Oncogene (2011) 30, 1449–1459 & 2011 Macmillan Publishers Limited All rights reserved 0950-9232/11 www.nature.com/onc ORIGINAL ARTICLE MUC1 enhances invasiveness of cells by inducing epithelial to mesenchymal transition

LD Roy1, M Sahraei1, DB Subramani2, D Besmer1, S Nath1, TL Tinder2, E Bajaj2, K Shanmugam2, YY Lee3, SIL Hwang3, SJ Gendler2 and P Mukherjee1

1Department of Biology, University of North Carolina Charlotte, Charlotte, NC, USA; 2Department of Biochemistry and Molecular Biology, Mayo Clinic School of Medicine in Arizona, Scottsdale, AZ, USA and 3Carolinas Healthcare Center, Charlotte, NC, USA

Increased motility and invasiveness of pancreatic cancer Introduction cells are associated with epithelial to mesenchymal transition (EMT). Snai1 and Slug are zinc-finger Pancreatic ductal (PDA) is the fourth transcription factors that trigger this process by repres- leading cause of cancer-related death in the United sing E-cadherin and enhancing and N-cadherin States (Chang et al., 2010). This disease remains a major expression. However, the mechanisms that reg- therapeutic challenge, as it is naturally resistant to ulate this activation in pancreatic tumors remain elusive. current chemotherapy/radiation treatments. Poor prog- MUC1, a transmembrane , is asso- nosis has been attributed to delayed diagnosis, early ciated with the most invasive forms of pancreatic ductal vascular dissemination and metastases to distant organs, (PDA). In this study, we show that over especially the liver and peritoneum (Burris et al., 1997). expression of MUC1 in pancreatic cancer cells triggers One important insight came from the discovery that the the molecular process of EMT, which translates to increased motility and invasiveness of cancer cells are increased invasiveness and . EMT was signifi- associated with epithelial to mesenchymal transition cantly reduced when MUC1 was genetically deleted in a (EMT) (Nieto, 2002; Thiery, 2002). In EMT, epithelial mouse model of PDA or when all seven tyrosines in the cells acquire fibroblast-like properties and show reduced cytoplasmic tail of MUC1 were mutated to phenylalanine intercellular adhesion and increased motility. This (mutated MUC1 CT). Using proteomics, RT–PCR and process is associated with the transcriptional repression western blotting, we revealed a significant increase in and functional loss of E-cadherin (Nieto, 2002; Thiery, vimentin, Slug and Snail expression with repression of 2002). Several transcription factors have been implicated E-Cadherin in MUC1-expressing cells compared with in this repression, including the zinc-finger of cells expressing the mutated MUC1 CT. In the cells the Snail/Slug family (Batlle et al., 2000; Cano et al., that carried the mutated MUC1 CT, MUC1 failed to 2000; Guaita et al., 2002), dEF1/ZEB1, SIP1 (Comijn co-immunoprecipitate with b-catenin and translocate to et al., 2001) and the basic helix-loop-helix E12/E47 the nucleus, thereby blocking transcription of the factor (Bolos et al., 2003). Snail was shown to repress associated with EMT and metastasis. Thus, functional the expression of E-cadherin and induce EMT in several tyrosines are critical in stimulating the interactions cancer cells (Batlle et al., 2000; Cano et al., 2000; Guaita between MUC1 and b-catenin and their nuclear translo- et al., 2002). Although several factors such as TGF-b cation to initiate the process of EMT. This study signifies and the estrogen receptor were shown to regulate Snail the oncogenic role of MUC1 CT and is the first to identify transcription (Fujita et al., 2003; Peinado et al., 2003), a direct role of the MUC1 in initiating EMT during the mechanisms that modulate the function of Snail pancreatic cancer. The data may have implications in have remained largely elusive. Recently, dual regulation future design of MUC1-targeted therapies for pancreatic of Snail by glycogen synthase kinase 3-b-mediated cancer. phosphorylation has been implicated in EMT (Zhou Oncogene (2011) 30, 1449–1459; doi:10.1038/onc.2010.526; et al., 2004). published online 22 November 2010 MUC1 is a transmembrane mucin glycoprotein that is overexpressed and aberrantly glycosylated in 490% of Keywords: MUCI; pancreatic cancer; epithelial to metastatic PDA (Lan et al., 1990). Although elevated mesenchymal transition; metastasis levels of MUC1 protein have been associated with higher metastasis and poor prognosis, its molecular role in metastasis remains unclear (Patton et al., 1995; Spicer et al., 1995; Schroeder et al., 2003; Tinder et al., 2008). Correspondence: Dr P Mukherjee, Department of Biology, University Multiple reports have appeared over the past decade of North Carolina Charlotte, 9201 University City Boulevard, demonstrating an incredible range of intracellular Charlotte, NC 28223, USA. E-mail: [email protected] signaling functions associated with the 72-amino-acid Received 28 July 2010; revised and accepted 26 September 2010; residue of the MUC1 cytoplasmic tail (MUC1 CT), published online 22 November 2010 which is reviewed in (Hollingsworth and Swanson, 2004; MUC1 initiates EMT in pancreatic cancer LD Roy et al 1450 Singh and Hollingsworth, 2006; Carson, 2008). MUC1 activating the EMT-associated proteins in the primary CT is a target for several kinases, including z-chain- tumors. Tumor lysates from PDA.MUC1 and PDA. associated protein kinase of 70 kD (ZAP-70), the MUC1KO tumors (n ¼ 5) were analyzed by proteomics. d-isoform of protein kinase C, glycogen synthase Several proteins were differentially expressed; however, kinase 3-b and the tyrosine kinases c-Src and Lck the most notable differences were in the levels of (Mukherjee et al., 2005; Hattrup and Gendler, 2008; vimentin and E-Cadherin. In the PDA.MUC1 tumors, Kufe, 2008). Phosphorylation of the tyrosine within the there was a significant fivefold increase in vimentin with TDRSPYEKV sequence by c-Src and Lck stimulates a concurrent decrease in E-Cadherin (Figures 1c and d). interactions between MUC1 and b-catenin whereas In contrast, the pattern was completely reversed in the phosphorylation of Ser by glycogen synthase kinase PDA.MUC1KO tumors with significantly higher (four- 3-b inhibits this interaction (Ren et al., 2002, 2006). The fold) levels of E-Cadherin and lower levels of vimentin same tyrosine residue is critically important for nuclear (Figures 1c and d). The data were the first indication of localization of MUC1, apparently because of the possible induction of EMT in the MUC1-expressing requirement for tyrosine phosphorylation to support tumors. Repression of E-Cadherin and induction of the appropriate interactions necessary for intracellular vimentin in the PDA.MUC1 tumors directly correlated trafficking. Binding to b-catenin appears to provide the with increased metastasis in the PDA.MUC1 mice. signals required for movement of MUC1 to the nucleus in tumor cells (Carson, 2008) thereby influencing MUC1 promotes induction of mesenchymal markers and transcription through T-cell factor/lymphoid enhancer functionally enhances the invasive capacity of pancreatic factor and/or other transcription factors (Hollingsworth cancer cells in vitro and Swanson, 2004). To further confirm the role of MUC1 in EMT, KCKO We have generated human pancreatic cancer cell lines and KCM cell lines were generated from the that express wild-type MUC1 or mutated MUC1 CT in PDA.MUC1KO and PDA.MUC1 mice, respectively. which all seven tyrosines are mutated to phenylalanine. MUC1 expression was confirmed by western blotting We have also generated mouse models of PDA that (Figure 1e) using MUC1 tandem repeat (TR, B27.29) either express human MUC1 (Tinder et al., 2008) or and MUC1 CT (CT2) antibodies with high expression genetically lack MUC1. Cell lines have been developed detected in KCM and no expression in KCKO cells from these mice to study the oncogenic role of MUC1 in (Figure 1e). B27.29 antibody detects the high-molecular- pancreatic cancer progression. In this study, we show weight extracellular domain (4200 kDa) and CT2 that EMT and secondary metastasis were significantly detects the low-molecular-weight CT domain (30 kDa). reduced in PDA mice that lack MUC1 compared with Both cell lines were subjected to an in vitro invasion PDA mice that express MUC1. In contrast, over- assay, results of which showed significantly higher expression of MUC1 in both human and mouse invasion index of the KCM compared with the KCKO pancreatic cancer cells initiated EMT by inducing the cells (P 0.01, Figure 1f), confirming MUC1 as the transcription factors Snail and Slug and repressing o major contributor of increased motility and invasive- E-Cadherin, which lead to increased invasion and ness. Furthermore, presence of MUC1 in KCM tumors metastasis. Furthermore, mutating the tyrosines in clearly repressed E-Cadherin expression (Figure 1d) MUC1 CT to phenylalanine completely blocked EMT and induced expression of transcription factors and and prevented metastasis. The data suggest direct proteins associated with the mesenchymal phenotype. oncogenic signaling through the tyrosines of MUC1 These included Snail, Slug, N-cadherin and vimentin CT in the induction of EMT and metastasis during (Figure 1g), which potentially contributed to the contact pancreatic cancer. inhibition and high invasion in vitro and in vivo (Figures 1b and f). Lastly, KCM cells expressed higher levels of vascular endothelial growth factor (VEGF) compared with KCKO cells (Figure 1g), another indicator of Results highly metastatic phenotype. Thus, we hypothesized that MUC1 induces EMT and initiates invasion in Lower incidence of metastasis in PDA.MUC1KO mice mouse pancreatic cancer cells. compared with PDA.MUC1 mice: evidence of EMT in PDA.MUC1 tumors PDA.MUC1 and PDA.MUC1KO mice were killed at Overexpression of MUC1 in human pancreatic cancer B36–40 weeks of age when all mice had primary cells promotes EMT and induces invasion: role of pancreatic tumors. When organs were evaluated for tyrosines in the MUC1 CT macroscopic lesions, 60% (8/13 mice) of the PDA. To further test our hypothesis and delineate the MUC1 mice developed lung metastasis with B40% potential mechanism, we stably infected human pan- (5/13) developing liver and B20% (3/13) developing creatic cancer cell lines, BxPC3 and Su86.86 with full peritoneum metastasis (Figure 1b). This was in contrast length MUC1 or mutated MUC1 CT. Cells expressing to PDA.MUC1KO mice wherein only 1 out of 10 mice full length MUC1 were designated BxPC3 MUC1 and developed metastasis in all three organs. To investigate Su86.86.MUC1 whereas cells expressing mutant MUC1 the mechanism by which MUC1 initiated higher CT were designated BxPC3.Y0 and Su86.86.Y0. Cells metastasis, we evaluated whether MUC1 has a role in infected with an empty vector containing the neomycin

