Posttreatment recurrence of colorectal cancer, the third most lethal cancer worldwide, is often driven by a subpopulation of cancer stem cells (CSC). The tight junction (TJ) protein claudin-2 is overexpressed in human colorectal cancer, where it enhances cell proliferation, colony formation, and chemoresistance in vitro. While several of these biological processes are features of the CSC phenotype, a role for claudin-2 in the regulation of these has not been identified. Here, we report that elevated claudin-2 expression in stage II/III colorectal tumors is associated with poor recurrence-free survival following 5-fluorouracil–based chemotherapy, an outcome in which CSCs play an instrumental role. In patient-derived organoids, primary cells, and cell lines, claudin-2 promoted colorectal cancer self-renewal in vitro and in multiple mouse xenograft models. Claudin-2 enhanced self-renewal of ALDHHigh CSCs and increased their proportion in colorectal cancer cell populations, limiting their differentiation and promoting the phenotypic transition of non-CSCs toward the ALDHHigh phenotype. Next-generation sequencing in ALDHHigh cells revealed that claudin-2 regulated expression of nine miRNAs known to control stem cell signaling. Among these, miR-222-3p was instrumental for the regulation of self-renewal by claudin-2, and enhancement of this self-renewal required activation of YAP, most likely upstream from miR-222-3p. Taken together, our results indicate that overexpression of claudin-2 promotes self-renewal within colorectal cancer stem-like cells, suggesting a potential role for this protein as a therapeutic target in colorectal cancer.

Significance: Claudin-2-mediated regulation of YAP activity and miR-222-3p expression drives CSC renewal in colorectal cancer, making it a potential target for therapy. Cancer Res; 78(11); 2925–38. ©2018 AACR.

Claudins are transmembrane proteins involved in tight junction (TJ) formation and function within healthy epithelia, endothelia, and other tissues (1). Claudins have been implicated in multiple biological processes, including the regulation of paracellular transport of solutes and water, the maintenance of epithelial tissue integrity, the maintenance of apico-basal polarity and the specification of morphology in embryos and tissues (2). Their expression is altered in several cancer types, and a tumor-promoting role has been suggested for several claudin isoforms (3), including in colorectal cancer (4, 5). In particular, claudin-2 is overexpressed in a significant proportion of human colorectal tumors (6, 7), and we recently demonstrated that this protein is instrumental to the tumor-promoting activity of the transcriptional regulator symplekin in colorectal cancer cells (8). Claudin-2 also promotes proliferation and anchorage-independent colony formation by human colorectal cancer cells in vitro, and increases tumor growth in xenografted mice (6).

Claudin-2 expression is restricted to the stem/progenitor cell compartment in the healthy intestinal epithelium (9), and its regulatory role in intestinal homeostasis was recently highlighted (10). We also detected enriched claudin-2 expression in mouse intestinal tumors displaying elevated expression of cancer stem cell (CSC) markers and Wnt target genes (11). In addition, claudin-2 was recently shown to promote anchorage-independent growth and to enhance resistance to 5-fluorouracil (5-FU; ref. 6). While several of these properties are among those that characterize CSCs (12), a putative role for claudin-2 in the regulation of these cells remains to be explored.

As many colorectal tumors overexpress claudin-2, we tested the hypothesis that high claudin-2 expression may contribute to the regulation of stem-like cells within these tumors. We found that claudin-2 overexpression was associated with reduced postchemotherapy disease-free survival in three independent cohorts of patients with colorectal cancer, and that claudin-2–overexpressing cells displayed elevated self-renewal in vitro and tumor-initiating frequency in vivo. Claudin-2 increased the proportion of ALDHHigh cancer stem-like cells in heterogeneous colorectal cancer cell populations, and this outcome was due to enhanced ALDHHigh cell self-renewal and to a higher conversion rate of ALDHLow into ALDHHigh cells. Claudin-2 selectively regulated the expression of 9 miRNAs in ALDHHigh cells, and miR-222-3p regulation was instrumental for the promotion of self-renewal by claudin-2. Finally, activation of YAP by claudin-2, most likely upstream from miR-222-3p was also essential to this effect. Our results indicate that elevated expression of claudin-2 promotes self-renewal within colorectal cancer stem-like cells, suggesting a potential role for this protein as a prognostic marker and/or a therapeutic target in colorectal cancer.

Further details concerning constructs/reagents, cell lines, and patient-derived tumor cell culture, CLDN2 expression/survival associations, in vivo experiments, RNA extraction, real-time PCR, Western blotting, ALDH activity-based cell sorting, IHC, sphere formation assays, and miRNA quantification and rescue experiments are provided in Supplementary Methods and Supplementary Table S1.

Association between CLDN2 expression and patient with colorectal cancer survival

The association between CLDN2 mRNA expression and disease outcome in patients with stage II/III colorectal cancer treated with 5-FU–based chemotherapy was analyzed in three independent datasets. Quantification of CLDN2 mRNAs from paraffin-embedded sections using qRT-PCR and correlation with patient survival was performed on a cohort treated at the Hospital Clínic of Barcelona in 1998–2005 (13, 14). Two other datasets were selected on the basis of the presence of CLDN2-specific probes, the number of patients (>150), and the availability of recurrence-free survival data. GSE24551–GPL11028 (15), was sourced and analyzed using the SurvExpress bioresource, (http://bioinformatica.mty.itesm.mx:8080/Biomatec/SurvivaX.jsp; ref. 16). GSE39582 (461 patients; ref. 17) was analyzed using the R2: Genomics Analysis and Visualization Platform (http://r2.amc.nl).

Patient-derived samples

All experiments including patient samples were performed in accordance with ethical principles for medical research involving human subjects as set by the Declaration of Helsinki. Patient-derived colon cancer cells (CPP) were derived detailed under Supplementary Methods from colorectal cancer biopsies obtained from Centre Hospitalier Universitaire (CHU) Carémeau (Nimes, France). The protocol was approved by the CHU Institutional Ethics Committee (human ethics agreement #2011-A01141-40, NCT01577511). Patient-derived organoids were grown from samples collected under human ethic agreement HREC/15/PMCC/112, Peter MacCallum Cancer Centre Human Research Ethics Committee (Melbourne, Australia). Signed informed consents were obtained from all patients prior to samples acquisition.

miRNA quantification

RNA including miRNAs were isolated using the miRNeasy Mini Kit (QIAGEN) and RNA quality was analyzed on the Agilent 2100 Bioanalyzer. Deep sequencing was run on an Ion Torrent PGM sequencer, using Ion 318 V2 chips and Ion PGM 200 V2 Sequencing Kit (Life Technologies). Sequence files were analyzed for quality control (FASTQC), aligned to the human genome (HG19) using the Torrent Suite and transferred to Partek Genomic Suite and Flow (Partek Incorporated) for mapping against miRBase V.21 and Ensembl Release 75. Reads were normalized to reads per million reads and miRNAs identified with at least 10 reads were used for further analysis on Partek Genomic suite.

Cell lines

Cell lines used in this study were obtained from the ATCC. Experiments were performed within a maximum of 10 passages after thawing, and the absence of Mycoplasma was confirmed monthly using the MycoAlert Mycoplasma Detection Kit (Lonza). All cell lines were authenticated using STR profiling (LGC standards).

In vivo experiments

Several xenograft models were used under agreements from the Monash Institute of Pharmaceutical Sciences Animal Ethics Committee (#MIPS_AEC2012.04 and #MIPS_AEC2013.01) to analyze the impact of claudin-2 on self-renewal.

To analyze the role of claudin-2 on primary tumor initiation, tumor incidence was monitored after subcutaneous injection of CPP1 or CMT93 cells to NOD/SCID mice. For serial transplantation, first-generation tumors were collected when reaching 500 mm3 and dissociated using the gentleMACS dissociator (Miltenyi Biotec). Viable (DAPI-negative) cells were serially transplanted into the flank of new NOD/SCID mice.

To assess the role of claudin-2 on metastasis initiation, luciferase-expressing CPP1 cells were injected into the cecum of NOD/SCID mice. Primary tumor growth and metastasis development were monitored for 6 weeks using Bioluminescence imaging (BLI). Forty-seven days after injection animals were euthanized and their liver was imaged ex vivo.

To characterize the impact of claudin-2 on postchemotherapy tumor recurrence, NOD/SCID mice were injected subcutaneously with T84 cells. When tumors reached 250 ± 25 mm3, mice were treated with FOLFIRI (40 mg/kg 5-FU + 15 mg/kg irinotecan + 30 mg/kg leucovorin, 2x/week, 3 weeks). Post-treatment residual tumors were dissociated and tumor cells re-implanted into second generation animals.

Statistical analysis

GraphPad Prism6 software was used for data analysis. After determining whether or not datasets for each experiment were normally distributed, a Student t test or a Mann–Whitney test was performed to analyze the difference between two groups of quantitative variables, as indicated in the figure legends. ANOVA was used for comparisons among three groups of quantitative variables, and pairwise comparisons were carried out using the Tukey post hoc test as indicated. Statistical significance of stem cell frequencies quantified using Extreme Limiting Dilution Analysis (ELDA) was assessed using the χ2 test, as described previously (18).

Elevated CLDN2 expression in human colorectal cancer is linked with posttreatment recurrence

CSCs play a key role in the poor clinical outcome of patients with colorectal cancer treated with adjuvant chemotherapy (19, 20). To determine whether claudin-2 overexpression may contribute to these outcomes, we first quantified prospectively the expression of CLDN2 mRNA in a homogenous population of primary tumor samples from patients with stage II/III colorectal cancer that were all treated with 5-FU–based adjuvant chemotherapy (14). CLDN2 mRNA was detected in 86 patient samples using qRT-PCR, and Kaplan–Meier analysis showed that high CLDN2 expression correlated with lower cancer-specific survival (P = 0.0329; Fig. 1A).

