High‐risk human papillomavirus infection associated with telomere elongation in patients with esophageal squamous cell carcinoma with poor prognosis

Telomere maintenance is crucial in carcinogenesis and tumor progression. The results of a previous study from the authors indicated that infection with high‐risk human papillomavirus (HR‐HPV) types 16, 18, and 58 was a risk factor for esophageal squamous cell carcinoma (ESCC) in the Shantou region of China. In the current study, the authors explored the association between HR‐HPV infection, telomere length (TL), and DNA methylation and their significance in the prognosis of patients with ESCC.


INTRODUCTION
Esophageal squamous cell carcinoma (ESCC) has a high mortality rate; its incidence varies greatly worldwide by geographic region and etiology. 1,2 Human papillomavirus (HPV), especially high-risk HPV (HR-HPV) such as HPV type 16 (HPV-16), HPV-18, and HPV-58, has been implicated as a risk factor for ESCC, especially in high-incidence regions. 3 We previously reported that HPV DNA or protein has been repeatedly found in ESCC tissue (approximately 50%-60%) from patients in the Shantou coastal area in southeastern China for 2 decades. [3][4][5][6][7][8] Although HPV continues to be found to be associated with ESCC, the underlying HPV oncogenic mechanisms and their clinical significance are not completely understood.
Telomeres, composed of tandem repeats of DNA and associated proteins, are important for chromosome integrity and stability to escape replicative senescence and for indefinite proliferation and are progressively shortened during each cell replication cycle. 9 Increasing evidence has confirmed that shortened telomere length (TL) and telomere dysfunction may be one of the earliest and most prevalent genetic alterations in the multistep process of malignant transformation, including in ESCC. [10][11][12][13] For example, a clinical study found a negative relation between increased tumor-to-normal telomere restriction fragment length ratio and telomerase activity, which was linked with poor prognosis in patients with ESCC. 13 Furthermore, telomerase activity was reported in an in vitro study of carcinogenesis of esophageal epithelial cells induced by HPV-18 infection. TL sharply decreased from normal fetal esophagus cells to passage 35 but was stably expressed in cells up to passage 100, and telomerase activity was consistent in cells from passages 35 to 100. 14,15 Maintenance of TL by human telomerase reverse transcriptase (hTERT) is critical to preserve the replicative potential of cancer cells. 9 Telomeric repeat-binding factor 2 (TRF2), as a telomere-binding protein, is required to prevent mammalian telomeres from activating DNA damage response, because inhibition of TRF2 could induce end-to-end chromosome fusion and growth arrest or apoptosis. 16 DNA methylation status, one of the genomic epigenetic conditions, in the telomeric region is altered in response to changes in TL. It is interesting to note that DNA methylation of transcriptionally repressive sequences in the hTERT or TFR2 promoter is associated with their upregulation in some cancers, including HPVimmortalized cells. 17,18 Moreover, subtelomeres, neighboring regions to telomeres, have features of constitutive heterochromatin, including H3K9 and H4K20 trimethylation, CBX3 binding, and DNA hypermethylation. 19 Mammalian telomeres and subtelomeres have features similar to pericentromeric regions, with a high density of DNA repeats. However, subtelomeres but not telomeres demonstrate high density of CpG sequences that are susceptible to methylation by DNA methyltransferases. 20 Recent evidence has suggested that subtelomeric methylation at certain regions was related to telomere shortening, thus indicating an association between subtelomeric chromatin structure and TL regulation in human hepatocarcinogenesis. 21 However, to the best of our knowledge, epigenetic modifications of human subtelomeres and their impact on the regulation of TL and telomere function remain uncharacterized in ESCC.
In our previous study, compared with cervical squamous cell carcinoma, ESCC samples demonstrated a relatively low HR-HPV DNA viral load, which did not affect the high presentation of HPV integration status and expression or association with tumor grade. 5 In the current study, we aimed to evaluate the role of HPV infection in TL and methylation patterns of telomeric-related genes (hTERT and TFR2) and subtelomeric DNA. We also examined the association with clinical pathological variables, including age, sex, smoking, alcohol consumption, lymph node metastasis, family history, pathologic stage, and outcome of ESCC in the Shantou coastal area of China, a region with a high incidence of ESCC.

Ethical Statement
This study was performed according to the Declaration of Helsinki and was approved by the Ethics Committee of the Affiliated Tumor Hospital of Shantou University Medical College and Peking Union Medical College Hospital. All patients were enrolled in the Shantou region of China and provided signed informed consent for the use of biological samples.

