LINE-1 Hypomethylation, DNA Copy Number Alterations, and CDK6 Amplification in Esophageal Squamous Cell Carcinoma

Esophageal squamous cell carcinoma (ESCC), the major histologic type of esophageal cancer in East Asian countries, is one of the most aggressive malignant tumors (1). Despite remarkable advances in multimodal therapies, including surgery, chemotherapy, radiotherapy, and chemoradiotherapy, the prognosis of patients remains poor, even for those whose carcinomas have been completely resected (2–4). To develop innovative strategies for treating ESCC, especially those that are molecularly targeted (5), it is of particular importance to increase our understanding of the cellular and molecular basis of this disease. Importantly, epigenetic changes, including alterations in DNA methylation, are reversible, and can thus be targets for cancer therapy or chemoprevention (6–8).

DNA methylation alterations occurring in human cancers include global DNA hypomethylation and site-specific CpG island promoter hypermethylation (9). Global DNA hypomethylation seems to play a crucial role in genomic instability, leading to cancer development (10–12). Because long interspersed nucleotide element-1 (LINE-1; L1 retrotransposon) constitutes a substantial portion (approximately 17%) of the human genome, the level of LINE-1 methylation is regarded to be a surrogate marker of global DNA methylation (13). In many types of human neoplasms, LINE-1 methylation has been shown to be highly variable (14–16), and LINE-1 hypomethylation is strongly associated with a poor prognosis (17–22). In a previous study of 217 curatively resected ESCC specimens (23), we demonstrated that LINE-1 hypomethylation is associated with a poor prognosis, suggesting that LINE-1 hypomethylation may be a biomarker that can be used to identify patients who will experience an inferior outcome. However, the mechanism by which LINE-1 hypomethylation (and thus global DNA hypomethylation) may confer a poor prognosis remains to be fully explored. Given the known relationship between genome-wide DNA hypomethylation and genomic instability (24–27), we hypothesized that global DNA hypomethylation might be an important influential factor for DNA copy number variations, leading to altered expression levels of oncogenes or tumor suppressor genes.

To test this hypothesis, we investigated LINE-1–hypomethylated and LINE-1–hypermethylated ESCC tumors using array-based comparative genomic hybridization (CGH array). LINE-1–hypomethylated tumors showed highly frequent genomic gains at various loci containing candidate oncogenes such as CDK6. Thus, we further examined the correlation between LINE-1 methylation levels, CDK6 amplification, CDK6 protein and mRNA expression, and clinical outcome.

All samples used in this study were collected at Kumamoto University Hospital (Kumamoto, Japan) between April 2005 and December 2011. LINE-1 methylation levels were quantified in 40 ESCC tumors for which freshly frozen specimens were available. Total RNA was obtained from 20 ESCC tumors, matched with 15 macroscopically normal esophageal tissues from the same patients. To assess the prognostic impact of CDK6 expression, we performed immunohistochemical staining of 129 ESCC samples. No patient received any preoperative treatment (chemotherapy, radiation therapy, or chemoradiotherapy). All 129 patients underwent curative resection. Tumor staging was performed according to the American Joint Committee on Cancer Staging Manual (seventh edition; ref. 28). Patients were observed at 1- to 3-month intervals until death or until March 31, 2013, whichever came first. Disease-free survival was defined as the period following surgical cancer treatment during which the patient survived with no sign of cancer recurrence. Among the 129 patients, 48 experienced disease recurrence during the follow-up period. This current analysis represents a new analysis of CDK6 on the existing ESCC database that has been previously characterized for LINE-1 methylation and clinical outcome (23). We have not examined CDK6 expression in any of our previous studies. Written informed consent was obtained from each subject, and the study procedures were approved by the institutional review board.

