DNA hypermethyation and silencing of PITX1 correlated with advanced stage and poor postoperative prognosis of esophageal squamous cell carcinoma

Esophageal squamous cell carcinoma (ESCC) is associated with the accumulation of genetic and epigenetic changes in the background mucosa. Dysregulated DNA methylation is known to lead to the inactivation of tumor suppressor genes and the activation of oncogenes. To identify the genes whose expression is perturbed by abnormal DNA methylation in ESCC, integrative transcriptomics by serial analysis of gene expression (SAGE) and methylome sequencing by methyl-DNA immunoprecipitation (MeDIP) analysis were performed. We found 159 genes with significantly decreased expression in ESCC compared to that in noncancerous esophageal mucosa. MeDIP-seq analysis identified hypermethylation in the promoter region of 56 of these genes. Using surgically resected tissues of 40 cases, we confirmed that the paired-like homeodomain 1 (PITX1) gene was hypermethylated in ESCC compared to that in normal tissues (P < 0.0001) by pyrosequencing. PITX1 overexpression in ESCC cell lines inhibited cell growth and colony formation, whereas PITX1 knockdown accelerated cell growth. A PITX1-transfected ESCC cell line, KYSE30, formed smaller tumors in nude mice than in mock-transfected cells. Hypermethylation of PITX1 was associated with tumor depth (P = 0.0011) and advanced tumor stage (P = 0.0052) and predicted poor survival in ESCC (hazard ratio, 0.1538; 95% confidence interval, 0.03159–0.7488; P = 0.0169). In this study, we found a novel tumor suppressor gene of ESCC, PITX1, which is silenced by DNA hypermethylation. Downregulation of PITX1 contributes to the growth and progression of ESCC. Hypermethylation of the PITX1 in ESCC correlated with tumor progression and advanced stage cancer, and may predict a poor prognosis.


INTRODUCTION
Esophageal squamous cell carcinoma (ESCC) is the major histological subtype of esophageal cancer and one of the most frequent fatal malignancies in the world, especially in Asian countries [1]. Environmental and lifestyle factors, such as smoking and alcohol consumption, are the risk factors for ESCC in Western countries [2], and the consumption of hot beverages has been shown to be a major risk factor in the East [3]. As with many other types of cancers, exposure of the esophageal epithelial mucosa to such conditions induces genetic mutations and epigenetic changes. These changes accumulate, induce mucosal dysplasia, and eventually, invasive squamous cell carcinoma.
CpG island hypermethylation occurs frequently in cancer cells and is associated with transcriptional repression and subsequent reduction or loss of gene function [4]. These events result in the loss of a benign phenotype and acquisition of malignant properties [5]. A previous global study using bead arrays to detect promoter-DNA methylation in ESCC demonstrated that a total of 37 CpG sites were differentially methylated between esophageal squamous cell tumors and normal mucosa [6]. The hypermethylation of some genes, such as APC, FHIT, PGP9.5, and TFLC1, is related to certain clinicopathological features of ESCC, including poor prognosis and treatment responses [7]. These findings suggest a biological significance for differentially methylated genes in ESCC. Recently, we reported that increased DNA methylation on the promoter of periplakin (PPL), a desmosomal protein, significantly altered ESCC stratification and led to a metastatic phenotype [8]. PPL had never been reported as a hypermethylated gene associated with ESCC, possibly due to the high methylation status of the PPL promoter region in the surrounding noncancerous mucosa. These results encouraged us to perform a genomewide DNA methylation analysis, combined with a complete expression profile, to identify epigenetically regulated molecular players in ESCC.
With the advent of next-generation sequencing technologies, methyl-DNA immunoprecipitation (MeDIP) analysis has become a powerful tool for comprehensive characterization of DNA methylation status in coding and non-coding regions of the genome. In this study, we assessed the genes displayed abnormal methylation in its promoter region in ESCC at a global level. We used MeDIP-seq and integrated the methylome profile with data from transcriptome sequencing by serial analysis of gene expression (SAGE) to find candidate genes with altered regulation and DNA methylation status. These candidates may be involved in the development and progression of ESCC. Herein, we found that paired-like homeodomain 1 (PITX1) was markedly suppressed by hypermethylation on the promoter in ESCC samples. We restored PITX1 expression in ESCC cell lines and found that PITX1 functions as a tumor suppressor in ESCC progression both in vitro and in vivo. Finally, we demonstrated for the first time that hypermethylation of PITX1 was a poor prognostic factor for patients with ESCC.

