EpCAM-associated claudin-7 supports lymphatic spread and drug resistance in rat pancreatic cancer


  • Florian Thuma,

    1. Department of Tumor Cell Biology, University Hospital of Surgery, Heidelberg, Germany
    Search for more papers by this author
    • Author contribution: FT and MZ contributed to conception and design, data assembly, analysis and interpretation; MZ wrote the manuscript.

  • Margot Zöller

    Corresponding author
    1. German Cancer Research Center, Heidelberg, Germany
    • Department of Tumor Cell Biology, University Hospital of Surgery, Heidelberg, Germany
    Search for more papers by this author

Correspondence to: Margot Zöller, Department of Tumor Cell Biology, Im Neuenheimer Feld 365, D-69120 Heidelberg, Germany, Tel.: 49–6221-565146, Fax: +49–6221-565199, E-mail: margot.zoeller@uni-heidelberg.de


Pancreatic cancer has a dismal prognosis because of early metastatic spread, a suggested feature of cancer-initiating cells (CIC). To control for a functional contribution of the pancreatic CIC-marker EpCAM, we explored metastasis formation by a stable EpCAM-knockdown (ASML-EpCkd) of the rat pancreatic adenocarcinoma line BSp73ASML (ASMLwt). As EpCAM associates with claudin-7, an ASML-claudin-7-knockdown (ASML-cld7kd) was included to differentiate between EpC- and EpC-cld7-mediated effects. The metastatic capacity of ASML-EpCkd and more pronounced ASML-cld7kd cells is strikingly reduced. EpC-associated cld7 interferes with EpC-mediated cell–cell adhesion and supports migration. This requires cld7 phosphorylation and formation of an EpC-cld7-tetraspanin-alpha6beta4 complex in glycolipid-enriched membrane domains (GEM), where cld7 associates via the tetraspanin-alpha6beta4 complex with phosphorylated ezrin. The association of cld7 with alpha6beta4 and cytoskeleton strongly stimulates tumor cell migration. However, EpC does not actively contribute. Instead, GEM-located cld7 associates with presenilin-2, which facilitates EpC cleavage and thereby tumor cell proliferation. Finally, the EpC-cld7 complex promotes drug resistance. Both EpC and cld7 support MAPK and JNK activation, such that in ASML-EpCkd and ASML-cld7kd cells an undue expansion of proapoptotic molecules is observed. Only cld7 promotes activation of the PI3K/Akt pathway by a strong downregulation of Pten. Accordingly, cisplatin treatment prolongs the survival time of ASML-cld7kd-bearing rats. Taken together, cld7 supports tumorigenic features of EpC by provoking EpC cleavage and thereby its cotranscription factor activity. On the other hand, only cld7 is directly engaged in motility and apoptosis resistance. Thus, at least in concern of migrating CIC, it is cld7 that acts as a CIC biomarker.




ASML with a stable knockdown of claudin-7


ASML with a stable knockdown of EpCAM


cancer-initiating cells






intracellular domain of EpCAM


glycolipid-enriched membrane microdomains






presenilin 2 N-terminal fragment




TNFα-converting enzyme/ADAM17


Western blot

There is hope that limitations in cancer therapy may be overcome by attacking a small population of cancer-initiating cells (CIC),[1] suggested to be essential for primary tumor and metastatic growth.[2] EpCAM (EpC) is a prominent CIC marker in colorectal, pancreatic, liver and breast cancer.[3, 4] However, similar to other CIC markers, there is limited information, whether EpC fulfills selective CIC-related tasks.[5]

EpC mediates homophilic cell–cell adhesion. An EGF-like and a thyroglobin domain are involved in anchoring actin microfilaments at the cell membrane via α-actinin, a process regulated by the cytoplasmic tail.[6] EpC also modulates adhesion, motility and invasion by interfering with E-cadherin via disrupting the link between β-catenin and F-actin,[7] by its engagement in Wnt/β-catenin signaling,[8] by controlling cell movement via downregulation of PKC[9] and by regulating MMP7 expression.[10, 11] More recently, evidence has been provided for functional activity of EpC in CIC.[12-14] EpC cross-linking triggers TACE (TNFα-converting enzyme) and PS2/NTF (presenilin 2 N-terminal fragment), the latter cleaving an intracellular peptide, EpIC, which forms a complex with β-catenin, FHL2 (Four-and-Half-LIM-only) and Lef-1, relocates to the nucleus and initiates, besides others, c-myc, cyclinA and E transcription.[12, 13] These findings have been extended to show EpIC-initiated transcription of additional reprogramming genes such as Oct4 and Nanog. In addition, the authors provide evidence for a contribution of EpIC to the process of epithelial–mesenchymal transition (EMT) with upregulation of vimentin, Snail, Slug and downregulation of E-cadherin.[14] The study was performed with a murine colon cancer and a human hepatoma line. There was no evaluation of cld7. However, it is known that hepatocyte progenitors express, besides EpC, cld7[15] and in colon and pancreatic cancer, EpC associates with claudin7 (cld7).[16]

Claudins, tight junction proteins,[17] also diffusely distribute in lateral membranes.[18] Palmitoylated claudins are partitioned into glycolipid-enriched membrane microdomains (GEM), where they interact with scaffold proteins creating a platform for signal transducing and a linkage to the cytoskeleton.[19] Claudins are PKA, PKC and MLCK targets,[20, 21] where cld phosphorylation prohibits integration into tight junctions, accompanied by loss of polarization.[22] In line with this, a cld7ko is lethal within 10 days after birth because of destruction of the intestine.[23] The authors speculate on the importance of a missing association with integrins and a striking upregulation of MMP9 contributing to gut destruction.[23] Notably, an EpCko also is associated with intestine destruction-promoted death within 1 week after birth, the authors demonstrating intestine destruction to be due to the missing association of EpC with cld7.[24] We reported that in tumor cells EpC preferentially associates with cld7[16] in GEM.[25] As GEM destruction by partial cholesterol depletion was accompanied by changes in apoptosis resistance and cell motility,[16, 26, 27] we hypothesized[28] that cld7-mediated GEM recruitment may account for CIC features of EpC, where GEM recruitment could facilitate EpIC generation via GEM-located TACE and PS2/NTF.[29, 30] We controlled our hypothesis by evaluating the impact of an EpC or cld7 knockdown on the highly metastatic rat pancreatic adenocarcinoma line ASML[31] (ASML-EpCkd and ASML-cld7kd). Expectedly, the metastatic capacity of both lines was strikingly reduced. However, cld7 rather than EpC apparently drives metastasis.