Patients and Tissue Samples
We selected 70 of 106 cases of ESCC according the modified inclusion criteria reported previously [5][6][7]22 : 1) all patients provided signed informed consent for the use of biological samples; 2) all patients had histological proof of ESCC; 3) all patients had undergone complete surgical resection (R0); 4) patients had received no neoadjuvant or adjuvant treatment; 5) full HPV infection data were available for all patients; 6) complete follow-up data were available for all patients; and 7) all patients lived in the Shantou region.
We obtained DNA samples of 70 cases of fresh ESCC and matched adjacent noncancerous tissues (> 5 cm away from malignant lesions, confirmed by microscopically negative margins) from patients who underwent surgery in the Department of Thoracic Surgery in the Affiliated Tumor Hospital of Shantou University Medical College from 2002 to 2006, as well as 50 cases of normal esophageal tissues from subjects who died of trauma (in the hospital or daily life; mainly traffic accidents and regular daily labor) or cardio-cerebrovascular disease. DNA samples were examined for TL by real-time polymerase chain reaction (PCR) and for hTERT, TFR2, or subtelomeric DNA methylation by methylationspecific PCR (MSP). All patient information including age, sex, and clinical characteristics; characteristics of HR-HPV-16, HR-HPV-18, and HR-HPV-58 infection; and tissue DNA extraction and storage are shown in Table 1.
Follow-up study for all patients with ESCC was performed at regular time intervals (3 months). Survival status was verified and updated from the records of the Tumor Registry in the Department of Thoracic Surgery at the Affiliated Tumor Hospital as of June 2012. Survival time was defined as the number of months from the completion of primary treatment until the patient demonstrated any clinical or radiological evidence of local or regional disease or distant metastasis at the time of the last follow-up.
Patients with ESCC were followed for a mean of 37.9 months (range, 2 months-84 months) according to a standardized protocol. In all, 29 patients who did not demonstrate disease recurrence were alive at the end of the follow-up period; 41 patients (58.6%) died due to hepatic Original Article or lung metastasis, disease recurrence, or further complications at 2 to 48 months after undergoing surgery for ESCC.

Methylation-Specific PCR
ESCC genome DNA was extracted with use of the TIA-Namp Genomic DNA Kit (Tiangen Biotech, Beijing, China) and modified by use of the BisulFlash DNA Modification Kit (Epigentek Group Inc, Farmingdale, NY) as previously described. 23,24 The modified DNA was used as a template for the methylation assay. The subtelomeric (7q, 18q, 9q, 17q, 21q, and XqYq) MSP procedure and primers were previously described 21,25 and hTERT and TRF2 PCR was previously described. 17,18 MSP replication was repeated twice for each sample. We used human placental genomic DNA (gDNA; Sigma-Aldrich Corporation, St. Louis, Mo) methylated in vitro with CpG Methyltransferase [M.SssI, New England Biolabs (UK) Ltd] and converted as a fully methylated (100%) methylation-positive control and the same unmethylated placental gDNA converted with sodium bisulfite as a methylation-negative control. PCR products were separated on agarose gels, and bands were observed by ethi-dium bromide staining. The methylation pattern represented the ratio of DNA methylation calculated as methylation/(methylation and unmethylation) 3 100%, which was quantified by densitometry with use of Image Analysis Software (Image J, National Institute of Health). Positive PCR products were ligated into the pGEM-T vector and confirmed by sequencing. In the current study, relatively high density ( 50%) of methylation patterns was considered hypermethylation, and relatively low density (< 50%) was considered hypomethylation. 26

