Background
Hepatocellular carcinoma (HCC) accounts for 85% to 90% of primary liver cancers and is the fifth most common cancer worldwide [
1,
2]. More than 250,000 deaths and 500,000 new cases occur globally each year [
1,
3]. One of the main reasons for the high mortality rate is the lack of effective treatments and the development of resistance to conventional chemotherapy and radiotherapy [
4]. In recent years, improved knowledge of signaling pathways regulating HCC growth and progression has led to the identification of several novel molecular targets. One of the most promising signaling pathways for molecular therapy of HCC appears to be the Hepatocyte Growth Factor (HGF)/c-Met cascade [
4‐
7].
HGF was first characterized as a factor that induces hepatocyte proliferation and as a motility factor of epithelial cells [
8‐
11]. HGF acts on c-Met, a high affinity tyrosine kinase receptor, and mediates several cellular behaviors including cell survival, proliferation, migration morphogenesis, and angiogenesis [
7‐
21]. Both c-Met and HGF are overexpressed during liver development and it is known that the signal elicited through binding of HGF to c-Met is one of the main stimuli for the G1-S progression in hepatocytes [
13]. In mice, deficiency of either HGF or c-Met expression causes embryonic lethality and reduced liver size; whereas, liver-specific deletion of c-Met induces hepatocyte necrosis and steatosis [
14,
15]. Furthermore, HGF/c-Met signaling is also essential for liver regeneration. A severe impairment of liver regeneration in the conditional c-Met mutant mice has been reported [
14].
In addition to regulating normal cellular function, c-Met is implicated in tumorigenesis. Aberrant c-Met signaling, c-Met mutations, c-Met amplification/overexpression, and autonomous growth control through autocrine signaling loops have been found to be associated with carcinogenesis [
7‐
21]. Perturbation of HGF/c-Met signaling is also involved in aggressive liver tumors and causes poor prognosis in HCC [
15,
17]. Recently, You et al. [
15] reported that c-Met represents a potential target of personalized treatment for HCC with an active HGF/c-Met pathway. While overexpression of c-Met has been found to be associated with decreased 5-year survival in patients with HCC [
17], deficiency of c-Met in hepatocytes has been reported to initiate tumorigenesis in liver [
18]. Stoelting et al. [
18] published that a defect in a c-Met-mediated signaling increases chemically-induced tumor initiation in liver. It also has been reported that the c-Met regulated gene expression signature characterizes a subset of HCC with aggressive phenotypic behavior and poor prognosis [
20]. Although inappropriate HGF/c-Met signaling is involved in all of these biological processes,
in vivo responses are rarely controlled by one signal; rather, interactions of multiple signaling pathways are involved. Recent studies have demonstrated additional roles for the HGF/c-Met signaling cascade in cancer through cross-talk with other signaling cascades, including integrins, class B plexins, proteoglycan CD44, G-protein coupled receptors, and many other receptor tyrosine kinases [
21]. Many of these combinatorial signal interactions lead to augmentation of HGF/c-Met signaling and also contribute to therapeutic resistance. Recently, it has been reported that c-Met interacts with Mucin 1 (MUC1) and catalyzes the phosphorylation of the MUC1 cytoplasmic C-terminus in pancreatic cancer cells [
22]. MUC1 is the best-characterized membrane-bound mucin that is expressed in most epithelial cells and is aberrantly overexpressed in various cancers, including breast, ovarian, lung, colon, and pancreatic carcinomas [
23,
24]. Although MUC1 expression correlates with high grade, metastasis potential, and poorer survival rate in breast cancer [
25], the studies about MUC1 expression level in HCC are controversial. In some studies elevated MUC1 levels have been reported, while in other reports no differences have been found [
26,
27]. It also has been published that the oncogenic effects of MUC1 are dependent on the cellular context [
28]. Furthermore, it is believed that different biological responses produced by MUC1 arise due to the particular repertoire of signaling molecules that interact with MUC1 [
29].
