Downregulation of MicroRNA-644a Promotes Esophageal Squamous Cell Carcinoma Aggressiveness and Stem Cell–like Phenotype via Dysregulation of PITX2

Esophageal squamous cell carcinoma (ESCC) is among the most lethal malignant head and neck tumors. It presents with a spectrum of aberrantly aggressive phenotypes (1) and despite recent advances in multimodality therapies, the prognosis remains poor (2, 3). Like other solid tumors, the pathogenesis of ESCC is a long process involving activation of oncogenes and/or inactivation of tumor suppressor genes. Recently, tremendous efforts have focused on identifying specific molecular markers associated with the progression of ESCC while simultaneous work has sought to better understand the etiology of the disease (3, 4). To date, however, promising molecular genetic alterations with clinical or prognostic significance in ESCC have remained elusive.

The Wnt/β-catenin signaling pathway is essential for embryonic development and aberrant activation has been associated with the progression of many cancer types, including ESCC (5, 6). Paired-like homeodomain transcription factor 2 (PITX2) is a member of the bicoid-related homeodomain family and a downstream effecter of Wnt signaling, which activates the expression of target genes required for cell proliferation and survival (7). In addition to its regulation by Wnt, PITX2 itself could directly activate Wnt ligand genes, activating the respective canonical Wnt/β-catenin signaling pathway, thus contributing to cancer progression (8). Overexpression of PITX2 has been frequently studied in human cancers, including thyroid, colorectal, and ovarian cancers (9–11). We recently found that PITX2 was frequently upregulated in ESCC tissues, as compared with the normal epithelial tissue. Furthermore, overexpression of PITX2 was correlated with poor prognosis and/or radiochemoresistance of the disease (12). Clearly, a better understanding of the tumor-specific regulation of PITX2 and the targets it regulates is crucial and has potentially important clinical applications. We therefore aimed to identify novel molecules capable of regulating PITX2 at the transcriptional and post-transcriptional levels in ESCC.

Recently, small noncoding RNA molecules, miRNAs, have emerged as key regulators of gene expression at the post-translational level (13), and deregulation of miRNAs has been implicated in the development and progression of nearly all tumor types. Multiple miRNAs have demonstrated critical roles during the development and/or progression of ESCC by regulating various critical genes (14, 15). Given the important oncogenic role of PITX2 in ESCC (12), we sought to determine whether PITX2 expression is regulated by specific miRNAs, with the hypothesis that the regulatory miRNAs could also be important in ESCC pathogenesis.

Herein, we provide evidence that miR-644a is a novel negative regulator of PITX2 expression in ESCC, and that expression of miR-644a was significantly lower in ESCC that recurred or otherwise had a poor prognosis. The levels of miR-644a regulate important traits in ESCC cells, including tumorigenesis, invasion, and the stem cell–like phenotype both in vitro and in vivo. Importantly, we reveal that depleting miR-644a levels in ESCC cells potently activates the Akt/GSK-3β/β-catenin signaling pathway through directly upregulating PITX2, ultimately increasing the malignancy of the tumor. The tumor-promoting downregulation of miR-644a in ESCC is caused by hypermethylation of its promoter.

Seven human ESCC cell lines (KYSE-30, TE-1, KYSE-510, KYSE-180, KYSE-140, KYSE-410, and KYSE-520) were obtained from DSMZ, the German Resource Center for Biological Material. The ESCC line Eca109 and the normal epithelial cell line NE-1 were kindly provided by S.W. Tsao and G. Srivastava (University of Hong Kong, Hong Kong, China). All cell lines were cultured in RPMI1640 supplemented with 10% FBS. A total of 190 formalin-fixed and paraffin-embedded (FFPE) ESCC tissue samples were obtained from the Department of Pathology of Sun Yat-Sen University Cancer Center (Guangzhou, China) between January 2007 and December 2008. The clinicopathologic characteristics of these patients are summarized in Supplementary Table S1. An additional panel of 20 fresh ESCC tissues and matched adjacent nontumor esophageal tissues were collected from Sun Yat-Sen University Cancer Center between January 2013 and June 2013, and stored in liquid nitrogen until further use. All ESCC specimens included in this study were obtained from patients with stage I–III ESCC disease during surgical resection. The samples contained matched tumors (percentage of tumor cells ≥ 70%) and corresponding normal mucosal tissue (>5 cm laterally from the edge of the cancerous region). None of the recruited patients received any preoperative treatment. Patients were monitored once every 3 months for the first 2 years after surgical resection, once every 6 months during the third and fourth years, and once a year after the fifth year postoperatively. Diagnostic examinations included esophagography, computed tomography, chest X-ray, abdominal ultrasonography, and bone scanning as required, to detect tumor recurrence and/or metastasis. Written informed consent was obtained from all patients before the study. This study was approved by Ethical Committee of Sun Yat-sen University Cancer Center (Guangzhou, China).