Figure 1.

Higher levels of CLDN2 are associated with posttreatment recurrence in patients with stage II and III colorectal cancer. Kaplan–Meier curves reflecting the difference of cancer-specific (A) and recurrence-free survival (B and C) among patients with chemotherapy-treated stage II and III colorectal cancer in relation with the claudin-2 mRNA expression levels (CLDN2) in their primary tumor sample. A, Cancer-specific survival rate in a cohort of 86 patients, high CLDN2-expressing (red line) versus low CLDN2-expressing group (blue line). B, Recurrence-free survival in 461 patients (GSE39582 dataset), analyzed using the R2 Genomics Analysis and Visualisation Platform (http://r2.amc.nl). Red line, high CLDN2-expressing group; blue line, low CLDN2-expressing group. C, Recurrence-free survival in 160 patients is indicated (GSE24551), analyzed using the SurvExpress bioresource (http://bioinformatica.mty.itesm.mx:8080/Biomatec/SurvivaX.jsp). Red line, high CLDN2-expressing group; blue line, low CLDN2-expressing group. D,CLDN2 gene expression level in low- and high-risk patients (data as in C). E, Claudin-2 immunostaining in tumor tissue sections from patients with stage II/III colorectal cancer. Images are representative of the observed staining pattern intensities (negative to strong) and subcellular distributions (membrane and/or diffuse).

Figure 1.

Higher levels of CLDN2 are associated with posttreatment recurrence in patients with stage II and III colorectal cancer. Kaplan–Meier curves reflecting the difference of cancer-specific (A) and recurrence-free survival (B and C) among patients with chemotherapy-treated stage II and III colorectal cancer in relation with the claudin-2 mRNA expression levels (CLDN2) in their primary tumor sample. A, Cancer-specific survival rate in a cohort of 86 patients, high CLDN2-expressing (red line) versus low CLDN2-expressing group (blue line). B, Recurrence-free survival in 461 patients (GSE39582 dataset), analyzed using the R2 Genomics Analysis and Visualisation Platform (http://r2.amc.nl). Red line, high CLDN2-expressing group; blue line, low CLDN2-expressing group. C, Recurrence-free survival in 160 patients is indicated (GSE24551), analyzed using the SurvExpress bioresource (http://bioinformatica.mty.itesm.mx:8080/Biomatec/SurvivaX.jsp). Red line, high CLDN2-expressing group; blue line, low CLDN2-expressing group. D,CLDN2 gene expression level in low- and high-risk patients (data as in C). E, Claudin-2 immunostaining in tumor tissue sections from patients with stage II/III colorectal cancer. Images are representative of the observed staining pattern intensities (negative to strong) and subcellular distributions (membrane and/or diffuse).

Close modal

We also performed a retrospective analysis of published datasets to determine whether CLDN2 mRNA expression level in primary colorectal tumors would be associated with specific disease outcomes. We used a large cohort of patients with colorectal cancer from the GSE39582 dataset (17), from which patients with stage II/III tumors, who are usually administered adjuvant chemotherapy, were selected using the R2 Genomics Analysis and Visualization Platform (Fig. 1B). High CLDN2 expression levels correlated with poorer probability of relapse-free survival in these patients (P = 0.049, n = 461). A stage-specific analysis showed the same trend for stage II (P = 0.058, n = 260) and stage III (P = 0.053, n = 201; Supplementary Fig. S1). Finally, using another homogenous cohort of patients with colon cancer with stage II/III colorectal cancer (GSE24551; ref. 15), we found that patients with high CLDN2 mRNA expression displayed significantly worse recurrence-free survival following adjuvant chemotherapy than those expressing low levels of claudin-2 (P = 0.0094, n = 160; Fig. 1C and D).

To determine whether these changes were detectable at protein level, we performed claudin-2 immunostaining on a tissue microarray (TMA) containing primary tumor sections from 24 patients with stage II and III colorectal cancer having undergone 5-FU–based adjuvant chemotherapy (Supplementary Table S2). Despite the small size of this cohort, our results (Fig. 1E; Supplementary Fig. S1) highlighted the positive association trend between claudin-2 expression and the presence of posttreatment recurrence in these patients (χ2 = 3.696, P = 0.0545). No correlation was detected between the cellular localization of claudin-2 staining (categorized into membrane, cytoplasmic/nuclear, or mixed pattern staining), and posttreatment recurrence (χ2 = 1.252, P = 0.2631; Supplementary Fig. S1).

These results demonstrate that increased CLDN2 expression within stage II/III colorectal tumors is associated with poor recurrence-free survival in patients treated with adjuvant chemotherapy.

Claudin-2 promotes the self-renewal of colorectal cancers in vitro

CSCs have frequently been identified as important drivers of postchemotherapy recurrence (12, 21). Using colorectal cancer cells isolated from colon tumor biopsies and established colorectal cancer cell lines, we therefore analyzed the impact of claudin-2 overexpression on self-renewal, a core functional characteristic of CSCs.

We first induced claudin-2 overexpression in a colorectal cancer cell line (SW480) expressing low levels of endogenous claudin-2, as shown in Fig. 2A. Claudin-2 overexpression strongly enhanced the ability of these cells to form colonospheres over multiple passages in suspension (Fig. 2B), an in vitro hallmark of self-renewing cells (22). The size of claudin-2–overexpressing colonospheres was enlarged compared with those of controls and claudin-2–overexpressing colonospheres survived for more than 1 month without passaging, in contrast with the rare spheres formed by control cells (Supplementary Fig. S2). We also analyzed the impact of claudin-2 overexpression on long-term self-renewal in patient-derived cells generated from a primary colon cancer sample (CPP1), using serial passaging followed by an ELDA assay. Claudin-2 expression increased the stem cell frequency of 10th generation cells from 1 in every 23.4 cells (confidence interval 1/11.5–1/47.4) to 1 in every 7 (1/3.7–1/13.4; Fig. 2C). Goodness-of-fit analysis highlighted the significant difference between control and claudin-2–expressing cells (Pearson χ2 = 5.25, P = 0.0219).

Figure 2.

Claudin-2 promotes self-renewal of colorectal cancer cells. Experimental modification of claudin-2 expression levels regulates self-renewal in several colorectal cancer cell lines. In B, D, and E, results are expressed as mean ± SEM, n = 3 (Student t test, **, P < 0.01). A, Claudin-2 expression in SW480 colorectal cancer cells was quantified by Western blot analysis in cells expressing a tetracycline-inducible claudin-2 (CLDN2) or control construct (CON), treated or not with doxycycline (DOX) for 24, 48, and 96 hours. Levels of β-actin expression were used as internal loading controls. Bottom, immunofluorescence detection of claudin-2 at the plasma membrane of cells expressing the claudin-2 construct (CLDN2), while claudin-2 expression is barely detectable in control cells (CON). Nuclei were counterstained with DAPI. Scale bars, 50 μm. B, Colonosphere formation was quantified in SW480 cells overexpressing Claudin-2 (CLDN2) or expressing a control vector (CON), at passage one (P1, top) or after four passages (P4, bottom). The number of colonospheres formed was quantified. C, ELDA was used to determine the self-renewal frequency of CPP1 patient-derived cells. The presence or absence of spheres was assessed seven days after seeding at 1,000, 100 or 10 or 1 cells/well (each concentration in 12 replicates) in an ultralow adherence 96-well plate and reported as the estimated sphere-forming frequency + confidence intervals (P = 0.0219, χ2, n = 3). D, Quantification of claudin-2 expression by Western blotting (top left) and immunofluorescent staining (bottom left, nuclei counterstained with DAPI) in DLD-1 cells expressing a claudin-2–specific shRNA (shCLDN2) or a control shRNA (shCON). Scale bars, 50 μm. Right, colonosphere formation was quantified in shCON and shCLDN2 DLD-1 cells. E, Colonosphere formation was quantified in T84 cells expressing a claudin-2–specific shRNA (shCLDN2) or a control shRNA (shCON) after one (P1) or two passages (P2) in suspension. It is shown that colonospheres are decreased when CLDN2 expression is downregulated. Analysis as mentioned in B. F, Three-dimensional colorectal cancer patient-derived colon organoids. Patient tumor cells from the colon were isolated, transfected with siRNA against CLDN2, and grown in Matrigel. Phase-contrast photographs (left) exemplify the development of three-dimensional organoids at day 3. The initiation of organoids (diameter ≥ 30 μm) was decreased after claudin-2–specific siRNA transfection (siCLDN2, 50 mmol/L), in comparison with controls (siCON). Scale bars, 50 μm. The mean incidence of organoids after three days is shown on the right. NT, nontransfected controls. siCON vs. siCLDN2, P = 0.0426, ANOVA with Tukey multiple comparisons test, n = 6.

Figure 2.