Measurement of TL by Quantitative Real-Time PCR
Mean TL was measured by quantitative PCR as described. 27 Briefly, the relative ratio of the copy number of telomere (T) repeats to that of a single gene (S; 36B4 gene) (T/S ratio) was established for each sample with standard curves. The ratio of each sample was then normalized to a reference DNA (human b-globin) to standardize the differences between runs. A reference DNA sample (transformed human epithelial kidney 293T cells) was paired with each measurement to control for interassay variability. A dilution series range from 1.56 to Abbreviations: ESCC, esophageal squamous cell carcinoma; HR-HPV, high-risk human papillomavirus; NT-TL, nontumor telomere length; T/NT-TL, ratio of tumor telomere length to nontumor telomere length; T-TL, tumor telomere length. a Data are shown as the medians (interquartile ranges). b P <.05 (statistically significant) when compared between indicated group and the first factor for each characteristic, which was calculated using the Mann- 100.00 nanograms (ng) (2-fold dilution; 7 points) in telomere and b-globin quantitative PCR was linear (correlation coefficient [R 2 ] 5 0.98). All analyses were performed with blinding to disease status. Each DNA sample was measured in triplicate. The mean standard deviation for these measurements was 6.8%. Telomere, single-gene, and b-globin PCR involved use of the LightCycler 480II (Roche, Basel, Switzerland). Water was used as a negative control. T/S values (x) were converted to kilobases (kb) (y) with the adjusted formula of y 5 1.53x 1 3.12 as described. 28 Statistical Analysis Statistical analyses involved the use of SPSS statistical software (version 16.0; SPSS Inc, Chicago, Ill). The normality of variables was assessed, and the mean HR-HPV E6 levels (determining the viral load) and the mean ratio of HR-HPV E2 to E6 gene levels (determining the degree of viral DNA integration) were log-transformed. Spearman correlation was used to analyze the association between clinicopathologic variables and TL expressed as the T/S value. The Mann-Whitney 2-sample U test was used to assess differences in TL by human esophageal carcinogenesis status. A P value < .05 was considered to be statistically significant. Survival was estimated by the Kaplan-Meier method, with comparison by log-rank statistics. For survival analysis, univariate and multivariate analysis involved a Cox proportional hazards model to identify the clinical variables (age, sex, smoking history, alcohol consumption, lymph node metastasis, tumor grade, family history, TL, and HPV-L1 positivity) associated with survival. Compared with NT-TL, T-TL was longer in those samples that were positive for HPV-L1 (represented as low-risk HPV and HR-HPV) or HR-HPV (P 5 .031 for HPV-L1 and P 5 .007 for HR-HPV) ( Fig. 1B and C). NE-TL, T-TL, and NT-TL did not differ with regard to infection with HPV-16, HPV-18, or HPV-58 (data not shown). The results of a previous study demonstrated relatively low HR-HPV-16, HR-HPV-18, and HR-HPV-58 copy number (2.55 copies/cell 6 3.19 copies/cell) and high HR-HPV integration (represented as ratio of E2 to E6/E7 genes) in ESCC. 5 In the current study, T-TL was found to be positively correlated with viral load (R, 0.410; P 5 .037) but was negatively correlated with the ratio of E2 to E6/E7 genes (r, 20.492; P 5 .01) (Figs. 1D and E). Relatively long T-TL and high T/NT-TL were frequently found with high viral load and integrated status (the integration form was defined by the ratio of E2 to E6/E7 genes of < 0.001). The episomal form was equally of E2 and E6/E7 gene copies. Ratios between 0.001 and the lower episomal ratios indicated the mixed form of HR-HPV in ESCC tissue. It was found that integrated HR-HPV infection noted in 10 patients with ESCC had the longest T-TL among all physical status types of HR-HPV infection. We also found high T/NT-TL in 32 of 35 ESCC samples with integrated or mixed forms of HR-HPV infection compared with 3 of 35 cases with episomal HR-HPV infection (Table 1).

HR-HPV
hTERT, TRF2, and Subtelomeric DNA Methylation Patterns in ESCC Patients Infected With HPV Given that hTERT, TRF2, and subtelomeric methylation patterns dynamically change and are associated with TL during carcinogenesis, we explored DNA methylation status by MSP assay in HPV-related ESCC specimens (Fig.  2). We found a hypermethylation ratio ( 50%) for hTERT, 7q, 9q, and XqYq but a hypomethylation ratio (< 50%) for TRF2, 17q, and 21q, as well as a wider distribution of the methylation ratio for TRF2, 7q, and 18q according to the standard deviation of the mean (Fig. 3A). HPV-L1 positivity increased the ratio of hTERT and 21q methylation, and HR-HPV positivity increased the ratio of hTERT, 18q, and 21q methylation and decreased the ratio of TFR2 methylation (Figs. 3B and 3C). However, HPV infection did not affect the ratio of 7q, 9q, 17q, and XqYq methylation. Furthermore, the ratio of HR-HPV E2 to E6/E7 genes was found to be negatively correlated with