Total RNA was extracted from ESCC tissues and cell lines with use of TRIzol reagent (Invitrogen). Real-time PCR (RT-PCR) was carried out using SYBR Green SuperMix (Roche) and ABI7900HT Fast Real-Time PCR system (Applied Biosystems). Either α-tubulin or U6 was used as an internal control. TaqMan probes were used to detect miR-644a, U6, PITX2, ITCH, and α-tubulin (GeneCopoeia). The remaining qPCR primers are listed in Supplementary Table S2.

The miR-644a expression vector (HmiR0349-MR03), control vector for miR-644a (CmiR0001-MR03), PITX2 coding sequence expression vector (EX-P0098-Lv105), and control vector for PITX2 (EX-NEG-Lv105) were purchased from GeneCopoeia. The target sequence of PITX2 used to construct a lentiviral shRNA was 5′-GCCGACTCCTCCGTATGTTTA-3′. Cells transfected with empty vector were used as controls. miR-644a mimic and antagomir-miR-644a were synthesized by Genepharma. Oligonucleotide transfection was performed using Lipofectamine 2000.

Vectors were packaged in 293FT cells using ViraPower Mix (GeneCopoeia). After culturing for 48 hours, lentiviral particles in the supernatant were harvested and filtered by centrifugation at 500 × g for 10 minutes, and then transfected into ESCC cells. The cells were cultured under puromycin (2 μg/mL) selection for 2 weeks, at which point real-time PCR was used to determine the level of miR-644a. Cell lines stably expressing miR-644a or negative control (NC) vector were designated as KYSE-1400/Eca109-Lv-miR-644a and KYSE-1400/Eca109-Lv-miR-NC cells, respectively.

Cell viability was measured by a 3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyl tetrazolium bromide (MTT) assay (Sigma). Briefly, cells were seeded in 96-well plates and cultured. Cell viability was examined by following standard procedures. Experiments were performed in triplicate.

To assay xenograft tumor growth, miR-644a–expressing cells (2 × 10⁶) or empty vector–transfected control cells were subcutaneously injected into the inguinal folds of 4-week-old BALB/c nude mice. For the tumor formation assay, three dosages (2 × 10⁵, 2 × 10⁴, or 2 × 10³) of the indicated cells in Matrigel (final concentration 25%) were inoculated subcutaneously into the inguinal folds of BALB/c nude mice. The mice were monitored daily for palpable tumor formation and tumors were measured using Vernier calipers, weighed, and photographed. All animal experiments were conducted according to the standard institutional guidelines of Sun Yat-Sen University Cancer Center (Guangzhou, China).

Ten 4-week-old BALB/c nude mice in each experimental group were injected with KYSE-140/Eca109-miR-644a or KYSE-140/Eca109-miR-control cells, respectively. Briefly, 2 × 10⁵ cells were injected intravenously through the tail vein into each mouse in a laminar flow cabinet. Six weeks after injection, the mice were sacrificed and examined.

In brief, paraffin-embedded sections were deparaffinized and incubated in retrieval buffer solution for antigen retrieval. Protein expression was visualized using a Dako Real Envision Kit (K5007, Dako) after staining with the primary antibody. Staining intensity was scored manually by two independent experienced pathologists as: 0 = no staining, 1 = weak staining, 2 = moderate staining, and 3 = strong staining. Tumor cells in five fields were selected randomly and scored on the basis of the percentage of positively stained cells (0%–100%). The final IHC score was calculated by multiplying the intensity score by the percentage of positive cells.