Claudin-2 promotes self-renewal of colorectal cancer cells. Experimental modification of claudin-2 expression levels regulates self-renewal in several colorectal cancer cell lines. In B, D, and E, results are expressed as mean ± SEM, n = 3 (Student t test, **, P < 0.01). A, Claudin-2 expression in SW480 colorectal cancer cells was quantified by Western blot analysis in cells expressing a tetracycline-inducible claudin-2 (CLDN2) or control construct (CON), treated or not with doxycycline (DOX) for 24, 48, and 96 hours. Levels of β-actin expression were used as internal loading controls. Bottom, immunofluorescence detection of claudin-2 at the plasma membrane of cells expressing the claudin-2 construct (CLDN2), while claudin-2 expression is barely detectable in control cells (CON). Nuclei were counterstained with DAPI. Scale bars, 50 μm. B, Colonosphere formation was quantified in SW480 cells overexpressing Claudin-2 (CLDN2) or expressing a control vector (CON), at passage one (P1, top) or after four passages (P4, bottom). The number of colonospheres formed was quantified. C, ELDA was used to determine the self-renewal frequency of CPP1 patient-derived cells. The presence or absence of spheres was assessed seven days after seeding at 1,000, 100 or 10 or 1 cells/well (each concentration in 12 replicates) in an ultralow adherence 96-well plate and reported as the estimated sphere-forming frequency + confidence intervals (P = 0.0219, χ2, n = 3). D, Quantification of claudin-2 expression by Western blotting (top left) and immunofluorescent staining (bottom left, nuclei counterstained with DAPI) in DLD-1 cells expressing a claudin-2–specific shRNA (shCLDN2) or a control shRNA (shCON). Scale bars, 50 μm. Right, colonosphere formation was quantified in shCON and shCLDN2 DLD-1 cells. E, Colonosphere formation was quantified in T84 cells expressing a claudin-2–specific shRNA (shCLDN2) or a control shRNA (shCON) after one (P1) or two passages (P2) in suspension. It is shown that colonospheres are decreased when CLDN2 expression is downregulated. Analysis as mentioned in B. F, Three-dimensional colorectal cancer patient-derived colon organoids. Patient tumor cells from the colon were isolated, transfected with siRNA against CLDN2, and grown in Matrigel. Phase-contrast photographs (left) exemplify the development of three-dimensional organoids at day 3. The initiation of organoids (diameter ≥ 30 μm) was decreased after claudin-2–specific siRNA transfection (siCLDN2, 50 mmol/L), in comparison with controls (siCON). Scale bars, 50 μm. The mean incidence of organoids after three days is shown on the right. NT, nontransfected controls. siCON vs. siCLDN2, P = 0.0426, ANOVA with Tukey multiple comparisons test, n = 6.

Close modal

Conversely, we used several approaches to downregulate claudin-2 expression in colorectal cancer cells expressing high endogenous levels of this protein. Inducible shRNA-mediated claudin-2 downregulation in DLD-1 cells significantly decreased their colonosphere-forming ability (Fig. 2D) and hampered colonosphere growth compared with controls (Supplementary Fig. S2). Similar results were obtained in T84 colorectal cancer cells (Fig. 2E) using shRNAs targeting a different region of the CLDN2 gene. Claudin-2 downregulation or overexpression did not alter the expression of claudin-1 (Supplementary Fig. S2), another claudin isoform previously shown to play a regulatory role in colorectal cancer (5), highlighting the claudin-2 specificity of this effect. To further corroborate whether claudin-2 expression regulates CSC-driven biological processes in primary cells, CLDN2-selective siRNAs were transfected into tumor organoids grown from surgically resected primary colorectal tumors. The initiation of colorectal cancer organoids was significantly decreased from three days after transfection with siCLDN2 (mean organoid number ± SEM = 19.64 ± 2.37) in comparison with siCON-transfected organoids (mean ± SE.M = 34.62 ± 5.30; P = 0.0426, ANOVA with Tukey test; Fig. 2F), suggesting that claudin-2 expression regulates organoid initiation in primary colorectal cancer cells.

Together these findings indicate that claudin-2 promotes self-renewal of colon cancer cells in vitro, a key characteristic of CSCs.

Claudin-2 promotes colorectal cancer self-renewal in vivo

An important feature of CSCs is their capacity to initiate tumors when injected at low concentration into immunocompromised animals (23). To determine whether claudin-2 regulates CSC self-renewal in vivo, patient-derived CPP1 cells were transfected to stably overexpress claudin-2 (Supplementary Fig. S3), and increasing numbers of these cells (50, 500, and 5,000) were subcutaneously grafted into immunocompromised NOD/SCID mice. Tumor initiation was readily detected in mice inoculated with claudin-2–overexpressing cells, with 3 of 5 mice injected with as little as 50 cells bearing detectable tumors by day 90 after injection (Fig. 3A). In contrast, no tumors were detected by day 90 following the injection of 50 control CPP1 cells. Tumor initiation was strongly delayed in mice injected with 500 or 5,000 control cells compared with claudin-2–expressing CPP1 (Fig. 3A), and tumor growth was reduced (Supplementary Fig. S3).

Figure 3.

Claudin-2 promotes colorectal cancer tumor initiation and self-renewal in vivo. Tumor incidence was quantified after subcutaneous xenografting of colorectal cancer cell lines stably overexpressing claudin-2 (CLDN2) or vector control cells (CON) into first- and second-generation NOD/SCID mice (n = 5 per group). A, Left, first-generation xenografts generated with patient-derived CPP1 cells at 50 or 500 or 5,000 cells/injection. Right, tumors collected from first-generation mice injected with 500 cells were harvested, pooled, and serially reinjected into second-generation animals, followed by regular monitoring of tumor intake and size. ‡ signifies that all mice were euthanized. B, Left, tumor incidence was quantified following inoculation of 10 or 100 or 1,000 CMT93 cells expressing a control or claudin-2–expressing construct. Right, representative photographs of tissue sections obtained from control or claudin-2–overexpressing xenografts and immunostained with a specific claudin-2 antibody. Photographs from two representative control tumors (CON) and two claudin-2–overexpressing tumors (CLDN2; 1,000 cells/injection) harvested at D37 are shown. Scale bar, 100 μm. C, Top, outline of chemotherapy and posttreatment reimplantation protocol of T84 cell xenografts. Bottom, claudin-2 expression in posttreatment tumor cells prior to reimplantation into second-generation animals (left), and tumor incidence in second-generation animals injected with decreasing amounts of control or claudin-2 shRNA-expressing T84 cells (right). *, Student t test, P < 0.05.

Figure 3.

Claudin-2 promotes colorectal cancer tumor initiation and self-renewal in vivo. Tumor incidence was quantified after subcutaneous xenografting of colorectal cancer cell lines stably overexpressing claudin-2 (CLDN2) or vector control cells (CON) into first- and second-generation NOD/SCID mice (n = 5 per group). A, Left, first-generation xenografts generated with patient-derived CPP1 cells at 50 or 500 or 5,000 cells/injection. Right, tumors collected from first-generation mice injected with 500 cells were harvested, pooled, and serially reinjected into second-generation animals, followed by regular monitoring of tumor intake and size. ‡ signifies that all mice were euthanized. B, Left, tumor incidence was quantified following inoculation of 10 or 100 or 1,000 CMT93 cells expressing a control or claudin-2–expressing construct. Right, representative photographs of tissue sections obtained from control or claudin-2–overexpressing xenografts and immunostained with a specific claudin-2 antibody. Photographs from two representative control tumors (CON) and two claudin-2–overexpressing tumors (CLDN2; 1,000 cells/injection) harvested at D37 are shown. Scale bar, 100 μm. C, Top, outline of chemotherapy and posttreatment reimplantation protocol of T84 cell xenografts. Bottom, claudin-2 expression in posttreatment tumor cells prior to reimplantation into second-generation animals (left), and tumor incidence in second-generation animals injected with decreasing amounts of control or claudin-2 shRNA-expressing T84 cells (right). *, Student t test, P < 0.05.

Close modal

To determine whether the claudin-2-driven self-renewal increase was maintained after passaging, cells from claudin-2-overexpressing xenografts (500 cell group) were dissociated once tumors reached a 500 mm3 volume, and reimplanted subcutaneously into new mice. Tumor initiation was detected in second-generation animals injected with 50 (1/5), 500 (4/5), and 5,000 (4/5) claudin-2–overexpressing cells (Fig. 3A), but serial transplantation of tumor cells was not successful from control tumor cells, as insufficient live cells could be recovered from the small control tumors (<35 mm3).

To determine whether the promotion of tumor initiation by claudin-2 extends beyond human colorectal cancer cells, we overexpressed claudin-2 in CMT93 mouse rectal carcinoma cells (Supplementary Fig. S3) and injected them into NOD/SCID mice (10, 100, or 100 cells/mouse) to determine their maximal engraftment efficiency (Fig. 3B). Thirty-seven days after injection, tumors had developed in mice from all claudin-2–overexpressing groups, including 20% (1/5) of mice inoculated with 10 cells (Fig. 3B). In contrast, tumors were only detectable in 4 of 5 mice inoculated with 1,000 control cells but not in those injected with 10 or 100 control cells. (Fig. 3B; Supplementary Fig. S3). Tumors were collected and claudin-2 immunostaining was used to validate claudin-2 overexpression in this tumor model (Fig. 3B).

To examine whether the metastasis-initiating ability of colorectal cancer cells is regulated by claudin-2, we performed intracecal injection of luciferase-expressing CPP1 cells (1 × 105/mouse) and monitored liver metastasis development for 6 weeks using bioluminescent imaging. All mice developed primary tumors, and 2 o 3 mice injected with claudin-2–overexpressing CPP1 cells also developed liver metastases, while none of the 5 control mice did (Supplementary Fig. S4). Two of 5 mice injected with claudin-2–overexpressing cells spontaneously died 2–3 weeks after injection and had to be excluded from the analysis.

Finally, to characterize the impact of claudin-2 on the recurrence potential of colorectal cancer cells after chemotherapy, NOD/SCID mice were injected subcutaneously with T84 cells expressing a control or a claudin-2–specific shRNA (n = 18/group, 1 × 106 cells/mouse). When tumors reached 250 ± 25 mm3, mice were treated for 3 weeks with FOLFIRI. Residual tumors were dissociated and live tumor cells reimplanted subcutaneously into second-generation animals (50, 500, 5,000 cells/mouse, 6 mice/group). Eight weeks later, cells expressing claudin-2 shRNA displayed reduced ability to initiate second-generation tumors at all concentrations compared with control cells (Fig. 3C).

Together, these findings demonstrate that claudin-2 promotes the primary and metastatic tumor-initiating capacity of colorectal cancer cells as well as their self-renewing ability after exposure to chemotherapy in vivo, which are both recognized properties of CSCs.