Differentially expressed genes (DEGs) in ESCC and correlation with abnormal promoter methylation
To uncover genes with disturbed expression and abnormal methylation in the promoter region, transcriptome and methylome analyses were performed and integrated. First, to search for differentially expressed genes (DEGs) between normal esophageal mucosa and ESCC, serial analysis of gene expression (SAGE)-seq was performed. Result of transcriptome is uploaded at GSE71411. In total, 59 genes had significantly increased expression in ESCC relative to normal mucosa (WAD score cutoff value = 1.7). Additionally, 159 genes had significantly decreased expression in ESCC. Next, to search for differentially methylated regions (DMRs) between normal esophageal mucosa and ESCC, methyl-DNA immunoprecipitation (MeDIP)-seq was performed. Result of methylome is uploaded at GSE71410. Hypermethylated regions in ESCC relative to normal mucosa totaled 123,020, and 79,436 hypo-methylated regions were found (log-fold change (logFC) cutoff value = 2). To integrate these data, the distance of each DMR from the transcriptional start site (TSS) of a gene was calculated, and a gene was designated as differently methylated if it had DMR from -10 kb to +2 kb from TSS. Integrated analysis of DEGs and DMRs identified 5 of 59 over-expressed genes with promoter hypo-methylation ( Table 1) and 56 of 159 downregulated genes with promoter hypermethylation (Table 2) in ESCC.
Among the cancer-specific hypermethylated genes, SCGB3A1 [9] and PPL [8] were already reported to have hypermethylation at its promoter, leading to reduced expression in ESCC. PITX1 [10][11][12][13][14][15], CST6 [16,17] and HOPX [18][19][20][21] were reported their downregulation in various types of human cancer, and PRSS27 [22] was known to be highly expressed in nonkeratinizing stratified squamous epithelia of human esophagus and its dysregulation was supposed to be related to characteristics of carcinoma. Thus, we initially started to evaluate results from the SAGE-seq analysis by quantitative RT-PCR using paired normal and ESCC frozen tissue samples from 32 patients on these 4 genes. The expression of these genes was significantly decreased in the cancer tissues compared with those in matched normal mucosa ( Figure 1A). To confirm the results of MeDIP-seq, DNA hypermethylationinduced gene silencing was also examined in the ESCC cell lines. Endogenous mRNA expression levels were determined in four ESCC cell lines (KYSE30, KYSE140, KYSE150 and KYSE270) and compared with the levels in cells treated with 5-aza-2′-deoxycytidine (5-aza-dC), an inhibitor of DNA methyltransferase. In all cell lines, treatment with 5-aza-dC significantly increased selected gene expression ( Figure 1B). These results indicate that PITX1, PRSS27, CST6, and HOPX selected from our integrative analysis are likely the targets of aberrant DNA hypermethylation, and DNA hypermethylation likely downregulated expression in ESCC. Among the differentially methylated genes uncovered here, PITX1 was selected for further analysis, because we were interested in its previously reported function as negative regulator for the telomerase reverse transcriptase (TERT), which plays an important role in the cell growth [23].

PITX1 protein expression was lower in ESCC than in normal esophageal mucosa
To investigate the protein expression and localization of PITX1 in the normal esophageal mucosa or ESCC, immunohistochemical analyses were performed. In the normal mucosa, PITX1 protein was detected in the nuclei of a majority of squamous cells, including the basal cell layer and spinous cells in the suprabasal cell layer. Cells in several stratified layers close to surface were not positive for PITX1 protein expression ( Figure 2A). Proliferating Ki67 + cells were found only in basal cell layer, and all of these cells expressed PITX1. In ESCC samples, the expression of PITX1 was dramatically decreased, while the expression of Ki67 had increased ( Figure 2B). The percentage of PITX1-and Ki67expressing cells was quantified. More than 80 % of nuclei in the normal mucosal layer expressed PITX1; however, the average percentage of PITX1 + cells in ESCC samples examined was 38.3%. The average percentage of Ki67 + cells of ESCC (67.0%) was higher than that of normal tissues (less than 10%). Double staining with anti-PITX1 and anti-Ki67 antibodies was difficult because the conditions for antigen retrieval for each protein were markedly different; however, these results indicated the emergence of ESCC-specific, PITX1 -Ki67 + proliferating cells, which were never detected in the normal tissues. In other words, the loss of PITX1 may be associated with cancer cell proliferation.