In this study, we hypothesized that the HGF/c-Met signaling pathway might play diverse roles in hepatocarcinogenesis, depending on the MUC1 status of the cells. To test this hypothesis, we first analyzed MUC1 and c-Met expression levels in HCC cell lines. In our previous studies, we characterized the differentiation status of HCC cell lines as “well-differentiated” and “poorly-differentiated”. Poorly-differentiated, highly motile and invasive HCC cell lines that display a mesenchymal phenotype were usually deficient in the expression of hepatocyte lineage markers. However, well-differentiated cell lines, which have limited motility and invasion ability and which display an epithelial phenotype, shared many feature with hepatocytes [
30,
31]. In this study we observed that poorly-differentiated HCC cell lines overexpressed both MUC1 and c-Met, whereas well-differentiated ones expressed little or no amount of the MUC1 and c-Met proteins. To support these data we also analyzed MUC1 and c-Met expression patterns in primary HCC tissues, as well as in normal and cirrhotic liver samples. We found that both c-Met and MUC1 expression were increased during hepatocarcinogenesis and correlated with the differentiation status of HCC tissues. When we tested the hypothesis that MUC1 might form a complex with c-Met in the HCC cells, we observed an interaction between MUC1 and c-Met that was down-regulated under HGF stimulation. We then demonstrated that activation and inhibition of HGF/c-Met signaling and silencing of MUC1 altered the activation of the c-Met target genes, and cellular motility and invasion.
Discussion
The HGF/c-Met pathway is an important regulator of tumor invasion and metastasis in various types of human cancer. Activating mutations and overexpression of c-Met have been associated with intrahepatic metastasis, vascular invasion, poor prognosis, and drug resistance in HCC [
19‐
21,
41]. The role of c-Met in the phosphorylation of MUC1 in pancreatic cancer progression has been recently reported [
22]. Remarkably, overexpression of MUC1 in cancer cells correlated with tumor metastasis, poor prognosis, and resistance to chemotherapeutic agents [
23,
42], which suggest that MUC1 and c-Met might work together. However, to our knowledge, there is no published study on the role of the potential crosstalk between MUC1 and c-Met in HCC. In this study, we investigated the possible roles of MUC1 and c-Met in hepatocarcinogenesis and how they interact in this process. We showed that expression patterns of c-Met and MUC1 correlated with each other and with the differentiation status of HCC cell lines and tumor tissues. Moreover, both c-Met and MUC1 expression levels were significantly higher in HCC tissues than in cirrhotic and normal liver samples. Furthermore, overexpression of c-Met and MUC1 was observed in poorly-differentiated and highly motile and invasive HCC cell lines. Correlated with these findings, we found MUC1 and c-Met to be expressed at higher levels in poorly-differentiated primary HCC tissues than in well- differentiated tumors. The overexpression of the MUC1 and c-Met proteins seen in cirrhotic liver and HCC tissues was not observed in normal liver biopsies. This suggests that the expression of MUC1 and c-Met increases during transformation of normal liver to HCC. While this paper was under preparation You et al. [
15] demonstrated that c-Met positive HCC cell lines display a mesenchymal phenotype compared to the c-Met negative cells, which have an epithelial phenotype [
15]. The results of You et al. [
15] support our observations that show poorly-differentiated HCC cells over-express c-Met, whereas well-differentiated ones do not. Additionally, herein we demonstrate MUC1 over-expression in the poorly-differentiated mesenchymal-like HCC cells. Since c-Met is a cell differentiation marker, selective expression of MUC1 and c-Met in poorly-differentiated cell lines might contribute to the epithelial-mesenchymal transition during hepatocarcinogenesis.
In addition to their complementary role in HCC development and progression, in this study we demonstrate an association between MUC1 and c-Met in Mahlavu cells. Singh et al. [
22] reported a finding supporting our observations; namely, they reported that c-Met is an interaction partner of MUC1 in pancreatic tumor cells. In addition to the MUC1/c-Met association, we also showed that HGF stimulation down-regulated the MUC1/c-Met interaction. When we treated HCC cells with HGF, we observed a remarkable reduction in MUC1 levels in a time-dependent manner, while c-Met expression was unaffected. Additionally, inhibition of c-Met activation restored MUC1 expression, which clearly demonstrated that HGF facilitated MUC1 down-regulation. Hence, we argue that in our experimental system the reduction in the MUC1/c-Met interaction is due to decreased MUC1 protein levels in response to HGF inductions. Contrarily, Singh et al. [
22] demonstrated that under HGF stimulation up to 120 min, MUC1 promoted endocytosis and increased turnover of c-Met. To understand the basis of the differences between HCC and pancreatic cancer cells we performed a time course experiment. We did not observe a noticeable down-regulation of c-Met between MUC1 over-expressing and MUC1 negative HCC cells during 16 h of HGF induction (Additional file
2 Figure S2). This is supported by our IHC analysis which demonstrated that the elevated levels of both MUC1 and c-Met expression in HCC tissues were correlated positively with each other; namely, no detectable c-Met down-regulation was observed in tissues that over-expressed MUC1. Our results also showed that silencing of MUC1 by siRNA upregulated HGF-mediated c-Met signaling network. Overall, our work is distinctive in demonstrating for the first time the effect of HGF induction both on MUC1/c-Met association and MUC1 protein levels without altering the c-Met expression pattern. Moreover, in addition to the modulation of MUC1 by c-Met, we have demonstrated the regulation of c-Met activity by MUC1.