Oncogene MUC1 initiates EMT in pancreatic cancer LD Roy et al 1451

Figure 1 PDA mice lacking MUC1 have significantly lower incidence of secondary metastasis associated with decreased induction of EMT proteins in the tumors. (a) Schematic representation of PDA, PDA.MUC1 and PDA.MUC1KO mice. PDA mice were mated to the human MUC1.Tg and mouse MUC1KO mice to get PDA.MUC1 mice and PDA.MUC1KO mice, respectively. (b) Percent of mice that developed metastasis in PDA.MUC1 versus PDA.MUC1KO mice. (c, d) Proteomics data on PDA.MUC1 and PDA.MUC1KO tumor lysates showing E-cadherin repression and vimentin upregulation. (e) Western blot analysis of MUC1 expression for KCM and KCKO cells using MUC1 TR and MUC1 CT monoclonal antibodies. (f) Transwell invasion assay showing significantly higher invasion index for KCM cells compared with KCKO cells (*Po0.001). (g) Gain of mesenchymal proteins in KCM cells versus KCKO cells. b-actin serves as control for equal protein loading. resistance were used as controls and designated real-time PCR arrays were used to explore transcription of BxPC3.Neo and Su86.86.Neo. The only difference 72 genes implicated in EMT and metastasis. Genes whose between MUC1 and Y0 constructs was that in Y0 transcription was altered by at least twofold compared construct, all seven tyrosines in MUC1 CT were with Neo cells were considered significant. Several genes replaced by phenylalanine. Expression of MUC1 was associated with tight junctions and epithelial phenotype confirmed in both cell lines by western blotting were significantly reduced in BxPC3 MUC1 compared (Figures 2a–d) using both the B27.29 and CT2 with the Neo cells. These included tight junction proteins antibodies. The small shift in electrophoretic mobility 1 and 3, occludin, laminin a5, b-and d-catenin, claudin noted in the Y0 cell lysate relative to MUC1 cells likely and E-cadherin (Figure 2g). In contrast, there was no such reflects a decrease in phosphorylation due to the change in the Y0 cells (Figure 2g). Conversely, genes mutations of the tyrosine residues in the Y0 cells associated with the mesenchymal phenotype and invasion (Figures 2b and d). An in vitro invasion assay showed were significantly increased in the BxPC3 MUC1 and significantly higher invasion index for MUC1 cells as decreased in the Y0 cells compared with Neo cells. These compared with Y0 and Neo cells (Po0.01, Figures 2e and included vimentin, Twist, Slug, Snail and goosecoid f). These data suggest that the increased invasiveness of (Figure 2h). Both functional and molecular studies the MUC1-expressing cells may be attributed to the indicate that MUC1 initiates the process of EMT via tyrosines in the MUC1 CT. To substantiate the signaling through the tyrosines in its CT and that no role of the MUC1 CT tyrosines as an inducer of EMT, external factor is necessary to stimulate these changes.

