p120 Catenin Suppresses Basal Epithelial Cell Extrusion in Invasive Pancreatic Neoplasia

Genetic and epigenetic alterations in genes encoding cell adhesion molecules are a hallmark of many epithelial cancers. For pancreatic cancer, homophilic cell adhesion has been categorized as one of 12 core signaling pathways (1). Homophilic cell adhesion in epithelial cells is mediated partly through cadherins and catenins in adherens junctions. Misexpression of the adherens junction protein p120 catenin has been identified in several types of human carcinomas (2). Studies have suggested variable roles for p120 catenin in the pathogenesis of epithelial cancers, to include tumor suppression and metastatic progression (3, 4). In human pancreatic cancer, misexpression of p120 catenin in primary tumors is significantly correlated with vascular invasion, metastasis, differentiation, pTNM stage, and poor survival (5–7). Reduction and cytoplasmic relocalization of p120 catenin has also been reported in 100% of solid pseudopapillary tumors of the pancreas (8). A study using a forward genetic screen in mice identified Ctnnd1 as a “candidate cancer gene” in Kras-driven pancreatic neoplasia (7). These data suggest that disruption of CTNND1 in pancreatic tumors has biologic relevance to disease, yet, the mechanisms by which p120 catenin contributes to the development and progression of pancreatic cancer are not understood.

Increased occurrence of metastasis with altered p120 catenin expression suggests that p120 catenin may play a role in metastatic progression of pancreatic cancer. A mechanism recently hypothesized to initiate metastasis by mediating invasion is basal epithelial cell extrusion (9, 10). Epithelial tissues maintain homeostatic cell numbers by extruding cells through a highly conserved mechanism involving production and secretion of the signaling lipid sphingosine 1-phosphate (S1P). Extracellular S1P binds to S1P receptor 2 (S1pr2) on neighboring cells, which induces contraction of an actomyosin band that extrudes the cell out of the epithelium while preserving barrier function (11–17). Mutations in the tumor suppressor APC and oncogenic Kras have been shown to shift the predominant direction of epithelial cell extrusion from apical to basal, where extruded cells invade the underlying epithelium and survive (18, 19). The results presented in this study show that p120 catenin restrains epithelial cell extrusion in the earliest stages of pancreatic neoplastic invasion, via a S1P/S1pr2–dependent mechanism.

For expression analysis, a labeling score of 0–2 corresponding to absent/low, medium, and high was assigned using IHC, and immunofluorescence staining or IHC was scored for predominant subcellular localization analysis. p120 catenin expression level and predominant subcellular localization were each scored by three independent observers blinded to lesion classification. In this study, predominant is defined as greater than or equal to 60% of the representative staining pattern or expression level.

Transgenic mouse strains Mist1^CreER/+ (C^iMist1; ref. 20), Ctnnd1^tm1Abre (p120^f; ref. 21), and lox-stop-lox; Kras^G12D (K; ref. 22) have been described previously. To perform lineage tracing, we used the Rosa26^mTmG (G; ref. 23) double fluorescent reporter allele. Transgenic strains Ptf1atm1(cre)Wri (C^Ptf1a; ref. 24), lox-stop-lox; KrasG12D (K; ref. 22), p53LoxP (P; ref. 25), and R26R-YFP (Y; ref. 26) were used to generate KPC^Ptf1aY mice, which were sacrificed between 6 and 8 weeks of age and maintained on a mixed genetic background. C^iMist1, p120^f, K, and G mice were maintained on a C57BL/6J background. To induce Cre recombination, mice were injected with 5 mg tamoxifen (Sigma, T5648) subcutaneously once per day for 3 consecutive days. Experimental pancreatitis was elicited as described previously (27). Mice were genotyped by PCR or Transnetyx. All pancreatic pathologies in transgenic mice and humans were classified by a pathologist. For NF-κB inhibition experiments, mice were injected intraperitoneally with 5 mg/kg/day SN50 (Santa Cruz Biotechnology, sc-3060). All animal studies were approved by the Animal Care and Use Committees at Johns Hopkins University (Baltimore, MD) and University of Texas Health Science Center at Houston (Houston, TX).