Ectopic PITX1 expression inhibited cell growth in vitro and in vivo
To clarify the significance of PITX1 defects in ESCC, PITX1-transfected and mock-transfected ESCC cell lines, KYSE30 and KYSE150, were prepared and examined for the effect of ectopic PITX1 expression on their growth. In both ESCC cell lines used, the ectopic expression of PITX1 suppressed cell growth compared with mock transfectants ( Figure 3A). Two clones of stably PITX1-transfected KYSE30 cells expressed PITX1 mRNA at different levels: clones 3 and 4 expressed PITX1 5-and 8-fold higher than the mock clones (clone 1 or 2), respectively. In PITX1-transfected KYSE30 cells, the growth inhibitory effect of PITX1 appeared to be dose-dependent on mRNA level ( Figure 3A and 3B). Furthermore, human normal epithelial cell HEEpiC was treated with control or PITX1-shRNA and examined for the effect of PITX1 suppression on its growth. Fortyeight hours after transfection, the growth of PITX1-shRNA treated cells was higher than that of control shRNA-treated cells ( Figure 3A). Since PITX1 has been reported as negative regulator for TERT, we examined the expression levels of TERT by quantitative RT-PCR, and then compared them between ESCC and paired normal tissues. In human tissue samples, expression levels of TERT in the cancer tissues were significantly higher than those in matched normal mucosa (P = 0.0388; Figure 3C). We next examined for the effect of ectopic PITX1 expression on the TERT transcription. Two clones of stably PITX1-transfected KYSE30 cells exhibited a significant decrease in TERT mRNA expression ( Figure  3B). In agreement with the results on TERT expression, the ectopic expression of PITX1 inhibited telomerase activity in both PITX1-transfected clones ( Figure 3D). In addition, treatment of KYSE30 cells with 5-aza-dC resulted in a 0.6-fold decrease in telomerase activity ( Figure 3E), concomitant with the induction of PITX1 expression ( Figure 1B). Stably PITX1-transfected KYSE150 cells also exhibited significantly decreased expression of TERT mRNA as compared with the mocktransfected KYSE150 cells ( Figure 3F). In a colony   Figure 3G). Finally, PITX1and mock-transfected KYSE30 cells were used in a xenograft model of tumor formation in BALB/c nude mice. PITX1 transfectants formed smaller tumors than mock transfectants ( Figure 3H). Collectively, these results demonstrate that expression of PITX1 inhibits cell growth of ESCC in vitro and in vivo.

DNA hypermethylation of PITX1 was associated with advanced tumor stage and poor survival in ESCC
MeDIP-seq analysis demonstrated hypermethylation at the PITX1 locus in ESCC compared to that in normal mucosa ( Figure 4A). Finally, to understand the significance of hypermethylation in the PITX1 gene, the DNA methylation levels of CpG islands were determined for the PITX1 promoter using pyrosequencing, and then they were compared between ESCC and paired normal tissues.
The average methylation level was 44.9 ± 10.7% in the tumor samples, compared with 35.2 ± 7.9% in the normal mucosa samples. A paired t-test indicated that the tumor DNA was significantly more methylated than that in the normal tissues ( Figure 4B). Methylation statuses of PITX1 in ESCC cell lines were also assessed by pyrosequencing. Hypermethylation of PITX1 was observed in all ESCC cell lines as compared to that in human normal epithelial cell HEEpiC ( Figure 4C). Because pyrosequencing reflects the methylation status of only 3 adjoining CpG motifs, the PCR products, containing 7 CpGs as shown in Figure  4A, were subjected to bisulfite sequencing. As expected, methylated CpGs were hardly seen in the HEEpiC cells, whereas it was clearly evident that the upstream region of the PITX1 gene was frequently hypermethylated in ESCC tissues ( Figure 4D). These results suggest that the loss of PITX1 by DNA hypermethylation is associated with cancer cell proliferation. To further clarify the clinical relevance of PITX1 hypermethylation in ESCC, we analyzed the methylation status of the PITX1 gene and clinicopathologic characteristics of 40 patients with Promoter, Gene body' means that CpG islands were located in the promoter region and gene body of indicated genes but ESCC-specific hyper-methylation was observed only in their promoters. 'Promoter' means that CpG island was located the promoter region of indicated genes and hyper-methylated in ESCC as compared in normal esophageal mucosa. 'Gene body' means that CpG island was located only in the gene body of indicated genes and ESCC-specific hyper-methylation was observed in their promoters. 'No' means that indicated genes have no CpG island. ESCC. We classified the samples into two groups: 1) percentage of methylated PITX1 DNA was ≥50% by pyrosequencing (hypermethylated) and 2) methylated DNA was less than 50 % (methylated). Of the 40 primary tumors studied, 13 were classified as hypermethylated (Table 3). Univariate analysis revealed no difference between the hypermethylated and methylated groups with respect to gender, tumor location, lymph node metastasis, or distant metastasis. However, there were significant differences between patients in the hypermethylated and methylated groups with respect to age (P = 0.0388), tumor depth (P = 0.0017), and tumor stage (P = 0.0053).
We then applied univariate analysis to evaluate the association between the methylation status of the PITX1 gene and patient prognosis. Kaplan-Meier curve analysis showed a significant difference between patients in the hypermethylated and methylated groups with respect to overall survival rate (hazard ratio, 0.1538; 95% confidence interval, 0.03159-0.7488; P = 0.0169; Figure 4B). These results suggest that the DNA hypermethylation of PITX1 is associated with cancer progression and predicts poor postoperative survival in ESCC.

