β-catenin activation down-regulates cell-cell junction-related genes and induces epithelial-to-mesenchymal transition in colorectal cancers

WNT signaling activation in colorectal cancers (CRCs) occurs through APC inactivation or β-catenin mutations. Both processes promote β-catenin nuclear accumulation, which up-regulates epithelial-to-mesenchymal transition (EMT). We investigated β-catenin localization, transcriptome, and phenotypic differences of HCT116 cells containing a wild-type (HCT116-WT) or mutant β-catenin allele (HCT116-MT), or parental cells with both WT and mutant alleles (HCT116-P). We then analyzed β-catenin expression and associated phenotypes in CRC tissues. Wild-type β-catenin showed membranous localization, whereas mutant showed nuclear localization; both nuclear and non-nuclear localization were observed in HCT116-P. Microarray analysis revealed down-regulation of Claudin-7 and E-cadherin in HCT116-MT vs. HCT116-WT. Claudin-7 was also down-regulated in HCT116-P vs. HCT116-WT without E-cadherin dysregulation. We found that ZEB1 is a critical EMT factor for mutant β-catenin-mediated loss of E-cadherin and Claudin-7 in HCT116-P and HCT116-MT cells. We also demonstrated that E-cadherin binds to both WT and mutant β-catenin, and loss of E-cadherin releases β-catenin from the cell membrane and leads to its degradation. Alteration of Claudin-7, as well as both Claudin-7 and E-cadherin respectively caused tight junction (TJ) impairment in HCT116-P, and dual loss of TJs and adherens junctions (AJs) in HCT116-MT. TJ loss increased cell motility, and subsequent AJ loss further up-regulated that. Immunohistochemistry analysis of 101 CRCs revealed high (14.9%), low (52.5%), and undetectable (32.6%) β-catenin nuclear expression, and high β-catenin nuclear expression was significantly correlated with overall survival of CRC patients (P = 0.009). Our findings suggest that β-catenin activation induces EMT progression by modifying cell-cell junctions, and thereby contributes to CRC aggressiveness.

Expression BeadChip (Illumina, San Diego, CA, USA). Hybridization of labeled-cRNA to BeadChip, washing, and scanning were performed according to Illumina Bead Station 500X manual. Extraction of mRNA expression data and statistical analysis of raw data were performed using software provided by the manufacturer (Illumina GenomeStudio v2011.1). Expression intensities were normalized by quantile normalization technique; and using normalized intensities, genes differentially expressed in HCT116-WT and HCT116-MT cells were determined by the integrated statistical method previously reported 14 . Our microarray data are deposited in Gene Expression Omnibus (GSE126845). Differentially expressed genes (DEGs) were selected as those with P < 0.05 and fold change > 1.5. Functional enrichment analysis was performed using BINGO 2.3 plugin for Cytoscape software (http://www.psb.ugent.be/cbd/papers/BiNGO/Home.html), DAVID software (Database for Annotation, Visualization an Integrated Discovery, v6.7; http://david.abcc.ncifcrf.gov) to identify gene ontology (GO) biological processes, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways represented by DEGs with statistical significance. cell cycle analysis. HCT116-P, HCT116-WT, and HCT116-MT cells were washed with PBS and fixed in 95% ethanol. Cells were then incubated at 4 °C for 24 h, washed again, and stained in a solution of PBS, propidium iodide, and RNase A in the dark at 37 °C for 20 min. All samples were analyzed by flow cytometry using FACS Calibur (BD Biosciences) system. Luciferase assays. Wnt/β-Catenin signaling activity was measured using TOP/FOP-Flash luciferase activity assay. TOP-Flash contained a wild type LEF/TCF-binding site, while FOP-Flash contained a mutant LEF/ TCF-binding site. HCT116 cells were transfected with reporter constructs (TOP-Flash or FOP-Flash). After 2 days, luciferase activity was measured using Dual-Luciferase Reporter Assay System Kit (Promega), according to manufacturer's instructions. Results are presented as relative luciferase activity.