For Western blots, total cellular protein was extracted from tissues or cells and separated by SDS-PAGE. Nuclear extracts were prepared using the Nuclear Extraction Kit (Active Motif), according to the manufacturer's instructions. Immunoprecipitation (IPs) were carried out using protein G agarose (Millipore). For ChIP analysis, ESCC cells (4 × 10⁷) were prepared by using the chromatin immunoprecipitation (ChIP) Assay Kit (Cell Signaling Technology) according to the manufacturer's instructions. The following primary antibodies were used: anti-PITX2 (Abcam); anti-α-tubulin, anti-GSK-3β, anti-phospho-GSK-3β (p-GSK-3β), anti-Akt, anti-phospho-Akt (p-Akt), anti-β-catenin, anti-Lef-1, anti-Axin-2, anti-Survivin, anti-acetylated-lysine (Cell Signaling Technology); anti-active-β-catenin (Millipore); anti-c-myc, anti-cyclin D1, anti-Histone H1 (Santa Cruz Biotechnology).

ESCC cells were treated with 5-aza-2′-deoxycytidine (5-aza-dC; 50 mmol/L and 100 mmol/L; Sigma-Aldrich) for 72 hours with a change of culture medium every 24 hours.

Genomic DNA was extracted and subjected to bisulfite treatment using the Epitect Bisulfite kit (Qiagen) according to the manufacturer's instructions. Bisulfite-treated DNA was then analyzed by bisulfite genomic sequencing (BGS) based on the following primers: forward, 5′-TGGAGTTGAGGAAAATTGG-3′; and reverse, 5′- ATTCTCATCCGAACTCCC-3′ (5 clones were picked for each sample).

Statistical analyses were performed using SPSS software (SPSS Standard version 16.0, SPSS Inc). Bivariate correlations between study variables were calculated by Pearson correlation coefficients. Differences between variables were analyzed by χ² or Fisher exact tests. Survival curves were plotted using the Kaplan–Meier method and compared with log-rank tests. Multivariate survival analysis was performed for all parameters found to be significant in univariate analysis using a Cox regression model. Comparisons between groups for statistical significance were performed with a two-tailed Student t test. Data are presented as the mean ± SD. P values <0.05 were considered significant.

Detailed methods on colony formation assay, wound-healing and invasion assays, luciferase reporter assay, immunofluorescent (IF) staining, sphere-forming assay, and flow cytometric analysis are reported in the Supplementary Data.

To investigate the potential miRNA regulators of PITX2, we first performed bioinformatic analysis by using the miRANDA predicting program to compare the predicted regulators with miRNAs downregulated in ESCC tumor tissues in published miRNA profiles (NCBI/GEO/GSE23142; ref. 16). The two screening algorithms overlapped on only one miRNA, namely, miR-644a (Fig. 1A). We, therefore, focused our subsequent efforts on miR-644a as a putative regulator of PITX2 in ESCC. We examined the levels of miR-644a and PITX2 expression in 20 pairs of fresh ESCC and nontumor esophageal tissues, as well as in 8 ESCC cell lines and a normal epithelial cell line, NE-1. The results showed that levels of miR-644a were significantly downregulated in ESCC tissues and cells as compared with the levels in nontumor tissues and NE-1 cells (Fig. 1B, i and C, i). Overall, there was a significant, inverse correlation between the levels of miR-644a and PITX2 mRNA either in ESCC tissues (r = −0.822, P < 0.001; Fig. 1B, ii and iii) or in ESCC cell lines (r = −0.960, P < 0.001; Fig. 1C, ii and iii).

We next examined the levels of miR-644a and PITX2 protein expression to determine their potential clinical and prognostic significance in a large cohort of 190 ESCC FFPE tissues. We found a significant inverse correlation between miR-644a levels and PITX2 protein expression (r = −0.669, P < 0.001; Fig. 1D, i). Furthermore, the level of miR-644a was significantly lower in ESCC patients with tumor recurrence and/or distant metastasis after surgery, compared with patients without tumor recurrence and/or distant metastasis (P < 0.001; Fig. 1D ii and iii). Expression level of miR-644a in ESCCs was negatively associated with tumor size (P = 0.031), T status (P = 0.004), N status (P < 0.001), and clinical stage (P < 0.001; Supplementary Table S2). Moreover, ESCC patients with high levels of miR-644a exhibited better overall survival (OS) and lower tumor recurrence rates than those with low miR-644a (Fig. 1E, i and ii). Higher PITX2 expression in ESCCs was positively associated with tumor size (P = 0.031), N status (P = 0.006), and clinical stage (P = 0.009, Supplementary Table S2) and the patients showed poorer OS and higher tumor recurrence rates than those with low PITX2 (Fig. 1E, iii and iv).