Genomic DNA was extracted from frozen esophageal cancer specimens using a QIAamp DNA Mini Kit (Qiagen) according to the manufacturer's protocol. Genomic DNA was modified with sodium bisulfite using an EpiTect Bisulfite kit (Qiagen). PCR and subsequent pyrosequencing for LINE-1 were performed as previously described by Ogino and colleagues, using the PyroMark Kit (Qiagen; ref. 14). This assay amplifies a region of LINE-1 element (position 305 to 331, accession no. X58075), which includes four CpG sites. Pyrosequencing reactions were performed using the PyroMark Q24 System (Qiagen). Bisulfite pyrosequencing consists of three steps; bisulfite conversion, PCR amplification, and pyrosequencing analysis. Unmethylated cytosine and methylated cytosine are differentiated by bisulfite treatment followed by PCR. In the pyrosequencing step, the cytosine:methylated cytosine ratio at each CpG site is measured as the ratio of thymine:cytosine. The cytosine content relative to the cytosine plus thymine content at each CpG site is expressed as a percentage. In this study, the average relative cytosine content at the four CpG sites was used as the overall LINE-1 methylation level in a given tumor. In published literature, we have validated our LINE-1 methylation pyrosequencing assay; we performed bisulfite conversion on five different DNA specimen aliquots and repeated PCR pyrosequencing five times using four macrodissected cancers. Bisulfite-to-bisulfite (between-bisulfite treatment) SDs ranged from 1.4 to 2.9 (median, 2.3), and run-to-run (between-PCR pyrosequencing run) SDs ranged from 0.6 to 3.3 (median, 1.2; ref. 29).

Paraffin-embedded tumor sections were dewaxed in xylene and ethanol, and antigen–epitope retrieval was performed using a streamer autoclave at 120°C for 15 minutes in antigen retrieval solution (pH9, Histofine; Nichirei Biosciences Inc.) and then allowed to cool. Endogenous peroxidase activity was blocked using 3% hydrogen peroxide. Primary antibodies against CDK6 (1:50 dilution; Santa Cruz Biotechnology Inc.) and cyclin D1 (1:50 dilution; Leica Biosystems Newcastle Ltd) were applied, and the slides were incubated at 4°C overnight. The secondary antibody used was a ready-for-use anti-mouse EnVision Peroxidase system (Dako Japan Inc.). The remaining procedure was performed using a Dako EnVision+ System (Dako Japan Inc). The stained slides were counterstained with hematoxylin and bluing reagent. One of the investigators, blinded to any other participant data, recorded nuclear CDK6 and cyclin D1 expression as absent, weak, moderate, or strong. In this study, tumors with weak to strong expression were defined as “positive,” whereas tumors not expressing these proteins were defined as “negative.” Among 129 ESCC tumors, 61 tumors (47%) and 52 tumors (40%) tested positive for CDK6 and cyclin D1, respectively.

Total RNA extraction, cDNA synthesis, and quantitative reverse transcription PCR (qRT-PCR) were carried out as previously described (30, 31). Total cellular RNA was extracted using the RNeasy Mini Kit (Qiagen), and cDNA was synthesized with the SuperScript III Transcriptor First Strand cDNA Synthesis System for RT-PCR (Invitrogen) according to the manufacturers' instructions. qRT-PCR was carried out using a LightCycler 480 II instrument (Roche). To determine the differences in the gene expression levels between specimens, the 2−ΔΔCt method was used to measure the fold changes among the samples. To carry out qRT-PCR, primers were designed using the Universal Probe Library (Roche) following the manufacturer's recommendations. The primer sequences and probes used for real-time PCR were as follows: CDK6, 5- TGATCAACTAGGAAAAATCTTGGA-3′, 5-GTCGACCGGTGCAATCTT-3′, and universal probe #2; glyceraldehyde-3-phosphate dehydrogenase (GAPDH), 5-AGCCACATCGCTCAGACAC-3′, 5-GCCCAATACGACCAAATCC-3′, and universal probe #60.

Array CGH was carried out using the Agilent oligonucleotide array–based CGH microarray kit (SurePrint G3 Human CGH Microarray Kit 2 × 400 K; Agilent Technologies) according to the manufacturer's protocol. Commercial male human genomic DNA was used as the control DNA. Following hybridization, washing, and drying steps, the microarray slides were scanned at 3 μm resolution, using the G2505C microarray scanner (Agilent Technologies). Features were extracted from the scanned images using the Feature Extraction software (version 1.5.1.0; Agilent Technologies). The extracted features were analyzed using the Agilent Cytogenomics software (version 2.0.6.0; Agilent Technologies).

