A Novel Wnt Regulatory Axis in Endometrioid Endometrial Cancer Free

Endometrial cancer is the most common gynecologic malignancy and ranks fourth in whole malignancies among women (1). Endometrioid endometrial carcinoma (EEC), accounting for ∼80% to 90% of the whole cases, originates from epithelial cells lining the endometrium and is often associated with estrogen stimulation, hormone receptor positivity, obesity, and favorable prognosis (2). Key mutational events have been characterized in EEC, but the underlying molecular mechanisms involving oncogenic or tumor suppressive factors remain poorly elucidated (3, 4). Recently, we have discovered a novel miR193–YY1–adenomatous polyposis coli (APC) regulatory axis that exerts functional roles in EEC development (5). The transcription factor YY1 plays an oncogenic function through epigenetic regulation of APC, a key molecule in regulating the Wnt/β-catenin signaling pathway, which plays an essential role in cancer progression. In this study, we investigated the tumor suppressive function of Protocadherin 10 (PCDH10) in EEC. PCDH10, a member of nonclustered protocadherin subfamily, was proposed as a tumor suppressor (6–8) in multiple cancers. Inactivation of PCDH10 because of promoter CpG hypermethylation has been detected in gastric, hepatocellular, colorectal, breast, cervical, lung, nasopharyngeal, esophageal, pancreatic, and bladder cancer (6, 9–16). Functional studies revealed that re-expression of PCDH10 inhibits cell growth, reduces clonogenicity, restrains cell invasion, and induces cell apoptosis, substantiating its tumor suppressor roles (10, 13, 14, 16). Moreover, methylation of PCDH10 manifests its significance through its association with clinical parameters. For example, in gastric cancer, methylation of PCDH10 was identified at early stages of carcinogenesis and linked with poor prognosis (10). In colorectal cancer, PCDH10 promoter methylation can be detected in plasma and the methylation rate in plasma is positively associated with that in tumor tissues in early-stage cancers (17). Although the involvement is extensively unraveled in a wide spectrum of cancers, the linkage between PCDH10 and EEC is unknown and the molecular mechanisms await exploration. Interestingly, recent studies added another layer of epigenetic control of PCDH10 expression. HOTAIR, a long noncoding RNA (lncRNA), recruits the repressive chromatin-remodeling complex PRC2 to PCDH10 genomic loci, leading to the silencing of PCDH10 expression (18, 19). This report suggested interactions between PCDH10 and the novel family of gene regulators, lncRNAs.

LncRNAs are RNA species more than 200 nucleotides in length and are recently discovered to constitute a large proportion of the whole transcriptome (20, 21). Increasing evidences suggest the importance of lncRNAs in numerous cellular processes impacting gene regulation, often through interacting with diverse chromatin complexes (22–24). In cancer, lncRNAs are now emerging as a prominent layer of transcriptional regulation. More and more lncRNAs are found to be deregulated in different contexts, in conjunction with their tissue-specific nature, making them promising therapeutic targets (25–27). Among them, matastasis-associated lung adenocarcinoma transcript 1 (MALAT1) is one of the most characterized. MALAT1 is a ∼7-kb long, nuclear retained and ubiquitously expressed long ncRNA (28, 29). Since its discovery as a prognostic factor for lung cancer metastasis (28), in the following decade MALAT1 has been shown to be broadly upregulated in a variety of cancer entities and to play critical roles in distinct cancer hallmark capabilities (26). For example, in liver cancer, MALAT1 is associated with risk of tumor recurrence after liver transplantation by modulating cell viability, motility, and invasiveness (30). In cervical cancer, MALAT1 promotes cell proliferation and invasion; knockdown of MALAT1 induces cell apoptosis (31). In addition, in colorectal cancer, MALAT1 harbors mutations and its processed 3′-end fragment affects tumor growth and invasion (32). A myriad of MALAT1 functions underscore its importance in tumor development and progression, however, study of MALAT1 function in EEC is still lacking. Furthermore, mechanistic studies revealed that MALAT1 could regulate alternative splicing, control the expression of cell-cycle regulators, and modulate E2F transcriptional activity under different cellular contexts (33–36). Notwithstanding the substantial advancements in our understanding of MALAT1 function, the transcriptional regulation of MALAT1 expression and the causes behind its deregulation in tumors are barely explored.