Material and Methods


ASML cells[31] were transfected with pSuperGFP-neo plasmid containing EpC or cld7 siRNA (for primers, see supporting information Table 1). Stable knockdown clones were established by cloning. Cells were maintained in RPMI 1640/10% fetal calf serum (FCS) with or without 0.5 µg/ml G418 (for Antibodies and chemicals, See supporting information Table 2).

Sucrose density gradient centrifugation

Cell lysates in 2.5 M sucrose were overlaid by a continuous sucrose gradient (0.25–2 M) and centrifuged (15 hr, 150,000g), collecting 12 1-ml fractions.

In vivo kinase assay

Immune complexes were suspended in lysis buffer containing a protease inhibitor mix. After centrifugation, beads were resuspended in 30 µl kinase assay buffer, 10 µCi [32P]γ-ATP and incubated (15 min, 37°C), stopping the reaction by 10 µl nonreducing 6× Laemmli buffer. Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) was followed by autoradiography.

Immunoprecipitation and Western blot

Lysates (30 min, 4°C, HEPES buffer, 1% Lubrol, 1 mM phenylmethanesulfonylfluorid (PMSF), 1 mM NaVO4, 10 mM NaF and protease inhibitor mix) were centrifuged (13,000g, 10 min, 4°C), mixed with antibody (1 hr, 4°C) and incubated with ProteinG-Sepharose (1 hr). For the analysis of released EpEX (EpC extracellular domain), culture supernatant were depleted of exosomes,[32] supernatant were 10 times concentrated and immunoprecipitated. Washed complexes/lysates, dissolved in Laemmli buffer, were resolved on 10–12% SDS-PAGE. After protein transfer, blocking, blotting with antibodies, blots were developed with ECL.


It followed routine procedures. For intracellular staining, cells were fixed and permeabilized. Samples were processed in a FACS-Calibur.


Snap-frozen sections (5 µm) were fixed, incubated with antibodies, washed and exposed to biotinylated secondary antibodies and alkaline phosphatase-conjugated avidin–biotin solution. Sections were counterstained with H&E. Cells on glass slides were fixed, permeabilized, blocked, incubated with primary antibody, fluorochrome-conjugated secondary antibody, blocked, incubated with second, dye-labeled primary antibody and washed. Slides were mounted in Elvanol. Digitized images were generated using a Leica DMRBE microscope.


Carboxyfluorescein succinimidyl ester (CFSE)-labeled cells were seeded on a cell monolayer in 96-well plates. After washing, adherent cells were stained with crystal-violet and/or lysed, evaluating (fluorescence) intensity photometrically. Adhesion is presented as percentage of seeded cells.


Cells, in the upper part of a Boyden chamber (RPMI/0.1% bovine serum albumin (BSA)), were separated from the lower part (RPMI/20% FCS) by 8-µm pore size polycarbonate membranes. After 16 hr, the lower membrane side was stained (crystal-violet), measuring OD595 after lysis. Migration is presented as % input cells. In an in vitro wound healing assay, a subconfluent monolayer was scratched with a pipette tip. Wound closure was controlled by light microscopy.


Cells (1 × 105) were grown for 48 hr in RPMI/10% FCS containing cisplatin. Survival was monitored by annexinV-APC/PI staining, MTT assay and 3H-thymidine uptake.

Soft agar assay

Tumor cells in 0.3% agar were seeded on a preformed 1% agar layer counting colonies after 3 weeks.

In vivo assays

BDX rats received tumor cells (1 × 106 or 5 × 106) intrafootpad (ifp) or intraperitoneally (ip). Phosphate buffered saline (PBS) or cisplatin (1 µg/g body weight) was given ip after 2 and 23 days. Rats were controlled weekly for local, draining lymph node or intraperitoneal tumor growth, short breathing or weight loss. Animals were sacrificed when draining nodes reached 2 cm diameter, ascites became obvious, rats lost >10% weight or latest after 120 days. Rats receiving 5 × 106 CFSE-labeled tumor cells, ifp, were sacrificed after 1 week. Animal experiments were Government-approved (Baden-Wuerttemberg, Germany).


p-Values <0.05 (two-tailed Student's t-test and Kruskal–Wallis test) were considered significant.


CIC accounting for metastasis formation, we evaluated by an EpCkd (ASML-EpCkd) in highly metastasizing ASML cells the contribution of the CIC-marker EpC.[33] In human gastrointestinal cancer[26] and ASMLwt cells,[16] EpC associates with cld7. Therefore, ASML-cld7kd cells served to control, whether EpC itself or cld7-associated EpC promotes metastasis.

EpC, cld7 and metastatic spread

Three ASML-EpCkd and four ASML-cld7kd clones have been generated. In vitro screening for stability, viability and functional activity not revealing any differences, we mostly proceeded with the clones shown in Figure 1a.

Figure 1.

In vivo growth of ASMLwt, ASML-cld7kd and ASML-EpCkd cells. (a) WB of ASMLwt, ASML-EpCkd and ASML-cld7kd cells with anti-EpC, anti-cld7 and anti-actin. (b–g) BDX rats received 1 × 106 tumor cells, intrafootpad. (b) Local growth and popliteal lymph node metastasis were controlled for 6 weeks (mean tumor diameter ± SD of five ASMLwt-bearing rats, tumor diameter only of one ASML-EpCkd and ASML-cld7kd rat with the highest tumor load; (c–g) rats were sacrificed, when becoming moribund or latest after 120 days. (c) Macroscopic appearance of the lung and immunohistochemistry of local tumor, popliteal node and lung including staining of C4.4A (ASML marker). (d) WB with anti-C4.4A; (e and f) ex vivo growth of dispersed popliteal node and lung cells in liquid cultures and soft agar; (g) WB after in vitro culture of the local tumor to control for in vivo stability of the EpCkd and the cld7kd (scale bar, c: 2.5 cm and e: 100 µm). ASML-EpCkd and ASML-cld7kd cells, though surviving, poorly grow in draining nodes, not in the lung.