Human esophageal cancer cell lines (TE-1, TE-4, TE-6, TE-8, TE-9, TE-10, TE-11, TE-14, and TE-15) were obtained from the Institute of Development, Aging, and Cancer, Tohoku University (Sendai, Japan) and were cultured in a medium supplemented with 10% FBS in 5% CO₂ atmosphere at 37°C.

CDK6-specific FISH probe was prepared using two human Bacterial Artificial Chromosomes (BAC) DNA clones RP11-809H24 and RP11-1102K14 by labeling with Cy3. As a control, a centromeric region-specific FISH probe for chromosome 7 was prepared using RP11-90C3. The images were captured with the CW4000 FISH application program (Leica Microsystems Imaging Solution).

For the statistical analyses, we used the JMP software (version 9; SAS Institute). For the survival analysis, the Kaplan–Meier method was used to assess the survival time distribution, and the log-rank test was used. We constructed a CDK6-adjusted Cox proportional hazard model to compute the HR according to the LINE-1 methylation status. Statistical differences were considered significant if P values were <0.05. All reported P values are two-sided.

We quantified the LINE-1 methylation levels of 40 fresh-frozen esophageal cancer tissues using a bisulfite PCR pyrosequencing assay, and obtained valid results in all 40 cases. Representative pyrograms for LINE-1 methylation levels are shown in Fig. 1A. LINE-1 methylation levels in ESCC fresh-frozen tissues were distributed widely and approximately normally. The distribution was as follows: mean, 51.0; median, 52.6; SD, 14.8; range, 24.4–79.7; interquartile range, 38.6–62.5 (all in 0–100 scale; Fig. 1B).

We have previously demonstrated that LINE-1 hypomethylation is associated with a poor prognosis in a study of 217 ESCC paraffin-embedded specimens (23). However, the biologic mechanism by which LINE-1 hypomethylation affects aggressive tumor behavior is not yet fully understood. We hypothesized that global DNA hypomethylation might be a crucial influential factor for DNA copy number variations. Thus, we carried out a CGH array analysis using three LINE-1–hypomethylated tumors (LINE-1 methylation level, 24.4% for case #1; 31.4% for case #2; and 33.5% for case #3) and three LINE-1–hypermethylated tumors (65.3% for case #4; 65.9% for case #5; and 79.2% for case #6). The analyzed cases were selected based on sufficient amount of freshly frozen tissues and sufficient DNA quality. Their clinical and pathologic characteristics are shown in Supplementary Table S1. Pyrograms for cases #1 and #4 are shown in Fig. 1A.

The average number of chromosomal aberrations in 6 ESCC tumors was 168.2 (range, 24–423). Interestingly, LINE-1–hypomethylated tumors showed higher frequency of chromosomal gain or loss compared with LINE-1–hypermethylated tumors. Figure 2A shows a summary of DNA copy number aberrations for LINE-1–hypomethylated tumors and LINE-1–hypermethylated tumors. The number of aberrations was significantly higher in LINE-1–hypomethylated cases (median, 310.0) than in LINE-1–hypermethylated cases (median, 26.3; P = 0.0076; Fig. 2B). A similar result was observed for total length of aberrations (P = 0.0013; Fig. 2B). Regions with an abnormal number of chromosome copies in all hypomethylated tumors are shown in Table 1. Genomic gains were much more common than losses, and genomic imbalance was not found in chromosome 13, 14, 15, or 16. Amplification of 7q21–22 containing a candidate oncogene CDK6 was observed in all LINE-1–hypomethylated tumors, whereas this locus was not amplified in all LINE-1–hypermethylated tumors. Previous studies have shown that CDK6 amplification is associated with CDK6 overexpression and poor survival in esophageal cancer, supporting that CDK6 might mark malignant esophageal cancer (32, 33). From these findings, we further hypothesize that global DNA hypomethylation in ESCC contributes to aggressive phenotype by genomic gain of oncogenes such as CDK6. Thus, the correlation between CDK6 aberrant expression (protein and mRNA levels) and LINE-1 hypomethylation in ESCCs was further investigated.