Tissues were fixed in 4% paraformaldehyde at 4°C, processed according to standard protocols, and embedded in paraffin. Antigen retrieval was performed using heat-mediated microwave methods and an antigen unmasking solution (Vector Laboratories, H-3300) for all antibodies except rat-anti-CD45, for which Retrievit 6 (BioGenex, BS-1006-00) was used. All sections were blocked in 10% FBS in PBST and primary antibodies were incubated overnight at 4°C. Secondary antibodies, from Jackson Immunoresearch, were used at 1:250 and incubated at room temperature for 2 hours for immunofluorescence and 30 minutes for IHC. For immunofluorescence, slides were stained with IHC-Tek Dapi counterstain solution (IHC World, IW-1404) and mounted in fluorescence mounting medium (Dako, S3023). For IHC, Vectastain Elite ABC Kit (Vector Laboratories, PK-6100) and DAB Peroxidase (HRP) Substrate Kit (Vector Laboratories, SK-4100) were used. Primary antibodies used in this study are described in Supplementary Table S1.

Acinar-to-ductal metaplasia (ADM), pancreatic intraepithelial neoplasia (PanIN)1, PanIN2/3, and fibrostroma were quantified using morphometric analysis on scanned hematoxylin and eosin (H&E) slides in ImageJ. Two sections per animal sampled at least 400 μm apart were analyzed. Quantification of pancreatic area excluded lymph nodes.

For quantification of CK19⁺ basal cell extrusion, CK19⁺ cells (excluding apically extruded CK19⁺ cells and normal pancreatic ducts) were counted in 1 scanned section per animal. For quantification of CK19⁺ apical cell extrusion, CK19⁺ cells that comprised a luminal pancreatic epithelial structure (lumen sized at least twice the diameter of a cell comprising the epithelial structure and excluding normal ducts) and its associated apically extruded CK19⁺ cells were counted in 1 scanned section per animal.

Pancreatic injury, defined as area containing metaplastic duct lesions and/or associated stroma, was quantified in 1 scanned H&E section per animal using morphometric analysis in ImageJ. Quantification of pancreatic area excluded lymph nodes.

DNA ploidy cell-cycle analysis of basally extruded single epithelial cells on 4-μm thick Feulgen-stained sections was accomplished using OTMIAS Version 2.0 Image Analysis Software by Olive Tree Media, LLC. Serial sections stained by CK19 IHC were used for the identification of isolated, basally extruded epithelial cells in Feulgen-stained sections. The internal reference control and isolated epithelial cells analyzed were located on the same Feulgen-stained section. An aneuploid peak is defined as any distinct peak with a DNA index > 1.25. Abnormal DNA content is defined as any aneuploid peak or any peak ≥5C.

Adult pancreatic cells were dissociated as described previously (28). RNA was isolated from sorted GFP⁺ cells using Arcterus PicoPure RNA isolation kit and gene expression was analyzed using Mouse exon microarray 1.0 ST (Affymetrix). For qPCR experiments, reverse transcription was accomplished using QuantiTect reverse transcription kit (Qiagen, 205311). Complementary DNA was amplified using TaqMan gene expression assays (Life Technologies).

CFPAC-1 and AsPC-1 cells were obtained from and authenticated by ATCC using morphology, karyotyping, and PCR-based approaches. Cells were maintained in DMEM, low glucose, GlutaMAX Supplement (Life Technologies, 10567-022) supplemented with 10% FBS (Sigma-Aldrich, F4135-500ML) and 1× penicillin–streptomycin–glutamine (Fisher Scientific, 10378-016) at 5% CO₂, 37°C. For culturing CFPAC-1 cells in Matrigel, a single-cell suspension was resuspended in 4% Matrigel (BD Biosciences, 356234) at a final concentration of 6 × 10⁴ cells/mL. Three hundred μL cells per well were placed in 24-well glass bottom plates (In Vitro Scientific, P24-1.5H-N) coated with a thin polymerized layer of Matrigel. For immunostaining, cells were fixed in 4% paraformaldehyde at 37°C for 20 minutes, permeabilized for 10 minutes with 0.2% Triton X-100, blocked with 10% FBS in PBST for 30 minutes, and incubated with primary antibodies for 2 hours at room temperature. Subsequently, cells were incubated with secondary antibodies (Jackson Immunoresearch) at room temperature for 2 hours. DAPI was used for nuclear staining and cells were mounted in fluorescence mounting medium (Dako, S3023).