In univariate analyses, miR-644a and PITX2 levels and TNM stage were significantly associated with OS and time to recurrence (TTR) in patients with ESCC (Supplementary Table S3). Further multivariate analysis revealed that miR-644a, PITX2, and TNM stage were independent prognostic indicators for both OS and TTR (Supplementary Table S3). We next divided the patients into four groups based on their levels of miR-644a and PITX2 expression. ESCC patients with high miR-644a and low PITX2 levels had the best OS and lowest frequency of recurrence. In contrast, the ESCC patients with low miR-644a and high PITX2 levels had the lowest OS and highest frequency of recurrence (P < 0.001; Fig. 1E, v and vi; Supplementary Table S3). Together, these data revealed that miR-644a and PITX2 are independent prognostic indicators in patients with ESCC and the combination of both factors has the strongest prognostic value.

Subsequently, we investigated the role of miR-644a in ESCC tumorigenesis and progression in vitro. Overexpression of miR-644a in KYSE-140 and Eca109 ESCC cells substantially reduced cell growth and colony formation and largely inhibited both cell migration and invasive capacities (Fig. 2A–D). Consistent with a role for miR-644a in tumorigenesis, knockdown of miR-644a with anti-miR-644a in the KYSE-410 ESCC cell line significantly increased cell growth, colony formation, migration, and invasion (Supplementary Fig. S1A–S1D).

We further studied the in vivo impact of miR-644a on ESCC cell growth and metastasis by injecting KYSE-140 or Eca109 cells containing either a control or miR-644a-overexpression vector into BALB/c nude mice, either subcutaneously or via the tail vein. We observed that overexpression of miR-644a significantly inhibited tumor growth in vivo (Fig. 2E, i–iii). Hematoxylin and eosin (H&E) staining revealed that tumors formed from control cells exhibited an extensive branch-like growth pattern that spread into the surrounding tissue, while miR-644a–overexpressing ESCC cells formed oval-shaped tumors with smooth margins and a noninvasive front. This suggests that the invasive behavior of ESCC cells could be suppressed by miR-644a in vivo (Fig. 2E, iv). In the mouse metastasis model, we did not detect tumor nodule formed in the liver of any mice examined. However, the number and size of metastatic colonies in the lungs were dramatically decreased in mice injected with miR-644a–overexpressing cells, compared with the control group (Fig. 2F).

On the basis of our in silico prediction, we cloned a fragment of the 3′UTR of PITX2 mRNA containing a site complementary to the seed region of miR-644a (Supplementary Fig. S2A). Further studies showed that overexpression of miR-644a in ESCC cells significantly suppressed the luciferase activity of PITX2 containing wild-type 3′UTR, but not that of PITX2-containing mutant 3′UTR (Supplementary Fig. S2B). Depletion of miR-644a increased the luciferase activity of PITX2 (Supplementary Fig. S3A). Concordantly, overexpression of miR-644a substantially decreased mRNA and protein levels of PITX2 in ESCC cells (Supplementary Fig. S2C and S2D), while downregulation of miR-644a dramatically increased their levels (Supplementary Fig. S3B and S3C). These results were verified by immunohistochemical analysis of PITX2 in xenograft tumors formed from miR-644a-overexpressing and control cells (Supplementary Fig. S2E).

To confirm that PITX2 is a functional target of miR-644a, stable miR-644a-ESCC (KYSE-140 and EC109) cells were transfected with pEZ-Lv105-PITX2 encoding the entire PITX2 coding sequence but lacking the 3′-UTR. As expected, overexpression of PITX2 in miR-644a-ESCC cells abrogated miR-644a-suppressed ESCC cell growth, colony formation, migration, and invasion abilities. Correspondingly, knockdown of PITX2 expression significantly inhibited ESCC cell growth, colony formation, migration, and invasion ability (Fig. 3A–D), which was consistent with the effect of miR-644a overexpression in ESCC cells. These data revealed that the effect of miR-644a in ESCC cells is affected by PITX2 levels.