Rats received ASMLwt, ASML-EpCkd or ASML-cld7kd cells ifp. ASMLwt-bearing rats became moribund after 6–7 weeks because of lung metastasis. At that time, the majority of ASML-EpCkd- and ASML-cld7kd-bearing rats not having developed a local tumor or popliteal node metastasis (Fig. 1b), rats were sacrificed after 12 weeks. ASML-EpCkd cells showed small local tumors and metastases in popliteal and/or paraaortic nodes, which only partly replaced the lymphatic tissue. Contralateral lymph nodes and lungs were tumor-free. The metastatic capacity of ASML-cld7kd cells was even more strikingly reduced. With one exception, draining nodes were macroscopically tumor-free. However, immunohistology, flow cytometry and Western blot (WB) revealed few ASML-cld7kd cells in draining nodes, but not lungs. This has been confirmed using the ASML marker C4.4A,[34] as lung epithelium expresses EpC and weakly cld7 and lymphatic vessels express cld7[35] (Figs. 1c and 1d, Table 1 and supporting information Fig. 1). Furthermore, tumor cell colonies grew in popliteal node cultures, although only after a stroma layer had formed. Soft-agar colony formation, too, confirmed the presence of ASML-EpCkd and ASML-cld7kd cells in draining nodes (Figs. 1e and 1f). Finally, as demonstrated by WB of ex vivo cultured cells, the phenotype of the ASML-EpCkd and ASML-cld7kd cells was maintained during the in vivo passage (Fig. 1g).

Table 1. In vivo growth of ASMLwt, ASML-EpCkd and ASML-cld7kd tumor cells: Tumor growth after intrafootpad application
Tumor lineaLocal growthbMetastasesb
  Draining lymph nodeDistant lymph nodesLung
  1. a

    Rats received 1 × 106 tumor cells, intrafootpad.

  2. b

    Values from ASMLwt-bearing rats were obtained when rats became moribund, and values from ASML-EpCkd- and ASML-cld7kd-bearing rats were collected when sacrificed after 120 days.

ASMLwt0.1–0.2 cm (10/10)>2 cm (10/10)>2 cm (10/10)>1,000 (miliary) (10/10)
ASML-EpCkd0.1–0.6 cm (7/11)0.4 (0.1–1) cm (6/11)0/110/11
ASML-cld7kd0.1–0.2 cm (2/10)0.2 cm (1/10)0/100/10

Thus, the metastatic capacity of ASML-EpCkd and ASML-cld7kd cells is strikingly reduced, but, without expanding, cells survive in the draining lymph node for 3 months. One possible explanation, why both the EpCkd and the cld7kd affect the metastasizing capacity could be that only cld7-associated EpC promotes metastasis. However, the unexpectedly stronger effect of a cld7kd points toward additional, cld7-selective activities, which we tried to unravel evaluating adhesiveness, motility and apoptosis resistance.

Cld7 interferes with EpC-mediated cell–cell adhesion

EpC-mediated homophilic cell–cell adhesion[6] could hamper metastasis formation. As EpC-associated cld7 interferes with EpC tetramer formation,[26] a cld7kd could strengthen adhesiveness. ASMLwt and more pronounced ASML-cld7kd cells adhere better to themselves than to ASML-EpCkd or EpC-negative AS cells. Adhesion of ASML-EpCkd cells to ASMLwt, ASML-EpCkd and ASML-cld7kd cells is slightly impaired. In line with PMA treatment supporting the EpC-cld7 association,[26] adhesion of ASMLwt cells to EpC-expressing ASML cells is reduced, but adhesion of ASML-cld7kd cells is not affected by PMA treatment. In addition, ASML-cld7kd cells agglomerate more efficiently than ASMLwt cells (supporting information Figs. 2A and 2B).

Pronounced adhesion and agglomeration of ASML-cld7kd cells are in line with cld7 interfering with EpC-tetramer formation/homophilic adhesion, where the stronger cell–cell adhesion of ASML-cld7kd cells could contribute to the poor recovery of ASML-cld7kd cells in the draining node. However, the rather weak differences in cell–cell adhesion hardly explain the striking reduction in metastatic growth of ASML-cld7kd and not that of ASML-EpCkd cells. One possibility could have been that EpC actively promotes motility[23] or that EpCAM gains motility-promoting activity by the association with cld7. The latter appeared likely as the PMA stimulation-supported EpC-cld7 association is accompanied by cld7 phosphorylation[16, 17] and formation of a larger complex, which includes Tspan8 and α6β4 and is located in GEM.[26, 27]

GEM location of the cld7-EpC complex and cld7-promoted motility by associating with PKC, α6β4 and the cytoskeleton

In advance of exploring a possible impact of the EpC-cld7 complex on motility, we evaluated the membrane subdomain localization of the EpC-cld7 complex as a GEM location because of the harboring of signal transduction molecules in these membrane domains[36] could have bearing on multiple functional activities of the EpC-cld7 complex.