Cell extracts were prepared according to standard protocols using cell lysis buffer (Cell Signaling Technology, 9803S) with protease inhibitor cocktail tablets (Roche, 4693159001) and 100 mmol/L phenylmethylsulfonylfluoride. Membranes were incubated with primary antibodies overnight at 4°C. After incubation at room temperature with the respective HRP-conjugated secondary antibody used at 1:5000 dilution for 1 hour, membranes were developed using either the SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific, 34080) or the SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific, 34095). Quantification of Western blot images was performed in ImageJ.

CFPAC-1 cells grown in Matrigel-formed spheres by day 2 and were transfected on day 2 using Lipofectamine RNAiMAX (Invitrogen, 56532). The final concentration of siRNA used per well was 5 pmol. siRNA against p120 catenin and control siRNA were obtained from Santa Cruz Biotechnology (sc-36139 and sc-37007, respectively). For S1pr2 agonist experiments, cells were treated with 10 μmol/L CYM-5520 (Sigma, SML1014-25MG) on day 4, 48 hours after siRNA transfection. CFPAC-1 spheres were fixed on day 6, 96 hours after transfection, for analysis. Day 0 is defined as the day of plating the cells in Matrigel. For in vitro SN50 experiments, AsPC-1 cells were transfected using Lipofectamine RNAiMAX in a 6-well plate with a final siRNA concentration of 75 pmol. Twenty-four hours after siRNA transfection, cells were treated with either dH₂0 or 18 μmol/L SN50 for 24 hours. Forty-eight hours after siRNA transfection, cells were harvested.

The Gene Expression Omnibus accession number for the microarray analysis reported in this article is GSE68090.

Data are presented as mean ± SEM and were analyzed in GraphPad Prism or Microsoft Office Excel. Statistical significance was assumed at a P value of ≤0.05. P values were calculated with the unpaired t test unless indicated otherwise. For interpretation of statistical results from unpaired t test, *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; and ****, P ≤ 0.0001.

Expression and localization of p120 catenin, a critical cytoskeletal regulator and component of adherens junctions, were rated in human normal pancreas, chronic pancreatitis, ADM, PanIN 1–3, and primary tumors and metastases. Examples of scores in each category are presented in Supplementary Fig. S1A. p120 catenin is expressed in all normal pancreatic cell types (Fig. 1A and B). Localization of p120 catenin was scored as predominantly membranous in 100% of normal pancreatic cell types (n = 22–33 patients for each pancreatic cell type; Fig. 1J). Punctate cytoplasmic and nuclear staining were also observed in normal pancreatic cells (Fig. 1A and B). p120 catenin expression was rated as high in >91% chronic pancreatitis and ADM (n = 25/26 chronic pancreatitis, n = 23/25 ADM; Fig. 1C and I). Expression levels of p120 catenin were significantly decreased in primary tumors when compared with PanIN lesions (n = 31 PanIN and n = 16 primary tumors; Fig. 1D–G and I). Moreover, p120 catenin expression was significantly lower in metastases when compared with primary tumors (n = 16 primary tumors and n = 16 metastasis; Fig. 1G–I). There was also a significant difference in predominant p120 catenin subcellular localization when comparing PanIN1-2 and PanIN3 (n = 26 PanIN1–2 and n = 5 PanIN3). Together, these findings demonstrate that altered p120 catenin expression and localization is a distinguishing hallmark of human pancreatic cancer progression.

Previously, using C^Pdx1; p120^f/f mice, we reported that p120 catenin is required for proper tubulogenesis and cell-type specification during pancreas development (29). To determine the function of p120 catenin in adult mouse pancreas in the absence of a confounding developmental phenotype, we crossed transgenic mice harboring floxed alleles of p120 catenin (p120^f; ref. 21) with Mist1^CreER/+ (C^iMist1) mice (20). p120^f/f, C^iMist1;p120^wt/wt, C^iMist1;p120^f/wt, and C^iMist1;p120^f/f mice displayed normal pancreatic histology 2–4 months after tamoxifen injection (Supplementary Fig. S1B). Similar to what we previously reported for C^Pdx1; p120^f/f mice 10 months of age (29), a subset of C^iMist1;p120^f/f mice 12 months arfter tamoxifen injection exhibited pancreatitis and ADM (Supplementary Fig. S1B).