As previous study suggested that PITX2 crosstalks with the Wnt/β-catenin pathway (7, 8), we sought to use miR-644a to probe the role of PITX2 in Wnt/β-catenin signaling. We demonstrated that miR-644a overexpression in ESCC cells reduced the pool of activated β-catenin (i.e., dephosphorylated at Ser37 and Ser41), nuclear β-catenin, and the typical downstream target genes of β-catenin (Fig. 3E). Furthermore, IF staining showed a clear redistribution of β-catenin from the nucleus to the cytoplasm after miR-644a overexpression (Fig. 3F). There was also a marked decrease in binding of β-catenin to the TCF sites of both the Lef-1 and c-myc promoters, examined by ChIP, in miR-644a–overexpressing ESCC cells (Fig. 3G). In addition, the TOP/FOP flash luciferase assay revealed that reporter expression was suppressed by miR-644a overexpression (Fig. 3H). Furthermore, we found that restoration of PITX2 levels in miR-644a-overexpressing ESCC cells largely blocked the inhibitory effect of miR-644a on the Wnt/β-catenin pathway (Fig. 3E–H). These findings indicated that miR-644a–mediated suppression of PITX2 expression inhibits the activation of Wnt/β-catenin signaling.

It has been shown that stem cell–like properties in cancer cells contribute to tumor recurrence and/or metastasis, and that activation of the Wnt/β-catenin pathway is thought to be a major driver of the development of stem cell–like properties (17, 18). Thus, we sought to determine whether the levels of miR-644a affect ESCC cell stem-like characteristics. Our results showed that overexpression of miR-644a in ESCC cells resulted in downregulation of many stemness-associated genes we examined by qRT-PCR (Fig. 4A and Supplementary Fig. S4A). Further functional assays demonstrated that miR-644a–overexpressing ESCC cells formed smaller and fewer spheres, with a reduced proportion of side population (SP) cells, and the cells were more sensitive to cisplatin and radiation, as compared with control cells (Fig. 4B–D and Supplementary Fig. S4B–S4D). However, when the levels of PITX2 were restored in miR-644a–overexpressing ESCC cells, the inhibitory effect of miR-644a on stem-like traits was partially blocked, whereas knockdown of PITX2 had a similar effect to miR-644a on ESCC stem-like traits (Fig. 4A–D and Supplementary Fig. S4A–S4D). These data revealed that miR-644a inhibits ESCC cell stem-like traits through regulation of the PITX2 gene.

In this study, we further assessed the in vivo effect of miR-644a on the tumorigenicity of ESCC cells in nude mice. Our results clearly showed that fewer and smaller tumors were formed after the injection of 2 × 10⁵, 2 × 10⁴, or 2 × 10³ miR-644a–overexpressing ESCC cells compared with the tumors from injection of control cells. Moreover, the first palpable tumor in the group injected with miR-644a–overexpressing cells occurred later than in the control group (Fig. 4E and Supplementary Fig. S4E).

Furthermore, we examined the levels of miR-644a and six markers associated with cell stemness (ABCG2, SOX2, Oct4, Nanog, Bmi1, and CD44) by RT-PCR in the 20 fresh ESCC patient samples. The levels of miR-644a expression were negatively correlated with the expressions of the 6 stemness-associated markers (P < 0.05, Supplementary Fig. S5).

Previous studies demonstrated that Akt activation inhibits GSK-3β activity, leading to activation of Wnt/β-catenin signaling and the development of stem-like traits and tumor aggressiveness (19, 20). Notably, after overexpression of miR-644a in ESCC cells, the levels of PITX2, p-Akt, p-GSK-3β (inactive form) and active-β-catenin were all reduced (Fig. 5A). There was a concomitant decrease in the cell invasion and sphere-forming abilities of the miR-644a–overexpressing cells (Supplementary Fig. S6). These malignant characteristics were substantially rescued, by PITX2 transfection into the miR-644a–overexpressing ESCC cells (Fig. 5A and Supplementary Fig. S6). These data suggested that miR-644a could inhibit the Akt/GSK-3β/β-catenin pathway through PITX2. To probe this interaction, we used the PI3K/AKT inhibitor, LY294002, to inactivate the Akt/GSK-3β/β-catenin pathway (Fig. 5A). We observed that LY294002 rescued both the cell invasion and sphere-forming abilities in miR-644a–overexpressing ESCC (Supplementary Fig. S6). Moreover, knockdown of PITX2 reduced p-Akt, p-GSK-3β and active-β-catenin levels in ESCC cells (Fig. 5A).