Irrespective of PMA treatment, EpC and cld7 colocalize with Tspan8 and α6β4 in ASML cells. However, with the exception of cld7-Tspan8, colocalization is mostly restricted to cell–cell contact sites. Instead, in PMA-treated ASMLwt cells, colocalization is stronger and shows a cell–cell contact independent patchy distribution. On the other hand, in ASML-cld7kd cells, colocalization of EpC with α6β4 and Tspan8 is weak. Although strengthened by PMA treatment, it remains mostly restricted to cell–cell contact sites. In ASML-EpCkd cells, colocalization of cld7 with α6β4 and Tspan8 resembles that of ASMLwt cells, particularly colocalization of cld7 with Tspan8 being detected independent of cell–cell contact sites even in unstimulated cells (supporting information Fig. 3). Coimmunoprecipitation confirmed the EpC-cld7-Tspan8-α6β4 association in ASMLwt and the cld7-Tspan8-α6β4 association in ASML-EpCkd cells, but EpC does not coimmunoprecipitate with α6β4 in ASML-cld7kd cells. Tspan8 coimmunoprecipitates with α6β4 independent of cld7 and EpC, coimmunoprecipitation being strengthened in PMA-treated cells (Fig. 2a). As tetraspanins are enriched in GEM, we next asked, whether EpC and cld7 are also recovered in GEM. After sucrose gradient centrifugation part of EpC, cld7 and Tspan8 are recovered in light density (GEM) fractions. Expectedly, after GEM destruction by partial cholesterol depletion (methyl-β-cyclodextrin, MβCD) of EpC, cld7 and Tspan8 are recovered in higher density sucrose fractions (supporting information Fig. 4). This finding allowed us to ask, whether the complex formation is strictly GEM-dependent. Coimmunoprecipitation of ASMLwt lysates after MβCD treatment showed that the EpC-cld7 association partly resists MβCD treatment, but the Tspan8-EpC association is hardly detectable. Accordingly, Tspan8 does not coimmunoprecipitate with EpC in MβCD-treated ASML-cld7kd lysates, but weakly coimmunoprecipitate with cld7 in ASML-EpCkd lysates (Fig. 2b). Coimmunoprecipitation of EpC and cld7 with Tspan8 in light sucrose density fractions confirmed that the complex is only recovered in GEM and that the EpC-Tspan8 association is strongly reduced in ASML-cld7kd lysates, but in ASML-EpCkd lysates the Tspan8-cld7 association mostly is maintained (Fig. 2c).

Figure 2.

GEM localization and cld7-promoted motility. (a) ASMLwt, ASML-EpCkd and ASML-cld7kd lysates were immunoprecipitated with anti-α6β4. Lysates and immunoprecipitates were blotted with anti-EpC, anti-cld7 and anti-Tspan8. (b) WB of lysates and immunoprecipitates of untreated and MβCD-treated cells with control IgG anti-EpC, anti-cld7 and anti-Tspan8 and (c) immunoprecipitates of sucrose gradient fractions with anti-Tspan8. WB of the lysates and immunoprecipitates with anti-Tspan8, anti-EpC and anti-cld7 are shown. (d) In vivo kinase assay and WB for ezrin, p-ezrin, PKC, p-PKC, EpC and cld7 of ASMLwt cell treated as indicated; the 19 and 32 kDa bands of the in vivo kinase assay correspond to cld7 and EpC; (e) ASMLwt cells treated as indicated were precipitated with anti-Tspan8, anti-α6β4, anti-EpC and anti-cld7. WB of cell lysates and the immunoprecipitates with anti-Tspan8, anti-EpC and anti-ezrin are shown; (f) ASMLwt, ASML-EpCkd and ASML-cld7kd cells were incubated with control IgG, anti-EpC, anti-α6β4 or anti-Tspan8, washed and seeded in the upper part of a Boyden chamber. The lower part contained RPMI/20%FCS. The percent migrating cells (mean ± SD, triplicates) after 16 hr is shown. (g) Subconfluent monolayers of ASMLwt, ASML-EpCkd and ASML-cld7kd cells were scratched. Medium was exchanged and contained PMA and/or control IgG, anti-Tspan8, anti-α6β4 or anti-EpC as indicated. The percent wound closure (mean ± SD, three assays) is shown. (f and g) Significant differences in migration compared to ASMLwt cells: s, significant antibody-mediated inhibition*. (h) Cells were seeded in RPMI/PMA, DMSO or STP in a Boyden chamber (upper part), the lower chamber contained RPMI/20% FCS; significant differences in the percent migrating cells (mean ± SD, triplicates) compared to ASMLwt cells*; in dependence on STP: s. coimmunoprecipitation of EpC with α6β4 and Tspan8 essentially requires cld7 and is only recovered in GEM. GEM-localized cld7 and ezrin are phosphorylated, likely via PKC. GEM-localized, phosphorylated cld7 associates via Tspan8 and α6β4 with phosphorylated ezrin. GEM-localized cld7 promotes in concert with α6β4 and Tspan8 tumor cell migration. An involvement of EpC in migration is suggested by the efficacy of an anti-EpC blockade. However, EpC does not contribute to the migratory phenotype as in the absence of cld7 migration is unaltered or reduced, whereas it is accelerated in the absence of EpC. We want to mention that repetition of these experiments with different ASML-EpCkd and ASML-cld7kd clones showed comparable results excluding clone-specific effects.

Thus, PMA strengthens GEM-localized cld7-EpC-Tspan8-α6β4 complex formation. The direct EpC-cld7 protein interaction[16] resists GEM destruction, but the cld7-Tspan8 association is reduced and the EpC-Tspan8 association is abolished, which is in line with EpC associating with Tspan8 via cld7.

We proceeded to search for GEM localization-dependent PKC activation and its impact on cld7 and ezrin phosphorylation. An in vitro kinase assay revealed phosphorylation of 19, 32 and 55 kDa proteins and several proteins between 85 and 130kDa. Phosphorylation was reduced after MβCD treatment and nearly abolished after staurosporin (STP) treatment. WB confirmed the 32-kDa protein to be EpC and the 19-kDa protein to be cld7 that phosphorylation was reduced after MβCD treatment of PMA-stimulated cells and was abolished in the presence of STP. PKC phosphorylation also was stronger in lysates of PMA-treated ASML cells, it was reduced in MβCD-treated cells and abolished in the presence of the PKC inhibitor STP. Ezrin phosphorylation, too, was reduced, particularly in MβCD-treated PMA-stimulated ASML cells (Fig. 2d). On the basis of the latter observation, we hypothesized that phosphorylated cld7-associated ezrin may link EpC to the cytoskeleton. Although ezrin poorly coimmunoprecipitated with cld7 in PMA-treated ASML cells, ezrin coimmunoprecipitated with Tspan8 and α6β4. Coimmunoprecipitation was abolished in STP-treated cells and strongly reduced in STP-treated cells after PMA stimulation (Fig. 2e).