As we observed mislocalized p120 catenin expression in human PanIN, before the onset of pancreatic cancer, we next sought to determine whether p120 catenin plays a functional role in PanIN formation and progression. To this end, we crossed C^iMist1;p120^f/f mice with lox-stop-lox-Kras^G12D (K) mice (22), which resulted in simultaneous ablation of p120 catenin and activation of oncogenic Kras^G12D in adult pancreatic acinar cells upon tamoxifen administration (Fig. 2A). The KC^iMist1 mouse model displays the full spectrum of murine premalignant ADM and PanIN1–3 lesions in a manner that faithfully recapitulates human premalignant pancreatic lesions (20). Survival analysis on cohorts of KC^iMist1; p120^wt/wt, KC^iMist1; p120^f/wt, and KC^iMist1; p120^f/f mice showed significant differences in overall survival (Fig. 2B). KC^iMist1; p120^f/f mice exhibited cachexia, which is also frequently observed in human pancreatic cancer patients (Fig. 2C and D). The gross appearance of KC^iMist1; p120^f/f pancreata was strikingly abnormal and enlarged when compared with KC^iMist1; p120^wt/wt pancreata (Fig. 2E–G).

As expected, p120 catenin was ubiquitously expressed in KC^iMist1; p120^wt/wt and KC^iMist1; p120^f/wt pancreata (Supplementary Fig. S2A and S2B). Minimal mosaic expression of p120 catenin was observed in KC^iMist1; p120^f/f pancreatic acini (Supplementary Fig. S2C). KC^iMist1; p120^f/f pancreata displayed marked acinar cell atrophy, pronounced inflammation, and contained stroma characterized by a unique cellular constitution that differs from the stroma in KC^iMist1; p120^wt/wt pancreata (Supplementary Fig. S2D–S2L). KC^iMist1; p120^f/f pancreata displayed less mucinous lesions than KC^iMist1; p120^wt/wt and KC^iMist1; p120^f/wt pancreata, as manifested by Alcian blue staining (Supplementary Fig. S1C). KC^iMist1; p120^f/f pancreata also showed areas of ductal dilation (Supplementary Fig. S2M–S2O). Disruption of contiguous basement membrane Laminin expression, a characteristic of human pancreatic cancer (30), was also seen in KC^iMist1; p120^f/f pancreatic lesions. This was accompanied by cells that escaped intact PanIN epithelium and invaded into the underlying tissue (Supplementary Fig. S2P–S2R).

Histologically, significant increases in ADM, low- and high-grade PanIN formation, and fibrostroma were observed in KC^iMist1; p120^f/f pancreata when compared with KC^iMist1; p120^f/wt and KC^iMist1; p120^wt/wt pancreata beginning 2 weeks after tamoxifen injection (Fig. 3A–C′ and Supplementary Fig. S1D and D′). Two months after tamoxifen injection, KC^iMist1; p120^f/f pancreata displayed significantly increased fibrostroma (75.69% ± 9.32% pancreatic area, n = 4 mice) when compared with KC^iMist1; p120^f/wt pancreata (0.55% ± 0.00% pancreatic area, n = 2 mice; P = 0.0058) and KC^iMist1; p120^wt/wt pancreata (1.09% ± 0.54% pancreatic area, n = 4 mice; P = 0.0002).

Previously, we and others have demonstrated that p120 catenin loss results in immune cell infiltration and inflammation, which is mediated in part by NF-κB activation (29, 31). We next queried whether NF-κB signaling contributed to the formation of ADM/PanIN/fibrostroma in KC^iMist1; p120^f/f mice. To this end, we treated KC^iMist1; p120^f/f mice with SN50, a potent inhibitor of NF-κB activation, and observed a significant reduction in ADM/PanIN/fibrostroma in SN50-treated KC^iMist1; p120^f/f mice when compared with controls (Fig. 3D–F). These data establish NF-κB signaling as a mechanism by which p120 catenin loss promotes increased ADM, PanIN, and inflammation in KC^iMist1; p120^f/f mice.