We measured CDK6 mRNA expression levels by qRT-PCR in 15 ESCCs tissues and matched normal esophageal mucosa. Cancer tissues showed significantly higher levels of CDK6 mRNA expression than matched esophageal mucosa (P = 0.0046; Fig. 3A). Then, CDK6 expression at the protein level in ESCCs was evaluated by immunohistochemistry. CDK6 immunoreactivity was absent in normal squamous cell epithelium (Fig. 3B-a). Among 40 tumors, 24 tumors (60%) showed CDK6-positive expression (Fig. 3B-b), and 16 tumors (40%) showed CDK6-negative expression (Fig. 3B-c). These observations of CDK6 overexpression may demonstrate that CDK6 has vital roles in ESCC tumorigenesis. To evaluate the impact of LINE-1 hypomethylation on CDK6 expression, we assessed the association between CDK6 expression and LINE-1 methylation levels in ESCC tissues. In 20 ESCCs for which LINE-1 data and RNA were available, CDK6 mRNA expression levels were inversely associated with LINE-1 methylation levels (P = 0.039; Fig. 3C). Tumors with CDK6-positive immunoreactivity exhibited significantly lower LINE-1 methylation levels (median, 42.3; mean, 45.1; SD, 13.1) than tumors with negative immunoreactivity (median, 60.1; mean, 60.0; SD, 12.8; P = 0.0001 by the paired t test; Fig. 3D). These results collectively suggest that LINE-1 methylation level (and therefore global DNA methylation level) might have an influence on CDK6 expression through genomic gain of CDK6.

We then tested this correlation between LINE-1 methylation levels and CDK6 expression levels using nine ESCC cell lines. LINE-1 methylation levels in nine ESCC cell lines were widely distributed (range, 22.8–67.0). The inverse relationship between LINE-1 methylation levels and CDK6 mRNA expression levels was observed (P = 0.028; Fig. 4A). Furthermore, to confirm the gene amplification of CDK6 in ESCCs, we applied the FISH analysis for TE11 cell line (black dot in Fig. 4B), which demonstrated lower levels of LINE-1 methylation and high levels of CDK6 mRNA expression in vitro. We counted the number of signals for CDK6 and chromosome 7 centromere (control) in 100 TE11 cells. In 91 of 100 cells, the number of signals for CDK6 was larger than that for chromosome 7 centromere (Supplementary Table S2), confirming the presence of CDK6 amplification in ESCC cells.

To understand the relationship between CDK6 and tumor malignant behavior in ESCC, we examined the prognostic role of CDK6 in 129 cases with available LINE-1 methylation data and CDK6 expression data. All cases received no preoperative treatment. The relationships between levels of LINE-1 methylation and clinical and pathologic features are summarized in Supplementary Table S3. The median follow-up time for censored patients was 2.7 years. Patients with tumor expressing CDK6 (n = 61) experienced lower disease-free survival rates than patients not expressing CDK6 (n = 68; log-rank P = 0.043, Fig. 5A). This result certainly implies that CDK6 may have oncogenic roles in ESCC and contribute to ESCC progression.

We next examined whether the influence of LINE-1 hypomethylation on patient prognosis was modified by CDK6 expression status. In this current study, LINE-1 hypomethylation was defined as “<55.5” according to our previous article (23). In Kaplan–Meier analysis, patients with LINE-1 hypomethylation (n = 31) experienced significantly shorter disease-free survival compared with LINE-1–hypermethylated cases (n = 98; log-rank P = 0.0005; Fig. 5B). In univariate Cox regression analysis, compared with LINE-1–hypermethylated cases, patients with LINE-1 hypomethylation experienced a significantly higher disease recurrence rate [HR, 2.69; 95% confidence interval (CI), 1.48–4.78; P = 0.0015; Fig. 5C]. In the CDK6-adjusted Cox model, the HR of LINE-1 hypomethylation for disease recurrence was decreased to 2.25 (95% CI, 1.22–4.04). Thus, this result shows the proportional reduction in the regression coefficient for LINE-1 methylation due to the inclusion of CDK6 expression in the Cox regression model. In the multivariate Cox model adjusted for sex (male vs. female), age at surgery (<64 vs. 65≤), tumor location (upper vs. lower), tumor stage (stage I vs. II, III), histologic grade (G1 vs. G2–4), and CDK6 expression (positive vs. negative), LINE-1 hypomethylation was found to be associated with a significantly higher recurrence rate (multivariate HR, 2.43; 95% CI, 1.19–4.89; P = 0.016).