Thus, cld7 profits from GEM-associated signaling molecules, particularly tetraspanin-associated PKC,[37] which promotes cld7 phosphorylation.[38] Via Tspan8 and α6β4 it also gains access to p-ezrin. These findings suggested that GEM-located cld7 and possibly—via its association with cld7—EpC might contribute to ASML cell motility.

Transwell migration in the presence of anti-Tspan8 and anti-α6β4, known to inhibit ASML cell migration,[39] revealed both antibodies inhibiting migration, but anti-EpC was inhibitory, too, which confirms that the GEM-localized EpC-cld7 complex contributes to ASML cell motility. However, transwell migration of ASML-cld7kd cells was unaltered or reduced compared to ASMLwt cells. Instead, ASML-EpCkd cells displayed increased migratory activity, which was inhibited by anti-Tspan8 and anti-α6β4. Antibody inhibition of poor migration of ASML-cld7kd cells is difficult to judge. But, if at all, it was weakly affected by anti-Tspan8 and anti-α6β4, but not by anti-EpC (Fig. 2f). Very similar findings accounted for in vitro wound healing: migration inhibition of ASMLwt cells by anti-Tspan8, anti-α6β4 and anti-EpC, accelerated migration of ASML-EpCkd and retarded migration of ASML-cld7kd cells (Fig. 2g and supporting information Fig. 5).

These findings indicate that cld7 is required for EpC recruitment into migration supporting GEM-located complexes and suggest that cld7-promoted motility might be a sequel of PKC-mediated cld7 phosphorylation in GEM, where it associates via Tspan8 and α6β4 with phosphorylated ezrin. In fact, ASMLwt and ASML-EpCkd migration is strongly impaired in the presence of STP, whereas poor migration of ASML-cld7kd cells is hardly affected (Fig. 2h).

Taken together, antibody blocking provided evidence for a contribution of cld7-associated EpC to ASML cell motility, but accelerated migration of ASML-EpCkd and poor migration of ASML-cld7kd cells argue against cld7-independent EpC activity in cell migration.

Taking the strong impact of cld7 on motility, it was surprising that we recovered only few ASML-EpCkd cells in the draining lymph node (Fig. 1). Thus, we controlled in a short-term assay the in vivo impact of cld7 on motility. Rats received CFSE-labeled tumor cells ifp. After 7 days CFSE-labeled ASML-cld7kd cells were hardly recovered in the draining node, but, unexpectedly, ASML-EpCkd cell recovery also was reduced compared to ASMLwt cells. However, the mean CFSE intensity of ASMLwt cells in the popliteal node was slightly lower than the mean intensity of ASML-EpCkd cells (supporting information Fig. 6).

The low recovery of ASML-cld7kd cells is in line with the in vitro demonstrated migration-supporting activity of cld7. Instead, the higher recovery of ASMLwt than of ASML-EpCkd cells in the draining node likely is motility-independent, but could well rely on proliferation-promoting activity of EpC.

Cld7-associated EpC cleavage supports proliferation

ASML cells express ADAM10 (data not shown) as well as ADAM17 (TACE) and PS2/NTF (Fig. 3a), where TACE and PS2/NTF were described to cleave EpC, released EpIC acting as cotranscription factor[12, 14] that might well support tumor growth.

Figure 3.

Cld7 facilitates EpC-promoted proliferation. (a) WB of ASMLwt, ASML-EpCkd and ASML-cld7kd lysates with anti-TACE and anti-PS2/NTF. (b) 3H-thymidine incorporation (triplicates, mean ± SD) of cells cultured for 16 hr with/without PMA. (c) Starved cells (using pools of ASML-EpCkd and ASML-cld7kd cells) were CFSE-labeled and cultured with/without PMA. CFSE dilution was evaluated by flow cytometry after 72 hr. Representative examples and mean values of triplicates are shown. (b and c) Significant differences compared to ASMLwt cells*. (d and e) ASMLwt and ASML-cld7kd cells were cultured for 24 hr with/without PMA. Where indicated, the medium contained the TACE inhibitor TAPI or the ADAM10 inhibitor GI254023X or the TACE and ADAM10 inhibitor GW-280264X. (d) Cell membrane expression of EpC was evaluated by flow cytometry. C4.4A staining served as control. Mean values of triplicates ± SD and representative examples are shown. Significant differences in the presence of ADAM10 and ADAM17 inhibitors are indicated*. (e) Culture supernatants were depleted of exosomes, concentrated and immunoprecipitated with anti-EpC. After SDS-PAGE and transfer immunoprecipitates were blotted with anti-EpC. (f) WB of untreated and PMA-treated cells with anti-EpC, anti-β-catenin, anti-c-myc and anti-actin. (g) Lysates of PMA-treated (1 and 12 hr) ASMLwt cells were immunoprecipitated with control IgG, anti-EpC and anti-cld7. Lysates and immunoprecipitates were separated by SDS-PAGE and after transfer blotted with anti-TACE, anti-PS2/NTF, anti-EpC and anti-cld7. In PMA-treated cells, cld7 associates with PS2/NFT, which might facilitate EpC cleavage, such that not only ASML-EpCkd but also ASML-cld7kd cell proliferation is impaired. Although cld7 does not associate with TACE, the strong association with PS2/NTF points toward the proximity of the EpC-cld7 complex to GEM-localized TACE facilitating EpC cleavage.

Figure 4.