Claudin-2 regulates phenotypic transitions between CSC and non-CSC states in vitro

We demonstrated above that claudin-2 overexpression enhances tumor initiation in vivo and colorectal cancer cell self-renewal in vitro, suggesting that this protein may increase the proportion and/or self-renewal potential of stem-like cells in colon cancer. To gain further mechanistic insight, we first investigated whether claudin-2 had an impact on the expression of stem cell markers such as BMI1, POU5F1 (Oct-4), and ALDH1A1. Levels of mRNA encoding BMI1, ALDHA1, and Oct-4 were significantly decreased in T84 cells where claudin-2 expression was constitutively downregulated (CLDN2-shRNA cells; Fig. 4A). Short-term claudin-2 downregulation using siRNA decreased the expression of some but not all of these markers in DLD-1, T84, and patient-derived CPP42 cells, suggesting that sustained claudin-2 downregulation may be required to transcriptionally regulate stem cell markers (Fig. 4B).

Figure 4.

Claudin-2 promotes the CSC phenotype in vitro. A, Expression of the mRNAs for CLDN2, BMI1, and Oct-4 in T84 cells stably transfected with control of claudin-2–specific shRNAs. Student t test, *, P < 0.01; mean ± SEM, n = 3. B, Expression of the mRNAs for CLDN2, BMI1, ALDH1A1, and Oct-4 in DLD-1 (left), T84 (middle), and patient-derived CPP42 cells (right) 24 hours after transient transfection with siRNA targeting CLDN2, expressed as fold change compared with their respective expression in cells expressing a β-galactosidase–specific control siRNA. Mean ± SEM from experimental replicates; data from one of two similar experiments; Student t test, *, P < 0.05; **, P < 0.01; ***, P < 0.001. C,CLDN2 mRNA expression in ALDHLow and ALDHHigh T84 cells (mean ± SEM). D, Percentages of ALDHLow (gray) and ALDHHigh cells (white) in untransfected T84 cells (top), as well as 3 days after transfection with siRNA selectively targeting CLDN2 (siCLDN2) or β-galactosidase (siCON). Data represent one of three similar experiments. E, Purified T84 ALDHHigh cells (white, top) or ALDHLow cells (gray, bottom) were independently transfected with a β-galactosidase–specific control siRNA (siCON) or with a claudin-2–specific siRNA (siCLDN2), and the percentages of ALDHHigh and ALDHLow cells were reanalyzed in each population 3 and 6 days after transfection. Data represent one of three similar experiments.

Figure 4.

Claudin-2 promotes the CSC phenotype in vitro. A, Expression of the mRNAs for CLDN2, BMI1, and Oct-4 in T84 cells stably transfected with control of claudin-2–specific shRNAs. Student t test, *, P < 0.01; mean ± SEM, n = 3. B, Expression of the mRNAs for CLDN2, BMI1, ALDH1A1, and Oct-4 in DLD-1 (left), T84 (middle), and patient-derived CPP42 cells (right) 24 hours after transient transfection with siRNA targeting CLDN2, expressed as fold change compared with their respective expression in cells expressing a β-galactosidase–specific control siRNA. Mean ± SEM from experimental replicates; data from one of two similar experiments; Student t test, *, P < 0.05; **, P < 0.01; ***, P < 0.001. C,CLDN2 mRNA expression in ALDHLow and ALDHHigh T84 cells (mean ± SEM). D, Percentages of ALDHLow (gray) and ALDHHigh cells (white) in untransfected T84 cells (top), as well as 3 days after transfection with siRNA selectively targeting CLDN2 (siCLDN2) or β-galactosidase (siCON). Data represent one of three similar experiments. E, Purified T84 ALDHHigh cells (white, top) or ALDHLow cells (gray, bottom) were independently transfected with a β-galactosidase–specific control siRNA (siCON) or with a claudin-2–specific siRNA (siCLDN2), and the percentages of ALDHHigh and ALDHLow cells were reanalyzed in each population 3 and 6 days after transfection. Data represent one of three similar experiments.

Close modal

As elevated enzymatic activity of aldehyde dehydrogenase has been reported as a robust marker to enrich CSCs (24), we examined whether claudin-2 expression differed in flow cytometry–enriched ALDHHigh versus ALDHLow cells. CLDN2 mRNA levels were 2.3-fold higher in ALDHHigh than in ALDHLow T84 cells (6.96 ± 0.15 vs. 2.84 ± 0.73, respectively; Fig. 4C). To determine whether claudin-2 expression impacts on the proportion of ALDHHigh cells within heterogeneous cancer cell populations, we transfected T84 cells with CLDN2-selective or with control siRNA. Three days later, control siRNA did not alter homeostatic proportions of ALDHHigh and ALDHLow cells (33% ± 4% and 66% ± 5.2%, respectively) compared with untransfected cells (35% ± 5% and 65% ± 5%; Fig. 4D). In contrast, claudin-2 downregulation significantly reduced the proportion of ALDHHigh cells (17% ± 2% ALDHHigh and 83% ± 6.5% ALDHLow cells; Fig. 4D).

Because the rate of phenotypic transitions between stem-like and non-stem cells was shown to modulate homeostatic proportions between these phenotypes in breast cancer (25), we analyzed whether claudin-2 could alter phenotypic transitions between ALDHHigh and ALDHLow cells in colorectal cancer cells. ALDHHigh and ALDHLow subpopulations were purified from T84 cells using flow cytometry, transfected independently with CLDN2-selective or control siRNA, and seeded as adherent layers in complete medium. Efficiency of claudin-2 downregulation was similar in ALDHHigh (86% inhibition compared with controls) and ALDHLow cells (79% inhibition). Claudin-2 downregulation was transient, with maximal reduction attained by 24 hours, followed by a progressive return to control levels (days 3–6; Supplementary Fig. S5).

The percentage of ALDHHigh and ALDHLow populations was then quantified 3 and 6 days after transfection to determine what proportion of the pure ALDHhigh or ALDHlow subpopulations seeded were able to convert to a different ALDH phenotype (Fig. 4E).

Transition from ALDHHigh toward ALDHLow phenotype was a rapid process. In controls 58% ± 5% ALDHLow cells were detected by day 3 after seeding purified ALDHHigh populations. This transition was faster in cells transfected with claudin-2 siRNA, where 72% ± 5.9% ALDHlow cells were detected by day 3. After 6 days, claudin-2 mRNA expression was back to control levels (Supplementary Fig. S5) and proportions of ALDHHigh and ALDHLow populations were no longer different in cells that had been transfected with control (28% ALDHHigh and 72% ALDHLow) or with claudin-2 siRNA (34% ALDHHigh and 66% ALDHLow; Fig. 4E). This result indicates that the transient downregulation of claudin-2 was responsible for enhancing the phenotypic transition from ALDHHigh to ALDHLow cells.

Conversely, we analyzed whether the transitions from ALDHLow to ALDHHigh was also controlled by claudin-2. Purified ALDHLow cells were able to reconstitute an ALDHHigh population, albeit at a slower rate than the ALDHHigh to ALDHLow conversion (Fig. 4D). ALDHLow cells transfected with control siRNA generated 15% ± 1.9% ALDHHigh cells by 3 days and 25% ± 10.2% by 6 days after transfection. Claudin-2 downregulation in ALDHLow cells reduced the rate of conversion into ALDHHigh cells, with only 2.5% ± 0.3% newly formed ALDHHigh cells at day 3. The proportion of ALDHLow cells converting to an ALDHHigh phenotype increased to 13% ± 3.4% by day 6, when the claudin-2 siRNA is no longer effective. Similar results were obtained with the DLD-1 cell line (Supplementary Fig. S6).

Taken together, these data demonstrate that claudin-2 enhances proportions of ALDHHigh stem-like cells by stabilizing the ALDHHigh cell phenotype and favoring phenotypic transitions from ALDHLow toward ALDHHigh subpopulations.

Claudin-2 regulates the self-renewal capacity of ALDHHigh cells

We demonstrated above that claudin-2 regulates homeostatic proportions of ALDHHigh cells within heterogeneous colon cancer cell populations. To establish whether this is reflected by an increased self-renewal frequency of ALDHHigh cells, we FACS-purified ALDHHigh T84 cells and transfected them in suspension with claudin-2-selective or control siRNA. 12 hours later these cells were seeded in an ELDA assay and the stem cell frequency of each cell population was quantified after 10 days as described previously (18). Claudin-2 downregulation significantly decreased the stem cell frequency from 1 in every 6.24 cells (confidence interval 1/3.31–1/11.8) to 1 in every 39.8 cells (1/21.2–1/74.8; χ2 test (χ2 = 14.6; *, P = 0.000132; Fig. 5A).

Figure 5.

Claudin-2 promotes CSCs self-renewal in vitro and tumorigenic capacities in vivo. A, Extreme limiting dilution analysis was used to determine the self-renewal frequency of ALDHHigh T84 cells transiently transfected with a claudin-2- (siCLDN2) or a β-galactosidase-specific (siCON) siRNA. The presence or absence of spheres was quantified seven days after seeding at 1,000, 100, or 10 or 1 cells/well (each concentration in 12 replicates) in an ultralow adherence 96-well plate and is reported as the estimated sphere-forming frequency + confidence intervals. P = 0.000132; χ2, representative replicate from three experiments. B, Tumor incidence measured 4 to 7 weeks after inoculation of NOD/SCID mice with purified ALDHHigh T84 cells transfected with siCLDN2 or siCON as described under Supplementary Methods, (500 cells/injection, n = 5 per group). C, Representative macroscopic appearance of T84 xenografts induced after subcutaneous inoculation of control siRNA–transfected (siCON) or siCLDN2-transfected (siCLDN2) ALDHHigh T84 cells and harvested after 7 weeks. Black dotted circle, lack of tumor initiation at the inoculation site. D, Scatter plot showing the expression level of CLDN2 mRNA in 10 patient-derived colorectal cancer cell populations, separated into two subgroups based on their stem cell frequencies (low or high, x-axis; mean ± SEM, unpaired t test, two-tailed, *, P < 0.05).