The exit of epithelial cells across the basement membrane of discernable epithelial structures, a process termed delamination, occurs in KPC^Pdx1Y and KC^iMist1Y PanIN mice and has been associated with epithelial-to-mesenchymal transition (EMT; ref. 32). As loss of p120 catenin in pancreatic cancer cell lines results in increased migration and invasion, also associated with EMT (33), we next sought to determine whether p120 catenin regulated delamination in PanIN mice. First, we examined expression of p120 catenin and E-cadherin in delaminated cells in two lineage-traced murine models of PanIN: KPC^Ptf1aY and KC^iMist1G (34, 35). Delaminated cells in KPC^Ptf1aY PanIN mice expressed decreased adherens junction proteins p120 catenin and E-cadherin when compared with their surrounding pancreatic epithelia (Supplementary Fig. S3A). KC^iMist1G; p120^wt/wt PanIN mice showed decreased p120 catenin and E-cadherin expression in non-epithelial delaminated cells, some of which have an elongated fibroblast cell morphology (Fig. 4A). These data show that decreased expression of adherens junction proteins p120 catenin and E-cadherin is a manifest feature of delaminated cells in PanIN mice.

To further investigate the role of p120 catenin in delamination in Kras-induced premalignant pancreatic neoplasia, we next examined delaminated cells in lineage-traced KC^iMist1G; p120^f/wt and KC^iMist1G; p120^f/fmice. Monoallelic and biallelic loss of p120 catenin resulted in remarkable abundant delamination of GFP⁺ cells that retained E-cadherin (Fig. 4B and C). These data show that p120 catenin is not required for the maintenance of E-cadherin localization to cell membranes in the context of oncogenic Kras. Both delamination, also termed basal epithelial cell extrusion (16), as well as apical epithelial cell extrusion, were significantly increased in KC^iMist1; p120^f/wt and KC^iMist1; p120^f/f pancreata when compared with KC^iMist1; p120^wt/wt pancreata (Fig. 4D–H). Abundant apical and basal epithelial cell extrusion was 100% penetrant in KC^iMist1; p120^f/wt and KC^iMist1; p120^f/f pancreata, with the phenotype evident in KC^iMist1; p120^f/f pancreata by 2 weeks after tamoxifen injection (Supplementary Fig. S3B). Quantification of extruded isolated CK19⁺ cells revealed 838 of 7,000 in KC^iMist1; p120^f/f pancreata, 76 of 7,000 in KC^iMist1; p120^f/wt pancreata, and 19 of 7,000 in KC^iMist1; p120^wt/wt pancreata (Supplementary Fig. S3C). Lineage tracing in KC^iMist1; p120^f/f pancreata showed extruded GFP⁺, CK19⁺ single cells were negative for Vimentin, a marker of mesenchymal differentiation and EMT (Supplementary Fig. S3D), suggesting that basal epithelial cell extrusion resulting from biallelic p120 catenin loss is not associated with incomplete EMT. Furthermore, treatment of KC^iMist1; p120^f/f mice with the NF-κB inhibitor SN50 significantly reduced basal epithelial cell extrusion when compared with controls (Fig. 4I). These data suggest that NF-κB signaling is a component of the regulatory network that controls basal epithelial cell extrusion in KC^iMist1; p120^f/f mice.

Biallelic loss of p120 catenin during pancreas development promotes apical epithelial cell extrusion, and luminal cell extrusion has also been reported with p120 catenin loss in developing kidney and MDCK cysts (29, 36). Given the abundant apical and basal extrusion observed in KC^iMist1; p120^f/f pancreata, we next sought to interrogate the extrusion behavior of adult pancreatic cells lacking p120 catenin in an experimental model of acute pancreatitis. Acute pancreatitis was induced in C^iMist1; p120^wt/wt, C^iMist1; p120^f/wt, and C^iMist1; p120^f/f mice (Fig. 5A). C^iMist1; p120^f/f pancreata showed significantly increased susceptibility to injury and inflammation when compared with C^iMist1; p120^wt/wt pancreata (Supplementary Fig. S4A–S4J). As activation of NF-κB signaling augments the severity of pancreatitis (37), we next assessed the role of NF-κB activation in C^iMist1; p120^f/f and control mice. Inhibition of NF-κB signaling significantly reduced injury in C^iMist1; p120^f/f pancreata to levels similar to C^iMist1; p120^wt/wt dH₂0–treated controls (Supplementary Fig. S4K and S4L), establishing NF-κB signaling as a direct mechanism by which p120 catenin loss promotes increased susceptibility to injury in C^iMist1; p120^f/f pancreata.