In this study, we found that PCDH10 is silenced in EEC cells/tumors through its promoter CpG hypermethylation. Ectopic expression of PCDH10 inhibits tumor growth and induces cell apoptosis both in vitro in EEC cells and in vivo in a xenograft tumor. Transcriptomic analysis revealed that PCDH10 expression downregulates MALAT1. Further mechanistic studies uncovered that MALAT1 expression is transcriptionally induced by Wnt/β-catenin signaling through a direct binding site of TCF4 in MALAT1 promoter region. Altogether, our results have uncovered a novel PCDH10–Wnt/β-catenin–MALAT1 regulatory axis that contributes to EEC development and progression.

EEC samples were acquired from the tissue bank of the Department of Obstetrics and Gynaecology, Prince of Wales Hospital. A total of 76 cases of primary endometrioid endometrial adenocarcinoma (EEC) and 45 cases of normal tissues (NE) were enrolled in this study. Normal endometrial tissue specimens were obtained from women who underwent a hysterectomy or endometrial curettage for endometrial-unrelated diseases, such as uterine myoma or prolapse. Clinical staging was performed according to International Federation of Gynecology and Obstetrics (FIGO) criteria. All specimens, clinical information, and procedures were approved by the Clinical Research Ethics Committee of The Chinese University of Hong Kong. Tissue microarray data of 142 endometrial cancer tumors, 19 metastatic lesions, and 18 endometrial hyperplasia samples from a prospectively collected Norwegian population-based bio bank as part of MoMaTEC trial (http://www.clinicaltrials.gov/ct2/show/NCT00598845) was used in PCDH10 and MALAT1 association analysis with clinical parameters, in accordance with approval from the Regional Committee for Medical and Health Research Ethics, Western Norway (NSD 15501). RNA was extracted from freshly frozen tumor samples with confirmed >80% tumor purity using the RNeasy Mini Kit (Qiagen) and hybridized to Agilent Whole Human Genome Microarrays 44 k (Cat. no. G4112F), according to the manufacturer's instructions (www.agilent.com) and described earlier (37). Arrays were scanned using the Agilent Microarray Scanner Bundle. Raw data were imported and analyzed in J-Express software (Molmine). Mean spot signal was used as intensity measure, and expression data were normalized using median over entire array.

Female athymic nude mice, ages 3 to 4 weeks, were used for tumor xenografts. A total of 5 × 10⁶ of empty vector control or PCDH10 stably expressing HEC-1-B cells were subcutaneously injected into the left and right flanks of the mice (n = 5 for each group). At 13 days after implantation, the tumor became apparent. Tumor dimensions were measured every 2 days and the tumor volume was estimated by using the formula V = D × d²/2, where D is the long axis and d the short axis of the tumor. Mice were sacrificed 43 days postinjection, and tumors were excised, weighed, and snap-frozen for RNA extraction, or paraffin-embedded for IHC staining. Si-NC or si-MALAT1 oligos were injected into the xenograft tumors for MALAT1 knockdown. Details can be found in Supplementary Information. All animal experiments were approved by The Chinese University of Hong Kong Animal Experimentation Ethics Committee.

Human endometrial cancer cell lines, AN3CA, HEC-1-B, HEC-1-A, KLE, and RL95-2, were obtained from the ATCC and cultured as recommended. To generate AN3CA or HEC-1-B cell line stably expressing PCDH10, the overexpression or vector control plasmid was transfected into the cells followed by G418 selection for 10 days. The resistant colonies were pooled together and amplified for further use. Details for viable cell counting, MTS assay, colony formation assay, wound healing assay and 5-Aza-2′-deoxycytidine (5-Aza) treatment, lithium chloride (LiCl) treatment, FACS analysis, TUNEL, and luciferase reporter assay are included in Supplementary Information.

The PCDH10 expression vector was a kind gift from Prof. Jun Yu (Chinese University of Hong Kong, CUHK, Hong Kong, China). Topflash reporter plasmid was a gift from Prof. Kingston Mak (CUHK). Human MALAT1 expression vector was a kind gift from Prof. Kannanganattu V. Prasanth (University of Illinois, Urbana IL). To construct the pGL3-MALAT1 wild-type reporter plasmid, a 516-bp fragment harboring TCF4 binding site was amplified from HEC-1-B genomic DNA and cloned into the KpnI and NheI site of the pGL3-basic vector (Promega) according to the manufacturer's instructions. The mutant plasmid was generated by mutating the TCF4 binding motif from CTTTGAA to CTTTGCG using the method described previously (38). Primers used are listed in Supplementary Table S5.

Genomic DNA was modified by sodium bisulfite as described in Supplementary Information.

siRNA against MALAT1 or control oligos were obtained from Ribobio. In each case, the concentration used for transient transfections was 100 nmol/L. Sequences of siRNA oligos are listed in Supplementary Table S5.