Relationship Between hTERT, TRF2, and Subtelomeric DNA Methylation and TL
Previous research has shown that subtelomeric methylation at certain regions was related to telomere lengthening or shortening in hepatocarcinogenesis and several cancer cells. 25,29 Therefore, we evaluated hTERT, TRF2, and subtelomeric DNA methylation changes in long ( 0.70) and short (< 0.70) T-TL groups as well as high ( 0.80) and low (< 0.80) ratios of T/NT-TL. High hTERT and 21q methylation was more prevalent with long compared with short T-TL and was positively correlated with T-TL. Conversely, long T-TL was associated with reduced 7q methylation (Figs. 3D, 3G, and 3H). TRF2, 8q, 17q, 18q, and XqYq methylation was not found to be related to T-TL. Unexpectedly, neither hTERT, TRF2, nor subtelomeric methylation was found to be associated with T/ NT-TL (data not shown). Table 1 presents the association between clinicopathologic variables and TL in patients with ESCC. Overall, T/NT-TL was higher in patients with ESCC with a history of smoking, HPV infection, or grade 3 tumor. The T/NT-TL ratio was lower for surviving than nonsurviving patients. We then assessed the impact of TL on survival by clinicopathologic factors. We considered 2 groups after comparing mean levels across quartiles: T-TL 0.70 and < 0.70 and T/NT-TL ratio 0.80 and < 0.80. Univariate Cox regression analysis revealed significantly worse outcomes for patients with grade 3 tumors (P for hazard ratio 5 .013), HR-HPV infection (P HR 5 .037), and a high T/NT-TL ratio (P HR 5 .014), with no effect noted for age, sex, smoking history, alcohol consumption, lymph node metastasis, family history, or HPV-L1 positivity (Table 3). This result was confirmed by Kaplan-Meier analysis (Fig. 4), which revealed significantly worse survival for patients with a T/NT-TL ratio 0.80 (P 5 .011), those with HR-HPV infection (P 5 .03), pathologic tumor grade (P 5 .033), and hTERT hypermethylation (P 5 .03) but not HPV-L1 positivity or a T-TL ratio 0.70.

HR-HPV-Related Positively High T/NT-TL Ratio Correlated With Poor Survival in Patients With ESCC
Furthermore, multivariate survival analysis with a Cox proportional hazards model revealed HR-HPV infection and a high T/NT-TL ratio as prognostic factors of survival (P 5 .027 and P 5 .043, respectively), independent of age, sex, smoking history, alcohol consumption, lymph node metastasis, and tumor grade. These findings suggest that a high T/NT-TL ratio as well as its confounders of HR-HPV infection and hTERT methylation might contribute to poor survival among patients with ESCC.