Figure 5.

Claudin-2 promotes CSCs self-renewal in vitro and tumorigenic capacities in vivo. A, Extreme limiting dilution analysis was used to determine the self-renewal frequency of ALDHHigh T84 cells transiently transfected with a claudin-2- (siCLDN2) or a β-galactosidase-specific (siCON) siRNA. The presence or absence of spheres was quantified seven days after seeding at 1,000, 100, or 10 or 1 cells/well (each concentration in 12 replicates) in an ultralow adherence 96-well plate and is reported as the estimated sphere-forming frequency + confidence intervals. P = 0.000132; χ2, representative replicate from three experiments. B, Tumor incidence measured 4 to 7 weeks after inoculation of NOD/SCID mice with purified ALDHHigh T84 cells transfected with siCLDN2 or siCON as described under Supplementary Methods, (500 cells/injection, n = 5 per group). C, Representative macroscopic appearance of T84 xenografts induced after subcutaneous inoculation of control siRNA–transfected (siCON) or siCLDN2-transfected (siCLDN2) ALDHHigh T84 cells and harvested after 7 weeks. Black dotted circle, lack of tumor initiation at the inoculation site. D, Scatter plot showing the expression level of CLDN2 mRNA in 10 patient-derived colorectal cancer cell populations, separated into two subgroups based on their stem cell frequencies (low or high, x-axis; mean ± SEM, unpaired t test, two-tailed, *, P < 0.05).

Close modal

To determine whether claudin-2 alters the tumor-initiating potential of ALDHHigh cells in vivo, purified ALDHHigh T84 cells were transiently transfected as described above and grown overnight to ensure optimal CLDN2 downregulation. As our results (Fig. 4) indicated that phenotypic transitions from ALDHHigh to ALDHLow population occurs rapidly, transfected ALDHHigh cells were repurified by FACS prior to subcutaneous injection into SCID/NOD mice (500 cells/injection) to minimize the dilution of the ALDHHigh population. Six weeks after injection, 5 of 5 mice inoculated with siCON-transfected ALDHHigh cells generated tumors while 2 of 5 mouse only exhibited a measurable tumor following inoculation of ALDHHigh expressing claudin-2 siRNA (Fig. 5B and C). These findings demonstrate that claudin-2 downregulation in T84 ALDHHigh CSC-like cells impairs their ability to initiate tumors in vivo.

We also quantified the endogenous claudin-2 expression levels and the stem cell frequency in 10 different patient-derived cell lines, and found that cell populations containing low proportions of self-renewing cells (<2 % of the total population, n = 5 patient-derived cell populations) expressed lower levels of claudin-2 mRNA (average 1/dCp = 0.060 ± 0.0006, n = 5) compared with those displaying higher proportions (1/dCp = 0.089 ± 0.009, n = 5; Fig. 5D). These data suggest that elevated claudin-2 expression is linked with a high self-renewal frequency in colorectal cancer cells.

Overall, our findings demonstrate that claudin-2 directly modulates the self-renewal and tumor-initiating ability of ALDHHigh colorectal CSCs.

Claudin-2 regulates the expression miR-222-3p to promote self-renewal in ALDHHigh cells

To gain mechanistic insights into how claudin-2 regulates the phenotype of colorectal CSCs, we quantified the effect of this protein on the expression levels of miRNAs, which have recently emerged as dynamic regulators of self-renewal (26, 27) and phenotypic plasticity (28) in CSCs.

We transfected ALDHHigh DLD-1 cells with claudin-2–selective (siCLDN2) or control (siCON) siRNA and quantified miRNA expression 48 hours later using next-generation sequencing. A total of 1,400 miRNAs were detected in ALDHHigh DLD-1 cells. These miRNAs were analyzed for their differential expression between siCON and siCLDN2 cells (> 1.5-fold up- or downregulation, P < 0.05; Fig. 6A). A total of 468/1,400 miRNAs were sufficiently abundant (>20 read counts) to be reliably and reproducibly quantified, 372 of which were expressed in both siCLDN2 and siCON ALDHHigh populations (Fig. 6B). Nine miRNAs were found to be differentially expressed in siCLDN2 ALDHHigh cells compared with controls, two downregulated (miR-1287-5p, miR-589-3p) and seven upregulated (miR-204-5p, miR-222-3p, miR-371b-3p, miR-372-3p, miR-373-3p, miR-629-5p, and miR-532-5p; Fig. 6C). To assess the relevance of these results in patient-derived cells, we quantified two robustly expressed miRNAs (miR-589-3p and miR-372-3p) using qRT-PCR in ALDHHigh CPP42 cells (high endogenous claudin-2 expression) in which claudin-2 expression was downregulated using selective siRNA transfection, and ALDHHigh CPP1 cells (low endogenous CLDN2 expression) that were made to stably overexpress CLDN2 (Supplementary Fig. S7). Similar to what we found in DLD-1 cells, CLDN2 downregulation in ALDHHigh CPP42 cells decreased miR-589-3p expression and increased miR-372-3p expression (Fig. 6D). Conversely, claudin-2 overexpression in ALDHHigh CPP1 cells increased miR-589-3p expression and decreased that of miR-372-3p (Fig. 6D). These results demonstrate that claudin-2 regulates the expression of selective miRNAs in ALDHHigh stem-like colorectal cancer cells. Interestingly, the top 10 most significantly represented molecular pathways in a KEGG pathway analysis of predicted gene targets for these 9 miRNAs included Wnt, Stem Cell, and Hippo signaling (Supplementary Table S3), indicating that claudin-2 expression in ALDHHigh cells is associated with a gene signature of CSCs and related pathways (29, 30).

Figure 6.

miR-222-3p is involved in claudin-2–mediated regulation of self-renewal in ALDHHigh cells. A, Volcano plot summarizing the differential expression of miRNAs [fold-change (x-axis) and P value (y-axis)] in ALDHHigh DLD-1 cells transfected with a claudin-2- (siCLDN2) or a β-galactosidase–specific (siCON) siRNA. Each dot represents one miRNA (red, significantly upregulated; green, significantly downregulated; gray, no significant differential expression; blue lines, threshold of significance). B, Venn diagram showing the number of miRNAs detected in siCLDN2 (n = 64), siCON (n = 32), or both siCLDN2 and siCON (n = 372) ALDHHigh DLD-1 cells (P < 0.05, n = 3). C, Heatmap representing differential expression of 9 miRNAs (P < 0.05, ± 1.5 fold change, > 20 read counts) in ALDHHigh DLD-1 cells transfected with a claudin-2- (siCLDN2) or a β-galactosidase–specific (siCON) siRNA. D, qRT-PCR expression levels for miR-589-3p and miR-372-3p in CPP42 cells (expressing siCLDN2 or siCON) and in CPP1 cells [overexpressing CLDN2 (CLDN2+) or a control vector (CON)]. Data from one representative experiment is shown (n = 2). E, Sphere-forming frequency of ALDHHigh DLD-1 cells transfected with β-galactosidase or claudin-2–specific siRNA with or without mimics or inhibitors for miR-222-3p, 371b-3p, or 589-3p as indicated, as determined using ELDA after seeding at 1,000, 100, 30, 10, and 3 cells/well in ultralow adherence plates (each concentration in 10–12 replicates; n = 3). Scoring was performed 9–11 days after transfection and data is reported as the estimated sphere-forming frequency + CIs, as well as estimated percentage of self-renewing cells. Statistical analysis of differences between groups was performed using the χ2 test. F, Expression of mRNAs encoding claudin-2 as well as miRNA target genes (predicted using Diana MicroT) was quantified using qRT-PCR 24 hours after transfection of ALDHHigh DLD-1 cells with β-galactosidase or claudin-2–specific siRNA with or without mimics or inhibitors for miR-222-3p, 371b-3p, or 589-3p. Data are expressed as fold change compared with β-galactosidase siRNA-transfected cells (n = 3, *, P < 0.05 compared with siBgal; #, P < 0.05 compared with siCLDN2 + control inhibitor, two-way ANOVA + Tukey multiple comparison test).

Figure 6.

miR-222-3p is involved in claudin-2–mediated regulation of self-renewal in ALDHHigh cells. A, Volcano plot summarizing the differential expression of miRNAs [fold-change (x-axis) and P value (y-axis)] in ALDHHigh DLD-1 cells transfected with a claudin-2- (siCLDN2) or a β-galactosidase–specific (siCON) siRNA. Each dot represents one miRNA (red, significantly upregulated; green, significantly downregulated; gray, no significant differential expression; blue lines, threshold of significance). B, Venn diagram showing the number of miRNAs detected in siCLDN2 (n = 64), siCON (n = 32), or both siCLDN2 and siCON (n = 372) ALDHHigh DLD-1 cells (P < 0.05, n = 3). C, Heatmap representing differential expression of 9 miRNAs (P < 0.05, ± 1.5 fold change, > 20 read counts) in ALDHHigh DLD-1 cells transfected with a claudin-2- (siCLDN2) or a β-galactosidase–specific (siCON) siRNA. D, qRT-PCR expression levels for miR-589-3p and miR-372-3p in CPP42 cells (expressing siCLDN2 or siCON) and in CPP1 cells [overexpressing CLDN2 (CLDN2+) or a control vector (CON)]. Data from one representative experiment is shown (n = 2). E, Sphere-forming frequency of ALDHHigh DLD-1 cells transfected with β-galactosidase or claudin-2–specific siRNA with or without mimics or inhibitors for miR-222-3p, 371b-3p, or 589-3p as indicated, as determined using ELDA after seeding at 1,000, 100, 30, 10, and 3 cells/well in ultralow adherence plates (each concentration in 10–12 replicates; n = 3). Scoring was performed 9–11 days after transfection and data is reported as the estimated sphere-forming frequency + CIs, as well as estimated percentage of self-renewing cells. Statistical analysis of differences between groups was performed using the χ2 test. F, Expression of mRNAs encoding claudin-2 as well as miRNA target genes (predicted using Diana MicroT) was quantified using qRT-PCR 24 hours after transfection of ALDHHigh DLD-1 cells with β-galactosidase or claudin-2–specific siRNA with or without mimics or inhibitors for miR-222-3p, 371b-3p, or 589-3p. Data are expressed as fold change compared with β-galactosidase siRNA-transfected cells (n = 3, *, P < 0.05 compared with siBgal; #, P < 0.05 compared with siCLDN2 + control inhibitor, two-way ANOVA + Tukey multiple comparison test).