Construction of β-catenin expression vectors and siRNAs for gene silencing. Coding sequences
of β-catenin were cloned into a pLECE3 vector conjugated with EGFP at C-terminus. For generation of a β-catenin Ser45del mutant construct, substitution mutagenesis was performed at 45 th amino acid using β-catenin wild-type (WT) construct. An EGFP-conjugated E-cadherin expression vector was purchased from Addgene (#Plasmid 20089, Watertown, MA, USA). For gene silencing experiments, cells were transfected with short interfering RNAs (siRNAs) targeting SNAI1, SNAI2, ZEB1, and TWIST1 (Bioneer, Daejeon, Korea). The siRNAs sequences are shown in Supplementary Table S1. Short hairpin RNAs (shRNAs) used in this study were obtained from a shRNA library of RNAi Consortium (TRC) provided by Yonsei genome center (Seoul, Korea). tissue microarray construction. Representative areas were selected based on microscopic examination of hematoxylin and eosin (H&E)-stained slides, and corresponding spots were marked on the surface of the paraffin block. Using a manual tissue microarrayer, selected areas were removed, and 3-mm tissue cores were placed into 6 × 5 recipient blocks. Tissues from invasive tumors were then extracted; two tissue cores were processed from each sample to minimize extraction bias and ensure each case was represented in the final tissue microarray (TMA) block. Each core was assigned a unique TMA location number that is linked to a database with other clinicopathological data. After microtome sectioning and H&E staining, final TMA blocks contained cores from 101/101 (100%) specimens.
Wound-healing and invasion assays. For wound-healing assays, 0.5 × 10 6 cells were seeded in 60-mm dishes and cultured until confluent. Using a (yellow) pipette tip, a straight scratch (wound) was generated, keeping the pipette tip at an angle of ~30°. Wound closure was then monitored over 48 h. For invasion assays, QCM 24-Well Cell Invasion assay kit was used (Millipore, Burlington, MA, USA), according to manufacturer's guidelines. Briefly, insert interiors were rehydrated with pre-warmed serum-free media, and 0.3 × 10 6 cells were added to each insert. Complete media (10% FBS) was added to the lower chambers, and cells were incubated for 48 h in a CO 2 incubator. Cells were stained with 0.1% crystal violet solution, and images were obtained using the 4× or 10× objective. tissue samples. A total of 101 stage III (metastasis to regional lymph nodes but not distant sites) CRC tissue samples that were treated with FOLFOX regimen were used in this study. This same cohort was analyzed in our previous study for molecular classification of CRCs 16 . All patients had undergone curative colorectal resection between 2006 and 2012, and none received neo-adjuvant chemotherapy. Specimens were obtained from the archives of Department of Pathology at Yonsei University in Seoul, Korea and from the Liver Cancer Specimen Bank of the National Research Resource Bank Program of the Korean Science and Engineering Foundation of the Ministry of Science and Technology. Patient data were collected retrospectively. All experimental protocols using patient tissues were approved by the Institutional Review Board of Yonsei University College of Medicine (approval number: 4-2018-0276) and conducted in accordance with all relevant policies, including obtaining informed consent from all subjects. Statistical analysis. Statistical analyses were performed using SPSS software, version 21.0.0.0 for Windows (IBM, Armonk, NY, USA). Mann-Whitney tests, Student's t-tests, Fisher's exact tests, or chi-square tests were used, depending on the purpose. Data are expressed as mean ± SD, and one-way analysis of variance (ANOVA) with a post hoc test (Bonferroni) was performed to compare multiple means. P-values < 0.05 were considered statistically significant.
Ethical approval and informed consent. Authorization for the use of tissues for research purposes was obtained from the Institutional Review Board of Yonsei University College of Medicine (approval number: 4-2018-0276).

Results
HCT-116 cells containing a β-catenin mutation show nuclear β-catenin localization and Wnt pathway activation. We utilized three HCT116 cell lines with differential β-catenin mutation status: parental HCT116 (HCT116-P) cells harboring one WT and one mutant allele (Ser45 del), HCT116-WT cells harboring one WT allele, and HCT116-MT cells harboring one mutant allele. As these lines are isogenic and differ only in their β-catenin mutation status, we considered this a proper model for investigating β-catenin mutation-induced phenotypes. Morphological analysis revealed clear differences. Specifically, HCT116-WT cells predominantly show epithelial-like morphology (round and organized sheet pattern), whereas HCT116-MT cells mostly display mesenchymal-like morphology (spindle-shaped and scattered pattern) (Fig. 1a). Both morphologies were heterogeneously observed in HCT116-P cells. Western blot and qRT-PCR analysis revealed similar levels of β-catenin mRNA and protein expression, respectively, in HCT116-WT and HCT116-MT cells, and these were about half the levels measured in HCT116-P cells, likely due to the single β-catenin allele in WT and MT lines ( Fig. 1b-d).