Oncogene MUC1 initiates EMT in pancreatic cancer LD Roy et al 1452

Figure 2 Significantly higher invasion and transcription of genes associated with a mesenchymal phenotype coupled with significantly lower transcription of genes associated with an epithelial phenotype in MUC1 overexpressing human pancreatic cancer cells as compared with control cells. Complete reversal is observed in the MUC1.Y0 expressing cells. MUC1 expression by western blot analysis: (a) MUC1 TR and (b) MUC1 CT expression of BxPC3 Neo, MUC1, and Y0; (c) MUC1 TR and (d) MUC1 CT staining of Su86.86 Neo, MUC1, and Y04200 kDa represents MUC1 TR domain and 30 kDa represents MUC1 CT domain. b-actin was used as loading control. (e, f) In vitro trans-well invasion assay for BxPC3 and Su86.86 cells, respectively. Compared to Neo and Y0 cells, MUC1-expressing cells have significantly higher invasion index (*Po0.001). (g, h) RT–PCR analysis of BxPC3 Neo, MUC1 and Y0 cells. (g) Transcription of genes generally associated with epithelial phenotype. (h) Transcription of genes generally associated with mesenchymal phenotype. Genes whose transcription was altered by at least twofold or more were considered significant. Average fold change is shown from three separate experiments. All experiments were repeated three times with a R-value (correlation coefficient) of 40.98. Clones from three independent infections were analyzed with similar results.

Loss of epithelial and gain of mesenchymal markers and metastasis may be blocked in human pancreatic in MUC1, but not in Y0 cells directly correspond to cancer cells. their transcription of metastasis-associated genes At the protein level, gain of mesenchymal markers such as Slug, Snail and vimentin and loss of E-cadherin Circulating tumor cells detected in mice bearing the expression was apparent in the MUC1 cells (Figures BxPC3 MUC1 but not Neo or Y0 tumors 3a–d). These alterations in protein expression compared To test tumor growth and metastasis in vivo, BxPC3 with Neo cells were not detected in the Y0 cells, MUC1, Y0 and Neo cells were subcutaneously injected implicating the importance of the tyrosines for effective into nude mice. All cells formed tumors within 15 days oncogenic signaling and induction of EMT (Figures post injection, with tumor growth being significantly 3a–d). Downregulation of E-cadherin was further higher in BxPC3 MUC1 versus Y0 and Neo tumors confirmed by confocal microscopy in BXPC3 MUC1 (Figure 4a; *P ¼ 0.02–0.05). BxPC3 MUC1 and Y0 but not in the Y0 cells (Figure 3e). It is well established tumors expressed high levels of MUC1 whereas Neo that Snail represses the expression of E-cadherin to tumors expressed negligible amounts of MUC1 induce EMT in several cancer cells (Batlle et al., 2000; (Figure 4b). We were unable to detect gross metastatic Cano et al., 2000; Guaita et al., 2002). Thus, the data lesions from these tumors; therefore, we evaluated levels clearly suggest that MUC1 has a critical role in inducing of circulating tumor cells in the blood of these mice. EMT in human pancreatic cancer cell lines via signaling Within 2 weeks of culture, we were able to detect through its CT tyrosines. Furthermore, genes associated colonies of tumor cells in the blood collected from with metastasis and angiogenesis such as VEGF, MMP- BxPC3 MUC1 but not from Neo or Y0 tumor-bearing 9, 3, and 2 and IL-6R were significantly increased mice (Figure 4c). The data signify the role of MUC1 CT in BxPC3 MUC1, but not in the Y0 cells compared tyrosines during pancreatic cancer growth and with Neo controls (Figure 3f). Therefore, we propose extravasation of the cells into the blood stream for that without functional tyrosines in MUC1 CT, EMT future metastasis.

Oncogene MUC1 initiates EMT in pancreatic cancer LD Roy et al 1453

Figure 3 Repression of E-cadherin and induction of Snail, Slug and vimentin in MUC1 cells coupled with significantly increased transcription of genes associated with metastasis. Complete reversal in MUC1-Y0 expressing cells. (a–d) Western blotting analysis of Slug, Snail, vimentin (a, c), and E-Cadherin (b, d) expression in BxPC3 and Su86.86. Neo, MUC1 and Y0 cells are shown. (e) Immunofluorescence staining and confocal microscopic image of E-cadherin levels in BxPC3 Neo, MUC1 and Y0 cells. Images were taken at  400 magnification. The experiments were repeated three times with three separate clones with similar results. b-actin was used as a loading control. (f) RT–PCR analysis of genes generally associated with metastasis and angiogenesis. Twofold or more difference was considered significant. Average fold change is shown from three separate experiments.

