A Complex of EpCAM, Claudin-7, CD44 Variant Isoforms, and Tetraspanins Promotes Colorectal Cancer Progression

Colorectal cancer is the third most frequent cancer in the Western world (1). Due to improved surgical techniques, including the excision of liver metastases, the 5-year survival rate has increased during the last two decades (2). Nonetheless, colorectal cancer still ranks fourth in cancer mortality (3). It relies on early metastatic spread accompanied by radiation and chemotherapy resistance (4, 5), the reasons for which are still not fully explored.

We recently noted that metastasizing gastrointestinal tumors of the rat frequently express a complex that includes the tetraspanin D6.1A (human: CO-029), EpCAM, claudin-7, and CD44 variant isoform v6 (CD44v6; refs. 6, 7). We assumed that this complex, rather than the individual molecules, might promote tumor progression. The following observations supported this hypothesis: (a) the CD44v6/CO-029/EpCAM/claudin-7 complex is located in tetraspanin-enriched membrane microdomains (TEM) that serve as a signaling platform (8, 9) and (b) the TEM-localized complex supports matrix adhesion, agglomeration, and apoptosis resistance distinctly to the individual components of the complex (6).

CD44, particularly CD44v6, expression has been described to serve in a wide range of tumor entities as a valuable marker for diagnosis, prognosis, and, in selected instances, also therapy (reviewed in refs. 10-12). However, particularly in colorectal cancer, contradictory findings have been reported that range from a positive to an inverse correlation between CD44v6 expression and prognosis in colorectal cancer (13-20). The same conflicting data were found for expression of CD44s (CD44 standard isoforms; refs. 21-23).

High CO-029 expression in gastrointestinal cancer is associated with a poor prognosis (24-29). The underlying mechanism has not yet been elucidated. It has been suggested that CO-029 is involved in cell proliferation and differentiation (27, 30-33). Tetraspanins mostly function as molecular facilitators that form complexes between themselves, additional transmembrane and signal-transducing molecules (8, 31-35). Although integrins are the preferential transmembrane partners (36-38), EpCAM and CD44 also associate with tetraspanins (6, 7, 9, 31, 39). With few exceptions, tetraspanin associations are of low stringency and are not due to a direct protein-protein interaction. Instead, associations are facilitated by localization in the TEM (reviewed in ref. 8).

EpCAM, a panepithelial, homophilic cell-cell adhesion molecule, is frequently up-regulated in colorectal cancer (reviewed in refs. 40, 41) and EpCAM-specific antibodies are used as a supportive drug in cancer therapy (42, 43). Although cell-cell adhesion should rather prohibit than support tumor cell dissemination (44, 45), there is evidence that EpCAM interferes with E-cadherin expression by disrupting the link between α-catenin and filamentous actin (46). EpCAM has also been shown to support cell motility (47-49) and overexpression can induce up-regulation of the proto-oncogene c-myc, thus supporting cell proliferation (50).

The fourth complex component, claudin-7, belongs to a family of proteins that span the membrane four times, but are structurally unrelated to the tetraspanin family. Claudins are essential components of tight junctions (51). According to their importance in maintaining cell polarity, loss of claudin expression in malignancy was frequently found to be associated with poor prognosis (52). This accounts for claudin-7 in breast (53, 54) and esophageal cancer (55). However, up-regulated claudin-7 expression was seen in renal (56), cervical (57), and gastric cancer (58, 59). This may be a consequence of claudin-7 being not only located in tight junctions, but also at the basolateral membrane and/or in vesicles (7, 60-62). In the gastrointestinal tract, basolaterally localized claudin-7 also colocalizes with EpCAM (7). The functional activities of claudins outside tight junctions are largely unknown.

Taken together, a complex of four molecules overexpressed in metastasizing rat tumor lines includes one molecule (CD44v6), of which expression by itself revealed conflicting results on colorectal cancer progression, two cell-cell adhesion molecules (EpCAM and claudin-7) that should hamper metastasis formation, and a tetraspanin (CO-029) known to act in concert with associated molecules. Thus, it becomes likely that the complex, but not the individual molecules, promote cancer progression. Here, we asked whether human colorectal cancer expresses this complex and whether expression of the complex correlates with tumor progression.