DISCUSSION
In this study, we demonstrated that the nuclear expression of PITX1, a member of the RIEG/PITX1 homeobox family, was suppressed in ESCC. Although downregulation of PITX1 was reported in various types of human cancer, including colon cancer cell lines, prostate and bladder tumors [10], lung [11], and gastric cancers [12], Barrett's-adenocarcinoma [13], oral tumors [14] and malignant melanoma [15], this report is the first to show a role for PITX1 in ESCC. Furthermore, these results point to a mechanism of PITX1 downregulation through elevated promoter methylation levels. Hypermethylation of PITX1 clearly correlated with advanced tumor size, advanced tumor stage, and a poor prognosis, suggesting the possibility of the methylation status of PITX1 as prognostic markers of ESCC. Protein localization of PITX1 in the esophageal benign and malignant mucosa, and its function of regulating the growth of ESCC, were also illustrated.
The results presented here clearly demonstrate that PITX1 is one of the targets of aberrant DNA hypermethylation in ESCC. A previous global study using bead arrays to detect promoter-DNA methylation uncovered 37 CpG sites that were differentially methylated between esophageal squamous cell tumors and normal mucosa [6]. PITX1, however, was not specifically listed in that report or other previous reports investigating hypermethylated genes in ESCC. This difference may be the result of the selected methodology for methylome analysis. In the method using bead arrays, bisulfite-converted genomic DNA is hybridized to probes designed against selected loci. Thus, by using bead arrays, DNA methylation patterns can be interrogated in a locus-targeted manner. In contrast, MeDIP-seq can  detect DMRs across the entire genome by sequencing of DNA captured by the MBD2 protein, resulting in a more comprehensive analysis of the entire genome. Another reason that PITX1 was not previously detected may be the high methylation status of the PITX1 promoter region in the normal/background mucosa; the lowest level of the PITX1 promoter methylation in all the normal tissue samples was 19% in our study. Although the difference was not dramatic, there was a statistically significant increase in the methylation levels on PITX1 in tumors. Thus, this method had advantage to detect ESCC-related changes those were not "all-or-none" type difference but significant additional changes to background mucosa. In fact, the presence of aberrant DNA hypermethylation in noncancerous esophageal mucosa of ESCC cases was reported in association with smoking history [24]. As shown in the case of PITX1, modest but significant additional DNA methylation to the esophageal mucosa may trigger the development of ESCC. Combined with our previous publications, we have now reported two new target genes of increased promoter-DNA methylation in ESCC: PITX1 and PPL [8]. In addition to these genes, our integrated analysis of DNA methylome and transcriptome identified 59 genes. Expression patterns of these genes may be disrupted by abnormal methylation in ESCC.
These results suggest that sequence-based methylome analysis is an effective tool for identifying novel methylation hotspots. This tool is especially powerful for the identification of genes with background methylation levels that are higher than typically observed, possibly due to exposure to various environmental risk factors.
Another noteworthy observation is that DNA hypermethylation of PITX1 correlated with advanced tumor size, advanced tumor stage, and a poor postoperative prognosis. ESCC is characterized by high malignancy and a poor prognosis. Although PITX1 methylation may correlate with total epigenetic alteration of many genes, which is associated with the phenotypes of clinically  advanced tumor cells, our studies clearly demonstrate that PITX1 methylation status of the tumor could be used as a clinical marker to predict prognoses. Multimodal therapy has been performed for ESCC, in combination of chemotherapy, radiation therapy and surgery. Since surgery for ESCC is the most invasive remedy, treatment without operation is sometime chosen. However, currently, there is no marker to support the selection of the most appropriate combination of treatment. At least, PITX1 methylation status of the tumor could be one of the useful references for clinical decision making of invasive surgery.
In this study, we have not analyzed the relationships between preoperative chemotherapy or irradiation efficacy and PITX1 methylation status. Further studies will refine the value of evaluation of PITX1 methylation status as a clinical prognostic marker of ESCC. One major tumor-suppressive mechanism is programmed cell death; induced by the shortening of telomeres through suppressed telomerase activity [25]. Previously, PITX1 was identified as a TERT suppressor through the direct binding of the PITX1 protein to the TERT promoter [23]. In addition to canonical function of telomere length regulation, TERT acts as a transcriptional modulator of Wnt/β-catenin and nuclear factor-κB (NF-κB) signaling pathways. In ESCC, TERT-mediated activation of the Wnt and NF-κB signaling pathways and subsequent enhanced expression of target genes may play a direct role in facilitating cancer-promoting functions, such as proliferation and resistance to apoptosis [26,27]. PITX1 is identified as a negative regulator of the RAS pathway through the RAS protein activator like 1 (RASAL1) gene, a member of the RAS GTPase-activating protein family [10]. Because activation of the RAS/mitogen-activated protein kinase (MAPK) pathway contributes to the tumorigenesis and progression of ESCC and PITX1 has been identified as a negative regulator of the RAS pathway via the downregulation of RASAL1 [10], the epigenetic silencing of PITX1 may result in the activation of the RAS/MAPK pathway, leading to the hyperproliferation of the esophageal epithelium. These molecular mechanisms may explain the findings that forced expression of PITX1 and knockdown of PITX1inhibited and accelerated cell growth, respectively. Overall, these studies support our in vitro and in vivo results that PITX1 is a tumor suppressor gene in ESCC.
Thus, our results are the first to indicate that PITX1 functions as a tumor suppressor gene in ESCC, and that hypermethylation of its promoter serves as a novel biomarker for predicting prognosis in ESCC.