Singh et al. [
22] showed that phosphorylation of the MUC1 cytoplasmic tail by c-Met activation enhances its interaction with p53, which leads to a reduction in the transcription of MMP1 and ultimately decreases invasion when MUC1 is overexpressed. In contrast, we have shown that MUC1 over-expressing HCC cell lines are highly motile and invasive. It has been reported that loss of p53 or the presence of abnormal forms of the p53 protein are common phenotypes in HCC cell lines including Mahlavu. Also, mutations of the p53 gene have been frequently detected in recurrent HCC patients [
43‐
45]. The differences between the results of Singh et al. [
22] and ours might be related to the p53 status of the HCC cells. To test this hypothesis, we evaluated functional p53 expressions in HCC cell lines examined in this study, and found no correlation between p53 status of HCC cells and MUC1 and/or c-Met co-expression or association (Additional file
3 Table S1). Therefore, the downstream signaling network of MUC1/c-Met association might be independent of p53 in HCC.
Previous studies have indicated that c-Met interacts with β-catenin at the inner side of the hepatocyte membrane in normal rat liver [
34]. After HGF stimulation, c-Met associated with β-catenin dissociates and translocates to the nucleus [
34,
35]. Besides being associated with c-Met, β-catenin also interacts with the cytoplasmic tail of MUC1 [
36]. Putting these findings together, we questioned whether MUC1 regulates c-Met activity via β-catenin. Since we did not observe any alteration in the MUC1/β-catenin interaction depended on HGF stimulation, we therefore examined the effect of HGF induction on β-catenin expression and phosphorylation. Our studies demonstrated that the phosphorylation of β-catenin at Ser-552, which mediates β-catenin migration to the nucleus [
37], was increased by HGF stimulation in a time-dependent manner. The interaction of c-Met RTK and β-catenin was reported by Monga SPS et al. in primary rat hepatocytes and they also identified the domain in β-catenin which is responsible for this interaction in rat hepatoma cells [
34,
46]. Although we demonstrated β-catenin/MUC1 and MUC1/c-Met interaction in poorly-differentiated HCC cells, we did not observe an interaction between β-catenin and c-Met in either Mahlavu or SNU-449 HCC cell lines in the presence or absence of HGF. The reason could be the deficiency of E-cadherin in poorly-differentiated HCC cell lines which we and others have reported previously [
30,
31,
47]. We believe that this is an important question that should be addressed in future investigations.
Nuclear localization of β-catenin is primarily associated with the induction of c-Myc expression [
34,
39]. It has been reported that a rise in the p-β-catenin level resulted in increased expression of c-Myc, which is a β-catenin target gene [
34,
39]. HGF-induced elevation of p-β-catenin and c-Myc levels observed in our study implies that β-catenin participates in HGF-mediated c-Met signaling in HCC cells. When HGF-induced c-Met activation was blocked by a c-Met specific inhibitor, β-catenin activation and concurrently c-Myc expression were suppressed. Interestingly, when MUC1 was silenced, HGF induced expression and activation of c-Met increased markedly compared to control cells. In accordance with c-Met activation, β-catenin activity and c-Myc expression increased in MUC1-silenced conditions. These results clearly demonstrate that MUC1 is a potential regulator of HGF/c-Met mediated β-catenin activation and of Myc expression in HCC cells. Aberrant activation of β-catenin signaling has been observed frequently in HCC [
30,
48,
49]; whereas, mutational activation of β-catenin signaling is found only in about 20-30% of HCCs [
49]. Our results suggest that MUC1/c-Met crosstalk is one of the important regulatory mechanisms involving aberrant activation of β-catenin signaling in HCC.