EpCAM and CO-029 expression have been associated with colorectal cancer progression (24, 26, 40, 41). However, opposing effects have been reported for CD44v6 (14-16, 19). Furthermore, the mechanism underlying the unexpected involvement of the cell-cell adhesion molecule EpCAM in tumor progression, as well as that of the tetraspanin CO-029, are unknown. We noted in a rat tumor model that (a) CD44v6 forms a TEM-located complex with D6.1A, EpCAM, and claudin-7 (6, 7), and that (b) complexes expressing rat tumor lines display functional activities different from tumor lines that express only individual components (6). Thus, we hypothesized that only coexpression of these molecules promotes colorectal cancer progression.

Snap-frozen tissues of primary colorectal cancer, normal colonic mucosa (n = 104), liver metastases, normal liver (n = 66), as well as peritoneal and adrenal metastases (n = 5), polyps and benign tumors (n = 6), were stained with anti-CD44v6, anti-EpCAM, anti–claudin-7, and anti–CO-029. Stainings were done using half the dose of the primary and the secondary antibodies required for optimal staining of nontransformed tissues. The reduction in the antibody concentration was necessary to circumvent unspecific staining of normal colonic mucosa and mucus-secreting tumor tissue. As a consequence, only tissues that expressed the molecules at high levels were stained, i.e., low level expression in normal colonic mucosa did not, or very weakly, become visible (Fig. 1A). Accordingly, we missed low-level expression of the four molecules in colorectal cancer and liver metastases. Nonetheless, distinct to strong staining of EpCAM, claudin-7, CO-029, and CD44v6 was seen in 73%, 53%, 81%, and 36% of colorectal cancers versus 10%, 3%, 1%, and 5% of corresponding tumor-free colorectal tissues, respectively. The same antibodies stained 74%, 38%, 75%, and 20% of liver metastases, and 8%, 0%, 6%, and 0% of tumor-free liver tissue from the same patients, respectively (Table 1; Fig. 1B). Distinct to strong staining of EpCAM, claudin-7, CO-029, and CD44v6 was seen in four of five, three of five, four of five, and two of five peritoneal metastases, respectively. Polyps and benign tumors (6) were distinctly to strongly stained by anti-EpCAM, weakly to distinctly stained by anti-CD44v6 and anti–claudin-7, and least frequently stained (three of six) by anti–CO-029. High expression of the four molecules was also seen in pancreatic cancer, but not in renal cell carcinoma (data not shown).

Up-regulated EpCAM and CD44v6 expression in benign and malignant tumors could indicate an association with proliferation rather than with tumor progression. Furthermore, using higher antibody concentrations, low-level EpCAM, CO-029, and CD44v6 expression was frequently observed in colorectal tissue (data not shown). Thus, expression of the molecules is increased, but not induced de novo in colorectal cancers.

Expression of the four molecules did not correlate with tumor-node-metastasis staging and tumor grading, with the exception of CD44v6, the expression of which was slightly down-regulated in grade 4 tumors (Table 2). Also, expression of the four molecules did not differ significantly between primary tumors and metastases (see Table 1). This was further supported by analyzing EpCAM, claudin-7, CO-029, and CD44v6 expression in 10 tissue samples, in which the primary tumor and metastases were concomitantly excised. Comparable expression levels of the four molecules were seen in primary tumors and metastases. Univariate analysis revealed a mean difference in expression levels of 0.1 for EpCAM (95% confidence limits, −0.126 to 0.326), 0.2 for claudin-7 (95% confidence limits, −0.050 to 0.450), 0.1 for CO-029 (95% confidence limits, −0.126 to 0.326), and 0.15 for CD44v6 (95% confidence limits, −0.091 to 0.391). Although the number of samples is low, the very narrow confidence limits for all four markers suggests that EpCAM, claudin-7, CO-029, and CD44v6 expression does not become up-regulated or down-regulated during colorectal cancer progression. Notably, too, the only pair of tissue samples that did not express claudin-7 was derived from a primary tumor and a peritoneal metastasis; The remaining nine metastatic tissues were liver-derived (Fig. 1C). Furthermore, the location of the primary tumor (colon versus rectum) had no effect on the expression of the four molecules (Table 3). With the exception of CD44v6 expression in liver metastases, expression of the individual molecules also did not correlate with disease-free survival (DFS; Table 4).