To determine β-catenin localization in each line, we first performed cell fractionation and western blot analyses, which revealed that HCT116-WT cells mostly show non-nuclear β-catenin expression, whereas HCT116-MT cells specifically display nuclear β-catenin expression. HCT116-P cells showed both nuclear and non-nuclear expressions (Fig. 1e). Subsequent immunofluorescence microscopy analysis confirmed that WT β-catenin in HCT116-WT cells mostly localizes on the plasma membrane, whereas mutant β-catenin in HCT116-MT cells is specifically found in the nucleus. HCT116-P cells display a mixed pattern of membranous, cytoplasmic, and nuclear localization (Fig. 1f). Wnt pathway activation was also strongly up-regulated in both HCT116-P and HCT116-MT cells, as compared to HCT116-WT cells (Fig. 1g). We further analyzed the expression of Wnt/β-catenin signaling target genes, CD44, MMP7, CCND1, CCNE2, BMP4, and AXIN2. qPCR analysis showed that the expression of BMP4, AXIN2, and CCNE2 was strongly up-regulated in HCT116-P and HCT116-MT cells compared to that of HCT116-WT cells. CCND1 expression was slightly up-regulated in HCT116-MT and HCT116-P cells. On the other hand, CD44 expression was higher in HCT116-WT than that in other cell lines, and MMP7 expression was specifically up-regulated in HCT116-P cells ( Supplementary Fig. S1). Overall, these results indicate that WT β-catenin is stably expressed on the cell surface where it participates in cell-cell junctions, whereas mutant β-catenin shows nuclear expression and functions as a transcriptional regulator of some of the reported Wnt/β-catenin signaling target genes according to the β-catenin mutation status in HCT116 cells.

cells expressing mutant β-catenin show enhanced expression of cell cycle-related genes and decreased expression of cell-cell adhesion pathway-related genes. Since HCT116-WT and
HCT116-MT cells show distinct patterns of β-catenin localization and cellular morphology, we performed microarray analysis on these lines to identify gene expression signatures that are dependent on β-catenin activation status and independent of other factors. We detected 1,507 DEGs (689 up-regulated, 818 down-regulated) displaying |fold change| >1.5 and P-value < 0.05 in HCT116-MT, as compared to HCT116-WT cells. Cytoscape GO analysis revealed HCT116-MT cells display significant up-regulation of cell cycle-related genes and significant down-regulation of cell-cell adhesion-related genes. Other cancer-related alterations (e.g., in cellular metabolism and antigen presentation genes) were also identified (Fig. 2a). Additional analysis using the ontology tool, DAVID, confirmed the up-and down-regulation, respectively, of cell cycle and cell-cell adhesion-related pathways in HCT116-MT cells (Supplementary Table S2). To narrow the list of candidate genes affected by β-catenin activation, we compared our 1,507 DEGs with known sets of 128 cell cycle-related genes and 134 cell adhesion-related genes from KEGG database. This enabled identification of 15 cell cycle-related and 22 cell adhesion-related candidate genes (Fig. 2b). Of these 37 genes, six were randomly chosen to validate our microarray data by qRT-PCR analysis, revealing consistent patterns of mRNA expression for each gene in both assays (Fig. 2c).
To confirm β-catenin activation promotes cell cycle up-regulation, we performed cell proliferation assays and flow cytometric analyses. As expected, HCT116-MT cells proliferated more rapidly than HCT116-WT cells, and HCT116-P cells showed an intermediate proliferation rate phenotype ( Supplementary Fig. S2a). We further detected a 16.9% decrease and 16.8% increase in G1 and G2/M phases, respectively, for HCT116-MT vs. HCT116-WT cells; and a 9.8% decrease and 7.7% increase in G1 and G2/M phases, respectively, for HCT116-P vs. HCT116-WT cells. These findings indicate that HCT116-MT and HCT116-P cells show strong and intermediate cell cycle activation, respectively, as compared to HCT116-WT cells ( Supplementary Fig. S2b).