Quantification of apical and basal epithelial cell extrusion revealed significant increases in C^iMist1; p120^f/f and C^iMist1; p120^f/wt pancreata when compared with C^iMist1; p120^wt/wt pancreata (Fig. 5B–D). These data suggest that p120 catenin regulates both apical and basal epithelial cell extrusion in adult mouse pancreas in the context of acute pancreatitis. Furthermore, C^iMist1G; p120^f/f mice treated with the NF-κB inhibitor SN50 show significantly reduced basal epithelial cell extrusion (Fig. 5E), demonstrating that activation of NF-κB regulates extrusion in the setting of p120 catenin loss and experimental pancreatitis.

We next sought to determine the fate of both apically and basally extruded epithelial cells in KC^iMist1; p120^wt/wt, KC^iMist1; p120^f/wt, and KC^iMist1; p120^f/f pancreata. As epithelial cells can extrude apically by a mechanism involving S1P/S1pr2 signaling and activation of cleaved caspase-3 (11, 14), we next examined cleaved caspase-3 expression in extruded CK19⁺ cells. Consistent with this previously reported mechanism of apical cell extrusion, we observed cleaved caspase-3 expression in apically extruded CK19⁺ cells, but not in basally extruded CK19⁺ cells in KC^iMist1; p120^wt/wt, KC^iMist1; p120^f/wt, and KC^iMist1; p120^f/f pancreata (Supplementary Fig. S5A–S5C), suggesting that basally extruded CK19⁺ cells exit intact pancreatic epithelium and remain viable.

Because we observed prominent nucleoli and nuclear enlargement of basally extruded epithelial cells in KC^iMist1; p120^f/f pancreata, we next analyzed the DNA content of isolated CK19⁺ cells in KC^iMist1; p120^f/f pancreata using OTMIAS Image Analysis Software on Feulgen-stained slides (Supplementary Fig. S6A–S6C). A population of pancreatic cells was observed with a DNA index of 1.5, which was indicative of aneuploidy (Supplementary Fig. S6C). In addition, 8.3% (21/253) pancreatic cells analyzed showed abnormal DNA content with a DNA index of ≥2.5 (Supplementary Fig. S6C). The histology of KC^iMist1; p120^f/f pancreata thus comprises a very unique phenotype with overall benign neoplasia and reactive stroma containing isolated epithelial cells that display features of malignancy including enlarged, hyperchromatic, and pleomorphic nuclei, prominent nucleoli, aneuploidy and occasional binuclear cells. These findings suggest that p120 catenin loss in the context of oncogenic Kras promotes formation of invasive pancreatic neoplasia (Supplementary Fig. S6D).

As we identified that loss of p120 catenin in cooperation with oncogenic Kras promotes invasion of epithelial cells displaying characteristics of malignancy in pancreatic neoplasia, we next examined expression of p120 catenin in isolated epithelial cells in human pancreatic ductal adenocarcinoma (PDA). Single malignant epithelial cells in human PDA are depicted in Supplementary Fig. S7A and S7B. Quantification of p120 catenin subcellular localization in 253 isolated epithelial cells from 17 patients with PDA showed a sparse 4.74% of cells with normal membrane labeling and 95.26% of cells with predominant cytoplasmic or absent p120 catenin localization (Supplementary Fig. S7C and S7D). These data show that altered p120 catenin subcellular localization is a distinctive feature of isolated malignant epithelial cells in human PDA.

As a means to identify a molecular basis by which p120 catenin ablation in cooperation with oncogenic Kras promotes epithelial cell extrusion, we performed whole transcriptome analysis on fluorescence-activated cell sorted (FACS) GFP⁺ pancreatic cells in KC^iMist1G; p120^wt/wt and KC^iMist1G; p120^f/f mice 2 weeks after tamoxifen injection (Supplementary Fig. S8A). IPA analysis showed 56 statistically significant differentially expressed pathways, several of which are related to actin cytoskeleton signaling, the inflammatory response, and cell adhesion and migration (Supplementary Fig. S8B). We next sought to validate these results by examining expression of select genes in each of these categories using IHC. Previously, we showed that actin cytoskeleton organization was disrupted in C^Pdx1; p120^f/f pancreata, and that this observation was associated with increased cytoplasmic PKCζ, a known modulator of actin cytoskeleton dynamics (29). Here, we similarly find that KC^iMist1; p120^f/f pancreata show increased cytoplasmic PKCζ when compared with KC^iMist1; p120^f/wt and KC^iMist1; p120^wt/wt pancreata (Supplementary Fig. S8C–S8E). Expression of adherens junction components E-cadherin and β-catenin was also reduced in KC^iMist1; p120^f/f pancreata (Supplementary Fig. S8F–S8K). In addition, IHC confirmed intrinsic activation of NF-κB in KC^iMist1; p120^f/f pancreata (Supplementary Fig. S8L–S8N).