Thus, we asked whether the coexpression of these markers might provide a diagnostic and/or prognostic variable. Analyzing the coexpression of pairs of these molecules significantly strengthened the difference between tumor/metastatic tissue and normal tissue, e.g., coexpression of EpCAM and claudin-7 was detected in 89% of primary tumors versus 13% of normal colon, and in 92% of liver metastases versus 6% of normal liver (data not shown). Coexpression of EpCAM, claudin-7, CO-029, and CD44v6 was seen in 76% of primary tumors and 3% of normal colon and in 64% of liver metastases and 0% of normal liver (Fig. 1D). Coexpression of the molecules also did not correlate with tumor staging and grading (data not shown), but significantly correlated with DFS (Table 5).

The high rate of CO-029 and EpCAM expression in colorectal cancer is well known. CD44v6 expression, although weaker than expression of the former two molecules, was still up-regulated in colorectal cancer. The use of a reduced amount of antibody probably strengthened the difference, with low-level expression in normal colonic tissue. Expression of claudin-7 in colorectal cancer, which to our knowledge had not yet been explored, was strongly increased as compared with normal tissue. Because coexpression of EpCAM, claudin-7, CO-029, and CD44v6, but not expression of the individual molecules, correlated with DFS, we speculated that the molecules may interact, and that their concerted activity promotes tumor progression. To support this hypothesis, complex formation was evaluated.

Flow cytometry of eight colorectal cancer lines revealed an expression profile resembling the one observed in freshly harvested tumor tissue. Four colorectal tumor lines (HT29, WIDR, SW948, and Lovo) expressed all four molecules, one line revealed low expression of CO-029 (SW480), and one line revealed low expression of EpCAM (Colo320DM). The latter and two additional lines (Colo205 and SW707) expressed CD44v6 at a very low level (Fig. 2A and B). The selective deficits in CD44v6 expression (Colo205 and SW707) and in CO-029 expression (SW480) allowed the evaluation of the contribution of these molecules in the supposed molecular colocalization/association.

EpCAM associated with claudin-7 irrespective of whether the lines expressed CD44v6 and/or CO-029. In lines expressing all four molecules (HT29 and SW948), the molecules colocalized and low level expression of CD44v6 (Lovo) sufficed for colocalization. In lines expressing very low levels of CD44v6 (Colo205 and SW707), CO-029 hardly or did not colocalize with claudin-7 (EpCAM). In lines expressing very low levels of CO-029 (SW480), colocalization of CD44v6 with EpCAM and claudin-7 was very weak. Notably, it is particularly the CO-029 tetraspanin that supports colocalization. Colocalization of the tetraspanin CD9, which also associates with CD44v (6), was included as a “specificity” control. There was no correlation between the colocalization of CD9 with CD44v6 versus EpCAM, claudin-7, or CO-029 (Fig. 3).

Because the colocalization of EpCAM/claudin-7 with CO-029 and CD44v6 was weak, if either CD44v6 or CO-029 expression was very low/negative, it could be hypothesized that both CD44v6 and the tetraspanin molecule were involved in complex formation. Alternatively, although not mutually exclusive, colocalization of the molecules, even at low expression levels, could be a result of enrichment in membrane microdomains that facilitates proximity.

Using coimmunoprecipitation, we first evaluated whether and which of the four molecules associate. We had previously shown that in the rat pancreatic adenocarcinoma line, BSp73ASML, coimmunoprecipitation of EpCAM and claudin-7 was stable after lysis with the strong detergent Triton X-100 (7). However, coimmunoprecipitation of EpCAM, claudin-7, and CD44v6 with D6.1A was only observed after lysis with mild detergents. It should be noted that the mild detergent did not suffice for complete extraction of the proteins, i.e., a considerable amount was recovered in the 20,000 × g pellet of the lysates (Fig. 4A). Similar to BSp73ASML lysates, EpCAM, claudin-7, CO-029, and CD44v6 coimmunoprecipitated in SW948 lysates, the cells expressing the four molecules at high levels (Fig. 4B). Even in Lovo and Colo205 lysates (low CD44v6 expression), EpCAM precipitates contained claudin-7 and CO-029. Although claudin-7 coimmunoprecipitated with EpCAM independent of CO-029 and CD44v6 expression, CD44v6 (data not shown) and CO-029 did not coimmunoprecipitate with EpCAM in the absence of claudin-7 (subclone of SW707; Fig. 4C).

These findings confirmed that claudin-7 directly associates with EpCAM and suggested that claudin-7–associated EpCAM only, or at least predominantly, becomes recruited towards CO-029 and CD44v6.