As we reported previously, poorly-differentiated HCC cells are highly motile and invasive under basal conditions; whereas, well-differentiated cells are not [
30]. Since MUC1 and c-Met are over-expressed and physically interact in poorly-differentiated HCC cells, we tested the role of this association on the invasive behavior of these HCC cells. It has been reported that MUC1 over-expressing pancreatic tumor cells have higher c-Met signaling activity and a greater tendency to metastasize when low levels of HGF are present in the tumor microenvironment [
22]. This is supported by the finding that the invasive ability of pancreatic tumor cells decreased under HGF stimulation [
22]. In this situation, the disruption of MUC1/c-Met interaction by HGF should decrease the invasive ability of HCC cells. However, the activation of HGF/c-Met signaling noticeably increased motility and invasiveness, despite MUC1 down-regulation of Mahlavu and SNU449 cells. As expected, the inhibition of the c-Met signaling network diminished HGF-induced cell motility and invasion by HCC cells. Interestingly, the knockdown of MUC1 increases invasion of Mahlavu and SNU 449 cells in response to HGF stimulation. It seems that HGF/c-Met mediated MUC-1 down-regulation or MUC1 silencing increased β-catenin activation and c-Myc expression, and this might confer a selective advantage for HCC cell invasion. This is supported by a report showing that elevated c-Myc expression via β-catenin phosphorylation, in response to HGF stimulation in colorectal carcinoma cells, is associated with a more tumorigenic and metastatic phenotype [
39]. Since poorly-differentiated HCC cells are more invasive than well-differentiated ones, we performed a few experiments regarding the role of MUC1 and c-Met cooperation on the differentiation of HCC cells (data not shown). Our data support that although c-Met and MUC1 co-expression is very important for cellular differentiation of HCC cells, presence of HGF in the microenvironment determines cellular fate. However, due to the limitations of IHC studies it is difficult to interpret the data in HCC tissues. Although we used sequential sections from one paraffin embedded tumor tissue obtained from each patient for MUC1 and cMet staining, we cannot conclude that over-expression of MUC1 and c-Met occurs in the same cells in HCC tissues. In addition as we described above, HGF levels in the microenvironment affect the behavior of MUC1 and c-Met positive cells. Further studies are needed to clarify the role of MUC1 and c-Met cooperativity in HCC cells, including HGF status of the microenvironment together with c-Met, β-catenin, and Myc activation status in HCC tissues.
Overall, our data suggests that MUC1 and c-Met are overexpressed in poorly-differentiated HCC cell lines and tissues. Under basal conditions, MUC1 and c-Met interact with each other. The activation of HGF/c-Met signaling targets MUC1 to reduce its protein level, and thus prevents the down-regulatory effects of MUC1 on HGF/c-Met signaling and in turn increases motility and invasiveness. In support of this model, the inhibition of HGF-induced c-Met activation restores MUC1 expression, which results in decreased cellular motility and invasiveness. Furthermore, the silencing of MUC1 increases HGF induced c-Met activation as well as the invasion of Mahlavu and 449 cells, showing that MUC1 down-regulation is an important regulator of c-Met activation in HCC.
Methods
Cell culture
Human hepatocellular carcinoma cell lines HuH-7, Hep3B, HepG2, SNU-449, SNU-475, and Mahlavu were cultured in DMEM supplemented with 10% FBS, 100U/ml penicillin, 0.1 mg/ml streptomycin, 2 mM L-glutamine and 1% MEM non-essential amino acids solution in a humidified 5% CO2 incubator at 37°C. HCC cell lines were kindly provided by Dr. Mehmet Öztürk (Bilkent University, Ankara, TR). Authentication of cell lines was done by DNA profiling at the University of Colorado Cancer Center (UCCC) DNA Sequencing & Analysis Shared Resource (CO, USA) using Applied Biosystem’s Identifiler kit (PN 4322288). Hepatocyte growth factor/scatter factor (HGF) was from R&D Systems (MN, USA). HGF (40 ng/ml) was used at specific time points after overnight starvation in DMEM with 2% FBS. For the inhibition of c-Met, SU11274 (Calbiochem 448101), was added to the cultures upon start of starvation. DMSO was used as solvent control of SU11274, which is dissolved in DMSO (Applichem).