Detection of EMT in vivo in BxPC3 MUC1 tumors compared with the Neo and Y0 tumors. VEGF Protein lysates from the tumors were analyzed for the expression was also determined by immunohistochem- epithelial and mesenchymal markers. Similar to the istry in the tumor sections with high levels detected in in vitro data in Figure 3, the BxPC3 MUC1 tumor lysate the BxPC3 MUC1 tumors versus Neo and Y0 tumors showed downregulation of E-Cadherin (Figure 4d) and (Figure 4g). Thus, BxPC3 MUC1-xenografted tumors upregulation of Slug, Snail and vimentin (Figure 4e) exhibited an aggressive phenotype consistent with the clearly indicating the initiation of EMT in vivo. This human tumors that express high MUC1. This included process was completely abrogated when the tyrosines loss of E-cadherin, induction of vimentin and upregula- were mutated even though high levels of MUC1 protein tion of metastatic and angiogenic factors. It is striking was being expressed in the Y0 tumors (Figure 4b). that simply abrogating the signal transduction events by mutating the seven tyrosines can completely reverse the process of EMT and significantly alter the tumor micro- Significantly higher levels of prometastatic and environment, subsequently leading to a less aggressive, proangiogenic growth factors in the BXPC3 less metastatic and less angiogenic phenotype. MUC1 tumor microenvironment compared with Y0 or Neo tumors To determine the effect of EMT on the tumor MUC1 interacts with b-catenin and translocates to the microenvironment itself, we tested the protein levels of nucleus in the BxPC3 MUC1 but not in the Y0 cells some of the known mediators of metastasis and To delineate the underlying mechanism by which the Y0 angiogenesis in the tumor lysates using the Ray Biotech tumors circumvent EMT, we tested the nuclear localiza- protein array (Ray Biotech, Norcross, GA, USA). tion of MUC1 CT and b-catenin in the MUC1 and Y0 Twofold or more increase was considered significant and cells. It is known that tyrosine residues in MUC1 CT are is represented (Figure 4f). Significant increases in levels critically important for nuclear localization of MUC1 of circulating VEGF (10-fold), insulin-like growth CT and b-catenin (Hollingsworth and Swanson, 2004). factor-1 (twofold), interleukin-6 (threefold) and its Although b-catenin was present in the nuclear extracts receptor (interleukin-6, 2.2-fold), stem cell factor (2.5 of both cells, the levels were significantly higher in fold), P- (20-fold) and epidermal growth factor BxPC3 MUC1 cells compared with the Y0 cells (2.8-fold) were observed in the BxPC3 MUC1 tumors (Figure 5a). Similar results were obtained with the

Oncogene MUC1 initiates EMT in pancreatic cancer LD Roy et al 1454

Figure 4 Higher tumor burden with loss of E-cadherin and gain of mesenchymal and metastatic proteins in BxPC3 MUC1 versus Y0 and Neo tumors. (a) In vivo tumor growth in nude mice. Significantly higher tumor burden in BxPC3. MUC1 versus Neo and Y0 tumors (*Po0.001). (b) MUC1 expression by western blotting using the MUC1 TR and CT antibodies. (c) Tumor cells cultured from whole blood of tumor-bearing mice. Circulating tumor cells detected in the blood of mice bearing the BxPC3 MUC1 but not BxPC3 Y0 or Neo tumors. (d) Expression of E-cadherin in primary tumors by western blotting. (e) Expression of Slug, Snail and vimentin in primary tumors by western blotting. (f) High levels of prometastatic and proangiogenic proteins detected in the BXPC3 MUC1 tumor lysate. Fold change in levels of various factors in BxPC3 MUC1 tumor lysate compared with Y0 and Neo. (g) Expression of VEGF and E-cadherin by . Images were taken at  400 magnification. Similar results with n ¼ 3 mice were obtained.

KCM and KCKO cells wherein higher levels of b- Discussion catenin were detected in the nuclear extract of the KCM versus the KCKO cells (Figure 4b). More importantly, Although it has been known for decades that MUC1 although both BxPC3 MUC1 and Y0 cells contained is aberrantly overexpressed in greater than 95% of equal levels of MUC1 (Figure 3a), there was no metastatic PDA and is associated with poor prognosis, detectable MUC1 CT in the nuclear extract of the Y0 its precise role has remained obscure. We show for the cells while high levels was detected in the BxPC3 MUC1 first time in both mouse and human pancreatic tumors cells (Figure 5a). The results were substantiated in that (1) overexpression of MUC1 initiates the process of Su86.86 MUC1 and Y0 cells (Figure 5c). The data EMT and augments metastasis and (2) lack of tyrosines clearly suggest that signaling through the tyrosines is in the CT of MUC1 abrogates this process. The first critically important for efficient nuclear translocation of evidence of EMT came from the proteomics data MUC1 CT and b-catenin. Purity of the nuclear extract showing repression of E-cadherin and induction of was confirmed by the presence of lamin and the absence vimentin expression in the PDA.MUC1 tumors when of IKK (Figures 5a–c). compared with the PDA.MUC1KO tumors. This As tyrosine phosphorylation is required for MUC1 correlated with significantly higher incidence of second- to bind b-catenin and translocate to the nucleus, we ary metastasis in PDA.MUC1 mice as compared with examined if b-catenin and MUC1 co-immunoprecipi- the PDA.MUC1KO mice. The PDA mice are unique in tated in these cells. b-catenin interacted with MUC1 that the pancreatic tumors arise spontaneously in an only in the BxPC3 MUC1 cells but minimally in the Y0 appropriate tissue background, within a suitable stromal cells (Figure 5d). The co-immunoprecipitation of and hormonal milieu, and in the context of a viable b-catenin and MUC1 was confirmed in the KCM and (Hingorani et al., 2003; Tinder et al., KCKO cells and in the Su86.86 cells (Figures 5e and f). 2008; Mukherjee et al., 2009). The tumor progression Thus, we confirm that binding of MUC1 to and the histopathology of the tumors in the PDA mice b-catenin appears to provide the signal required for mimic the human disease with the initial development of movement of MUC1 to the nucleus and that the pancreatic intraepithelial neoplastic lesions progressing tyrosines in MUC1 CT have to be functional for the to in-situ and invasive adenocarcinoma. two proteins to interact. The process of EMT and increased invasiveness in

Oncogene MUC1 initiates EMT in pancreatic cancer LD Roy et al 1455

Figure 5 MUC1 interacts with b-catenin and translocates to the nucleus in the MUC1-expressing cells. Complete reversal in the Y0 cells. Protein expression of MUC1 CT and b-catenin in the nuclear extracts of (a) BxPC3 MUC1 and Y0. (b) KCM and KCKO and (c) Su86.86 MUC1 and Y0 cells. Lamin and IKK used as positive and negative control for nuclear extracts. Co-immunoprecipitation of MUC1 CT and b-catenin in both directions from (d) BxPC3 MUC1 and Y0 (e) KCM and KCKO and (f) Su86.86 MUC1 and Y0 cells. Non-specific pull down was not detected using an IgG control antibody (data not shown). Note: Neo cells were not included in the data as they express minimal levels of endogenous MUC1 and we found little to no b-catenin or MUC1 CT in the nucleus (data not shown).