To corroborate the hypothesis, HEK293T cells were transfected with rat EpCAM, claudin-7, and D6.1A cDNA, or combinations thereof. EpCAM coimmunoprecipitated with claudin-7 in HEK293T cells expressing EpCAM and claudin-7 (Fig. 5A). Instead, D6.1A coimmunoprecipitated with EpCAM only in HEK293T cells that expressed claudin-7 (Fig. 5B). Finally, D6.1A also coimmunoprecipitated with claudin-7 in HEK293T cells that did not express EpCAM (Fig. 5C). Thus, claudin-7 is most important for complex formation between EpCAM, D6.1A/CO-029, and CD44v6.

Tetraspanins form a web in glycolipid-enriched membrane domains (TEM), and CD44v6 associates with D6.1A preferentially in TEM (8). Because EpCAM associates with CO-029 and CD44v6 only in the presence of claudin-7, it is likely that claudin-7 is required for the recruitment of EpCAM into TEM. The hypothesis was controlled by sucrose density gradient centrifugation of transiently transfected HEK293T cells. EpCAM hardly localized in TEM. Claudin-7 was recovered in TEM and non-TEM fractions. In HEK293T cells that coexpressed EpCAM and claudin-7, EpCAM was partly recruited into TEM. A considerable amount of D6.1A is recovered in TEM. However, in HEK293T cells that coexpressed EpCAM and D6.1A, EpCAM mostly remained in the dense fraction. On the contrary, a strong shift of EpCAM towards the TEM fraction was noted in HEK293T cells that coexpress EpCAM, claudin-7, and D6.1A (Fig. 6A). Thus, EpCAM becomes recruited into TEM mostly via claudin-7. Our interpretation was confirmed using human colorectal cancer lines that expressed all four molecules (SW948) or CO-029 at a very low level (SW480), or were missing claudin-7 (SW707 subline). EpCAM was recovered in TEM in SW948 and SW480 lysates, but not in the SW707 subline lysate (Fig. 6B). These findings confirm that EpCAM becomes incorporated into a TEM-located complex with CO-029 and CD44v6 via its claudin-7 association.

EpCAM, claudin-7, CO-029, and CD44v6 coexpression correlated with the early appearance of liver metastasis in colorectal cancer. To reassure the in vivo relevance of complex formation, coimmunoprecipitation of EpCAM, claudin-7, CO-029, and CD44v6 were evaluated in freshly harvested tissues of colorectal cancer and liver metastases.

In normal colonic mucosa, claudin-7 and CO-029 coimmunoprecipitated, albeit weakly with EpCAM. Accordingly, anti–CO-029 coimmunoprecipitated EpCAM and claudin-7. CD44v6 was not detected. Instead, in the tumor tissue, claudin-7, CO-029, and CD44v6 coimmunoprecipitated with EpCAM, anti–CO-029 precipitated EpCAM, claudin-7 and CD44v6, and anti-CD44v6 precipitated EpCAM, claudin-7 and CO-029. In the absence of claudin-7, anti-CD44v6 did not precipitate EpCAM or CO-029 and coimmunoprecipitation of CO-029 with EpCAM was weak. In the absence of claudin-7 and CD44v6, anti-EpCAM hardly precipitated CO-029 and vice versa (Fig. 7). Thus, complex formation between EpCAM, claudin-7, CO-029, and CD44v6 was observed in vivo and coexpression of the four molecules was invariably accompanied by complex formation.

The functional consequences and the underlying molecular mechanisms of complex formation between EpCAM, claudin-7, CO-029, and CD44v6 in colorectal cancer progression has not yet been explored. However, because coexpression was more frequently observed in primary colorectal cancer than in liver metastases, and because only coexpression in liver metastases, but not in the primary tumor, correlated with DFS, it was tempting to speculate that complex formation might rather provide a survival advantage of isolated tumor cells than directly support the process of tumor cell dissemination. In fact, a CD44v-knockdown in the BSp73ASML model was accompanied by a striking decrease in apoptosis resistance.⁶