DISCUSSION
Our previous study suggested a significant association between HR-HPV expression and integration with ESCC tumor grade with risk of ESCC compared with cervical cancer. 5 In the current study, we identified the relationship between hTERT, TRF2, and subtelomeric DNA methylation patterns and TL in patients with ESCC related to HPV infection. The results of the current study revealed that 1) TL gradually decreased from NE mucosa to adjacent NT tissue to T tissue; 2) HR-HPV infection viral load and integration could elongate the TL in esophageal tumor tissue; 3) a high ratio of hTERT DNA methylation was positively correlated with HPV integration and TL; and 4) a high T/NT-TL ratio as well as its confounders of HR-HPV infection and hTERT methylation might contribute to poor survival among patients with ESCC in the Shantou region of China.
Telomere shortening is widely accepted as being crucial to the initiation of carcinogenesis by inducing chromosomal instability through the breakage-fusionbridge cycle of affected cells. 9,30 However, the effect of HPV on TL and its significance in clinical pathological parameters and outcome in patients with ESCC was to the best of our knowledge largely unknown. As Original Article previously described, 13 we detected a decrease in TL from the NE to adjacent noncancerous and tumor tissue in patients with ESCC, although widespread TL abnormalities were shown in nearly all examined tissue. HPV infection, especially HR-HPV infection and integration, could inhibit the shortened TL in tumor tissue of ESCC. Previous studies have found the integration of HPV DNA at the c-Myc locus in genital 31 and lung tumors 12 to be positively associated with c-Myc oncogene activation by promoting hTERT transcription and  [1] or negative [-]) and (D) long tumor TL (T-TL) ( 0.7 or not). Boxes and whiskers represent the 25th to 75th and 10 th to 90th percentiles, respectively; the median is the central line in each box. *P <.05, evaluated by the Mann-Whitney 2-sample U test. Correlation analysis of the hTERT and TRF2 methylation ratio with the ratio of (E and F) HR-HPV E2 to E6/E7 genes and (G and H) TL is shown.
maintaining TL. In addition, c-Myc upregulation could also be indirectly mediated by inactivation of p53 via elevated and consistent expression of HPV E6 protein after DNA integration. Hence, this evidence and the current study findings suggest that maintaining an optimal TL through hTERT transcription by HPV DNA integration may play an important role in cell proliferation and immortalization in HPV-related carcinogenesis. 14,15,32   Telomeres and subtelomeres are generally heterochromatic. The increased presence of a cluster of CpG sites in the hTERT, TRF2, and subtelomeric promoter region of genes prompted us to investigate the importance of DNA methylation and modifications in telomerase activation and length. Recently, certain regions of subtelomeric methylation patterns related to TL underwent dynamic changes during hepatocarcinogenesis and in other human cancer cell lines. 21,25 However, to the best of our knowledge, no studies to date have focused on the hTERT, TRF2, and subtelomeric DNA methylation patterns of telomerase dysfunction in ESCC with repeated detection of HPV. MSP assay revealed subtelomeric methylation status with a wide distribution of methylation ratios on 8 chromosomes in ESCC specimens. It is important to note that both hTERT and 21q methylation patterns revealed hypermethylation in HPV-infected cells and ESCC samples and a positive association with tumor TL. Conversely, a low level of 7q was associated with long telomeres. Nonetheless, 8q (27 kb), 9q (41 kb), 17q (3 kb-13 kb), 18p (2 kb), and XqYq (40 kb), which are closer to telomeres than 7q (104 kb), were not found to be related to TL in ESCC, which suggests that the distance between the subtelomeric target sites and telomeric chromatin was unlikely to be a critical factor, at least among patients with ESCC residing in the Shantou region of China. In addition, patterns of hTERT, TRF2, and subtelomeric methylation might be related to each other in ESCC. For example, methylation of TRF2 was negatively associated with hTERT or 21q, but there was positive concordance between 7q and 9q or 18q. Except for TRF2, the ratio of hTERT, 18q, and 21q methylation was increased in patients with HPV-infected ESCC. Moreover, subtelomeric regions far away from telomeres (eg, 8q, 17q, 18q, and 21q) tended toward more frequent hypomethylation than did 7q, 9q, and XqYq in the current study, which was consistent with data reported in hepatocarcinogenesis. 21 Although the subtelomeric methylation status in different regions is unclear, the results of the current study reveal a diverse role of telomeric chromatin in specific regions for telomere regulation in esophageal carcinogenesis.
A growing body of evidence has suggested an association between shortened telomeres and an increased risk of various human cancers. Indeed, an optimal ratio of tumor-to-normal telomere restriction fragment length was found to be an independent prognostic factor of overall survival in patients with neuroblastoma, 33 hepatocellular carcinoma, 34 non-small cell lung cancer, 35 colorectal carcinoma, 36 and ESCC. 13 The current study, which used Cox regression and Kaplan-Meier analyses, revealed poor survival for patients with high pathologic tumor grade, HR-HPV infection, hTERT hypermethylation, and a high T/NT-TL ratio ( 0.80). In addition, the T/NT-TL ratio was found to be an indicator of worse prognosis and independent of tumor grade, HPV infection, age, sex, smoking history, alcohol consumption, lymph node metastasis, and family history. The results of a previous study indicated that elevated levels of the main human telomerase components such as telomerase activity, hTERT, human telomerase RNA component, telomerase protein 1, c-Myc, TRF1, and TRF2 contributed to the regulation of telomerase expression in tumor tissues, which might result from progressive shortening of telomeres in esophageal carcinogenesis.
A limitation of the current study is that hTERT expression or its activity and their correlation with HPV expression was not detected in ESCC tissue because of a lack of direct evidence of HPV maintaining TL. However, some sporadic clues indicate that hTERT promoter hypermethylation is correlated with its upregulation in HPV-related cervical cancer cells 17 and transcriptional activation of hTERT by E6 oncoprotein is required for evidence of HPV-16/HPV-18-infected lung tumorigenesis. 12 In addition, in the current study, HR-HPV expression was found to be positively correlated with the pathological grade of ESCC and hTERT expression, and telomerase activity was markedly stimulated in tumor tissues, suggesting the critical role of HR-HPV infection in the elongation of TL in patients with ESCC.
The results of the current study provide a unique finding that HPV infection, especially HR-HPV infection and integration, is associated with telomerase elongation in patients with ESCC with a poor prognosis, a finding that appears to be related to hTERT, TRF2, and subtelomeric DNA methylation. A future larger cohort study is needed to validate the clinical observations of the current study, and more in-depth investigation is needed to examine whether the prevention of HR-HPV infection will have an impact on ESCC.