An RNA antisense probe was in vitro transcribed corresponding to 6871 to 7224 bp of MALAT1 (RefSeq accession, NR_002819; ref. 29). The DNA template for this sequence was PCR amplified and subcloned into the EcoRV site of the pUC57 vector (GeneScript). The plasmid was then linearized using BamHI enzyme and in vitro transcribed to synthesize RNA probe using DIG RNA Labeling Kit (Roche Molecular Biochemicals) with T3 RNA Polymerase (Ambion), according to the manufacturer's instructions. Details for in situ hybridization (ISH) were included in Supplementary Information.

Western blotting analysis was performed according to our previous protocol (5). The following dilutions of antibodies were used: anti-β-catenin (1:1,000, 9526; Cell Signaling Technology), anti-α-tubulin (1:5,000, sc-23948; Santa Cruz), anti-caspase-3 (1:300, sc-7148; Santa Cruz), and anti-caspase-9 (1:2,000, ab25758; Abcam). For immunofluorescence staining of cultured cells, the following dilutions were used: anti-Ki67 (1:200, sc-15402; Santa Cruz). For Ki67 and DAPI quantification, counting was conducted on at least 10 randomly chosen fields using Image-Pro Plus 6.0 software.

Chromatin immunoprecipitation (ChIP) assays were carried out as previously described (39). Five micrograms of antibodies against β-catenin (9562; Cell Signaling Technology) or equal amount isotype IgG (Santa Cruz Biotechnology) as a negative control was used for each 2 × 10⁷ cell per ChIP. Immunoprecipitated genomic DNA was resuspended in 15 μL of water. PCR was then performed with 1 μL of DNA as a template and products were analyzed by qRT-PCR on a 7900HT system (Life Technologies). Primers for PCRs are listed in Supplementary Table S5.

Preparation of transcription libraries for sequencing on Illumina GA2x platform was carried out using mRNA-seq Sample Preparation Kit (Part no. 1004898 Rev. D) according to the manufacturer's standard protocol. Read mapping to genome with splice-aware aligner Sequenced were described in Supplementary Information.

Data were analyzed using SPSS 17.0 software package (SPSS, Statistical Product and Service Solution Chicago). The difference of PCDH10 mRNA expression between tumor and adjacent nontumor tissues was analyzed by the Mann–Whitney U test. For tissue microarray data, the nonparametric Mann–Whitney U test was applied to assess the relationship between MALAT1 or PCDH10 expression levels and clinical parameters. For analyzing the association of MALAT1 ISH scoring with clinical parameters, a Pearson χ² test was used. The correlation between PCDH10 and MALAT1 expression using tissue microarray data was carried out using the Pearson correlation and a simple linear regression model. The difference in tumor growth rate between two groups of mice was determined by repeated measures analysis of variance. All quantitative data, like mRNA expression, MTS and luciferase activity data were obtained as triplicates. Data were shown as mean ± standard deviation (SD). Statistical significance between two groups was assessed by the Student t test. All tests were 2-sided, and P < 0.05 was considered statistically significant.

PCDH10 is reported to be silenced in multiple human cancers (6, 9–16). To investigate its potential involvement in EEC, we first examined its mRNA expression in five EEC cells lines, AN3CA, HEC-1-A, HEC-1-B, RL95-2, and KLE as well as microdissected EEC tumor samples using normal endometrial (NE) tissue as controls. Results showed that PCDH10 was significantly downregulated in 76 tumor samples and all five cell lines as compared with 45 normal controls (Fig. 1A and Supplementary Table S1). Interestingly, no significant difference is seen between secretory versus proliferative endometrium but its level is much higher in the postmenopausal endometrium (Supplementary Fig. S1A). Next, we tested whether such downregulation was caused by its promoter hypermethylation. In line with the previous findings from other cancer types, aberrant hypermethylation in a CpG island (+8 to −328 bp upstream transcriptional start site, TSS) of PCDH10 promoter was detected in EEC tumors while not observed in NE tissues by bisulfite genomic sequencing (Fig. 1B). This finding was further strengthened by analyzing the genome-wide methylation data generated on HumanMethylation 450 array by The Cancer Genome Atlas (TCGA) project. Our analysis results from a cohort of 208 EEC patients and 34 normal controls showed a marked hypermethylation on the above region of PCDH10 promoter in EEC samples compared with the normal controls (Fig. 1C and Supplementary Fig. S1B and S1C). Interestingly, mining the data we identified a second CpG island downstream of the TSS (+1194 to +3159), which is also significantly hypermethylated in EEC samples. Furthermore, we downloaded gene expression data generated from an Illumina GA RNA sequencing platform from TCGA and explored the correlation between PCDH10 expression levels with the methylation intensity. Expectedly, an evident anticorrelation was detected in about 40% of the samples (Fig. 1D). These samples exhibit high methylation level and a low expression; on the other hand, some samples with low methylation level exhibit either high or low expression level, resulting in an L shape, which is typical for genes repressed by methylation (40). Consistently, when treated with demethylation agent, 5-Aza, the promoter hypermehtylation was markedly reduced (Fig. 1E) and PCDH10 expression was restored in EEC cells (Fig. 1F). Collectively, the above results demonstrate that PCDH10 is downregulated in EEC cell lines and some tumor samples through its promoter hypermethylation.