Chemoresistance was evaluated by Annexin/propidium iodide uptake after culturing colorectal cancer lines for 1 to 3 days in the presence of 25 μg/mL of cisplatin. Less than 15% of HT29 and SW498 cells were stained by Annexin/propidium iodide after 24 h and only ∼30% of these cells were apoptotic after 72 h. The death rate was slightly increased in Lovo and SW707 cells that expressed CD44v6 at a low or very low level. Instead, >80% of SW480 cells that expressed CO-029 at a very low level were apoptotic after 72 h. 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide staining after culturing cells for 72 h in the presence of 0.3 to 35 μg of cisplatin confirmed the high apoptosis resistance of HT29 and SW948 cells that expressed all four molecules at a high level, strongly reduced apoptosis resistance in cells expressing CD44v6 at a very low level (Colo205 and SW707), and a further reduction in apoptosis resistance of SW480 cells (CO-029 low; Fig. 8A). To further support the importance of the EpCAM/claudin-7/CO-029/CD44v6 complex in apoptosis resistance, SW948 and HT29 cells were transfected with EpCAM or claudin-7 short interfering RNA. The transfection rate was in the range of 80% to 85% (data not shown). The intensity of EpCAM and claudin-7 expression was reduced by a factor of 10 to 15 (Fig. 8B). When EpCAM or claudin-7 RNAi-transfected SW948 and HT29 cells were cisplatin-treated, apoptosis resistance was significantly reduced, with claudin-7 down-regulation exerting a stronger effect than EpCAM down-regulation (Fig. 8C). Finally, the importance of the TEM localization of the complex was controlled by methyl-β-cyclodextrin treatment, which suffices to destroy glycolipid-enriched membrane microdomains by partial cholesterol depletion (35). Methyl-β-cyclodextrin–treated cells were cultured for 24 h in the presence of 25 μg/mL of cisplatin. Although apoptosis of Colo205, SW707, and SW480 cells, which do not express the complex (or at a very low level), increased by a factor of 1.2 to 1.5, apoptosis of HT29 and SW948 cells increased by a factor of 4 (Fig. 8D). These findings point towards a significant contribution of the TEM-localized EpCAM/claudin-7/CO-029/CD44v6 complex to apoptosis resistance.

Taken together, colorectal cancer frequently expresses a complex of EpCAM, claudin-7, CO-029, and CD44v6 that is located in TEMs. Complex formation and recruitment of EpCAM into TEMs require claudin-7. The complex, but not the individual molecules, promote colorectal cancer progression that may involve increased apoptosis resistance.

High-level EpCAM and CO-029 expression has been associated with colorectal cancer progression (24, 26, 40, 41). The contribution of CD44v6 is controversially disputed (14, 15, 17, 19). Attempting to clarify the latter controversy, we hypothesized the possibly that rather than individual CD44v6 molecules, CD44v6 in concert with associated molecules, might support colorectal cancer progression. We noted in a rat tumor model that CD44v6, CO-029, EpCAM, and claudin-7 form a complex in TEM, membrane microdomains enriched in glycolipids and tetraspanins (8). A corresponding complex was detected in human colorectal cancer, and expression of the complex but not of the individual molecules, inversely correlates with DFS. Thus, human colorectal cancer progression may be supported by the TEM-located complex rather than by the individual molecules.

Before discussing our interpretation, we should briefly comment on the immunohistologic staining protocol. Using the recommended dose of antibodies, high background staining of shock-frozen colorectal cancer and colonic mucosa sections was observed. Reducing the concentration of the primary and the secondary antibodies provided a window in which background staining was eliminated. However, weak staining of colonic mucosa with EpCAM and CD44v6 was lost (40, 63). Nonetheless, the vast majority of colorectal cancers and liver metastases were still strongly stained by anti-EpCAM, anti–claudin-7, and anti–CO-029, which—at least for EpCAM—is in line with the suggested 100-fold to 1,000-fold up-regulation of expression in cancer tissue (40). Staining of CD44v6 was weak with few exceptions. This could be indicative of a lower affinity of the antibody or, more likely, a less pronounced increase in CD44v6 expression.

We also want to comment on the mild lysis conditions that were required to detect interactions between CO-029, CD44v6, and EpCAM/claudin-7 because, exclusively, the EpCAM/claudin-7 interaction represents a direct protein-protein interaction that was not destroyed by stringent detergents. After lysis in mild detergents and 20,000 × g centrifugation, a considerable amount of the proteins were recovered in the pellet, which implies that the proteins were either incompletely solubilized or may be contained in larger complexes. Although the recovery of the proteins in the 20,000 × g pellet prohibits quantitation of the relative amount of complex-associated versus free proteins, ongoing studies, in which cells were transfected with increasing amounts of cDNA, indicate a highly efficient association of claudin-7 with EpCAM. Quantitation of the relative amount of complex-associated protein for all four components has not yet been done. This aspect requires further evaluation and should be kept in mind, although it has no effect on our interpretation of complex-mediated activities.