Given that cyclin D1 is the positive regulatory partner of CDK6, we further examined the relationship between cyclin D1 expression, LINE-1 methylation, and patient outcome. No relationship was observed between cyclin D1 expression and level of LINE-1 methylation (P = 0.94; Supplementary Fig. S2A and S2B). We also found no association between status of cyclin D1 expression and disease-free survival (log-rank P = 0.60; Supplementary Fig. S2C). Furthermore, the prognostic impact of LINE-1 hypomethylation was not reduced by cyclin D1 expression (cyclin D1-adjusted HR, 2.74; 95% CI, 1.51–4.88).

We previously demonstrated the correlation between LINE-1 hypomethylation (i.e., global DNA hypomethylation) and poor prognosis in ESCCs (23). In this current study, to clarify the mechanism by which LINE-1 hypomethylation affects ESCC aggressive behavior, we investigated LINE-1–hypomethylated and LINE-1–hypermethylated tumors using CGH array. We found that LINE-1–hypomethylated tumors presented highly frequent genomic gains at various loci containing candidate oncogenes including CDK6. The association between CDK6 amplification, CDK6 expression, and LINE-1 methylation levels in ESCCs was also observed. CDK6 expression was significantly associated with unfavorable prognosis in 128 patients with ESCC. Moreover, we have shown the proportional reduction in the regression coefficient for LINE-1 methylation due to the inclusion of CDK6 expression in the Cox regression model. Collectively, CDK6 may be an intermediate factor explaining the relationship between LINE-1 hypomethylation and poor prognosis. Global DNA hypomethylation in ESCC might contribute to the acquisition of tumor aggressive behavior through genomic gains of oncogenes such as CDK6 (Fig. 5D).

LINE-1 represents a major repetitive element and occupies approximately 17% of the human genome. Thus, the level of LINE-1 methylation is regarded to be a surrogate marker of global DNA methylation (13). Tumor LINE-1 hypomethylation has been associated with inferior prognosis in not only ESCC but also many different human tumor types such as colon cancer, glioma, ovarian cancer, lung cancer, and gastric cancer (17–22). In addition, LINE-1 hypomethylation was associated with clinically aggressive disease in patients with prostate cancer and gastrointestinal stromal tumors (16, 34). Nonetheless, the mechanism by which tumoral LINE-1 hypomethylation correlates with malignant phenotype or patient prognosis in human cancers remains to be fully explored. Experimental evidence supports the relationship between LINE-1 hypomethylation (i.e., global DNA hypomethylation) and genomic instability. A study using Nf1^+/−p53^+/− mice has shown that introduction of a hypomorphic Dnmt1 allele causes genome-wide DNA hypomethylation that leads to significant increases in the loss of heterozygosity rate and tumor development (35). Another study has shown that genomic hypomethylation in Apc^Min/+ mice leads to increases in microadenoma formation through loss of heterozygosity at the Apc locus (36). Thus, we carried out an array CGH analysis to assess the potential implication of LINE-1 hypomethylation in genomic instability. Interestingly, LINE-1–hypomethylated tumors showed highly frequent genomic gains at various loci containing numerous candidate oncogenes. Our finding is likely consistent with aforementioned experimental results.