In our study, we also found that the levels of miR-644a were negatively correlated with the expressions of PITX2, p-AKT, p-GSK-3β, and nuclear β-catenin in 190 cases of FFPE ESCC tissues (P < 0.001, Fig. 5B; Supplementary Table S4). These results were confirmed in 20 fresh ESCC specimens, in which a negative correlation between miR-644a levels and the expressions of PITX2 (r = −0.790, P = 0.007), p-AKT (r = −0.857, P = 0.002), p-GSK-3β (r = −0.744, P = 0.014), and nuclear β-catenin (r = −0.763, P = 0.010, Fig. 5C) was observed. Furthermore, the expression levels of PITX2, p-AKT, p-GSK-3β, and nuclear β-catenin in xenograft tumors that developed from the miR-644a–overexpressing ESCC cells were significantly lower than those from the control cells (Supplementary Fig. S7).

As the phosphorylation and activity of Akt are regulated by its acetylation status, which regulate the downstream effectors (20, 21), we therefore investigated whether miR-644a is involved in the acetylation of Akt, GSK-3β, and β-catenin. These proteins were immunoprecipitated using the corresponding antibodies and their acetylation statuses were determined using anti-acetylated-lysine antibodies. The results showed that β-catenin was not acetylated in either miR-644a–overexpressing or control cells, whereas GSK-3β was constitutively acetylated under both conditions (Fig. 5D). In contrast, although Akt was constitutively acetylated in ESCC cells, the levels of acetylated Akt were dramatically upregulated after overexpression of miR-644a; this could be largely inhibited by enforced expression of PITX2 (Fig. 5D). These results revealed that miR-644a repression of PITX2 can regulate the levels of acetylated Akt in ESCC cells.

Finally, we explored the molecular mechanisms responsible for the downregulation of miR-644a in ESCC. The miR-644a coding sequence is located in the intron of the ITCH gene, and miR-644a expression correlated with ITCH mRNA expression in ESCC cell lines (r = 0.967, P<0.001) and ESCC tissues (r = 0.963, P < 0.001, Fig. 6A), suggesting that miR-644a is cotranscribed with ITCH mRNA. We next examined whether reduced miR-644a expression in ESCC was the result of promoter hypermethylation. Analysis of the ITCH promoter region using the UCSC genome browser (http://genome.ucsc.edu/) indicated a CpG island located between −285 bp and −56 bp relative to the transcription start site (Fig. 6B). Furthermore, BGS PCR showed that the CpG islands within the ITCH promoter were hypermethylated in 2 miR-644a low–expressing ESCC cell lines (KYSE-140 and Eca109) and 3 ESCC tumor samples. In contrast, the CpG islands were mostly not methylated in the control NE-1 and miR-644a high–expressing ESCC cells (KYSE-410) and 3 paired normal esophageal tissues (Fig. 6C). Moreover, miR-644a expression was significantly increased in ESCC cell lines with low levels of miR-644a after treatment with the methylase inhibitor 5-aza-dC, but not in NE-1 and KYSE-410 cells (Fig. 6D). These results indicated that reduced miR-644a expression in ESCC is attributable to hypermethylation of its promoter.

We previously reported that PITX2 was frequently overexpressed in ESCCs and associated with ESCC patients' chemoresistance and poor prognosis (12). However, the regulatory mechanisms of PITX2 in ESCC remained elusive. In the current study, we first identified that miR-644a functioned as an important negative regulator of PITX2 in ESCC cells and tissues. Next, we provided evidence that decreased expression of miR-644a in ESCC is important in the acquisition of an aggressive and/or poor prognostic phenotype. Thus, we conclude that miR-644a level appears to have the potential to predict certain outcomes for ESCC patients. Specifically, miR-644a expression could be used as an additional tool in identifying those ESCC patients at increased risk of tumor invasion and/or metastasis. These findings underscore a potentially important role of miR-644a downregulation in the development and/or progression of ESCC.