CD44v6 was originally defined as a metastasis-associated molecule (64) which is expressed by basal cryptic cells in the gut (63, 65). Due to the reduced amounts of antibody, the low expression on cryptic cells in the gut was not detected. Nonetheless, CD44v6 expression was seen in benign polyps as well as in the majority of colorectal cancer tissues and liver metastases, which implies up-regulated expression in those tissues. However, fewer metastatic than primary tumor tissues were stained. The latter feature, as well as staining of polyps, argues against CD44v6 promoting colorectal cancer progression by itself. The findings are in line with several reports on CD44v6 being associated with dysplasia, but not colorectal cancer progression (15-17, 19, 63). On the other hand, transfection of nonmetastasizing tumor cells with CD44v6 cDNA supports metastatic spread (64, 66), and a CD44v-knockdown in the highly metastatic BSp73ASML line that expresses CD44v4-v7, D6.1A, EpCAM, and claudin-7 at a high level, significantly impairs metastasis formation.⁶ These observations argue for a contribution of CD44v6 in gastrointestinal tumor progression. Complex formation of CD44v6 with CO-029, EpCAM, and claudin-7 could provide a possible explanation.

Here, we report for the first time that claudin-7 expression is up-regulated in colorectal cancer and liver metastasis. Our finding that claudin-7 expression, although not at a statistically significant level, is lower in grade 4 than grade 1 tumors, is also in line with reports on claudin-7 up-regulation in gastric cancer that becomes weaker in late stage tumors (57, 58).

Claudin-7 is not exclusively recovered in tight junctions (7, 60-62, 67) and has been suggested to serve different functions depending on its localization in defined membrane subdomains (61). However, those functions are not defined and it has not yet been reported that claudin-7 could be recovered in TEM and that it could recruit EpCAM into these microdomains, which obviously is of functional and clinical importance. Only expression of the TEM-localized EpCAM/CO-029/CD44v6/claudin-7 complex correlated inversely with DFS and a high level of apoptosis resistance. TEM are known as signaling platforms (8, 35, 36, 68). We have previously shown that the functional activities of the EpCAM/CO-029/CD44v6/claudin-7 complex, which cannot be ascribed to individual molecules, essentially depends on TEM localization (6). Taking the clinical importance of complex formation, it has become important to define the molecule(s) that initiate(s) complex formation. CO-029 associates with CD44v6 in TEM (6). However, EpCAM hardly colocalizes/coimmunoprecipitates with CO-029 or CD44v6 in the absence of claudin-7. Instead, claudin-7, which also coimmunoprecipitated with CO-029 and CD44v6 in cells not expressing EpCAM, interacts with EpCAM in a direct protein-protein interaction and recruits EpCAM into the TEM-located complex. Thus, complex formation is strongly supported by claudin-7.

The fact that claudin-7 suffices for complex formation does not imply that functional activity of the complex is also dominated by claudin-7, rather it seems that a crosstalk between the four molecules is required. A cell line with very low CO-029 expression revealed very low apoptosis resistance as compared with cell lines expressing the complex. A CD44v4-v7 knockdown is accompanied by a striking reduction in apoptosis resistance,⁶ and a transient knockdown of EpCAM, and more pronounced of claudin-7, was also accompanied by increased apoptosis susceptibility. Furthermore, apoptosis resistance depends on TEM localization. Dissolving TEM/the TEM-localized complex by methyl-β-cyclodextrin treatment led to a significant decrease in apoptosis resistance in tumor lines expressing the complex. TEM provide a signaling platform, in which tetraspanins (D6.1A/CO-029) function as a scaffold (8, 68). In addition, claudin-7 serine phosphorylation depends on TEM localization (6, 7), and CD44v6 phosphorylation varies depending on TEM localization (69). However, the direct link between apoptosis resistance and complex-mediated signal transduction, as well as additional, exclusively complex-mediated functions, remains to be elucidated.

Primary colorectal cancers as well as liver metastases frequently express EpCAM, claudin-7, CO-029, and CD44v6. Coexpression of the four molecules is inevitably accompanied by complex formation, which depends on claudin-7, its association with EpCAM, and the recruitment into TEM. Complex-mediated functions, at least apoptosis resistance, require all four components of the complex. Importantly, expression of the complex, but not the solitary expression of EpCAM, claudin-7, CO-029, or CD44v6, is accompanied by a poor prognosis with reduced DFS.