Patients
Paired ESCC and normal tissue specimens were obtained from 40 patients who had undergone esophagectomy or esophagogastrectomy with confirmed diagnosis of ESCC. The patients were selected from those who underwent surgery between January 2013 and August 2015 at the National Center for Global Health and Medicine (NCGM). Quantitative PCR for expression analysis was performed using 32 paired frozen samples from these patients. This study was approved by the research ethics committees of the NCGM (121 or 1484), and informed consent was obtained from all the patients before the samples were collected.
The characteristics of ESCC are summarized in Table 3. Among the 40 paired ESCC samples examined in this study, paired specimens from a 62-year-old man with ESCC at pT3, pN2, and pM0 (stage IIIB) were subjected to methylome analysis and transcriptome analysis using a next-generation sequencing (NGS) system.

SAGE-seq analysis
SAGE-seq libraries were generated from 5 μg of total RNA using SOLiD SAGE Kit with Barcoding Adaptor Module and multiplexed using SOLiD RNA Barcoding Kit Module 1-16 (Life Technologies, Carlsbad, CA) according to the manufacturer's instructions. Briefly, total RNA was treated with DNase I, and then bound to Dynabeads Oligo(dT) EcoP magnetic beads for cDNA synthesis. The double-stranded cDNA synthesized on the beads was digested with Nla III and ligated to the adaptor containing an EcoP15I restriction enzyme recognition site. The adapter-ligated cDNA was digested with EcoP15I to release the tag containing the adaptor sequence and 27 bp of unique sequence from a single transcript. Each library was amplified in a PCR reaction by 8 cycles.