Impact of cld7 and EpC on anchorage independence and apoptosis resistance. (a–c) ASMLwt, ASML-EpCkd and ASML-cld7kd cells were cultured for 48–72 hr in the presence of titrated amounts of cisplatin. (a) Mitochondrial respiratory activity (MTT assay after 48 hr; relative activity, mean ± SD, triplicates, after normalization to cultures without cisplatin). (b) 3H-thymidine incorporation (48 hr, mean ± SD, triplicates); (c) apoptosis after 48 and 72 hr (%AnnexinV-APC/PI stained cells, mean ± SD, triplicates); the amount of cisplatin required for 50% reduction of proliferation and mitochondrial integrity or 30% apoptosis and significant differences compared to ASMLwt cells are indicated. (d) Soft agar colony formation (3 weeks): morphological appearance (scale bar: 200 µm) and number of colonies (mean ± SD, three experiments); significant differences compared to ASMLwt cells*. (e) Soft agar colony formation (3 weeks) in the absence/presence of cisplatin; relative reduction in colony formation and the amount of cisplatin required for 50% reduction is indicated; p-values are shown. (a–e) Repetition with different clones showed comparable results. (f) Rats received 5 × 106 ASMLwt, ASML-cld7kd and ASML-EpCkd cells and PBS or 1 µg/g cisplatin (days 2 and 23), intraperitoneally; survival time and rate and mean ± SD survival time of five rats per group; significant differences compared to ASMLwt-bearing rats and between untreated and cisplatin-treated rats are shown. EpC, but more pronounced cld7 contribute to anchorage independence and drug resistance. Accordingly, cisplatin treatment considerably prolongs the survival time of ASML-cld7kd-bearing rats.

Proliferative activity of unstimulated ASML-EpCkd and ASML-cld7kd cells was only slightly reduced, but both lines hardly responded to PMA treatment and few cells progressed through more than three cycles within 72 hr (Figs. 3b and 3c). These data suggest EpIC contributing to ASMLwt cell proliferation under stress, where the EpC-cld7 association in GEM could strengthen EpC cleavage. Available EpIC-specific antibodies not recognizing rat EpIC, we searched for increased recovery of EpEX. EpC expression is reduced in ASML-cld7kd cells compared to ASMLwt cells. However, in starved and overnight PMA-treated ASML-cld7kd cells EpC surface expression was increased in ASML-cld7kd, but not in ASMLwt cells. EpC expression was slightly increased in ASMLwt cells, when cultured in the presence of TAPI (ADAM17 inhibitor); it was strongly upregulated in PMA-treated ASMLwt cells. Slightly increased EpC expression was also seen in TAPI-treated ASML-cld7kd cells, but PMA treatment did not strengthen the effect. An ADAM10 inhibitor, which efficiently prevented C4.4A cleavage that was included as control, exerted no significant effect on EpC recovery in untreated and PMA-treated ASMLwt and ASML-cld7kd cells (Fig. 3d). Instead, after 24-hr culture in the presence of PMA, significantly less EpC was recovered in exosome-depleted supernatant of ASML-cdl7kd than ASMLwt cells (Fig. 3e), indicating impaired EpC cleavage in the absence of cld7. Repeating the experiment in the presence of the TACE inhibitor TAPI and the ADAM10 inhibitor GI254023X[40] confirmed inhibition of EpC cleavage preferentially by TAPI rather than GI254023X (Fig. 3e). In line with cld7 contributing to EpC cleavage, PMA-stimulated ASML-cld7kd and ASML-EpCkd showed a reduction in β-catenin and only a minor increase in c-myc compared to ASMLwt cells (Fig. 3f). Additionally, cld7 coimmunoprecipitates PS2/NTF after PMA treatment. Neither EpC nor cld7 coimmunoprecipitates TACE (Fig. 3g).

We concluded that PMA treatment promotes EpC cleavage and that cld7 contributes to EpIC-initiated proliferation by recruiting EpC into the proximity of GEM-localized TACE and PS2/NTF,[29, 30] which becomes strengthened by the cld7-PS2/NTF association.

These findings could explain the retarded growth rate of ASML-EpCkd cells. Poor in vivo ASML-cld7kd cell progression suggests additional defects in ASML-cld7kd cells.

Cld7 promotes anchorage independence and apoptosis resistance via Pten

Highly apoptosis-resistant ASMLwt cells require >10 µg/ml cisplatin for a 50% reduction in proliferation. Instead, 1/2.6 µg cisplatin suffice for a 50% reduction in ASML-cld7kd and ASML-EpCkd cell proliferation (Fig. 4a), increased cisplatin sensitivity also being seen in the MTT assay (Fig. 4b). Less than 2.5 µg/ml cisplatin sufficed for 50% ASML-cld7kd cell death, whereas 15 and 35 µg/ml cisplatin induced 50% ASML-EpCkd and ASMLwt cell death (Fig. 4c). High soft agar cloning efficacy of ASMLwt cells was impaired in ASML-EpCkd and, more strongly, in ASML-cld7kd cells and 0.2 µg/ml cisplatin induced a further 50% reduction (Figs. 4d and 4e).

The strong effect of cisplatin on ASML-cld7kd cells prompted us to evaluate drug resistance in vivo. To guarantee tumor growth and to circumvent an impact of migratory activity, rats received a high dose of tumor cells (5 × 106), intraperitoneally, and in 3-week intervals 1 µg/g cisplatin. Cisplatin treatment prolonged the survival time of ASMLwt-bearing rats from 25 to 40 days, had not significant effect on ASML-EpCkd-, but nearly doubled that of ASML-cld7kd-bearing rats (Fig. 4f). The in vivo growth pattern of the three lines differed significantly. ASMLwt cells grew miliary in the omentum, lung metastases being only seen in cisplatin-treated rats surviving >30 days. ASML-EpCkd and ASML-cld7kd cells formed large tumors tightly attached to intraperitoneal organs. They did not settle or grow in the lung. Particularly, ASML-cld7kd cells grew in the thymus. ASML-EpCkd- and most pronounced ASML-cld7kd-bearing rats were burdened by tumor cell-loaded ascites and pleural effusions (Table 2).