Close modal

To determine whether the identified miRNAs play a role in the effect of claudin-2 on self-renewal, we used the DianaMicroT database to select candidate miRNAs that exhibit an enriched number of putative targets compared with what would be randomly expected within a manually curated list of 777 self-renewal-related genes (Supplementary Tables S4 and S5). Three miRNAs (miR-222-3p, miR-371b-3p, and miR-589-3p) were selected for further validation as described under Supplementary Methods. To determine whether preventing claudin-2–driven miRNA regulation would alter the impact of claudin-2 on self-renewal, ALDHhigh DLD-1 cells were FACS-purified and transfected with CLDN2 siRNA in the presence or not of a miR-589-3p mimic, of miR-222-3p, or miR-371b-3p inhibitors, or of their respective controls. An ELDA assay was performed to determine whether any of these treatments could block the inhibition of self-renewal observed in cells transfected with claudin-2 siRNA. We found that the miR222-3p inhibitor only was able to prevent the reduction of self-renewal induced by CLDN2 siRNAs in ALDHhigh DLD-1 cells (Fig. 6E). CLDN2 and miRNA target gene expression were quantified to confirm treatment efficacy (Fig. 6F). Expression of predicted miR-222-3p target genes such as SOX10 and cKIT was decreased in cells transfected with CLDN2-specific siRNA (Fig. 6F), in agreement with the increased expression of miR-222-3p observed in these cells (Fig. 6C). Expression of SOX10 and KIT returned to control levels in cells treated concomitantly with the CLDN2 siRNA and miR-222-3p inhibitor but not with CLDN2 siRNA and other miRNA inhibitors, mimics, or controls, confirming the ability of the miR-222-3p to selectively restore miR-222-3p target gene expression despite the maintained downregulation of claudin-2 levels (Fig. 6F). These findings indicate that decreased miR-222-3p expression is essential for the promotion of self-renewal by claudin-2.

Activation of YAP is necessary for claudin-2–mediated regulation of self-renewal

Because the Hippo-YAP signaling pathway is under the control of multiple protein complexes involved in cell–cell junctions and cell polarity (31), and in view of the recently identified link between miR-222 and YAP-TEAD in gastric cancer (32), we examined whether this pathway was involved in the regulation of self-renewal by claudin-2. We first examined YAP localization and quantified the expression of YAP target genes after experimental modulation of claudin-2 in colorectal cancer cells. While YAP was mostly found in the nucleus of colorectal cancer cells before or after claudin-2 downregulation (Fig. 7A) or overexpression (Fig. 7B), expression of target genes (AXL, CTGF) was significantly decreased following claudin-2 downregulation (Fig. 7C) and increased in claudin-2–overexpressing cells (Fig. 7B). Decrease AXL expression following claudin-2 downregulation was only slightly reduced by the miR-589-3p mimic and miR-222-3p inhibitor (Fig. 7C), suggesting that miR-222-3p, 371b-3p and 589-3p are not essential mediators of AXL regulation by claudin-2. To establish whether YAP could act upstream from miR-222-3p in mediating claudin-2-driven self-renewal, control or claudin-2–overexpressing CPP1 cells were treated with two compounds that inhibit YAP via different mechanisms, Verteporfin (33) and simvastatin (34). miR-222-3p expression was decreased in claudin-2–overexpressing cells, and that this effect was inhibited by verteporfin and simvastatin treatment (Fig. 7D). Treatment with these compounds also reversed the promoting effect of claudin-2 overexpression on CPP1 self-renewal (Fig. 7E). Collectively, these results demonstrate that increased YAP activity acts upstream from the regulation of miR-222-3p expression by claudin-2, and indicates that this pathway is instrumental for the regulation of self-renewal by claudin-2.

Figure 7.

YAP acts upstream from miR-222-3p to mediate claudin-2–promoted self-renewal. A, Immunofluorescent staining of claudin-2 and YAP in DLD-1 colorectal cancer cells transfected with β-galactosidase (control) or claudin-2 specific siRNA. Nuclei were stained using DAPI. Scale bars, 50 μm. B, Decreased mRNA expression for the key YAP target gene AXL in ALDHhigh DLD-1 cells transfected with β-galactosidase or claudin-2–specific siRNA with or without mimics or inhibitors for miR-222-3p, 371b-3p, or 589-3p as indicated, as quantified using qRT-PCR. P < 0.05 compared with siBgal (*) or to siCLDN2 (#), ANOVA with Tukey post hoc test. C, YAP immunostaining and expression of YAP target genes in CPP1 patient-derived cells expressing a control or claudin-2–encoding vector. *, P < 0.05 compared with control CPP1 cells, Student t test. Expression of claudin-2 in these cells is shown in Supplementary Fig. S2. D, Expression of miR-222-3p in control or claudin-2–expressing CPP1 cells treated with vehicle, verteporfin, or simvastatin as indicated, quantified using the TaqMan miRNA assays. Data from a representative experiment is expressed as fold change compared with vehicle-treated control CPP1 cells (n = 3 technical replicates). E, Sphere-forming frequency in control (CT) or claudin-2 expressing CPP1 cells, determined using ELDA after seeding at 1,000, 100, 10, and 1 cell(s)/well in ultralow adherence plates (each concentration in 12 replicates) and treating with vehicle, verteporfin, or simvastatin as indicated. Scoring was performed 7–9 days after seeding and data is reported as the estimated sphere-forming frequency + confidence intervals, as well as estimated percentage of self-renewing cells. Statistical analysis of differences between groups was performed using the χ2 test.

Figure 7.

YAP acts upstream from miR-222-3p to mediate claudin-2–promoted self-renewal. A, Immunofluorescent staining of claudin-2 and YAP in DLD-1 colorectal cancer cells transfected with β-galactosidase (control) or claudin-2 specific siRNA. Nuclei were stained using DAPI. Scale bars, 50 μm. B, Decreased mRNA expression for the key YAP target gene AXL in ALDHhigh DLD-1 cells transfected with β-galactosidase or claudin-2–specific siRNA with or without mimics or inhibitors for miR-222-3p, 371b-3p, or 589-3p as indicated, as quantified using qRT-PCR. P < 0.05 compared with siBgal (*) or to siCLDN2 (#), ANOVA with Tukey post hoc test. C, YAP immunostaining and expression of YAP target genes in CPP1 patient-derived cells expressing a control or claudin-2–encoding vector. *, P < 0.05 compared with control CPP1 cells, Student t test. Expression of claudin-2 in these cells is shown in Supplementary Fig. S2. D, Expression of miR-222-3p in control or claudin-2–expressing CPP1 cells treated with vehicle, verteporfin, or simvastatin as indicated, quantified using the TaqMan miRNA assays. Data from a representative experiment is expressed as fold change compared with vehicle-treated control CPP1 cells (n = 3 technical replicates). E, Sphere-forming frequency in control (CT) or claudin-2 expressing CPP1 cells, determined using ELDA after seeding at 1,000, 100, 10, and 1 cell(s)/well in ultralow adherence plates (each concentration in 12 replicates) and treating with vehicle, verteporfin, or simvastatin as indicated. Scoring was performed 7–9 days after seeding and data is reported as the estimated sphere-forming frequency + confidence intervals, as well as estimated percentage of self-renewing cells. Statistical analysis of differences between groups was performed using the χ2 test.

Close modal

In this study, we found that claudin-2 overexpression was linked with poor outcome in patients with chemotherapy-treated stage II/III colorectal cancer and promoted the self-renewal properties of patient-derived colorectal cancer cells and cell lines in vitro and in vitro. High claudin-2 expression stabilized the phenotype of self-renewing colorectal cancer cells, slowing down their differentiation toward non-stem-like states and promoting the phenotypic transition of non-stem cells toward a stem-like phenotype. Claudin-2 differentially regulated the expression of nine miRNAs in ALDHHigh stem-like cells, and we found that activation of YAP and downstream repression of miR-222-3p were involved in the promoting effect of claudin-2 on self-renewal.

Previous studies have shown that claudin-2 expression is upregulated during progression of colorectal cancer (6, 7). However, these studies did not directly assess claudin-2 expression levels in relation with treatment and disease outcomes. Our analysis of multiple independent datasets indicates that, in patients with stage II/III colorectal cancer undergoing adjuvant chemotherapy, the high expression of CLDN2 significantly correlates with enhanced recurrence and poorer prognosis. Accordingly, the ability of claudin-2 to promote postchemotherapy recurrence was confirmed using serial transplantation of colorectal cancer cells following chemotherapy. These findings build on recent in vitro studies suggesting that increased claudin-2 expression decreased the sensitivity to 5-FU (6) and are consistent with the recently reported association between claudin-2 expression and poor clinical outcome in stage IV colorectal cancer (35). In addition, claudin-2 was reported to be a negative prognostic factor and a marker of liver metastasis and early recurrence in breast cancer (36, 37).