MUC1-expressing pancreatic tumors was confirmed in nucleus to render its oncogenic signal (Wen et al., 2003). cell lines generated from the PDA mice and validated We show that lack of tyrosines in the MUC1 CT in two human pancreatic cell lines. The underlying regulates the interaction between MUC1 CT with molecular mechanism in all cell lines pointed to the b-catenin, and therefore its efficient in translocation to induction of transcription factors, Snail, and Slug, the nucleus (Figure 5). It is also known that b-catenin which stimulated the expression of vimentin and binds to the SXXXXXSSL motif in the CT of MUC1 repressed E-cadherin expression. E-cadherin has a key (Yamamoto et al., 1997) and changes in tyrosine role in the establishment and maintenance of adherent phosphorylation of MUC1 CT correlated with differ- junctions and loss of the same results in contact ences in cell adhesion (Yamamoto et al., 1997; Quin and inhibition and cell motility. Interestingly, the molecular McGuckin, 2000). Thus, we can speculate that the process of EMT was completely abrogated when all underlying mechanism for EMT might be the require- seven tyrosines in the CT of MUC1 were mutated to ment of the tyrosines adjacent to the b-catenin-binding phenylalanine. Note that the cytoplasmic tail of MUC1 domain (possibly at the TDRSPYEKV site) to be Y0 shows a shift in electrophoretic mobility relative to phosphorylated. The c-Src and d-isoform of protein MUC1 WT; this likely reflects a decrease in phospho- kinase C have been shown to phosphorylate the rylation due to mutation of the tyrosine residues and tyrosines at that site and increase the interaction therefore change in the overall charge. This has been between b-catenin and MUC1 (Li and Kufe, 2001; previously reported in other cell types (Thompson et al., Ren et al., 2002). We therefore propose that MUC1 may 2006). We do not believe that the change in the directly influence the transcriptional co-activator status electrophoretic mobility is due to size as the D between of b-catenin, and upregulate several genes that are seven tyrosines and seven phenylalanines is B47 associated with EMT such as Snail, Slug, Twist, daltons. Recent evidence shows that the phosphoryla- vimentin and goosecoid (Figure 2d), possibly through tion of the tyrosines in the MUC1 CT is critical for interactions with T-cell factor/lymphoid enhancer factor the binding of b-catenin to MUC1 CT and that the and/or other transcription factors. Snail in turn MUC1 CT-b-catenin complex can be translocated to the represses E-cadherin message and protein levels in

Oncogene MUC1 initiates EMT in pancreatic cancer LD Roy et al 1456 MUC1-expressing cells (Figures 1–4), perhaps through b-catenin-MUC1 interaction and translocation to the direct binding to the three E-boxes in the E-cadherin nucleus, leading to activation of the EMT-associated promoter (Guaita et al., 2002). E-cadherin has a key role transcription factors Snail and Slug. This in turn enables in the establishment and maintenance of adherence the tumors to acquire a highly aggressive phenotype junctions, repression of which leads to downregulation creating a prometastatic microenvironment in vivo. of tight junction proteins 1 and 3, occludin, laminin Future studies will focus on identifying the particular a5, and b-and d-catenins (Figure 2c). Dissolution of tyrosine/s within MUC1 CT and determining whether pancreatic epithelial cell junctions is hence brought inhibiting its phosphorylation using small molecule about by a concerted activation of mesenchymal kinase inhibitors may lead to a novel treatment modality proteins and repression of epithelial proteins in vitro for pancreatic cancer. and in vivo, leading to EMT and metastasis exemplified by the presence of circulating tumor cells in BxPC3 MUC1 tumors (Figure 4c) and enhanced metastasis in Materials and methods the PDA.MUC1 mice (Figure 1b). These molecular changes lead to an in vivo micro- Mouse model environment favoring increased angiogenic and proin- PDA mice were generated in our laboratory on the C57BL/6 flammatory cytokines in the MUC1 tumors (Figures 4f background by mating the P48-Cre with the LSL-KRASG12D and g), leading to increased metastasis. The glycoprotein mice (Hingorani et al., 2003) and further mated to the P-selectin, an adhesion molecule involved in the initiation MUC1.Tg mice to generate PDA.MUC1 mice (Tinder et al., of inflammatory process and induction of interleukin-6 2008; Mukherjee et al., 2009) or to the MUC1KO mice (Spicer et al., 1995) to generate PDA.MUC1KO mice (Figure 1a). and its receptors (Kaikita et al., 1995; Dymicka- Primary tumors were dissociated using collagenase IV Piekarska et al., 2007) was increased by 20-fold in (Worthington Biochemical Corp, Lakewood, NJ, USA) and MUC1 versus Neo tumors (Figure 4f). VEGF, a potent several lines of cells were generated in our laboratory. These inducer of angiogenesis and promoter of tumor progres- cells are designated KCKO for cells lacking MUC1 and KCM sion (Justinger et al., 2008) was increased by 10-fold in for cells expressing human MUC1. the MUC1 tumors (Figure 4f). Lastly, stem cell factor, epidermal growth factor and insulin-like growth factor-1, Invasion assays factors known to induce growth and proliferation in Cells were serum-starved for 24 h before plating for the pancreatic cancer cells (Bardeesy and DePinho, 2002; invasion assay. Cells in serum-free media (50 000 cells) were Mroczko et al., 2005; Wolpin et al., 2007) were increased plated over transwell inserts (BD Biosciences, San Jose, CA, by 2–3-fold in MUC1 tumors (Figure 4f). Thus, it is USA) precoated with reduced growth factor matrigel, and indeed striking that the significant induction of these were permitted to invade toward the serum contained in the factors in the MUC1 tumors was completely abrogated bottom chamber for 48 h. Percent invasion was calculated as absorbance of samples/absorbance of controls  100. when the tyrosines in MUC1 CT were deemed non- functional (Figure 4f). This once again signifies the critical requirement of the MUC1 CT tyrosines in Western blots and antibodies Equal quantities of tumor lysate were loaded on SDS–PAGE pancreatic cancer oncogenesis and metastasis. gels. MUC1 CT antibody CT2, was made in Mayo Clinic The regulation of MUC1 CT phosphorylation and Immunology Core (Schroeder et al., 2001). MUC1 TR parameters that affect its association with b-catenin are (B27.29) antibody was acquired from Biomira Inc. (Edmon- not fully understood, although interaction of MUC1 ton, Alberta, Canada). All other antibodies (Snail, Slug, with ErbB1 (Schroeder et al., 2001) has been shown N-cadherin, vimentin, VEGF, E-cadherin, b-catenin and b- to enhance the binding between c-src, b-catenin and actin) were purchased from Santa Cruz Biotechnology, (Santa MUC1 (Hollingsworth and Swanson, 2004) and Cruz, CA, USA) and were used according to manufacturer’s increase ERK1/2 phosphorylation and NFkB activation recommendations. Co-immunoprecipitation using CT2 (5ml.) (Schroeder et al., 2001; Thompson et al., 2006). and b-catenin (5ml.) was carried out as previously reported (Al Preliminary data from our laboratory show that the Masri and Gendler, 2005) using 1 mg of protein lysate prepared in 1% Brij buffer followed by denaturing western active form of NFkB-p65 subunit translocation to the blotting. nucleus is negligible in the Y0 cells compared with the MUC1 cells (unpublished data). Our study exemplifies the functional consequences Cloning of MUC1 WT and MUC1 Y0 vectors MUC1 Y0 was created using the Quick Change mutagenesis (EMT and metastasis) of signaling through MUC1 CT, kit (Stratagene, La Jolla, CA, USA) (Thompson et al., 2006). and for the first time delineates the role of this protein in Briefly, primers based on the MUC1 sequence were designed a unique mouse model of PDA that lacks MUC1. The containing single-base alterations resulting in mutation of the data define the role of MUC1 signaling (through its tyrosine residues (Y) in MUC1 CT to phenylalanine (F) as CT tyrosines) in the initiation of EMT perhaps by shown schematically.