Table 2. In vivo growth of ASMLwt, ASML-EpCkd and ASML-cld7kd tumor cells: Tumor growth under cisplatin treatment after intraperitoneal application
Tumor lineaTreatmentSurvivalp value (tumor)p value (treatment)Solid local tumorSolid metastasisAscitesPleural effusion
  1. a

    Rats received 5 × 106 tumor cells, intraperitoneally.

ASMLwtNone0/525 + 0  4–5 cm, miliary mostly omentumLung, 0/55 ml, 5/5None
 Cisplatin0/539.6 + 10.8 0.01671–2 cm, miliary mostly omentumLung, 2/50.5 ml, 2/55/5
ASML-EpCkdNone0/545.6 + 16.90.0260 5 cm, large nodules, attached to intraperitoneal organsNone20–40 ml, 4/5None
 Cisplatin0/553.8 + 17.9 ns0.3–2 cm, large nodules, attached to intraperitoneal organsThymus, 1/510 ml, bloody 4/51/5
ASML-cld7kdNone0/536.8 + 7.40.0075 2 cm, large nodules, attached to intraperitoneal organsThymus, 2/5>50 ml, 5/52/5
 Cisplatin1/572.6 + 33.5 0.04780.5–2 cm, large nodules, attached to intraperitoneal organsThymus, 4/45–10 ml, bloody 4/43/4

How does cld7 contribute to drug resistance? ASML-EpCkd and ASML-cld7kd cells cultured in the presence of cisplatin showed slightly increased caspase8, but stronger upregulation of cleaved caspase9 and caspase3 than ASMLwt cells. Only PMA-treated ASMLwt cells showed reduced cleaved caspase9 and caspase3 (Fig. 5a), the minor impact on caspase8 expression excluding a major contribution of EpC and cld7 to receptor-mediated apoptosis. Evaluating whether EpC and cld7 may protect from mitochondrial apoptosis revealed stronger upregulation of proapoptotic BID, BAK, BAX and Smac/Diablo in cisplatin-treated ASML-EpCkd and, particularly, ASML-cld7kd than ASMLwt cells (Fig. 5b). This is in line with reduced MAPK (ras, ERK1/2) and JNK (JNK, c-jun) pathway activation in PMA-treated ASML-EpCkd and ASML-cld7kd cells (Fig. 5c), which can explain higher cisplatin sensitivity. The more striking loss of ASML-cld7kd cell apoptosis resistance relies on upregulated Pten expression and reduced Pten phosphorylation. Accordingly, Akt activation and downstream antiapoptotic protein expression, mostly not differing between PMA-treated ASMLwt and ASML-EpCkd cells, were significantly reduced in ASML-cld7kd cells (Fig. 5d). Thus, by Pten upregulation, activation of the PI3K/Akt pathway is severely impaired in ASML-cld7kd cells.

Figure 5.

Apoptosis induction in ASML-cld7kd and ASML-EpCkd cells: ASMLwt, ASML-cld7kd and ASML-EpCkd cells were cultured with/without PMA or 2 µg/ml cisplatin. (a) WB of lysates with anti-caspase8, anti-caspase9, anti-cleaved caspase9, anti-caspase3, anti-EpC, anti-cld7 and anti-actin. (b) Flow-cytometry analysis of BID, BAK, BAX and Smac/Diablo expression. (c and d) WB of lysates with the indicated antibodies; (a–d) left: representative examples, right: mean values of three experiments presenting (a, c, and d) the ratio of phosphorylated to nonphosphorylated protein or nonphosphorylated protein to actin or (b) the percent stained cells; significant differences compared to ASMLwt: *. Slightly increased apoptosis susceptibility of ASML-EpCkd cells corresponds to reduced MAPK and JNK pathway activation. The striking loss in apoptosis resistance of ASML-cld7kd cells correlates to upregulation of Pten and impaired PI3K/Akt pathway activation.

We conclude that slightly increased apoptosis susceptibility of ASML-EpCkd cells likely relies on impaired MAPK and JNK activation allowing for pronounced proapoptotic molecule expression, while rescued Pten expression in ASML-cld7kd cells promotes a striking loss in apoptosis resistance by downregulation of antiapoptotic molecules.


ASML cells expressing the CIC markers CD24, CD44/CD44v6, EpC and CD133, are slowly cycling, anchorage-independent, apoptosis-resistant and highly metastatic.[41] To control for functional engagement of EpC in metastasis, we explored the effect of an EpCkd and included an ASML-cld7kd, as cld7 recruits EpC into GEM, important signaling platforms.[36] Pointing toward the EpC-cld7 complex being essential, both ASML-EpCkd and ASML-cld7kd cells largely lost metastasizing capacity. However, cld7 rather than EpC acts as driver.

EpC and cld7 in primary tumor and metastatic growth

ASML cells migrate toward the draining node, settle, grow and progress toward distant lymph nodes and lung.[31] ASML-EpCkd and ASML-cld7kd cells mostly remain at the injection site and form small tumors after long latency. Lymphatic, distinct to vascular vessels, express cld7.[35] Whether this has bearing on the lymphatic-restricted metastatic route of ASML cells and the impaired capacity of ASML-cld7kd cells to progress through the lymphatics remains to be explored. Few ASML-EpCkd and ASML-cld7kd tumor cells reaching the draining node show impaired proliferation and do not reach the lung, or do not survive, which cannot be differentiated by the experiments. Irrespective of this open question, ASML-EpCkd and ASML-cld7kd cells hardly metastasize.

Nonetheless, there are discrete differences between ASML-EpCkd and ASML-cld7kd cells. (i) ASML-EpCkd cells slowly grow in the draining node; ASML-cld7kd cells do not expand or become apoptotic; (ii) after intraperitoneal application, mostly ASML-cld7kd cells develop survival-limiting ascites and pleural effusions; (iii) ASML-cld7kd cells grow in the thymus and (iv) only ASML-cld7kd tumors respond to cisplatin. These distinct activities of EpC and cld7 versus the EpC-cld7 complex may, at least partly, rely on EpC-cld7 recruitment into GEM and cooperation with additional GEM-located molecules.