The role of miR-644a in human cancers remains poorly studied. Only one report has documented a significant association between the levels of miR-644 and patients' survival in acute myeloid leukemias (22). In the current study, a series of in vitro and in vivo assays were employed to investigate the role of miR-644a in regulating the characteristic malignant phenotype of ESCC. Our results clearly demonstrated that miR-644a could substantially inhibit ESCC cells growth and/or aggressiveness by inactivating the Wnt/β-catenin signaling through repressing PITX2. Various studies have documented that the Wnt/β-catenin signaling pathway is constitutively activated in many types of human cancer, and plays an important role in triggering aggressive cellular features that constitute a malignant phenotype (23–25). It was reported that PITX2 and the Wnt/β-catenin pathway exert positive-feedback regulation, while aberrant activation of Wnt/β-catenin signaling is strongly involved in ESCC tumorigenesis and/or progression (8, 26). These results collectively suggest that the Wnt/β-catenin axis is one of the critical downstream pathways activated by PITX2 accumulation in ESCC malignancy that we now propose is miR-644a-mediated.

To date, however, the molecular mechanism by which PITX2 regulates Wnt/β-catenin signaling have not been elucidated. PITX2 was previously shown to interact directly with the Wnt gene promoters, thus inducing the activation of the canonical WNT/β-catenin pathways in human ovarian cancer cells (8). Notably, PITX2 may function as a transcription factor and is thus expected to be localized in the cell nucleus. However, consistent with our previous study (12), we found that PITX2 was predominantly located in the cytoplasm of ESCC cells, which suggests that, in addition to its potential role as a transcription factor, the function and localization of PITX2 may be tumor tissue specific. In ESCC cells, PITX2 activates the WNT/β-catenin signaling through other unclear mechanisms. It is known that Akt activation can inhibit GSK-3β activity, leading to nuclear β-catenin accumulation and ultimately triggering tumor progression (27). We therefore investigate whether PITX2 can activate Akt/GSK-3β signaling, thus affecting the Wnt/β-catenin pathway in ESCC cells. Our results demonstrate that in ESCC, the Akt/GSK-3β pathway is involved in the miR-644a–mediated regulation of the Wnt/β-catenin signaling. Ectopic overexpression of PITX2 in ESCC cells could not only reverse miR-644a–inhibited Akt/GSK-3β/β-catenin activity, but also substantially suppressed miR-644a–induced Akt acetylation. These results provided evidence that in ESCC, PITX2 activates the Wnt/β-catenin pathway through the Akt/GSK-3β signaling cascade to exert its downstream oncogenic effect. Clearly, further studies are needed to clarify the precise molecular mechanisms governing how PITX2 regulates the activity of Akt/GSK-3β/β-catenin.

Recently, the cancer stem cell (CSC) hypothesis has attracted significant attention, and the development of stem cell–like properties in cancer cells has been recognized as a hallmark of disease progression, while the unlimited self-renewal capacity of CSCs is thought to contribute to tumor progression and recurrence (28, 29). Accumulating evidence supports the involvement of some miRNAs in the regulation of CSC function (30, 31); in particular, reduced let-7 expression was associated with tumor-initiating cells after the initiation of chemotherapy in breast cancer, and is thought to result in the maintenance of their undifferentiated status and proliferative potential (32). Similarly, miR-429 contributes to hepatocyte self-renewal, malignant proliferation, tumorigenicity, and chemoresistance, making it an attractive potential target to inactivate tumor-initiating cells in the liver (33). We show that miR-644a functions as an upstream activator of the Wnt/β-catenin pathway in ESCC, a pathway well characterized in the development of stem cell–like properties in a wide variety of cancer types, including ESCC (25, 26). Therefore, we hypothesized that the levels of miR-644a could influence the stemness of ESCC cells via regulation of PITX2. Our results demonstrated a novel role for miR-644a in suppressing ESCC tumorigenesis by modulating ESCC cell's stem-like traits through PITX2. We thus propose that miR-644a downregulation in ESCC cells, which increases PITX2 expression, results in aberrant activation of the Wnt/β-catenin pathway and a resultant stem cell–like phenotype. This might promote ESCC cell survival following adjuvant therapy, such as radiochemotherapy, thus mediating resistance to therapy and resulting in tumor recurrence and/or metastasis.