Although EMT signature was not statistically significant in the pathway analysis results of our microarray data, we further analyzed expression of four main EMT factors, SNAI1 (SNAIL-encoding gene), SNAI2 www.nature.com/scientificreports www.nature.com/scientificreports/ (SLUG-encoding gene), ZEB1, and TWIST1 by normalizing them to GAPDH expression. All of the four EMT factors were found to be up-regulated in HCT116-MT cells compared to HCT116-WT cells (SNAI1, 1.22 folds; SNAI2, 2.11 folds; ZEB1, 2.10 folds; TWIST1, 1.26 folds) ( Supplementary Fig. S3a). We also compared our 1,507 DEGs with a set of 200 EMT-related genes obtained from KEGG database, which allowed us to obtain a list of 28 EMT-related genes, including strong up-regulation of VIM (vimentin) expression (7.57-fold increase) ( Supplementary Fig. S3b). These findings suggest that expression of some EMT-related genes is significantly altered by β-catenin mutation.

HCT116-MT cells show dual loss of tight and adherens junctions, whereas HCT116-P cells dis-
play only tight junction impairment. We next investigated β-catenin mutation-mediated alterations in cell-cell adhesion genes and found that, of the 22 adhesion-related genes identified by GO analysis, CLDN7 and CDH1 were most strongly down-regulated by β-catenin activation. CLDN7 encodes Claudin-7 and CDH1 encodes E-cadherin, which are well-known structural mediators that form TJ and AJ, respectively. We therefore evaluated their expression and localization in our three cell lines, under the conditions of low, moderate, and high cell density, as cell-cell junction molecules are known to show density-dependent expression 17 . Western blot analysis revealed a slight increase in β-catenin expression with increasing cell density in all HCT116 cell lines. For E-cadherin, HCT116-P and HCT116-WT cells show a pattern of gradually increasing cell density-dependent expression, whereas it is barely detectable in HCT116-MT cells. Claudin-7 is also highly expressed and increases with cell density in HCT116-WT cells; however, expression is much lower and unaffected by cell density in HCT116-P cells and is barely detectable in HCT116-MT cells (Fig. 3a,b). β-catenin, E-cadherin, and Claudin-7 mRNA expression patterns mirror their protein expression patterns ( Supplementary Fig. S4). These findings suggest impairment of Claudin-7-mediated TJ in HCT116-P cells and of both TJ and E-cadherin-dependent AJ in HCT116-MT cells. www.nature.com/scientificreports www.nature.com/scientificreports/ To test cell-cell junction status in our cell lines, we performed immunofluorescence microscopy analysis. In HCT116-P and HCT116-WT cells at low density, E-cadherin and β-catenin showed poor colocalization and mostly nuclear β-catenin expression. At moderate cell density, β-catenin began to colocalize with E-cadherin on the cell surface, which was increasingly evident and organized at high density, indicating proper AJ formation (Fig. 3c,d). In HCT116-MT cells, however, E-cadherin was barely detectable at any cell density (Fig. 3e). For Claudin-7, HCT116-WT cells showed increasingly clear and organized membranous expression in a cell density-dependent manner, with gradual TJ formation. However, Claudin-7-mediated TJ appeared to be poorly formed in HCT116-P, irrespective of cell density, and protein expression was barely detected in HCT116-MT cells at any density (Fig. 3f-h).