Asa defective S1P/S1pr2–mediated cell extrusion has been shown to shift the predominant direction of cell extrusion from apical to basal (19), we next queried whether p120 catenin regulated S1P/S1pr2 signaling. We performed qPCR on FACS-sorted GFP⁺ pancreatic cells at 1 month after tamoxifen injection, which showed an overall decrease in expression of genes mediating S1P/S1pr2 signaling in KC^iMist1G; p120^f/f mice when compared with KC^iMist1G; p120^wt/wt mice (Fig. 6A). The expression of genes involved in the biosynthetic pathway of S1P, Sphk1 and Sphk2, was significantly decreased −5.83 fold and −3.96 fold, respectively, suggesting that p120 catenin regulates biosynthesis of S1P in a mutant Kras-dependent context.

We next sought to understand how p120 catenin regulated expression of S1P/S1pr2 pathway members. As we observed activation of NF-κB in KC^iMist1; p120^f/f mice, we hypothesized that this regulation may occur though NF-κB signaling. Similar to our observations in KC^iMist1; p120^f/f mice, AsPC-1 cells treated with p120 catenin siRNA showed significantly reduced expression of S1pr2 (Fig. 6B and C). Inhibition of NF-κB activation with SN50 completely restored expression of S1pr2 in p120 catenin siRNA-treated AsPC-1 cells, suggesting that p120 catenin regulates expression of S1pr2 through activation of NF-κB (Fig. 6B and C).

We next investigated the relationship between p120 catenin loss and epithelial cell extrusion mediated by S1P/S1pr2 signaling using an epithelial, Kras mutant, human pancreatic cancer cell line. CFPAC-1 cells express both p120 catenin and S1pr2 and form spheres (CFPAC-1 spheres) that can extrude cells basally when grown in Matrigel (Fig. 6D and E). The percentage of CFPAC-1 spheres extruding epithelial cells basally increased significantly following p120 catenin knockdown (Fig. 6F). Furthermore, the specific S1pr2 agonist CYM-5520 (38) significantly decreased the frequency of basal extrusion in p120 catenin-deficient spheres to levels similar to those observed in p120 catenin–expressing spheres (Fig. 6D). The ability of restored S1P/S1pr2 signaling to rescue the p120 catenin loss of function phenotype indicates a direct mechanism by which p120 catenin loss promotes increased basal epithelial cell extrusion.

The critical role of the cell–cell adhesion apparatus during tumorigenesis of epithelial cancers is firmly established (1, 39). The association of p120 catenin with E-cadherin at epithelial cell membranes is crucial for formation and maintenance of adherens junctions (40). p120 catenin loss or mislocalization can destabilize E-cadherin and affect the adhesive repertoire of the cell and its signal transduction status. In KC^iMist1; p120^f/f mice, simultaneous p120 catenin and E-cadherin loss likely destabilizes cell adhesion and promotes migration and invasion.

Studies have shown that p120 catenin can function as a bona fide tumor suppressor in murine oral cavity, esophagus, and forestomach (3). In contrast, targeted knockout of p120 catenin in murine salivary gland (21), intestine (31, 41), brain (42), mammary gland (43), kidney (36), eye (44), epidermis (31, 45), and lung (46) is not sufficient to promote formation of cancer. It is therefore likely that the requirement for p120 catenin in development and progression of cancer is tissue-specific. For pancreas, ablation of p120 catenin in pancreatic progenitor cells in C^Pdx1; p120^f/f mice (29) and somatic knockout of p120 catenin in adult pancreatic acinar cells in C^iMist1;p120^f/f mice does not result in development of pancreatic cancer, so evidence for a bona fide tumor suppressor role for p120 catenin in pancreas is lacking.