The decrease of PCDH10 expression in EEC tumors implies that it may play a tumor-suppressive role in EEC tumorigenesis. To test this notion, we first performed gain-of-function study by overexpressing PCDH10 in two EEC cell lines, AN3CA or HEC-1-B. Successful restoration of PCDH10 expression (Fig. 2A) was found to inhibit EEC cell proliferation in a dose-dependent manner as assessed by both viable cell counting (Fig. 2B) and MTS assay (Fig. 2C). Furthermore, overexpression of PCDH10 in the 2 cell lines impeded their abilities to grow in an attachment-independent manner; the colonies formed in PCDH10 overexpressing cells were remarkably fewer in number and smaller in size than those in the empty vector group (Fig. 2D). PCDH10 has been found to be involved in suppression of cell migration in several cancer types (10, 16). To answer whether this could be extended to EEC cells, we generated AN3CA and HEC-1-B cell lines stably expressing PCDH10 or empty vector (Supplementary Fig. S2A). Interestingly, results from wound healing assay demonstrated that PCDH10 overexpression had no detectable effect on cell motility in both cell lines (Fig. 2E).

To determine how cell proliferation is modulated by PCDH10, we investigated its effect on cell cycle. A significant delay in G₁ and G₂ progression was only observed in HEC-1-B but not in AN3CA cells (Fig. 2F). However, a marked increase in the number of subG₁ cells was detected in both AN3CA and HEC-1-B cells (Fig. 2F), suggesting that PCDH10 expression may have led to cell apoptosis. To confirm this thought, Annexin V–PI double staining revealed that overexpression of PCDH10 induced a reduction in the living cell population (Annexin V−/PI−, LL) and an accompanying increase in the early apoptotic population (Annexin V+/PI−, LR; Fig. 2G). Consistently, TUNEL assay also showed the number of apoptotic cells was significantly increased upon PCDH10 expression (Fig. 2H). Finally, an increased level of active caspase-3 and caspase-9 proteins was detected in PCDH10-expressing cells (Supplementary Fig. S2B). Collectively, these results lead us to conclude that PCDH10 is a proapoptotic factor in EEC cells.

To gain insights into the underlying mechanism of PCDH10 function in EEC cells, we performed a genome-wide analysis to globally characterize PCDH10-affected transcriptomic changes. PolyA⁺ RNAs were extracted from PCDH10-overexpressing HEC-1-B or the vector control cells and subjected to RNA-sequencing analysis. The majority of sequenced reads can be mapped to coding regions (CDS and UTRs, >70%) and much fewer in introns, intergenic, and noncoding regions (Supplementary Fig. S3A), suggesting a great specificity for expressed mRNA and rejection of genomic DNA and unspliced pre-mRNA. As a result, we found a total of 1,233 and 679 genes were up- and downregulated in PCDH10 overexpressing HEC-1-B cells with respect to control cells (Supplementary Fig. S3B and Supplementary Dataset). To validate the RNA sequencing findings, 12 genes (6 upregulated and 6 downregulated genes) were randomly selected and subjected to qRT-PCR examination. The results agreed favorably with the RNA-seq data (Supplementary Fig. S3C).

Subsequent gene ontology (GO) analysis of the above downregulated list of genes revealed an enrichment for GO items including “Cell cycle phase,” “Cell cycle process,” “Mitotic cell cycle,” “M phase,” and “Nuclear division” (Fig. 3A and Supplementary Dataset). Such alterations are in keeping with the cell-cycle arrest and cell-proliferation inhibition observed upon PCDH10 expression (Fig. 2). The upregulated genes are enriched for GO categories related to “Glycoprotein,” “Cell membrane,” “Cell adhesion,” etc. (Fig. 3B). Although apoptosis related GO terms were not enriched, many proapoptotic factors, including BAD, PLEKHF1, and FIS1, were upregulated (Supplementary Fig. S3D and Supplementary Dataset), supporting the observed impact of PCDH10 on cell apoptosis.