To investigate the involvement of Wnt pathway in loss of cell-cell junction, we performed immunofluorescence microscopy analysis, western blot, and qPCR using HCT116-WT cells after Wnt3a treatment. Immunofluorescence microscopy and western blot analysis showed that 2-hour treatment of Wnt3a strongly induced concomitant translocation of β-catenin, SNAIL, ZEB1, and TWIST1 into the nucleus; and 4-hour treatment of Wnt3a led to downregulation of E-cadherin and Claudin-7 expressions (Supplementary Fig. S5a,b). Wnt3a treatment also induced transcriptional activation of downstream target genes of Wnt/β-catenin signaling, such as CD44, AXIN2, and CCND1 ( Supplementary Fig. S5c). E-cadherin binds to both WT and mutant β-catenin, and β-catenin is released from the cell membrane and degraded by loss of E-cadherin. It is well-documented that E-cadherin sequesters β-catenin to the cell membrane, and thereby inhibits EMT progression 18,19 . Since there are no reports regarding the binding of E-cadherin with mutant β-catenin, we tested if E-cadherin binds to both WT and mutant β-catenin. Considering HCT116-MT cells exclusively show nuclear β-catenin expression, β-catenin binding to TCF4 was also tested. Immunoprecipitation was performed using a β-catenin antibody, and the binding of β-catenin with E-cadherin or TCF4 was analyzed in HCT116-WT and HCT116-MT cells. As expected, WT β-catenin was mainly bound to E-cadherin, while mutant β-catenin was mostly bound to TCF4 (Fig. 4a). To objectively compare binding affinity of WT and mutant β-catenin to E-cadherin, we generated EGFP-conjugated β-catenin expression constructs (a β-catenin-WT and β-catenin-S45del construct) (Fig. 4b). HCT116-P cells were transfected with β-catenin-WT or β-catenin-S45del vector. β-catenin-S45del vector showed stronger β-catenin expression compared to β-catenin-WT vector, due to resistance of mutant β-catenin to proteasomal degradation (Fig. 4c). mRNA expressions from both constructs were similar (Fig. 4d). Importantly, both synthetic WT and mutant β-catenin showed similar cellular localization (membranous, cytoplasmic, and nuclear) (Fig. 4e), and immunoprecipitation with an EGFP antibody showed that synthetic WT and mutant β-catenin bind to both TCF4 and E-cadherin, indicating that E-cadherin similarly binds to WT and mutant β-catenin (Fig. 4f). Next, we investigated if overexpression of E-cadherin in HCT116-MT cells could lead to the sequestration of mutant β-catenin to the cell membrane (Fig. 4g). Immunofluorescence microscopy analysis and immunoprecipitation with a β-catenin antibody

HCT116 cells show differential migratory and invasive activity according to their β-catenin mutation status.
To evaluate migratory activity of HCT116 cells with differential β-catenin mutation status and dysregulation of cell-cell junctions, we performed wound-healing assays, measuring gap healing at 24 and 48 h after initial scratch. We found that wound-healing rate is significantly higher for HCT116-P than HCT116-WT cells, and the highest overall for HCT116-MT cells (Fig. 5a,b). We then performed invasion assays and found HCT116-MT and HCT116-P cells are highly invasive compared to HCT116-WT cells, with HCT116-MT cells showing the highest invasive activity (Fig. 5c,d). This suggests that β-catenin activation and loss of cell-cell junctions are directly associated with mesenchymal-like features of HCT116 cells. www.nature.com/scientificreports www.nature.com/scientificreports/

β-catenin/ZEB1 axis plays important roles in loss of cell-cell junction in both HCT116-P and HCT116-MT cells. Although both HCT116-P and HCT116-MT cells display mesenchymal-like behaviors,
they show distinct cell morphology and motility patterns, which are mainly attributable to β-catenin mutation status (Figs. 1a and 5a-d). To more precisely elucidate EMT status in HCT116-P and HCT116-MT cells, we measured mRNA expression of SNAI1, SNAI2, ZEB1, TWIST1, VIM, and CDH2 (N-cadherin coding gene) under conditions of increasing cell density. Using qRT-PCR, we detected SNAI1 mRNA expression levels that are 2and 3-fold higher, respectively, in HCT116-P and HCT116-MT cells than in HCT116-WT cells, at all densities (Fig. 6a). SNAI2 mRNA was also up-regulated by about 2.5-fold in HCT116-MT compared to HCT116-WT cells, although the lowest expression was detected in HCT116-P cells (down 6-fold vs. HCT116-MT). Compared to HCT116-WT cells, ZEB1 mRNA expression was slightly up-regulated in HCT116-P cells, whereas about 2-fold up-regulation was detected in HCT116-MT cells. In contrast, TWIST1 expression was strongly up-regulated only in HCT116-P cells in a cell density-dependent manner. CDH2 expression was slightly higher in HCT116-MT cells, compared to that of HCT116-P and HCT116-WT cells. Notably, Vim expression was selectively and strongly up-regulated in HCT116-MT cells. Similar protein expression patterns were also detected for each EMT marker (Fig. 6b). These data suggest that HCT116-MT cells show more progressed EMT than HCT116-P cells, in terms of both motility and EMT marker expression.