The inflammation suppression function of p120 catenin is well documented (47). Loss of p120 catenin generates a microenvironment disposed to chronic inflammation in normal, injured, and neoplastic pancreas. We show that formation of ADM/PanIN/fibrostroma in KC^iMist1; p120^f/f mice and injury in C^iMist1G; p120^f/f mice is mediated, in part, through activation of NF-κB. Dexamethasone, which has anti-inflammatory and immunosuppressant effects (including inhibition of NF-κB activation), has been shown to reduce ADM and PanIN formation in mice (32), suggesting a potential link between inflammation and premalignant lesion formation.

We show loss of p120 catenin in delaminated cells associated with EMT and non-EMT in PanIN mice. While loss of p120 catenin precedes non-EMT–associated delamination in KC^iMist1G; p120^f/f pancreata, it is unclear whether loss of p120 catenin precedes or occurs after delamination associated with EMT in KPC^Ptf1aY and KC^iMist1G; p120^wt/wt pancreata. Actin cytoskeleton remodeling events are necessary for EMT, and loss of cytoplasmic p120 catenin before occurrence of delamination may prevent cytoskeletal restructuring required for delamination associated with EMT. Biallelic loss of p120 catenin results in significant downregulation of genes mediating S1P/S1pr2 signaling in PanIN mice and AsPC-1 cells, an observation that we demonstrate in vitro is mediated partly through activation of NF-κB. Activation of S1pr2 in vitro completely rescues increased basal epithelial cell extrusion seen with p120 catenin loss. These data suggest that p120 catenin loss in the context of oncogenic Kras may promote neoplastic epithelial cell invasion in part by altering S1P/S1pr2 signaling through activation of NF-κB (Fig. 7).

Few mechanisms for non-EMT–associated delamination are described, yet the evidence that EMT is not absolutely required for invasion, dissemination, and metastasis is emerging (48–50). In summary, we have created a model in which cooperating genetic insults have unraveled a new mechanism for neoplastic epithelial cell invasion in premalignant pancreatic cancer. The evidence that p120 catenin regulates epithelial cell extrusion through activation of NF-κB is compelling, as p120 catenin loss affects this biologic process in injured and neoplastic pancreata. Further studies are needed to clarify the metastatic potential associated with monoallelic and biallelic p120 catenin loss in different pancreatic cell types and how these characteristics determine the evolution of PDA.

M. Younes is a president and has ownership interest (including patents) in the Olive Tree Media, LLC. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.M. Hendley, A. Maitra, S.D. Leach, J. Bailey

Development of methodology: A.M. Hendley, Y.J. Wang, M. Younes, J. Bailey

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.M. Hendley, Y.J. Wang, J. Alsina, I. Ahmed, K.J. Lafaro, H. Zhang, N. Roy, S.G. Savidge, Y. Cao, M. Hebrok, A. Reynolds, M.G. Goggins, C.A. Iacobuzio-Donahue, J. Bailey

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.M. Hendley, K.J. Lafaro, H. Zhang, Y. Cao, A. Maitra, M. Younes, C.A. Iacobuzio-Donahue, S.D. Leach, J. Bailey

Writing, review, and/or revision of the manuscript: A.M. Hendley, K. Polireddy, K.J. Lafaro, A. Maitra, M.G. Goggins, M. Younes, S.D. Leach, J. Bailey

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.M. Hendley, Y. Cao, M.G. Goggins, S.D. Leach, J. Bailey

Study supervision: S.D. Leach, J. Bailey

The authors thank Melissa Pruski, Kanchan Singh, Neal C. Jones, Ka Liu, Anzer Habibulla, Mara Swaim, Jeffrey Roeser, and Qingfeng Zhu for their indispensable help in maintaining mouse colonies, genotyping, and technical assistance. The authors thank Florencia McAllister for helpful intellectual discussions. The authors also thank Conover Talbot Jr. for crucial aid with analysis of microarray gene expression.

This work was supported by NIH P01 CA134292, RO1 DK097087, and R21 CA158898 (S.D. Leach). J.M. Bailey was supported by an AACR/Pancreatic Cancer Action Network Pathway to Leadership Award and by Texas Medical Center Digestive Diseases Center grant (DK56338).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.