Among all the potential targets influenced by PCDH10, MALAT1 caught our attention because of its known oncogenic role in other cancers (26). The RNA-seq data revealed that MALAT1 RNA level was decreased by ∼70% in PCDH10-expressing cells (Fig. 3C and D and Supplementary Dataset). To independently validate the result, qRT-PCR assay also showed that MALAT1 was indeed significantly downregulated upon PCDH10 overexpression in both AN3CA and HEC-1-B cells (Fig. 3E).

The impact of PCDH10 on MALAT1 in expression level prompted us to ask whether MALAT1 is functionally downstream of PCDH10. To test this notion, we examined the effect of MALAT1 knockdown in AN3CA and HEC-1-B cells by using 2 siRNAs oligos targeting it (siMALAT1A and siMALAT1B) and a nontargeting sequence as a control (siNC). Successful decrease of MALAT1 (Supplementary Fig. S4A and S4B) led to a significant delay in cell proliferation compared with the siNC-treated cells as revealed by cell-counting assay (Fig. 3F) and MTS assay (Fig. 3G). Conversely, when we overexpressed MALAT1 through the transient transfection of a plasmid (Fig. 3H), cell proliferation was enhanced compared with the cells transfected with an empty vector plasmid (Fig. 3I). Furthermore, TUNEL assay showed that siMALAT1 treatment induced a striking increase in apoptotic cell population by more than 5 folds compared with siNC treatment (Fig. 3J). Altogether, the above findings suggested that MALAT1 knockdown phenocopied PCDH10 overexpression effect in EEC cells and implicated MALAT1 as a downstream effector of PCDH10. To strengthen this finding, we found that overexpression of MALAT1 reversed the inhibitory effect of PCDH10 on EEC cell growth (Fig. 3K).

Next, we further explored the molecular mechanism underlying PCDH10 suppression of MALAT1. Little is known about how PCDH10 downregulation promotes tumorigenesis, but PCDHγ, another member of PCDH superfamily, was previously found to negatively regulate Wnt/β-catenin signaling (41, 42). The Wnt/β-catenin signaling pathway plays pivotal roles in programming developmental gene expression; constitutive activation of this pathway is involved in diverse cancer types (43). The abnormal activation was also found to be a common event in EEC, underscoring its clinical significance (44, 45). We therefore speculated that loss of PCDH10 has contributed to the activation of the Wnt signaling pathway, which subsequently induces MALAT1 expression. To test this notion, we first examined the effect of PCDH10 restoration on Wnt signaling. Compared with the vector control, enhanced expression of PCDH10 caused a sharp decrease of several known Wnt targets, including LEF1, TCF1, and c-MYC in both AN3CA and HEC-1-B cell (Fig. 4A). Consistently, PCDH10 overexpression repressed the activity of a Wnt signaling reporter, Topflash luciferase reporter, by ∼40% when compared with the control (Fig. 4B). Collectively, our data implicate that PCDH10 negatively modulates canonical Wnt/β-catenin signaling.

Having established the connection between PCDH10 and Wnt/β-catenin signaling, we further asked whether PCDH10 influences MALAT1 expression through this pathway. Given that Wnt/β-catenin signaling transcriptionally activates the expression of a myriad of genes, we evaluated its effects on MALAT1 transcription. Activation of Wnt signaling by treatment of the EEC cells with LiCl led to a mild but significant increase of MALAT1 expression as well as other known targets, c-MYC, TCF1, or LEF1 (Fig. 4C), suggesting Wnt activation indeed induces MALAT1 transcription.

To further ask whether the regulation is directly through TCF/β-catenin binding to MALAT1 promoter, we searched for TCF binding sites by analyzing publically available TCF4 ChIP-sequencing (seq) data generated from various cell lines (HeLa-S3, MCF-7, HCT-116, HEK293, and PANC-1; ref. 46) by ENCODE. In all the 5 cell lines, two TCF4 binding peaks proximal to the TSS of MALAT1 were detected (Fig. 4D and Supplementary Fig. S4C), suggesting a possibility for the direct binding of TCF4 on MALAT1 promoter. Indeed, a consensus binding motif of TCF4 was found in the promoter region (+78 to +88 bp) using rVISTA 2.0 tools (Fig. 4D and Supplementary Fig. S4C; ref. 47). To experimentally test TCF association with this site, we cloned the genomic region, harboring it into a luciferase reporter (wild type, WT) and transfected it into EEC cell to test its response to Wnt activation. Results showed that the reporter activity was induced upon Wnt activation by LiCl treatment in a dose-dependent manner; however, the response was lost when we mutated the TCF binding site by altering the consensus sequence (Mut; Fig. 4E). These results suggested that Wnt signaling probably activates MALAT1 transcription through the identified TCF4 binding site. Furthermore, when examining the effect of PCDH10 on the reporter activity, we found PCDH10 overexpression suppressed the WT but not the Mut reporter activities (Fig. 4F), suggesting PCDH10 regulates MALAT1 transcription through inhibiting Wnt signaling. To strengthen the above findings, we performed ChIP to detect β-catenin enrichment on MALAT1 promoter. Expectedly, as shown in Fig. 4G, a specific and robust enrichment of β-catenin (∼8-fold) was found on the identified TCF4 site, which is comparable to its enrichment on a known target, c-MYC; and the enrichment was significantly diminished (∼4-fold) upon PCDH10 overexpression.