Next, we sought to investigate the relationship between mutant β-catenin, EMT factors, and loss of cell-cell junction. To identify EMT transcription factors directly affected by β-catenin, we ablated β-catenin expression in HCT116-P and HCT116-MT cells using lentiviral-encoded short hairpin RNA (shRNA). Notably, HCT116-P and HCT116-MT cells with repressed β-catenin expression commonly showed epithelial-like morphology (Fig. 6c), and the growth rate of HCT116-P and HCT116-MT cells with ablated β-catenin expression was much lower than their counterpart cells with normal β-catenin expression (Fig. 6d). Then, we measured the expression of EMT factors by western blot. Complete knockdown of β-catenin barely affected the expression of SNAIL and SLUG in both cell lines, whereas complete loss of ZEB1 with decrease of TWIST1 expression and complete loss of ZEB1 with increase of TWIST1 expression were observed in HCT116-P and HCT116-MT cells, respectively. Importantly, knockdown of β-catenin led to strong increase of E-cadherin and Claudin-7 expression (Fig. 6e). These findings suggest that ZEB1, a direct target gene of β-catenin, might be functionally important in terms of the mutant β-catenin mediated downregulation of cell-cell junction molecules in both HCT116-P and HCT116-MT cells. Next, we analyzed the expression of E-cadherin and Claudin-7 in HCT116-P and HCT116-MT cells after knockdown of SNAIL, SLUG, ZEB1, or TWIST1 by siRNA. Western blot and qPCR analysis showed that ZEB1 knockdown strongly increases the expression of E-cadherin and Claudin-7 in both cell lines, while TWIST1 knockdown slightly up-regulates the expression of E-cadherin and Claudin-7 only in HCT116-P cells. Downregulation of SNAIL or SLUG barely affected expression of E-cadherin and Claudin-7 (Fig. 6f,g).

Claudin-7 down-regulation in HCT116 cells is critical for acquisition of mesenchymal-like fea-
tures. In order to confirm whether mesenchymal-like features of HCT116-MT and HCT116-P cells directly result from β-catenin activation-mediated down-regulation of cell-cell junction molecules, we generated stable Claudin-7 or E-cadherin knockdown HCT116-WT cells using shRNA. Protein knockdown was confirmed by western blot, and we found that Claudin-7 knockdown promotes decreased E-cadherin expression and vice versa (Fig. 7a). Critically, both Claudin-7 and E-cadherin knockdown cells showed mesenchymal-like morphology, displaying multiple lamellipodia at their leading edges to facilitate migration (Fig. 7b). Immunofluorescence analysis further revealed disturbed cell-cell interactions and interdependent loss of Claudin-7 and E-cadherin expression in stable knockdowns (Fig. 7c). Wound-healing was significantly accelerated in both Claudin-7 and E-cadherin www.nature.com/scientificreports www.nature.com/scientificreports/ knockdown cells, showing more rapid closure in E-cadherin knockdowns (Fig. 7d,e). Enhanced invasion was also observed in both knockdowns, with a higher activity observed in Claudin-7 knockdown cells (Fig. 7f,g). These findings indicate interdependence between AJ and TJ formation, and suggest that dysregulation of TJs is sufficient to induce mesenchymal-like features in HCT116 cells.

Loss of E-cadherin leads to dramatic up-regulation of migratory and invasive activity in HCT116-P cells.
Our data indicate that Claudin-7 is down-regulated in HCT116-P compared to HCT116-WT cells and does not increase with cell density. Furthermore, levels of membranous E-cadherin in HCT116-P cells are similar to those observed in HCT116-WT cells. We therefore tested whether Claudin-7-mediated TJs are indeed impaired in HCT116-P cells, and if additional loss of E-cadherin further up-regulates migratory and invasive activity of these cells. To this end, we generated stable shRNA-knockdowns of Claudin-7 or E-cadherin in HCT116-P knockdown cells. Protein knockdown was confirmed by western blot. We further found that Claudin-7 knockdown promotes a slight decrease in E-cadherin expression and vice versa (Fig. 8a). Additionally, no clear morphological changes were observed in Claudin-7 knockdowns, whereas E-cadherin knockdown cells showed disturbed cell polarization and adjoining (Fig. 8b). These morphological and expressional changes were also observed during immunofluorescence microscopy analysis (Fig. 8c). We also found that E-cadherin www.nature.com/scientificreports www.nature.com/scientificreports/ knockdown strongly induces HCT116-P cell migration, whereas migration is only slightly enhanced by Claudin-7 knockdown (Fig. 8d,e). Consistent with this, invasive activity was strongly up-regulated in E-cadherin knockdowns, but was induced to a much lesser extent in Claudin-7 knockdown cells (Fig. 8f,g). These findings indicate that AJs are minimally affected by further loss of TJs in HCT116-P cells due to pre-existing TJ impairment, and loss of AJs leads to a more advanced EMT phenotype in those cells.