In recent years, high-resolution array-based CGH has been applied to identify target oncogenes and tumor suppressor genes through defining recurrent gains and losses in various cancers (37). Two studies with CGH array analysis for ESCC samples have revealed high-level amplifications at 3q27.1, 7p11, 8q21.11, 8q24.21, 11q13.3, 11q22, 12q15–q21.1, 18q11.2, and 19q13.11–q13.12, and homozygous deletions at 4q34.3–q35.1 and 9p21.3 (38, 39). Bandla and colleagues conducted a comprehensive analysis of DNA copy number abnormalities in both ESCCs and esophageal adenocarcinoma. They reported a number of oncogenes that were amplified in both histologic types, including CDK6, MCL1, EGFR, SMURF1, KRAS, ERBB2, CCNE1, VEGFA, MET, and IGF1R (32). Among these genes, CDK6 amplification has been associated with CDK6 overexpression and a poor survival in esophageal cancer (33), supporting that CDK6 might mark esophageal cancer with aggressive biologic behavior. We also found that CDK6 expression was related with a poor prognosis in ESCCs. These results certainly imply that CDK6 has oncogenic roles in ESCC and contributes to ESCC progression. Our finding of the correlation between LINE-1 hypomethylation, CDK6 amplification, and CDK6 aberrant expression suggests that LINE-1–hypomethylated tumors might acquire malignant aggressive behavior by increasing genomic amplification and by altering the expression levels of candidate oncogenes represented by CDK6. However, other mechanisms may also be involved. In addition to its role as a surrogate marker for global DNA methylation, LINE-1 methylation status by itself likely has biologic effects, because retrotransposons such as LINE-1 elements can provide alternative promoters and contribute to noncoding RNA expression, which regulates the functions of a number of genes.

CDK6, which belongs to a family of serine–threonine kinases, is an important regulator of G₁–S-phase progression. Together with its binding partner cyclin D1, CDK6 forms active complexes that promote cellular proliferation by phosphorylating and inactivating the retinoblastoma protein (40). CDK6 amplification has been previously reported in human neoplasms including esophageal cancer, in which (in most cases) it is correlated with an adverse prognosis (33). Our current study revealed a relationship between CDK6 expression and LINE-1 hypomethylation (i.e., global DNA hypomethylation) and poor prognosis, suggesting that aberrant expression of CDK6 might be epigenetically regulated, and mark an aggressive type of esophageal cancer. Importantly, the current study found no relationship between cyclin D1 expression and LINE-1 hypomethylation, suggesting that, with regard to G₁/transition regulation, cyclin D1 is regulated by an entirely different mechanism from that of CDK6.

Interestingly, CDK6 is capable of blocking cell differentiation apart from the classical role in promoting cell proliferation (41, 42). This uniqueness makes it an attractive candidate for the development of small-molecule inhibitors. PD-0332991 is a recently developed specific inhibitor targeting CDK4/6-specific phosphorylation of retinoblastoma protein, and is currently in phase I clinical trials for advanced cancers (43). This small-molecule inhibitor has been shown to potently suppress proliferation and anchorage-independent growth of esophageal cancer cell lines (33). Hereafter, CDK6 expression in the resected specimens might attract increasing attention as a biomarker for patient selection. In this respect, our findings may have clinical implications.

In summary, LINE-1–hypomethylated ESCC tumors presented highly frequent genomic gains at various loci containing CDK6. In addition, we found that LINE-1 methylation levels were associated with CDK6 expression in ESCCs and that the prognostic impact of LINE-1 hypomethylation in patients with ESCC might be attenuated by CDK6 expression. Collectively, these findings may suggest that global DNA hypomethylation in ESCC might contribute to the acquisition of aggressive tumor behavior through genomic gains of oncogenes such as CDK6. Future studies are needed to confirm our findings, and also to examine other potential mechanism(s) by which genome-wide DNA hypomethylation affects tumor behavior.

No potential conflicts of interest were disclosed.

Conception and design: Y. Baba, H. Baba

Development of methodology: Y. Baba, H. Baba

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Baba, A. Murata, H. Shigaki, K. Miyake, H. Baba

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Baba, M. Watanabe, T. Ishimoto, K. Sakamaki, H. Baba

Writing, review, and/or revision of the manuscript: Y. Baba, M. Watanabe, H. Baba

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Shigaki

Study supervision: M. Watanabe, M. Iwatsuki, N. Yoshida, E. Oki, M. Nakao

This work was supported in part by a Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (JSPS; grant number 24659617).

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.