Cld7 and motility

ASML-EpCkd and particularly ASML-cld7kd cells reach the draining node with low efficacy. EpC-mediated cell–cell adhesion[6] might hamper metastasis formation, yet, the association with cld7 prohibits tetramer formation.[26] This can explain that few ASML-cld7kd cells reach the draining node. In addition, cld7 promotes motility. GEM-localized phosphorylated cld7 associates via Tspan8 and α6β4 with ezrin, whereby cld7-associated EpC also gains access to the cytoskeleton. In ovarian cancer, too, PKCε-promoted cld4 phosphorylation weakens tight junctions.[42] A Cld1 mutation not allowing phosphorylation hampers melanoma cell motility.[21] PKC also contributes to motility-promoting integrin β4 phosphorylation.[43] Thus, PKC activation can strengthen motility by concomitant cld7 and α6β4 phosphorylation. Unexpectedly, also fewer CFSE-labeled ASML-EpCkd than ASMLwt cells were recovered in the draining node 1 week after ifp application. As the CFSE staining intensity is reduced in ASMLwt compared to ASML-EpCkd cells, we consider the observation possibly independent of motility but rather related to impaired proliferative activity of ASML-EpCkd cells.

Taken together, phosphorylated cld7 prevents EpC-mediated cell–cell adhesion and actively promotes motility via the Tspan8 and α6β4 association (supporting information Fig. 6A).

EpC and lymphatic spread

Intraperitoneally injected ASMLwt cells grow miliary and hardly produce ascites. ASML-cld7kd-bearing rats develop large amounts of ascites and pleural effusion, which could relate to pronounced EpC tetramer formation that provokes lymphatic vessel plugging. Thus, ovarian cancer cells accumulate at lymphatic vessels in the peritoneum and block lymphatic clearance.[44] Few ASML-cld7kd cells leaving the peritoneal cavity grow in the thymus, which resembles thymic metastasis of EpC- or EpC-plus cld7-transfected AS cells[27] and parathymic node settlement of EpC-expressing ovarian cancer cells.[44] Homophilic adhesion of ASML-cld7kd cells with thymic epithelial EpC may promote proliferation, as described for early EpC+ thymocytes interacting with EpC+ thymic epithelium.[45]

We interpret these findings in the sense that depending on the partner cell, metastasizing cells can receive via homophilic EpC tetramer-binding growth-promoting signals. This is an EpC genuine metastasis-promoting activity, which may be hampered by cld7 (supporting information Fig. 6B).

Growth promotion by EpC and cld7

If cld7 only contributes to migration, intraperitoneal ASML-cld7kd growth should be unimpaired/accelerated because of EpIC cotranscription factor activity.[12] We confirmed c-myc upregulation and increased proliferative activity in PMA-stimulated ASMLwt, but not ASML-cld7kd cells. Our data suggest that this deficit in ASML-cld7kd cells is due to the failure of EpC recruitment into GEM, where EpC cleavage becomes facilitated by the proximity of TACE and PS2/NTF[29, 30] and the cld7 association with PS2/NTF. The importance of the GEM location is further strengthened by the observation that an ADAM10 inhibitor, ADAM10 being rather equally distributed in the cell membrane,[46] hardly affected EpC cleavage and that EpC cleavage was not strengthened by PMA treatment, which drives cld7-associated EpC into GEM. Thus, cld7 supports proliferation-promoting activity of EpC, which can explain the slower progression of ASML-cld7kd cells.

Although not central for metastasis, EpC-cld7-expressing tumor cells will profit from GEM-located cld7 that supports growth-promoting activity of EpC (supporting information Fig. 6C).

Cld7 and drug resistance

ASML-cld7kd cells are more cisplatin-susceptible than ASMLwt and ASML-EpCkd cells. Comparable caspase8 activation argues against an impact of cld7 and/or EpC on receptor-mediated apoptosis. Instead, caspase9 and caspase3 activation was pronounced in cisplatin-treated ASML-EpCkd and more strongly ASML-cld7kd cells. In ASML-EpCkd cells caspase9 and caspase3 activations likely are initiated via pronounced proapoptotic molecule activation due to mitigated MAPK and JNK pathway activation.[47] However, strongly reduced ASML-cld7kd cell drug resistance rather relies on impaired PI3K/Akt pathway activation, required for antiapoptotic protein stabilization. Increased Pten expression and reduced Pten phosphorylation may well be the initial trigger. By dephosphorylating PIP3, Pten acts upstream of PI3K as the central phosphatase in blocking survival.[48] Pten upregulation in ASML-cld7kd cells severely impairs BAD phosphorylation, Bcl-Xl/Bcl2 expression and cisplatin resistance.

With regard to the apoptosis resistance of metastasizing tumor cells mostly cld7, independent of EpC, supports activity of the antiapoptotic PI3K/Akt pathway by downregulating Pten expression. The linkage between cld7 and Pten repression remains to be explored (supporting information Fig. 6D).


EpC and CD44/CD44v6 are pancreatic CIC markers.[3] We expected an EpCkd, similar to a CD44v6kd, to impair metastasis.[41] Although this has been the case, EpC is mostly involved via associated cld7 that triggers motility and apoptosis resistance. Our findings do not exclude EpC bioactivity affecting tumorigenicity. However, without question, cld7 acts as a CIC biomarker, in particular supporting migrating CIC. We currently establish ASML-EpC rescue clones with/without mutations in the EpC-cld7-binding site[26] to precisely define the impact of the EpC-cld7 association versus association-independent activities of EpC and cld7 in metastasis. Irrespective of this open question, in view of the bioactivity of cld7/cld7-EpC complexes, claudin-7 should be taken into account as a therapeutic target in pancreatic cancer.


This work was supported by the Deutsche Krebshilfe (MZ). The authors thank Dr. Stefan Rose-John for the generous supply of commercially not available ADAM10 inhibitors, Drs. Uwe Galli and Theron Johnson for help with primer design and Christine Niesik for help with animal experiments and immunohistology.