Together, these data demonstrate that MALAT1 genomic locus is under the direct transcriptional regulation of Wnt/β-catenin signaling and PCDH10 suppresses MALAT1 expression through impairing β-catenin binding to its promoter. To further answer the question of how PCDH10 affects β-catenin binding, we tested whether it has any impact on β-catenin expression or its nuclear translocation. The results on the other hand indicated that overexpression of PCDH10 had no impact on the total β-catenin expression and however caused slight decrease of nuclear β-catenin level (Supplementary Fig. S5).

Finally, we evaluated the function of the PCDH10–MALAT1 regulatory axis in vivo first using a xenograft tumor model established by injecting HEC-1-B cells stably expressing PCDH10 or empty vector into nude mice. Measuring the tumor size every 3 days, we found PCDH10 overexpression markedly delayed tumor growth from the very beginning when compared with vector control (n = 5; P < 0.01; Fig. 5A). Tumor size and mass were evidently smaller at the end of evaluation (Fig. 5B and C) when PCDH10 overexpression was also confirmed by qRT-PCR (Fig. 5D). TUNEL assay on the paraffin sections revealed that PCDH10 overexpression caused a severe cell apoptosis in vivo (Fig. 5E), which was in agreement with its proapoptotic effect found in vitro. To illustrate the effect of MALAT1, siRNA oligos were injected into the HEC-1-B xenograft tumor to knockdown MALAT1 in vivo. A significant delay in tumor growth was observed during a 10- day measurement course (n = 5; P < 0.01; Fig. 5F). Tumor size and mass were also decreased in the end stage tumors (Fig. 5G–I). These results demonstrated MALAT1 decrease led to a comparable phenotype as PCDH10 overexpression. Consistently, we also detected a reduced level of both MALAT1 and c-MYC expression in PCDH10 overexpressing tumors (Fig. 5J), suggesting the existence of the PCDH10-Wnt/β-catenin-MALAT1 axis in vivo in the xenograft tumors.

Next, we validated the above findings in EEC clinical samples. First, we examined the MALAT1 expression in 32 EEC samples by In situ hybridization (ISH) on the paraffin sections using a MALAT1-specific probe (29). As expected, MALAT1 signal was mainly detected in the nuclei of endothelial cells and a much higher level was found in EEC samples as compared with the 9 normal tissues (Fig. 5K and Supplementary Table S2). In addition, a strong MALAT1 signaling seemed to be associated with low histologic grade (grade 1/2 vs. 3; P = 0.028; Fig. 5L). However, we did not observe an association between MALAT1 level and FIGO stage (Supplementary Fig. S6A). Like PCDH10, MALAT1 expression seems to be regulated during the menstrual cycle; it is higher in proliferative compared with secretory and postmenopausal samples (Supplementary Fig. S6B). To further evaluate potential association of MALAT1 with other clinical parameters, we explored tissue microarray data performed on an independent validation set of 142 EEC and 18 hyperplasia samples (Supplementary Table S3 and ref. 45). The results revealed that high expression of MALAT1 was linked with hyperplasia (P = 0.017), menopausal status (P = 0.028), no recurrence (P = 0.032), and low metastasis potential (P = 0.041; Table 1). PCDH10 on the other hand only exhibited association with hyperplasia (P = 0.013; Supplementary Table S4). Further exploring the TAGA gene expression data from a large cohort of EEC samples (n = 253), we also detected a reverse association between PCDH10 and MALAT1 expression levels with a statistical significance (P = 0.0324; Fig. 6M). This anticorrelation was however not found from analyzing the tissue microarray data probably because of its smaller sample size. In addition, to further validate the regulation between Wnt/β-catenin and MALAT1, we stained β-catenin by immunohistochemistry (IHC) on a subset (n = 37) of EEC tumors. The expression level of MALAT1 by ISH staining is found strongly correlated with the total or nuclear level of β-catenin by IHC staining (Fig. 6N and O). However, no correlation was detected between PCDH10 expression and β-catenin signal (total or nuclear level; data not shown). Altogether, these analyses confirm the presence of PCDH10–Wnt/β-catenin–MALAT1 regulation in clinical samples.