Using limiting dilution assays in vitro and in vivo, we determined that elevated claudin-2 expression in colorectal cancer cells increased the self-renewing capacity of CSCs and skewed phenotypic transitions in favor of the ALDHHigh phenotype. This discovery of a promoting role for claudin-2 on self-renewal complements its previously reported function on the promotion of tumor growth (6) and the inhibition of cell polarity and terminal differentiation (8). We corroborated these results in several animal models of primary and metastatic tumor initiation, and our findings are thus consistent with a promoting role of claudin-2 on CSC-like cell self-renewal and with the reported role of chemoresistant CSCs in tumor progression, metastasis initiation (20, 23), and posttreatment recurrence (38). Interestingly, two other claudin isoforms (claudin-1 and/or -4) were also reported as biomarkers of progression and recurrence in oral squamous cell carcinoma (39) and lung adenocarcinoma (40), suggesting that these isoforms may also promote self-renewal in these models.

In addition, miRNA expression profiling was used to identify that claudin-2 selectively regulated the expression of several miRNAs in ALDHHigh CSCs. miRNAs are important modulators of CSCs (41) and claudin-1 was recently shown to control miRNA dynamics in breast cancer (42). Our findings corroborate the ability of claudins to regulate miRNA expression and provide important insights into mechanisms that contribute to the control of self-renewal. Indeed, selective inhibition of miR-222-3p was found to prevent the decrease in self-renewal detected upon claudin-2 downregulation in ALDHhigh colorectal cancer cells, indicating that claudin-2 represses miR-222-3p in CSCs and that this contributes to its promotion of self-renewal. Interestingly, other recent reports suggested that miR-222 promotes the proliferation (43), migration, and invasion (44) of colorectal cancer cells. However, these studies did not specifically analyze the 3p isoform of miR-222 and this apparent paradox may be explained by the existence of multiple naturally-existing miR-222 isoforms with distinct functions (45).

Our study also established that YAP activation was essential to the promotion of self-renewal by claudin-2 in ALDHhigh CSCs. Key YAP target genes such as AXL and CTGF were activated by claudin-2 and pharmacologic inhibitors of this pathway reversed YAP target gene activation and prevented the claudin-2-mediated self-renewal increase. In addition, pharmacologic inhibition of YAP was found to rescue the claudin-2-induced miR-222-3p repression, suggesting that YAP acts upstream from miR-222-3p in mediating the effect of claudin-2 on colorectal CSCs. Although our results do not indicate if miR-222-3p is a direct target of YAP-mediated transcription, YAP has been shown to regulate the expression of other miRNAs (46) and to regulate miRNA biogenesis in cancer cells (47).

Our findings of a role for YAP are consistent with the well-described role of YAP activity in promoting stemness and tumor initiation (48), and with the role of AXL as a key mediator of chemotherapy-induced invasion and poor clinical outcome in colorectal cancer (49). In contrast, our results may appear paradoxical in light of the described role of junction proteins in the inhibition of YAP activity in healthy and polarized epithelia (31). However, in polarized tissues, cell junction proteins form highly organized membrane complexes and usually drive contact inhibition by activating Hippo signaling and sequestering YAP away from the nucleus, thereby blocking its transcriptional activity (31). In the context of colorectal tumor cells, where polarity and contact inhibition are at least partially lost, our results indicate that overexpression of claudin-2 enhances the YAP activity without strongly affecting its nuclear localization, supporting the observation by others that regulation of YAP in tumor cells can be independent from Hippo (48).

In conclusion, our findings indicate that high expression of claudin-2 in colorectal tumors promotes CSC self-renewal and hampers their transition toward more differentiated phenotypes. This provides a credible rationale for the potential development of claudin-2–selective antibodies as prospective tools to decrease the self-renewal potential of colorectal CSCs. A proof of concept for therapeutic usage of claudin-specific antibodies has recently been established, with the claudin-18.2 antibody IMAB362 currently undergoing phase II clinical trials in gastrointestinal, esophageal, and ovarian cancer (NCT01630083, NCT020054351, and NCT01197885 at www.clinicaltrials.gov). From a mechanistic perspective, our findings unravel a novel mechanism implicating the claudin-2/YAP/miR-222-3p axis and linking the cell membrane to the regulation of self-renewal in colorectal cancer cells.

A.F. Hill is a consultant/advisory board member for Caldera Health (NZ). E.K. Sloan is a associate professor (adjunct) at University of California. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Paquet-Fifield, S.L. Koh, L. Cheng, A. Puisieux, J. Pannequin, E.K. Sloan, F. Hollande

Development of methodology: S.L. Koh, L. Cheng, R. Nasr, J. Pannequin, A.F. Hill, F. Hollande

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Paquet-Fifield, S.L. Koh, L. Cheng, L.M. Beyit, C.E. Shembrey, C. Moelck, C. Behrenbruch, M. Papin, M. Gironella, S. Guelfi, R. Nasr, F. Grillet, M. Prudhomme, J.-F. Bourgaux, A. Castells, J.-M. Pascussi, A.G. Heriot, A.F. Hill, E.K. Sloan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Paquet-Fifield, S.L. Koh, L. Cheng, L.M. Beyit, C. Moelck, M. Papin, M. Gironella, R. Nasr, A. Castells, M.J. Davis, A.F. Hill, E.K. Sloan, F. Hollande

Writing, review, and/or revision of the manuscript: S. Paquet-Fifield, S.L. Koh, L. Cheng, C. Moelck, A. Castells, A.G. Heriot, A. Puisieux, A.F. Hill, E.K. Sloan, F. Hollande

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.L. Koh

Study supervision: F. Hollande

This work was supported by the National Health and Medical Research Council (NHMRC) of Australia (#1049561, #1069024) and the French Association pour la Recherche contre le Cancer (ARC)(#5046).

The authors wish to thank Drs. Hyun-Jung Cho and Paul McMillan (BOMP imaging facility, University of Melbourne) and Dr. Vanta Jameson (Flow Cytometry facility, University of Melbourne) for their technical support. They are also very grateful to Prof. K. Harvey for helpful discussions and for sharing Hippo/YAP pathway reagents. Biospecimens and data used for TMA immunostaining experiments were obtained from the Victorian Cancer Biobank, Victoria, Australia, with appropriate ethics approval. The Victorian Cancer Biobank is supported by the Victorian Government.

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.