Immunoprecipitation
Total cell lysates for immunoprecipitation (IP) and immunoblotting (IB) were prepared from HuH-7, Hep3B, HepG2, SNU-449, SNU-475, and Mahlavu cells with modified RIPA buffer (50 mM Tris-Cl pH 7.4, 150 mM NaCl, 1 mM EDTA pH 8.0, 1% NP-40, 1x protease inhibitor cocktail (Roche, 11836153001) 1 mM NaF, 1 mM Na3VO4. Protein concentrations of samples were determined by the BCA assay following the manufacturer’s instructions (Pierce, IL, USA). 1000 μg of total lysate was used to analyze the interaction between c-Met and MUC1 in Mahlavu cells. Samples were incubated with 4 μg anti-MUC1 (sc-7313) or anti-c-Met (c-28) (sc-161) antibodies for 2 h at 4°C, and then Gamma-Bind Sepharose beads (Amersham 17-0886-01) were added to the mixture and further incubated overnight at 4°C. IP samples were then washed three times with IP washing buffer (50 mM Tris-Cl pH 7.4, 150 mM NaCl, 1 mM EDTA pH 8.0, 1% NP-40, 0.1x protease inhibitor cocktail (Roche) 0.1 mM NaF, 0.1 mM Na3VO4). Samples were re-suspended in 2x loading dye, boiled for 5 min at 95°C, and bound proteins were analyzed by immunoblotting as described below.
Immunoassays
Total protein and cytosolic extracts were prepared by using modified RIPA buffer and Fermentas Proteojet Cytoplasmic and Nuclear Protein Extraction Kit (K0311), respectively. For immunoblotting equal volumes of total or cytosolic lysates were loaded onto an SDS polyacrylamide gel for electrophoretic analysis. The proteins in the gel were transferred onto PVDF membranes (Pierce), which were first blocked with Tris-buffered saline with 0.1% Tween-20 (TBST) containing 5% nonfat dry milk for 1 h at room temperature. The membrane then was blotted with primary antibodies against phospho-Met (Y-1234/1235) cell signaling 3129, MUC1 (VU4H5) sc-7313, MUC1 cell signaling 4538, phospho-p44/42 Erk1/2 (MAPK) (Thr202/Tyr204) cell signaling 9101, MAPK (ERK1) (C-16) sc-93, β-catenin (E-5) sc-7963, phospho-β-catenin cell signaling 9566S), vimentin BD-550513, c-Myc (sc-40) calnexin (sc-11397), cytokeratin-18 (sc-51582), lamin A/C (sc-7293) in TBST containing 3% NFDM, and Met (sc-161) in phosphate buffer saline containing 0.1% Tween-20 and 3% bovine serum albumin overnight at + 4°C. Proteins were detected by HRP-conjugated anti-rabbit (Pierce) and anti-mouse secondary antibodies (Pierce), with visualization by the ECL detection system (Pierce). The specific bands were recorded on X-ray film. Equal loading and transfer were confirmed by repeat probing for calnexin and Coomassie Blue Staining of proteins in gels. Band intensities were quantified as pixels by using ImageJ software (NIH). For quantitative determination of MUC1 antigen in conditioned media the Access Family of Immunoassay Systems (MUC1 (CA-15-3), IM2397 Beckman Coulter) was used. The MUC1 epitope (located within the 20-residue tandem repeat domain, SAPDTRPA) is recognized by B27.29 and DF3 monoclonal antibodies in this system.
To silence MUC1, Mahlavu cells were transfected with 500nM siRNA (ON-TARGETplus SMARTpool siRNA L-004019-01-0020 Dharmacon) for 72 h using Fugene HD Transfection Reagent (Ref 04 709 705 001). Non-targeting siRNA (control siRNA-A: sc-37007) was used as scramble control. After 72 h incubation, cells were harvested or replated for subsequent experiments.
Motility and invasion assay
In-vitro motility and invasion assays were performed as described [
28]. The migration of Mahlavu cells was measured by Biocoat Cell Environment control inserts (8-μm pore size; BD Biosciences). Invasion assays with the same cell line were carried out using Matrigel Invasion Chambers (BD Bioscience). Briefly, cells transiently transfected with 500 nM MUC1 siRNA (as described above) or cells pretreated overnight with 1.5 μM SU11274 in DMEM with 5% FBS were placed in upper chambers. Untreated and control siRNA treated cells were used as controls. 5,000 cells were inoculated into each chamber. After 24 h incubation at 37
oC, the medium was removed and cells were fixed and stained with Diff Quick (Siemens Healthcare Diagnostics). Cells on the upper portion of the membrane were wiped off with a cotton-tipped swab and cells that had traversed through the membrane were counted using a bright-field inverted microscope. Total cell numbers were counted for each chamber. Experiments were performed in at least triplicates. Bars represent fold differences in mean migrating or invading cell numbers. Fold differences were calculated by dividing the experimental results by the control results.