Colorectal cancer and surrounding normal colon tissue (n = 104), liver metastases, and normal liver (n = 66), peritoneal and adrenal gland metastases (n = 5), polyps, and benign tumors (n = 6) were collected during surgery from 171 patients. Both primary tumor and metastatic tissue was contemporarily collected from 10 patients. Tissue was snap-frozen in liquid nitrogen. Tumor tissue was derived from 53 female and 112 male patients with a mean age of 64.1 years, ranging from 35 to 88 years. Histologic type and tumor grading are shown in the supplement (Table S1). Informed consent on tissue collection was obtained from each patient. Tissue collection was approved by the University Ethics Review Board.

The long-term colorectal cancer lines Colo205, Colo320DM, HT29, Lovo, SW480, SW707, SW948, and WIDR were maintained in RPMI 1640/10% FCS/nonessential amino acids/10 mmol/L sodium pyruvate. The origin and description of these lines have recently been summarized (29). The metastasizing rat pancreatic adenocarcinoma line BSp73ASML (70) was cultured in RPMI 1640/10% FCS. HEK293 and HEK293T cells (71) were stably or transiently transfected with the rat cDNA of CD44v4-v7 (64), EpCAM (72), claudin-7 (7), D6.1A (30), or combinations thereof using the pcDNA3.1(+) vector that carries either the neomycin or hygromycin resistance cDNA. The cDNA inserts were controlled by sequencing. Expression was controlled by flow cytometry. Mock, EpCAM, claudin-7, and D6.1A-transfected HEK293 cells were maintained in RPMI/10% FCS/500 μg/mL G418 or 150 μg/mL of hygromycin. Short interfering RNA (1.25 mmol/L; Hs_TACSDT1_5_HP validated siRNA/EpCAM knockdown and Hs_CLDN7_2_HP siRNA/claudin-7 knockdown) was mixed with 0.25 μL of HiPerFect transfection reagent (Qiagen, GmbH) in tissue culture medium without serum for 5 to 10 min at room temperature and then dropwise added to 1 × 10⁵ tumor cells which were seeded into 96-well plates. Cells were incubated with the transfection complexes under their normal growth conditions for 16 h. Transfection efficiency and gene silencing were monitored after 16 h using the RNAi Human/Mouse Starter Kit (Qiagen). Confluent cultures were trypsinized and split.

The following antibodies were used: anti-human CD44v6 (vFF18; ref. 73), anti-panCD44 (25-32), anti-CD9 (MEM-62; ref. 74), anti–CO-029 (24), anti-EpCAM (HEA125; ref. 75), anti–claudin-7, which cross-reacts with rat claudin-7 (guinea pig antiserum, raised against the COOH-terminal domain; ref. 76), anti-TfR (transferrin receptor, OKT9); anti-rat CD44v6 (A2.6), anti-EpCAM (D5.7), anti-D6.1A (D6.1; ref. 77), anti-panCD44 (Ox50), and anti-TfR (Ox26). Hybridoma culture supernatants were purified by passage over protein G-Sepharose and, where indicated, were labeled with biotin or FITC. Anti–E-cadherin, dye-labeled secondary antibodies, and streptavidin were obtained commercially (BD/PharMingen and Dianova).

Tumor cells (1 × 10⁵) were stained according to routine protocols. Before staining, trypsinized cells were allowed to recover for 2 h at 37°C in RPMI 1640/10% FCS. For intracellular staining (claudin-7), cells were fixed and permeabilized. Apoptosis was evaluated by staining with Annexin-FITC and propidium iodide. Samples were analyzed using a FACSCalibur (BD).

Cryostat sections (5 μm) of snap-frozen tissue were fixed in chloroform/acetone (1:1) for 4 min, and nonspecific binding was blocked by an avidin-biotin blocking kit solution (Vector Laboratories) plus 2% bovine serum albumin and 2% serum corresponding to the origin of the primary antibody. Sections were stained with anti-EpCAM, anti-CD44v6, anti–CO-029, anti-CD9, and anti–claudin-7 (1 h at room temperature). Before staining with anti–claudin-7, tissue was fixed (4% paraformaldehyde, 30 min, 4°C) and permeabilized (0.1% Triton X-100, 4 min, 4°C). After washing, tissue was incubated with biotinylated secondary antibodies (30 min) and alkaline phosphatase conjugated avidin-biotin complex solution (5-20 min). Sections were counterstained with Mayer's hematoxylin. For negative controls, the primary antibody was replaced with normal mouse or guinea pig immunoglobulin. Staining was assessed semiquantitatively according to the following scores: no stained cells, −; weak staining of <25% of cells, ±; moderate staining of ∼50%, +; moderate to strong staining of >50%, ++, strong staining of >75%, +++.