In this study, we identified PCDH10 as a tumor suppressor downregulated in EEC. Although the downregulation of PCDH10 has been reported in a wide range of cancers (6, 9–16), as far as we know, this is the first study to demonstrate its association with EEC. Our results clearly showed that PCDH10 promoter is hypermethylated through both genomic bisulfite sequencing analysis of locally collected EC samples and analyzing TCGA data from worldwide EEC samples. Thus, loss of PCDH10 expression through its promoter hypermethylation seems to be a common event occurring in many tumors. Exploiting the TCGA datasets, ours is the first to report an anticorrelation between PCDH10 promoter CpG methylation levels with its mRNA expression in ∼40% of EEC tumors examined, thus providing solid evidence to support that the hypermethylation causes PCDH10 silencing in these tumors. This type of methylation-expression correlation analysis has not been reported in any other PCDH10 studies. Notably, in addition to the previously known promoter CpG island, we also identified a second CpG island downstream of TSS that is hypermethylated in EECs. It will be interesting to explore whether it also contributes to PCDH10 silencing in the future. Consistently, recent integrative genomic analysis of endometrial cancers did not reveal copy number alterations or mutations on PCDH10 gene, reinforcing the significance of epigenetic downregulation (3). Interestingly, in addition to promoter hypermethylation, recent studies uncovered alternative epigenetic events partaking in its downregulation in breast cancer and pancreatic cancer (18, 19). In these reports, PRC2 complex were found to be recruited, mediated by HOTAIR, to bind the promoter region of PCDH10, resulting in an increase in repressive chromatin mark, H3K27me3. Such cooperative action of DNA methylation and histone modifications in orchestrating target gene expression has been extensively revealed in developmental and pathologic processes, including cancer (48). For example, in our recent work, we demonstrated APC promoter, in addition to under regulation of DNA methylation, is subjected to YY1-mediated EZH2 recruitment and subsequent H3K27me3 modification, leading to its inactivation (5). Indeed, we also noticed a substantial subset of EEC samples (n = 58) with low PCDH10 expression do not bear promoter hypermethylation when analyzing TCGA data (Fig. 1C and D), suggesting additional mechanisms may exist to repress PCDH10 expression in these samples.

Further functional studies revealed PCDH10 could be a potential modulator of EEC carcinogenesis both in vitro and in vivo. Restoration of PCDH10 successfully reduced cell growth and induced cell apoptosis, which is in keeping with what was uncovered in gastric, lung, and esophageal cancers. Notably, we could not recapitulate all functional features of PCDH10 demonstrated in other studies (10, 13, 14, 16). For example, in gastric cancer cells, PCDH10 was shown to exert significant impacts on cell migration (10), however, this was not observed in the two EEC cell lines investigated in our study. The impact of PCDH10 on cell cycle also differs between two EEC cell lines (Fig. 2F). These discrepancies may reflect cell-type– and cancer-type–specific roles of PCDH10.

Despite the considerable advancements in our understanding of the tumor suppressive functions of PCDH10 in distinct cancer types, nothing is known about the underlying molecular basis. Our findings, for the first time, uncovered the transcriptomic influence exerted by PCDH10 through RNA-seq. Moreover, we identified MALAT1 as a functional downstream target of PCDH10. MALAT1, as one of the most characterized lncRNAs, has been broadly associated with cancer development and engaged in diverse facets of cancer progression. Our report, nonetheless, represents the first demonstrating its functionality in EEC. In keeping with previous reports, MALAT1 plays an oncogenic role in EEC cells and tumors. In addition, we demonstrated that MALAT1 is positively associated with hyperplasia, and negatively with metastasis, suggesting its predictive value as a molecular biomarker. Of note, the reverse association between MALAT1 and the metastasis status in EEC is distinct to the findings from other cancer types. In lung cancer, high expression of MALAT1 is linked with metastasis and MALAT1 regarded as a critical regulator of metastasis phenotype (28, 36). In bladder cancer, high MALAT1 expression level is also connected with high-grade and high-stage carcinoma (49). These findings suggest that MALAT1 function is cancer-type dependent. Linkage between high MALAT1 expression with hyperplasia, to some extent, reflects its roles in cell proliferation. Endometrial hyperplasia is the result of excessive proliferation of the endometrium cells and is considered as a remarkable risk factor for the development of EEC or even coexists with EEC (45). The positive correlation of MALAT1 with hyperplasia and low-grade EEC implies that dysregulation of MALAT1 is an early event in EEC development. The lower MALAT1 expressions in metastasis and tumors leading to recurrences or progressions point to other carcinogenic mechanisms as prevailing in the late stages of development as well as the most aggressive endometrial cancers.