Oncogene MUC1 initiates EMT in pancreatic cancer LD Roy et al 1457 Successful mutations were confirmed with DNA sequencing. phosphate buffered saline before sonication in buffer B (20 mM MUC1 Y0 and MUC1 WT were cloned into the pLNCX.1 HEPESpH7.9,25%glycerol,400mM NaCl, 1.5 mM MgCl2, vector consisting of the neomycin resistance gene for retroviral 0.2 mM EDTA, 0.5 mM dithiothreitol) to obtain the nuclear infection. Note: cells designated as BxPC3 MUC1 or fraction, which was then spun at 13 000 r.p.m. for 20 min, 4 1C, Su89.86.MUC1 represents cells expressing full length MUC1 for western blot analysis using CT2 and b-catenin antibodies. that consists of the extracellular domain, the transmembrane domain and wild-type cytoplasmic tail domain. Cells desig- In vivo tumor growth nated as BxPC3.Y0 or Su89.86. Y0 represents cells expressing Two-month old nude mice were injected with 5x106 tumor cells full length MUC1 that consists of the extracellular domain, the into the flank of the mice. Tumors were allowed to grow for 2 transmembrane domain and mutant cytoplasmic tail domain months. Mice were palpated every second day starting at 7 in which only the seven tyrosines are replaced by phenyl- days post tumor injection until sacrifice. Tumor weight was alanine. BxPC3.Neo and Su86.86.Neo represents cells that calculated according to the following formula: grams ¼ (length only express the empty vector and therefore represents the in centimeters  (width)2)/2 (Simpson-Herren and Lloyd, endogenous levels of MUC1 in these cells. 1970). On killing the mice, the tumor portions were prepared for lysates and fixed for immunohistochemistry. Cell culture and retroviral infection BxPC3 cells and Su86.86 (American Type Culture Collection, Immunohistochemistry Manassas, VA, USA) are two human pancreatic cancer cell Tumor sections were formalin fixed and paraffin embedded lines that express very little endogenous MUC1. For retroviral blocks were prepared by the Histology Core at The Mayo infection, GP2-293 packaging cells (stably expressing the gag Clinic and 4-micron thick sections were cut for immunostain- and pol proteins) were co-transfected with the full length ing. VEGF expression was determined using the antibodies MUC1 construct or the Y0 construct or an empty vector (Santa Cruz Biotechnologies) at 1:50 dilution followed by the expressing the VSV-G envelope protein as previously described appropriate secondary antibody (1:100 dilution, DAKO, (Thompson et al., 2006). Cells were selected with 0.5 mg/ml Carpenteria, CA, USA). Sections were developed using 3,30- G418, beginning 48 h post infection. Three independent diaminobenzidine as the chromogen and hematoxylin as the infections of the constructs were carried out with similar counterstain. Slides were examined under microscopy and results. Expression of the constructs was stable throughout the pictures taken at  400 magnification. span of experiments. Cells infected with vector alone were used as control and designated Neo. For MUC1 and Y0-infected Isolation of viable circulating human pancreatic cancer cells cells, MUC1 þ ve cells were sorted using the FACSAria (BD Blood samples (0.5–1.0 ml) were obtained by cardiac puncture Biosciences). For Neo-infected cells, MUC1-ve cells were from individual tumor-bearing animals using heparin as the sorted. anticoagulant and processed immediately. Erythrocyte-free nucleated cell fractions were obtained using the RBC lysis Confocal microscopy buffer (Stemcell Technologies, Inc., Vancouver, British Cells were plated on chamber slides and grown to the desired Columbia, Canada). This fraction was immediately expanded confluency. Cells were washed, fixed in À20 1C ethanol, in culture in Dulbecco’s modified Eagle’s medium as permeabilized with 0.5% Tween 20 and stained with previously described (Glinskii et al., 2003). E-cadherin antibody. Nuclei were stained with To-pro-3 (Invitrogen, Carson City, CA, USA). Pictures were taken at Protein array  400 using a confocal microscopy (Carl Zeiss International, Tumor lysates were analyzed using the RayBio Custom Thornwood, NY, USA). protein array kit (RayBiotech, Norcross, GA, USA). Assay was conducted according to the manufacturer’s instructions Real-time PCR in the Taqman low-density custom made array (Das Roy et al., 2009). Chemiluminescence was detected using format a EpiChemi3 Darkroom imaging system and LabWorks Total RNA was extracted according to standard protocol densitometry software (UVP Bioimaging, Upland, CA, USA). using the Qiagen RNeasy mini-kit protocol (Invitrogen). Data were corrected for background signal and normalized to Complementary DNA was constructed using TaqMan Reverse positive controls using RayBio Analysis Tool software. Transcription cDNA kit from Applied Bioscience (Foster City, CA, USA). Each low-density custom array card was Proteomics configured for 72 different genes in triplicates. The SDS2.2 For the electrophoretic separation of each sample, 30 mgof software (Carlsbad, CA, USA) was used for qualitative protein was loaded on a 10% bis-Tris NuPAGE gel (Invitro- analysis. Arrays were performed independently at least three gen) and proteomic analysis conducted as previously reported times for each cell line; values were obtained for the threshold (Hwang et al., 2006). A Nano Acuity UPLC system connected cycle (Ct) for each gene and normalized using the average of to an LTQ-Orbitrap hybrid MS system (Thermo Fisher, four housekeeping genes on the same array (HPRT1, RPL13A, Pittsburg, PA, USA) with Nanospray interface was used. Data GAPDH, ACTB). Change (DCt) between Neo, MUC1 and Y0 were acquired using Xcalibur and searched by Bioworks was found by: DCt ¼ Ct (MUC1 or Y0)ÀCt (neo) and fold software (Thermo Fisher) using IPI mouse v.3.18 fasta change by: fold change ¼ 2(ÀDCt). Values are provided as fold database. Search results were entered into Scaffold software change. (Proteome Software, Portland, OR, USA) for compilation, normalization and comparison of spectral counts. Data Isolation of nuclear protein sets were imported into R program (www.r-project.org) for Cells were washed in cold phosphate buffered saline and then statistical computing. resuspended in buffer A (10 mM HEPES pH 7.5, 10 mM KCl, 0.1 mM EDTA, 0.1 mM ethylene glycol tetraacetic acid and 0.1% Statistical analysis Nonidet-P40) on ice for 20 min. Lysates were spun at 6000 r.p.m., Statistical analysis was performed with SPSS 10 software 1min, 41C and the resulting pellets were washed twice in (IBM Corporation, Somers, NY, USA). P-values were generated