MeDIP-seq analysis
Genomic DNA was extracted from tissue specimens and sheared for the MeDIP assay. The MeDIP assay was performed by using the MethylMiner™ Methylated DNA Enrichment Kit (Life Technologies) according to the manufacturer's instructions. MeDIP specificity was confirmed by MeDIP-PCR with primer sets of the active-chromatin locus glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and imprinted locus H19 (data not shown). The MeDIPed DNA was prepared for NGS-seq library construction using the SOLiD Fragment Library Construction Kit and multiplexed by using SOLiD™ Fragment Library Barcoding Kit Module 1-16 (Life Technologies) according to the manufacturer's instructions. Briefly, each MeDIPed DNA was endrepaired, barcoded by adapter ligation, amplified by 15 PCR-cycles, and purified.

Sequencing
The size of the constructed library was quantified by the 2100 Bioanalyzer system with the High Sensitivity DNA kit (Agilent Technologies, Santa Clara, CA) and the SOLiD Library TaqMan Quantitation Kit (Life Technologies). Library DNA was subjected to emulsion PCR, enrichment, and deposition onto bead according to the supplier's protocol. Finally, DNA was sequenced using the AB 5500 Genetic Analyzer (Life Technologies).

Informatics
All short reads from the AB 5500 Genetic Analyzer were aligned to human genome version 37.3 (hg19) using the Tophat2 program with default parameters. We defined a region as a peak, if the region had more than 10 RP100M (Reads per 100 million of mapped reads) for an interval of more than 20 nucleotides [28]. The weighted average difference (WAD) score was calculated based on fold change values and weighted with degrees of expression values [28]. Sequences corresponding to differentially expressed genes (DEGs) in ESCC compared with normal mucosa were detected by a WAD score of 1.7 for SAGEseq. Sequences corresponding to differentially methylated regions (DMRs) in ESCC relative to normal tissue were detected using Iida's methods with log-fold change (logFC) 2 for MeDIP-seq. For each gene, the distance from the transcription start site (TSS) was calculated using the NCBI Entrez annotation (hg19) as a reference. Genes having peaks of −10 kb to +2 kb from TSS were designated as DNA methylated genes.

Reverse-transcription and quantitative PCR
Total RNA was isolated from tissues using RNA Bee RNA Isolation Reagent (Tel-Test, Inc., Friendswood, TX). After the RNA was treated with DNase I, doublestranded cDNA was synthesized using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA). Quantitative PCR was performed using ABI TaqMan probes (Applied Biosystems) as described previously. Threshold cycle numbers were determined using the Sequence Detector software and transformed as described by the manufacturer, with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as the calibrator gene. The TaqMan Gene expression assay IDs used in this study for the genes as follows: PITX1, Hs00267528_m1; PRSS27, Hs01029754_m1; CST6, Hs01029754_m1; HOPX, Hs04188695_m1; TERT, Hs00972656_m1; GAPDH, Hs00266705_g1.

Ectopic expression of PITX1
To construct the expression vector of the PITX1 gene, a DNA fragment encoding the full-length ORF (SC122251, OriGene Technologies, Rockville, MD) was digested with the restriction endonucleases SalI and XmaI, and then the fragment was ligated into the pIRES2-EGFP vector (CLONTECH). The resulting vector was transfected into KYSE30 and KYSE150 cells using lipofectamine LTX reagent (Life Technologies). Stably transfected cells were isolated using the MoFlo TM XDP Cell Sorter (Beckman Coulter, Inc.).

Transient transfection of shRNA
To knockdown the expression of the PITX1 gene, four unique 29 mer shRNA constructs in plasmids were obtained from OriGene Technologies (TG310405). In these four 29mers (A, B, C, and D), only construct A efficiently suppressed the PITX1 expression; therefore, this construct was used. For control experiments, a noneffective 29-mer scrambled shRNA cassette in same plasmid was used. They were transfected into HEEpiC cells using lipofectamine 3000 reagent (Life Technologies) according to the manufacturer's instructions.

Immunohistochemical analysis
Formalin-fixed, paraffin-embedded sections of surgical specimens from patients with ESCC were deparaffinized and rehydrated. Antigen retrieval was performed in 10 mM sodium citrate buffer (pH 6.0) using an autoclave for 5 min. Sections were stained with hematoxylin (KY and YIK), and 27-1406 (YIK) from the National Center for Global Health and Medicine, and MEXT-Supported Program for the Strategic Research Foundation at Private Universities for Waseda University (to TD).