of HEPICD by proteasome represents a newly described means to regulate cell surface availability and functionality of HEPCAM. For example, onset of migration was accompanied by significant increase of intracellular speckles of HEPCAM (16) along with de novo expression of the EMT marker vimentin (Fig. 4B). It is conceivable that HEPCAM becomes actively retrieved from membranes of tumor cells via endocytosis and BACE1/proteasome action to warrant or even induce a mesenchymal shift in phenotype. Comparable erasure of Epcam was observed during differentiation processes of embryonic stem cells (51,52) and might be the result of transcriptional polycomb silencing (55) as well as post-translational degradation.
Regulated intramembrane proteolysis of mouse Epcam at pH 4.5 resulted in complete cleavage of full-length protein, which was not the case at pH 7 (31). Although cleavage at pH 4.5 appeared total, cleavage at pH 7 affected ϳ20% of molecules, which is in line with results presented here for human EpCAM. One possible explanation for these differences in cleavage rates comes from mapping of cleavage sites onto three-dimensional structures of HEPCAM. Both ␣-sites are located in a groove formed by two monomers in their native states, i.e. a cis-dimer (34). Cleavage of HEPCAM by cathepsin L within the thyroglobulin domain was observed in solution, even though the corresponding site appears inaccessible in cis-dimers (34). Possibly, monomers occur at a certain rate and allow for cleavage, or accessibility to ␣-sites in cis-dimers at the membrane might require conformational changes induced by external yet unknown ligands of HEPCAM or by sheddases themselves. In contrast, the acidic milieu of endo-and lysosomes warrants disruption of cis-dimers (34,60) and provides BACE1 with a fully accessible cleavage site in monomers. Consequently, we suggest that cleavage at the plasma membrane affects only a proportion of EpCAM molecules and might reveal dependent on transient conformational change through ligand(s), whereas endocytosis and BACE1 cleavage of monomers will result in complete cleavage. ␥-Secretase cleavage sites are all accessible even in the dimeric state of HEPCAM, which is in line with the highly efficient processing of CTF fragments to HEPICD.
This study reports for the first time on the existence of A␤-like fragments of HEPCAM. Accordingly, function(s) of HEPCAM A␤-like fragments are as yet unexplored and clearly deserve(s) future attention. In further analogy to the amyloidogenic pathway of APP, endocytosis and processing of HEPCAM by BACE1 might have an impact on normal and pathophysiological conditions. Initially, EpCAM received its name from a postulated function i the formation of homophilic cell-cell adhesions (3,4). These seminal results were obtained after ectopic expression in murine fibroblasts and L153S mammary carcinoma cells. Upon transfection of HEPCAM into these cells, which lacked expression of typical adhesion molecules such as E-cadherin, morphological changes with increasing intercellular contact and aggregation of cells were observed (4). However, when studied in carcinoma cells bearing an epithelial phenotype and expressing classical cadherins, HEPCAM paradoxically displayed opposite effects and weakened adhesion mediated by classical cell adhesion molecules (5). Knock-out mice are likewise inconsistent with respect to function(s) of mEpcam in normal cells. Whereas the first published knock-out reported on embryonic lethality (6), following two publications, described a perinatal lethality because of severe intestinal leakage, which resembled congenital tufting enteropathy (7,8). Here again, even similar phenotypes could not be assigned to a unified molecular mechanism. Guerra et al. (7) claimed malformation of adherens junctions in enteric mucosa, which was based on mislocalization of E-cad- herin and ␤-catenin. Involvement in adherens junctions and E-cadherin integrity was supported by Epcam knock-out models in zebrafish and claw frog (10,11,13). In contrast, Lei et al. (8) claimed normal localization of E-cadherin and ␤-catenin and formation of adherens junctions but loss of tight junctions due to the reduction of claudin proteins, with special emphasis on claudin 7, a functional interaction partner of Epcam in rat carcinoma cells (61,62). Hence, the role of Epcam in cell adhesion appears rather indirect, and consensus on molecular mechanisms was not achieved. Although HEPCAM is commonly introduced as a cell adhesion molecule (CAM), it has no structural similarity to any CAM family, and its direct role in adhesion of cancer cells remained unchallenged over decades. We show that neither inhibition of cleavage by ADAMs, BACE1, and ␥-secretase proteases nor cellular knock-out or knockdown of HEPCAM in carcinoma cell lines affected the cell-matrix or cell-cell adhesion using standard calcein assays. In fact, neither homophilic nor a possible heterophilic adhesion was affected by cellular knock-out of HEPCAM. Lack of impact on cell-matrix is in accordance with published work (3,4), but the lack of impact on cell-cell adhesion is clearly contradictory to the standard perception of HEPCAM as a CAM. Proof of a cell-cell adhesion function of HEPCAM originates from work in mesenchymal cells (4), which is obviously not its natural cell type, although claudin-7 was reported to recruit Epcam to ADAM protease and ␥-secretase to foster migration and mesenchymal transition via HEPICD formation (63). Inhibition of cleavage was performed throughout a 24-h period before measurement of adhesion and resulted in almost complete lack of HEPICD formation. Hence, neither reduced degradation of HEPCAM nor lack of signaling through HEPICD influenced adhesion of carcinoma cells. Furthermore, HEPCAM is frequently down-regulated in carcinoma cells undergoing EMT (20), questioning the validity of roles deduced from its ectopic expression in mesenchymal cells. Effects of HEPCAM on tight or adherens junctions observed in mice could be specific to normal epithelia or remain below the detection limit in carcinoma cells expressing high levels of classical cadherins. Withdrawal of calcium resulted in substantially decreased cellmatrix and cell-cell adhesion of carcinoma cells, however, without any influence of HEPCAM, which would have been expected from a calcium-independent adhesion molecule. Finally, atomic force microscopy experiments were used to monitor protein-protein interactions of the extracellular domain HEPEX. Although the known adhesion protein desmoglein 3 displayed clear binding properties, HEPEX interactions appeared only marginally superior to BSA controls. Hence, EpCAM interactions and functions in adhesion appear either particularly weak or are entirely lacking.
Eventually, we could not detect any difference in adhesion of carcinoma cells whatsoever with the technologies applied, and thus, we have to conclude that the role(s) of HEPCAM as a direct homophilic adhesion molecule in carcinoma cells are questionable. Therefore, we propose to also question the name of EpCAM as a cell adhesion molecule. We suggest to the scientific community to evade this controversy by renaming as "epithelial cell-activating molecule" based on its demonstrated functions in proliferation and differentiation.