Clinicopathologic and molecular features of CRCs with nuclear β-catenin expression. Our findings suggest that β-catenin mutations promote its nuclear expression, leading to Claudin-7 and E-cadherin down-regulation. We therefore examined the association between nuclear β-catenin expression and expression of Claudin-7/E-cadherin in 101 stage III CRC tissues by IHC. We exclusively detected membrane-associated β-catenin in crypt cells of normal colon mucosa, whereas CRCs showed heterogeneous nuclear accumulation of β-catenin that can be categorized as follows: 1) no nuclear expression, similar to normal tissues, 33/101 samples; 2) low nuclear expression (1-29% of tumor cells), 53/101 samples; and 3) high nuclear expression (≥30% of tumor cells), 15/101 samples ( Supplementary Fig. S6a,b).
We then assessed clinicopathologic features of CRC tissues showing negative, low, and high nuclear β-catenin expressions (Supplementary Table S3), and found that those with high expression showed significantly larger tumor size (P = 0.028) and decreased prevalence of microsatellite instability-high (MSI-H) (P = 0.013). We could not find statistical significance between high nuclear β-catenin expression and expression of Claudin-7 (P = 0.474) and E-cadherin (P = 0.093), probably due to the small case number of CRCs with high nuclear β-catenin expression ( Supplementary Fig. S6c and Supplementary Table S3). In the same context, no significant relationship was found between the mean percentage of nuclear β-catenin expression in CRCs showing metastasis or no metastasis after operation (P = 0.078) (Supplementary Fig. S6d). Notably, Kaplan-Meier survival analysis showed that overall survival (P = 0.009) of CRC patients with negative and low nuclear β-catenin expression is significantly longer than that of patients with high nuclear β-catenin expression ( Supplementary Fig. S7). www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
As a key downstream WNT pathway effector, β-catenin plays known roles in proliferation, survival, and EMT of CRC cells. EMT is particularly important due to its involvement in loss of cell-cell adhesion and gain of migratory and invasive features, which render some cancers incurable by current therapeutic regimens 20 . Indeed, most cancer-related deaths are attributable to metastatic diseases. Loss of cell-cell junctions is critical for EMT progression; and although its clinical significance has been reported in CRC 21 , relationship between loss of cell-cell junction and β-catenin-mediated EMT is poorly understood. Therefore, elucidation of mechanisms underlying β-catenin-mediated EMT is critical for understanding metastasis in a subset of CRCs.
In most CRCs, β-catenin stabilization results from mutations in APC. However, as APC is involved in various cellular functions, independent of β-catenin 22 , it is difficult to assess specific consequences of β-catenin activation in APC mutant cells. Therefore, in this study, we sought to elucidate the mechanism of β-catenin activation-mediated EMT using isogenic HCT116 cell lines with differential β-catenin mutation status. To validate our strategy, we indeed evaluated the localization and activation status of β-catenin in two APC mutant CRC cell lines (DLD-1 and LoVo), two CRC cell lines with no Wnt pathway relevant mutations (RKO and HCT8), a hepatocellular carcinoma (HCC) cell line with an AXIN1 mutation (Hep3B), a HCC cell line with a CTNNB1 mutation (HepG2), and a CRC cell line with a CTNNB1 mutation (LS174T). Immunofluorescence microscopy and western blot analysis showed that CRCs with APC mutations more clearly display nuclear β-catenin localization compared , and E-cadherin (stained in red) in HCT116-P cells with stable shRNA-mediated knockdown of Claudin-7 or E-cadherin knockdown. (d) Wound-healing assay performed on HCT116-P cells with stable knockdown of Claudin-7 or E-cadherin. (e) Quantification of relative migration distances for the wound-healing assay knockdown shown in. (d) Statistical significance between 0 and 24 hours, and 24 and 48 hours is shown. (f) Invasion assay performed on HCT116-P cells with stable knockdown of Claudin-7 or E-cadherin. (g) Quantification of the relative number of knockdown stained cells in (f) by ImageJ software. Error bars in (e,g) represent the SD of the mean of results from three independent experiments. Statistical significance between shControl, shCLDN7, and shCDH1 is shown. All assays were carried out in triplicate.