In colocalization studies, tumor cells grown on glass slides were fixed and permeabilized as described above. After washing and blocking (0.2% gelatin and 0.5% bovine serum albumin in PBS), cells were incubated with the primary antibody (60 min, 4°C) and Cy2-labeled anti-mIgG (60 min, 4°C) or with dye-labeled (FITC or TxR) primary antibodies. Free binding sites were blocked by incubation with an excess of mIgG. Thereafter cells were incubated with anti–claudin-7 (1 h, 4°C), biotinylated anti–guinea pig IgG (1 h, 4°C) and Cy3-labeled streptavidin (1 h, 4°C). After washing, slides were mounted in Elvanol. Digitized images were generated using a Leica DMRBE microscope equipped with a SPOT CCD camera (Diagnostic Instruments Inc., Software SPOT2.1.2). Double-stained samples images were digitally overlaid.

Cells or lysates were labeled with 0.2 mg/mL of water-soluble Biotin-X-NHS (Merck Biosciences) in 25 mmol/L of HEPES (pH 7.2), 150 mmol/L of NaCl, 5 mmol/L of MgCl₂ (30 min, 4°C), followed by quenching with PBS/200 mmol/L of glycine (pH 7.2) for 15 min.

Homogenized tissues and cells were lysed (1 h, 4°C) in 4 mL of ice-cold lysis buffer [25 mmol/L HEPES, 150 mmol/L NaCl, 5 mmol/L MgCl₂ (pH 7.2)] containing 1% Brij98 or 1% CHAPS. Lysis buffers contained a protease inhibitor cocktail (Boehringer Mannheim) and 2 mmol/L of phenylmethylsulfonyl fluoride. After centrifugation (10 min, 20,000 × g), lysates (1 mL) were precleared by incubation with 1:10 volume Protein G-Sepharose and protease inhibitor cocktail (2 h, 4°C). Precleared lysates were incubated overnight at 4°C with 2 μg of antibody or control IgG. ProteinG-Sepharose was added for 4 h. Immune complexes were washed four to six times. Immunoprecipitated proteins were analyzed by SDS-PAGE, followed by Western blotting.

Lysates and immunoprecipitated proteins were resolved on 12% or 15% SDS-PAGE under nonreducing conditions. The proteins were transferred to Hybond enhanced chemiluminescence at 30 V overnight. After blocking (5% fat-free milk powder, 0.1% Tween 20), immunoblotting was done with the indicated (biotinylated) antibodies, followed by rabbit anti-mouse, anti-guinea pig, or streptavidin-HRP. Blots were developed with the enhanced chemiluminescence detection system. Densitometric analysis was done with NIH Image 1.60 software.

Cells were lysed as described above. After centrifugation (10 min, 20,000 × g) lysates were adjusted to 40% sucrose in HEPES buffer in a total volume of 4.5 mL. Lysates were layered on 1.3 mL of 50% sucrose and overlaid with 2.3 mL of 30%, 2.3 mL of 20%, and 1.3 mL of 5% sucrose. After centrifugation (200,000 × g, 16 h), 12 fractions (1 mL) were collected from the top of the tubes. Fractions were resuspended in lysis buffer.

Tumor cells (5 × 10⁴) were seeded in flat-bottomed 96-well plates. After settling, serial dilutions of cisplatin (starting concentration, 35 μg/mL) in RPMI/10% FCS were added. Alternatively, cells were incubated with 100 to 500 Gy and seeded thereafter in 96-well plates. Where indicated, cells were pretreated with methyl-β-cyclodextrin (10 mmol/L, 30 min, 37°C) to destroy glycolipid-enriched membrane microdomains. Cells were cultured for 24 to 72 h. Living cells were stained with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide. The color reaction (550 nm) was measured in an ELISA reader.

According to the specific question, associations between quantitative and ordered variables were quantified by Spearman rank correlation. The Jonckheere-Terpstra test for trend was used to investigate a trend in proportions. The Wilcoxon rank sum test was used for two-group comparisons of quantitative variables. The signed rank test was used to compare paired quantitative observations. All tests were two-sided and were done at the 0.05 level, 95% confidence intervals were calculated for mean score differences.

We thank J. McAlear for invaluable help with editing.