Previous studies have supported the notion that some aberrantly expressed miRNAs could be used as an ideal target in cancer therapy; in general, therapeutic modulation of miRNAs is achieved by inhibiting oncogenic miRNAs, or by reconstituting tumor-suppressive miRNAs (34, 35). Multiple approaches have also been in rapid development to open novel avenues for exploiting miRNAs in preclinical practice (36). For instance, restoring the expression of silenced miRNAs, such as Let-7, miR-16, and miR-31, has been shown to have strong anticancer effects in various preclinical models (37). These data, together with our current findings, suggest that increasing the levels of miR-644a in ESCC cells by miR-644a replacement might be an efficacious approach for cancer treatment, leading to the inactivation of certain oncogenic pathways, specifically the Akt/GSK-3β/β-catenin signaling, thus generating strong suppressive effects on ESCC stemness and/or progression.

In our study, one critical question was raised: how is miR-644a downregulated in ESCCs? It is well established that promoter methylation and genomic deletion are the most common pathologic mechanisms for aberrant expression of certain miRNAs in human cancers (38). In the current study, we provided several lines of evidences that the promoter hypermethylation of miR-644a is a key regulator of miR-644a expression and that this mechanism appears to be used to decrease expression of miR-644a in ESCCs. It is noteworthy here that 5-aza-dCv, a widely used methylase inhibitor, could significantly increase the levels of miR-644a in ESCC cells with low endogenous miR-644a. Previous studies have suggested that the reversal of some epigenetic processes, such as DNA demethylation, could prevent or reverse the malignant phenotype of human cancers by reactivating tumor suppressor genes (39, 40). The potential anticancer activities of DNA demethylation agents, including 5-aza-dC, have been studied extensively and represent a promising avenue for cancer prevention and/or therapy (41, 42). Furthermore, 5-aza-dC has shown prominent antitumor activity in vivo and is being evaluated for its potential clinical significance (43, 44). These data imply that in addition to the miR-644a replacement therapy strategy in ESCC, demethylation of the miR-644a promoter by methylase inhibitors is another approach that could prove effective as an ESCC treatment.

In summary, our report describes, for the first time, the expression dynamics of miR-644a in human ESCCs. Decreased expression of miR-644a, caused by its promoter hypermethylation, may be important in the tumorigenesis and acquisition of an aggressive/poor prognostic phenotype of ESCC. In addition, functional and mechanistic studies of miR-644a provided in this study suggest a mechanism by which downregulation of miR-644a could support ESCC cells' aggressiveness and/or stem cell–like phenotype by activating Akt/GSK-3β/β-catenin signaling through PITX2, an activity that might be responsible, at least in part, for the development and/or progression of human ESCCs. Thus, miR-644a could be employed as a new prognostic marker and/or be developed as a therapeutic for treatment of ESCC.

No potential conflicts of interest were disclosed.

Conception and design: D. Xie

Development of methodology: Z.-H. Chen, Y. Xu, Z.-S. Zheng, S. Ye

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.-X. Zhang, Z.-H. Chen, Y. Xu, J.-W. Chen, H.-W. Weng, M. Yun, C. Chen, B.-L. Wu, E.-M. Li, J.-H. Fu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.-X. Zhang, Z.-H. Chen, Z.-S. Zheng

Writing, review, and/or revision of the manuscript: J.-X. Zhang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Xu, H.-W. Weng, M. Yun, C. Chen, B.-L. Wu, E.-M. Li, J.-H. Fu

Study supervision: S. Ye, D. Xie

This work was supported by the National Natural Science Foundation of China (nos. 81401991, 81225018, and 81572359), the Natural Science Foundation of Guangdong (no. S2012010009466), the Foundation of Guangdong Esophageal Cancer Institute (no. M01513), and the Natural Science Foundation of China-Guangdong Joint Fund (no. U1301227).

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.