Oncogene MUC1 initiates EMT in pancreatic cancer LD Roy et al 1458 using the one-way analysis of variance and significance was Acknowledgements confirmed using the Duncan and Student–Newman–Keul test. Values were considered significant if Po0.05. We acknowledge the funding provided by NIH CA R01CA118944. We thank Dr Jennifer Curry for critically reviewing the manuscript. We thank Cathy S Madison and Conflict of interest Carole M Viso for their technical help with the confocal microscopy. We also acknowledge all the technicians in the Drs Mukherjee and Gendler work have been funded by the animal facility and in the histology core facilities at The Mayo NIH. All other authors declare no conflict of interest. Clinic.

References

Al Masri A, Gendler SJ. (2005). Muc1 affects c-Src signaling in PyV Hwang SI, Lundgren DH, Mayya V, Rezaul K, Cowan AE, Eng JK et al. MT-induced mammary tumorigenesis. Oncogene 24: 5799–5808. (2006). Systematic characterization of nuclear proteome during Bardeesy N, DePinho RA. (2002). Pancreatic cancer biology and apoptosis: a quantitative proteomic study by differential extraction genetics. Nat Rev Cancer 2: 897–909. and stable isotope labeling. MolCellProteomics5: 1131–1145. Batlle E, Sancho E, Franci C, Dominguez D, Monfar M, Baulida J Justinger C, Schluter C, Oliviera-Frick V, Kopp B, Rubie C, Schilling et al. (2000). The transcription factor snail is a repressor of MK. (2008). Increased growth factor expression after hepatic and E-cadherin in epithelial tumour cells. Nat Cell Biol pancreatic resection. Oncol Rep 20: 1527–1531. 2: 84–89. Kaikita K, Ogawa H, Yasue H, Sakamoto T, Suefuji H, Bolos V, Peinado H, Perez-Moreno MA, Fraga MF, Esteller M, Cano Sumida H et al. (1995). Soluble P-selectin is released into A. (2003). The transcription factor slug represses E-cadherin the coronary circulation after coronary spasm. Circulation 92: expression and induces epithelial to mesenchymal transitions: a 1726–1730. comparison with snail and E47 repressors. J Cell Sci 116: 499–511. Kufe DW. (2008). Targeting the human MUC1 oncoprotein: a tale of Burris III HA, Moore MJ, Andersen J, Green MR, Rothenberg ML, two proteins. Cancer Biol Ther 7: 81–84. Modiano MR et al. (1997). Improvements in survival and clinical Lan MS, Batra SK, Qi WN, Metzgar RS, Hollingsworth MA. (1990). benefit with gemcitabine as first-line therapy for patients with Cloning and sequencing of a human pancreatic tumor mucin advanced pancreas cancer: a randomized trial. J Clin Oncol 15: cDNA. J Biol Chem 265: 15294–15299. 2403–2413. Li Y, Kufe D. (2001). The Human DF3/MUC1 carcinoma-associated Cano A, Perez-Moreno MA, Rodrigo I, Locascio A, Blanco MJ, del antigen signals nuclear localization of the catenin (ctn). Barrio MG et al. (2000). The transcription factor snail controls Biochem Biophys Res Commun 281: 440–443. epithelial-mesenchymal transitions by repressing E-cadherin expres- Mroczko B, Szmitkowski M, Wereszczynska-Siemiatkowska U, sion. Nat Cell Biol 2: 76–83. Jurkowska G. (2005). Hematopoietic cytokines in the sera of Carson DD. (2008). The cytoplasmic tail of MUC1: a very busy place. patients with pancreatic cancer. Clin Chem Lab Med 43: 146–150. Sci Signal 1: pe35. Mukherjee P, Basu GD, Tinder TL, Subramani DB, Bradley JM, Chang BW, Siccion E, Saif MW. (2010). Updates in locally advanced Arefayene M et al. (2009). Progression of pancreatic adenocarcino- pancreatic cancer. Highlights from the ‘2010 ASCO Annual ma is significantly impeded with a combination of and Meeting’. Chicago, IL, USA. June 4–8, 2010. JOP 11: 313–316. COX-2 inhibition. J Immunol 182: 216–224. Comijn J, Berx G, Vermassen P, Verschueren K, van Grunsven L, Mukherjee P, Tinder TL, Basu GD, Gendler SJ. (2005). MUC1 Bruyneel E et al. (2001). The two-handed E box binding zinc finger (CD227) interacts with lck tyrosine kinase in Jurkat lymphoma cells protein SIP1 downregulates E-cadherin and induces invasion. Mol and normal T cells. J Leukoc Biol 77: 90–99. Cell 7: 1267–1278. Nieto MA. (2002). The snail superfamily of zinc-finger transcription Das Roy L, Pathangey LB, Tinder TL, Schettini JL, Gruber HE, factors. Nat Rev Mol Cell Biol 3: 155–166. Mukherjee P. (2009). -associated metastasis is sig- Patton S, Gendler SJ, Spicer AP. (1995). The epithelial mucin, MUC1, nificantly increased in a model of autoimmune arthritis. Breast of milk, mammary gland and other tissues. Biochim Biophys Acta Cancer Res 11: R56. 1241: 407–423. Dymicka-Piekarska V, Matowicka-Karna J, Gryko M, Kemona- Peinado H, Quintanilla M, Cano A. (2003). Transforming growth Chetnik I, Kemona H. (2007). Relationship between soluble P- factor beta-1 induces snail transcription factor in epithelial cell lines: selectin and inflammatory factors (interleukin-6 and C-reactive mechanisms for epithelial mesenchymal transitions. J Biol Chem protein) in . Thromb Res 120: 585–590. 278: 21113–21123. Fujita N, Jaye DL, Kajita M, Geigerman C, Moreno CS, Wade PA. Quin RJ, McGuckin MA. (2000). Phosphorylation of the cytoplasmic (2003). MTA3, a Mi-2/NuRD complex subunit, regulates an domain of the MUC1 mucin correlates with changes in cell-cell invasive growth pathway in breast cancer. Cell 113: 207–219. adhesion. Int J Cancer 87: 499–506. Glinskii AB, Smith BA, Jiang P, Li XM, Yang M, Hoffman RM et al. Ren J, Bharti A, Raina D, Chen W, Ahmad R, Kufe D. (2006). MUC1 (2003). Viable circulating metastatic cells produced in orthotopic oncoprotein is targeted to mitochondria by heregulin-induced but not ectopic prostate cancer models. Cancer Res 63: 4239–4243. activation of c-Src and the molecular chaperone . Oncogene Guaita S, Puig I, Franci C, Garrido M, Dominguez D, Batlle E et al. 25: 20–31. (2002). Snail induction of epithelial to mesenchymal transition in Ren J, Li Y, Kufe D. (2002). Protein kinase C delta regulates function tumor cells is accompanied by MUC1 repression and ZEB1 of the DF3/MUC1 carcinoma antigen in beta-catenin signaling. expression. J Biol Chem 277: 39209–39216. J Biol Chem 277: 17616–17622. Hattrup CL, Gendler SJ. (2008). Structure and function of the cell Schroeder JA, Adriance MC, Thompson MC, Camenisch TD, surface (tethered) . Annu Rev Physiol 70: 431–457. Gendler SJ. (2003). MUC1 alters beta-catenin-dependent Hingorani SR, Petricoin III EF, Maitra A, Rajapaske V, King C, tumor formation and promotes cellular invasion. Oncogene 22: Jacobetz MA et al. (2003). Preinvasive and invasive ductal 1324–1332. pancreatic cancer and its early detection in the mouse. Cancer Cell Schroeder JA, Thompson MC, Gardner MM, Gendler SJ. (2001). 4: 437–450. Transgenic MUC1 interacts with EGFR and correlates with MAP Hollingsworth MA, Swanson BJ. (2004). Mucins in cancer: protection kinase activation in the mouse mammary gland. J Biol Chem 276: and control of the cell surface. Nat Rev Cancer 4: 45–60. 13057–13064.

Oncogene MUC1 initiates EMT in pancreatic cancer LD Roy et al 1459 Simpson-Herren L, Lloyd HH. (1970). Kinetic parameters and growth toward immunosuppression in a mouse model of spontaneous curves for experimental tumor systems. Cancer Chemother Rep 54: pancreatic adenocarcinoma. J Immunol 181: 3116–3125. 143–174. Wen Y, Caffrey TC, Wheelock MJ, Johnson KR, Hollingsworth MA. Singh PK, Hollingsworth MA. (2006). Cell surface-associated mucins (2003). Nuclear association of the cytoplasmic tail of MUC1 and in signal transduction. Trends Cell Biol 16: 467–476. beta-catenin. J Biol Chem 278: 38029–38039. Spicer AP, Rowse GJ, Lidner TK, Gendler SJ. (1995). Delayed Wolpin BM, Michaud DS, Giovannucci EL, Schernhammer ES, mammary tumor progression in Muc-1 null mice. J Biol Chem 270: Stampfer MJ, Manson JE et al. (2007). Circulating insulin-like 30093–30101. growth factor axis and the risk of pancreatic cancer in four Thiery JP. (2002). Epithelial-mesenchymal transitions in tumour prospective cohorts. Br J Cancer 97: 98–104. progression. Nat Rev Cancer 2: 442–454. Yamamoto M, Bharti A, Li Y, Kufe D. (1997). Interaction of the Thompson EJ, Shanmugam K, Hattrup CL, Kotlarczyk KL, Gutierrez DF3/MUC1 breast carcinoma-associated antigen and beta-catenin A, Bradley JM et al. (2006). Tyrosines in the MUC1 cytoplasmic tail in cell adhesion. J Biol Chem 272: 12492–12494. modulate transcription via the extracellular signal-regulated kinase Zhou BP, Deng J, Xia W, Xu J, Li YM, Gunduz M et al. (2004). 1/2 and nuclear factor-kappaB pathways. Mol Cancer Res 4: 489–497. Dual regulation of snail by GSK-3beta-mediated phosphorylation Tinder TL, Subramani DB, Basu GD, Bradley JM, Schettini J, Million in control of epithelial-mesenchymal transition. Nat Cell Biol 6: A et al. (2008). MUC1 enhances tumor progression and contributes 931–940.

Oncogene