In this study, we also provide novel insights into how PCDH10 acts on MALAT1 through modulating Wnt/β-catenin signaling. We identified MALAT1 as a direct transcriptional target of Wnt/β-catenin in EEC cells. A TCF4 binding site was identified on its promoter region, which was proved to mediate Wnt effect on MALAT1. According to ChIP-seq data from ENCODE, the TCF binding on this site is present in multiple cancer cell lines, including HeLa-S3, MCF7, HCT116, and PANC-1; thus, Wnt regulation of MALAT1 likely occurs as a general pathway in various cancers. In addition to TCF binding, in SK-N-SH cell, cyclic AMP-responsive element binding (CREB) transcription factor was reported to be associated with MALAT1 promoter region (50). Recent study revealed MALAT1 is also under the regulation of one miRNA, miR-9 (51). Considering its length and high abundance in many cell entities, we speculate that MALAT1 is under tight control of many factors. Indeed, ChIP-seq and chromatin state segmentation data from ENCODE project revealed binding peaks of many transcription factors and histone marks upstream its promoter (29, 46).

Aberrant activation of the Wnt/β-catenin signaling pathway occurs commonly in EEC and presumably occurs early, leading to EEC initiation (44). It is often caused by mutations in β-catenin or Wnt antagonists (e.g., APC), resulting in nuclear accumulation of β-catenin and augmented transcription of downstream targets (4). However, enhanced Wnt/β-catenin signaling is also observed in EEC tumors that do not carry such mutations, suggesting an alternative mechanism underlying Wnt activation. Recently, protocadherin family members were implicated in crosstalks with the canonical Wnt/β-catenin signaling pathway. In Wilms' tumor (nephroblastoma), PCDHγ members are able to repress β-catenin/TCF-mediated transcription and suppress β-catenin/TCF reporter activity (42), which is also supported by evidence from colon cancer (41). In line with these findings, our current study also uncovered a link between PCDH10 and Wnt signaling, suggesting an alternative means through PCDH10 silencing to activate Wnt/β-catenin signaling. Exactly how PCDH10 suppresses Wnt/β-catenin signaling activity is still unclear. From our results (Supplementary Fig. S5), PCDH10 re-expression did not affect the total cellular β-catenin level and caused slight decrease of nuclear β-catenin level, suggesting PCDH10 effect on β-catenin transcriptional activity may not be directly through regulating its expression or nuclear translocation. This was also supported by the findings from EEC samples where no significant correlation between β-catenin (total or nuclear) and PCDH10 levels was observed. Thus, PCDH10 effect on Wnt/β-catenin signaling may be mediated through other factors in this pathway such as TCF or LEF. These questions warrant future investigation on this aspect.

Collectively, we have identified a novel regulatory axis, PCDH10-Wnt/β-catenin-MALAT1 in the effort of elucidating the tumor suppressive function of PCDH10 (Fig. 6), rendering this the first report to show PCDH10 and lncRNA function in EEC and to elucidate the transcriptional regulation of MALAT1. It will be of great interest to explore the clinical significance of PCDH10 and MALAT1 in the future.

No potential conflicts of interest were disclosed.

Conception and design: Y. Zhao, H. Wang

Development of methodology: Y. Zhao, Y. Yang, J. Trovik, L. Zhou, H. Wang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Zhao, Y. Yang, J. Trovik, T.-S. Lau, E.A. Hoivik, H.B. Salvesen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Zhao, Y. Yang, J. Trovik, K. Sun, P. Jiang, T.-S. Lau, E.A. Hoivik, H.B. Salvesen, H. Sun, H. Wang

Writing, review, and/or revision of the manuscript: Y. Zhao, J. Trovik, E.A. Hoivik, H.B. Salvesen, H. Sun, H. Wang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Yang, K. Sun, H.B. Salvesen

Study supervision: H. Sun, H. Wang

This work was supported by an HMRF grant to H. Wang from the Food and Health Bureau of the Hong Kong Special Administrative Region, China (Project Ref. 01120446), a 973 grant from the Ministry of Science and Technology of China (2014CB964700 to H. Wang), and two GRF grants to H. Wang (476310 and 476113) from Research Grants Council (RGC) of the Hong Kong and Private Fund form Dept O&G, CUHK.

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.