Cell adhesion and proliferation assay
SiRNA or SU11274 pretreated and untreated and/or HGF stimulated Mahlavu cells were plated on the 96 E-Plate (Roche). Adhesion and proliferation were monitored in a real-time cell electronic sensing RT-CES system (xCeLLigence-Roche Applied Science) for 96 h. These experiments were performed in at least triplicate.
Histopathology
Tissue samples were obtained from 42 patients with HCC and a cirrhotic history and 26 patients with only cirrhosis. They all had received transplants in Dokuz Eylul University, Izmir, Turkey. Normal donor liver biopsies were used as controls. The study was approved by the Ethics Committee of Dokuz Eylul University Medical School. Written informed consents were obtained from patients before liver transplantation or liver biopsy sampling. All tissue samples were fixed in formalin and embedded in paraffin. Archival materials of the patients were reevaluated by a certified pathologist (ÖS) for the confirmation of the diagnosis and to choose the most appropriate tissue block for immunohistochemistry. The histopathological analyzes of all patients were carried out by the WHO histopathological classification of liver and intrahepatic bile ducts [
29]. Standard 5 μm tissue sections were taken on lysine-coated slides.
Immunohistochemical procedure
Sections were deparaffinized in xylene and then rehydrated. Immunostaining was performed using an automated immunohistochemical stainer according to the manufacturer’s guidelines (Biogen, Lab vision autostainer 360). The antigen retrieval was performed by treatment of proteinase K for 20 min at 37oC. Endogenous peroxidase activity was blocked by incubation with 3% H2O2 for 15 min at room temperature. Tissues were incubated for 9 min with avidin-biotin blocking solution (SkyTek Lab.), and then primary antibodies anti-MUC1 (sc-7313) or anti-c-Met (sc-161) were applied at 1:100 dilutions and incubated for 35 min. The sections were stained with 3, 3-diaminobenzidine tetrahydrocloride (DAB), a chromogen stain (brown in color), and counterstained with hematoxylin.
Evaluation of staining
All staining was semi-quantitatively evaluated by a certified pathologist (ÖS). Expression of c-Met was defined as membranous and/or cytoplasmic when more than 10% of the hepatocytes stained positive for c-Met. The extent of staining was scored as one positive (10% to 25% of cells were positively stained), two positive (25%-50% of cells were positively stained) and three positive (more than 50% of the cells were positively stained). Staining with MUC1 antibody was defined as cytoplasmic and canalicular staining in hepatocytes and the extent of staining was again scored semi-quantitatively. If less than 10% of hepatocytes expressed MUC1 antibody this staining was scored as one positive staining between 10% and 30% of cells, and staining in more than 30% of cells were regarded as two and three positive, respectively. The intensity of MUC1 and c-Met immunostaining was semiquantitatively graded as follows: none (0), weak (+1), moderate (+2), and intense (+3).
Statistical analysis
All data for motility and invasion assays were expressed as mean ± S.E. Statistical analysis was performed using the GraphPad Prism and Statistical Package for Social Sciences 15.0 (SPSS Inc., Chicago, IL, USA). Statistical methods included Analysis of variance (ANOVA), Mann–Whitney U test, and ×2-test. ANOVA was used in the case of comparison of multiple groups. Mann–Whitney U test and ×2-test were used for the evaluation between two points as appropriate. Overall survivals of cells were computed using the Kaplan-Meier method and comparison between groups were analyzed using the log-rank test. Correlation between two groups was assessed by Pearson’s correlation analysis. p < 0.05 was considered statistically significant.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
GB and PK carried out western blot and co-immunoprecipitation experiments and drafted the manuscript with NA. PK and MC both participated in experimental procedures related with MUC1 silencing. PK also carried out motility and invasion experiments. ÖS performed the immunohistochemical analysis on tissue samples and analyzed them. SK provided tissue samples and clinical data. NA supervised project, made substantial contributions to conception and design of the study, analysis and interpretation of data, and wrote the main manuscript. EE gave technical support and conceptual advice and edited the manuscript for intellectual content. CK performed cell line authentication experiments and critically edited the manuscript. All authors read and approved the final version of manuscript.