1.
Lal-Nag
M
,
Morin
PJ
. 
The claudins
.
Genome Biol
2009
;
10
:
235
.
2.
Gupta
IR
,
Ryan
AK
. 
Claudins: unlocking the code to tight junction function during embryogenesis and in disease
.
Clin Genet
2010
;
77
:
314
25
.
3.
Morin
PJ
. 
Claudin proteins in human cancer: promising new targets for diagnosis and therapy
.
Cancer Res
2005
;
65
:
9603
6
.
4.
Darido
C
,
Buchert
M
,
Pannequin
J
,
Bastide
P
,
Zalzali
H
,
Mantamadiotis
T
, et al
Defective claudin-7 regulation by Tcf-4 and Sox-9 disrupts the polarity and increases the tumorigenicity of colorectal cancer cells
.
Cancer Res
2008
;
68
:
4258
68
.
5.
Singh
AB
,
Sharma
A
,
Smith
JJ
,
Krishnan
M
,
Chen
X
,
Eschrich
S
, et al
Claudin-1 up-regulates the repressor ZEB-1 to inhibit E-cadherin expression in colon cancer cells
.
Gastroenterology
2011
;
141
:
2140
53
.
6.
Dhawan
P
,
Ahmad
R
,
Chaturvedi
R
,
Smith
JJ
,
Midha
R
,
Mittal
MK
, et al
Claudin-2 expression increases tumorigenicity of colon cancer cells: role of epidermal growth factor receptor activation
.
Oncogene
2011
;
30
:
3234
47
.
7.
Kinugasa
T
,
Huo
Q
,
Higashi
D
,
Shibaguchi
H
,
Kuroki
M
,
Tanaka
T
, et al
Selective up-regulation of claudin-1 and claudin-2 in colorectal cancer
.
Anticancer Res
2007
;
27
:
3729
34
.
8.
Buchert
M
,
Papin
M
,
Bonnans
C
,
Darido
C
,
Raye
WS
,
Garambois
V
, et al
Symplekin promotes tumorigenicity by up-regulating claudin-2 expression
.
Proc Nat Acad Sci U S A
2010
;
107
:
2628
33
.
9.
Rahner
C
,
Mitic
LL
,
Anderson
JM
. 
Heterogeneity in expression and subcellular localization of claudins 2, 3, 4, and 5 in the rat liver, pancreas, and gut
.
Gastroenterology
2001
;
120
:
411
22
.
10.
Ahmad
R
,
Chaturvedi
R
,
Olivares-Villagomez
D
,
Habib
T
,
Asim
M
,
Shivesh
P
, et al
Targeted colonic claudin-2 expression renders resistance to epithelial injury, induces immune suppression, and protects from colitis
.
Mucosal Immunol
2014
;
7
:
1340
53
.
11.
Bonnans
C
,
Flaceliere
M
,
Grillet
F
,
Dantec
C
,
Desvignes
JP
,
Pannequin
J
, et al
Essential requirement for beta-arrestin2 in mouse intestinal tumors with elevated Wnt signaling
.
Proc Nat Acad Sci U S A
2012
;
109
:
3047
52
.
12.
Adorno-Cruz
V
,
Kibria
G
,
Liu
X
,
Doherty
M
,
Junk
DJ
,
Guan
D
, et al
Cancer stem cells: targeting the roots of cancer, seeds of metastasis, and sources of therapy resistance
.
Cancer Res
2015
;
75
:
924
9
.
13.
Planque
C
,
Rajabi
F
,
Grillet
F
,
Finetti
P
,
Bertucci
F
,
Gironella
M
, et al
Pregnane X-receptor promotes stem cell-mediated colon cancer relapse
.
Oncotarget
2016
;
7
:
56558
73
.
14.
Giraldez
MD
,
Lozano
JJ
,
Cuatrecasas
M
,
Alonso-Espinaco
V
,
Maurel
J
,
Marmol
M
, et al
Gene-expression signature of tumor recurrence in patients with stage II and III colon cancer treated with 5′fluoruracil-based adjuvant chemotherapy
.
Int J Cancer
2013
;
132
:
1090
7
.
15.
Sveen
A
,
Agesen
TH
,
Nesbakken
A
,
Rognum
TO
,
Lothe
RA
,
Skotheim
RI
. 
Transcriptome instability in colorectal cancer identified by exon microarray analyses: Associations with splicing factor expression levels and patient survival
.
Genome Med
2011
;
3
:
32
.
16.
Aguirre-Gamboa
R
,
Gomez-Rueda
H
,
Martinez-Ledesma
E
,
Martinez-Torteya
A
,
Chacolla-Huaringa
R
,
Rodriguez-Barrientos
A
, et al
SurvExpress: an online biomarker validation tool and database for cancer gene expression data using survival analysis
.
PLoS One
2013
;
8
:
e74250
.
17.
Marisa
L
,
de Reynies
A
,
Duval
A
,
Selves
J
,
Gaub
MP
,
Vescovo
L
, et al
Gene expression classification of colon cancer into molecular subtypes: characterization, validation, and prognostic value
.
PLoS Med
2013
;
10
:
e1001453
.
18.
Giraud
J
,
Failla
LM
,
Pascussi
JM
,
Lagerqvist
EL
,
Ollier
J
,
Finetti
P
, et al
Autocrine secretion of progastrin promotes the survival and self-renewal of colon cancer stem-like cells
.
Cancer Res
2016
;
76
:
3618
28
.
19.
Dawood
S
,
Austin
L
,
Cristofanilli
M
. 
Cancer stem cells: implications for cancer therapy
.
Oncology
2014
;
28
:
1101
7
,
10
.
20.
Todaro
M
,
Francipane
MG
,
Medema
JP
,
Stassi
G
. 
Colon cancer stem cells: promise of targeted therapy
.
Gastroenterology
2010
;
138
:
2151
62
.
21.
Islam
F
,
Gopalan
V
,
Smith
RA
,
Lam
AK
. 
Translational potential of cancer stem cells: a review of the detection of cancer stem cells and their roles in cancer recurrence and cancer treatment
.
Exp Cell Res
2015
;
335
:
135
47
.
22.
Kreso
A
,
O'Brien
CA
. 
Colon cancer stem cells
.
Current Protoc Stem Cell Biol
2008
;
Chapter 3:Unit 3.1
.
23.
Kreso
A
,
Dick
JE
. 
Evolution of the cancer stem cell model
.
Cell Stem Cell
2014
;
14
:
275
91
.
24.
Huang
EH
,
Hynes
MJ
,
Zhang
T
,
Ginestier
C
,
Dontu
G
,
Appelman
H
, et al
Aldehyde dehydrogenase 1 is a marker for normal and malignant human colonic stem cells (SC) and tracks SC overpopulation during colon tumorigenesis
.
Cancer Res
2009
;
69
:
3382
9
.
25.
Gupta
PB
,
Fillmore
CM
,
Jiang
G
,
Shapira
SD
,
Tao
K
,
Kuperwasser
C
, et al
Stochastic state transitions give rise to phenotypic equilibrium in populations of cancer cells
.
Cell
2011
;
146
:
633
44
.
26.
Deng
L
,
Shang
L
,
Bai
S
,
Chen
J
,
He
X
,
Martin-Trevino
R
, et al
MicroRNA100 inhibits self-renewal of breast cancer stem-like cells and breast tumor development
.
Cancer Res
2014
;
74
:
6648
60
.
27.
Wang
H
,
Sun
T
,
Hu
J
,
Zhang
R
,
Rao
Y
,
Wang
S
, et al
miR-33a promotes glioma-initiating cell self-renewal via PKA and NOTCH pathways
.
J Clin Invest
2014
;
124
:
4489
502
.
28.
Zhu
Y
,
Luo
M
,
Brooks
M
,
Clouthier
SG
,
Wicha
MS
. 
Biological and clinical significance of cancer stem cell plasticity
.
Clin Translat Med
2014
;
3
:
32
.
29.
Hao
J
,
Zhang
Y
,
Jing
D
,
Li
Y
,
Li
J
,
Zhao
Z
. 
Role of Hippo signaling in cancer stem cells
.
J Cell Physiol
2014
;
229
:
266
70
.
30.
Reya
T
,
Clevers
H
. 
Wnt signalling in stem cells and cancer
.
Nature
2005
;
434
:
843
50
.
31.
Gumbiner
BM
,
Kim
NG
. 
The Hippo-YAP signaling pathway and contact inhibition of growth
.
J Cell Sci
2014
;
127
:
709
17
.
32.
Li
N
,
Yu
N
,
Wang
J
,
Xi
H
,
Lu
W
,
Xu
H
, et al
miR-222/VGLL4/YAP-TEAD1 regulatory loop promotes proliferation and invasion of gastric cancer cells
.
Am J Cancer Res
2015
;
5
:
1158
68
.
33.
Brodowska
K
,
Al-Moujahed
A
,
Marmalidou
A
,
Meyer Zu Horste
M
,
Cichy
J
,
Miller
JW
, et al
The clinically used photosensitizer Verteporfin (VP) inhibits YAP-TEAD and human retinoblastoma cell growth in vitro without light activation
.
Exp Eye Res
2014
;
124
:
67
73
.
34.
Wang
Z
,
Wu
Y
,
Wang
H
,
Zhang
Y
,
Mei
L
,
Fang
X
, et al
Interplay of mevalonate and Hippo pathways regulates RHAMM transcription via YAP to modulate breast cancer cell motility
.
Proc Nat Acad Sci U S A
2014
;
111
:
E89
98
.
35.
Mezheyeuski
A
,
Strell
C
,
Hrynchyk
I
,
Guren
TK
,
Dragomir
A
,
Doroshenko
T
, et al
Treatment-related survival associations of claudin-2 expression in fibroblasts of colorectal cancer
.
Virchows Arch
2017 Nov 13
.
[Epub ahead of print]
.
36.
Kimbung
S
,
Kovacs
A
,
Bendahl
PO
,
Malmstrom
P
,
Ferno
M
,
Hatschek
T
, et al
Claudin-2 is an independent negative prognostic factor in breast cancer and specifically predicts early liver recurrences
.
Mol Oncol
2014
;
8
:
119
28
.
37.
Tabaries
S
,
Dupuy
F
,
Dong
Z
,
Monast
A
,
Annis
MG
,
Spicer
J
, et al
Claudin-2 promotes breast cancer liver metastasis by facilitating tumor cell interactions with hepatocytes
.
Mol Cell Biol
2012
;
32
:
2979
91
.
38.
Colak
S
,
Medema
JP
. 
Cancer stem cells–important players in tumor therapy resistance
.
FEBS J
2014
;
281
:
4779
91
.
39.
DE Vicente
J
,
Fernandez-Valle
A
,
Vivanco-Allende
B
,
Santamarta
TR
,
Lequerica-Fernandez
P
,
Hernandez-Vallejo
G
, et al
The prognostic role of claudins -1 and -4 in oral squamous cell carcinoma
.
Anticancer Res
2015
;
35
:
2949
59
.
40.
Chae
MC
,
Park
CK
,
Keum
DY
,
Hwang
I
,
Kwon
KY
,
Jang
BC
. 
Prognostic significance of claudin 4 in completely resected adenocarcinoma of the lung
.
Korean J Thorac Cardiovasc Surg
2014
;
47
:
262
8
.
41.
Garofalo
M
,
Croce
CM
. 
Role of microRNAs in maintaining cancer stem cells
.
Adv Drug Deliv Rev
2015
;
81
:
53
61
.
42.
Majer
A
,
Blanchard
AA
,
Medina
S
,
Booth
SA
,
Myal
Y
. 
Claudin 1 expression levels affect miRNA dynamics in human basal-like breast cancer cells
.
DNA Cell Biol
2016
;
35
:
328
39
.
43.
Liu
S
,
Sun
X
,
Wang
M
,
Hou
Y
,
Zhan
Y
,
Jiang
Y
, et al
A microRNA 221- and 222-mediated feedback loop maintains constitutive activation of NFkappaB and STAT3 in colorectal cancer cells
.
Gastroenterology
2014
;
147
:
847
59.e11
.
44.
Gao
H
,
Cong
X
,
Zhou
J
,
Guan
M
. 
MicroRNA-222 influences migration and invasion through MIA3 in colorectal cancer
.
Cancer Cell Int
2017
;
17
:
78
.
45.
Yu
F
,
Pillman
KA
,
Neilsen
CT
,
Toubia
J
,
Lawrence
DM
,
Tsykin
A
, et al
Naturally existing isoforms of miR-222 have distinct functions
.
Nucleic Acids Res
2017
;
45
:
11371
85
.
46.
Shen
S
,
Guo
X
,
Yan
H
,
Lu
Y
,
Ji
X
,
Li
L
, et al
A miR-130a-YAP positive feedback loop promotes organ size and tumorigenesis
.
Cell Res
2015
;
25
:
997
1012
.
47.
Yu
FX
,
Guan
KL
. 
Transcription and processing: multilayer controls of RNA biogenesis by the Hippo pathway
.
EMBO J
2014
;
33
:
942
4
.
48.
Zanconato
F
,
Cordenonsi
M
,
Piccolo
S
. 
YAP/TAZ at the Roots of Cancer
.
Cancer Cell
2016
;
29
:
783
803
.
49.
Dunne
PD
,
McArt
DG
,
Blayney
JK
,
Kalimutho
M
,
Greer
S
,
Wang
T
, et al
AXL is a key regulator of inherent and chemotherapy-induced invasion and predicts a poor clinical outcome in early-stage colon cancer
.
Clin Cancer Res
2014
;
20
:
164
75
.