LINE-1 hypomethylation in gastric cancer, detected by bisulfite pyrosequencing, is associated with poor prognosis

A total of 247 consecutive patients with gastric cancer who were undergoing resection at Kumamoto University Hospital between April 2005 and December 2009 were enrolled in this study. Nineteen patients were excluded for reasons of unavailability of adequate tissue samples. Because 22 patients received preoperative treatment, they were excluded from this study. Thus, we initially quantified LINE-1 methylation in 206 cancer specimens and obtained valid results in 203 (99 %) of the cases. Thus, a total of 203 gastric cancers were finally included in this study, and 74 cases were randomly chosen from these 203 cases to evaluate LINE-1 methylation level in normal matched mucosa. Patients were observed at 1- to 3-month intervals until death or 30 June 2011, whichever came first. Tumor staging followed the American Joint Committee on Cancer Staging Manual (7th edition) [17]. Overall survival was defined as the time between the date of the operation and the date of death. In our cohort, the 3-year overall survival rates of patients treated by gastrectomy were 91.9 % for stage I, 79.0 % for stage II, 56.8 % for stage III, and 19.5 % for stage IV. These rates are similar to those from the Japanese Gastric Cancer Association nationwide registry (94.1 % for stage I, 78.4 % for stage II, 53.2 % for stage III, and 22.4 % for stage IV), certainly supporting the absence of bias in our database. Written informed consent was obtained from each subject, and the study procedures were approved by the institutional review board. The term “prognostic marker” was used throughout this article according to the REMARK Guidelines [18].

Hematoxylin and eosin (H&E)-stained slides of the tumors were reviewed, and areas of tumors and histologically normal gastric mucosae adjacent to tumors were marked by one pathologist (Y.B.). H&E-stained tissue sections of the largest cross-sectional slice (depending on tissue and tumor size; on average, large tumor tissue 10 μm × 1 section) from each case were scraped off slides for DNA extraction. Genomic DNA was extracted from the tumor and normal epithelium. 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 et al., using the PyroMark kit (Qiagen) [14, 19, 20]. This assay amplifies a region of LINE-1 element (position 305–331 in accession no. X58075), which includes four CpG cites. The PCR conditions were 45 cycles of 95 °C for 20 s, 50 °C for 20 s, and 72 °C for 20 s, followed by 72 °C for 5 min. The biotinylated PCR product was purified and made single-stranded to act as a template in a pyrosequencing reaction, using the Pyrosequencing Vacuum Prep Tool (Qiagen). Pyrosequencing reactions were performed in the PyroMark Q24 System (Qiagen). The nucleotide dispensation order was ACT CAG TGT GTC AGT CAG TTA GTC TG. The non-CpG cytosine in LINE-1 repetitive sequences has been documented to be rarely methylated. Thus, complete conversion of cytosine at a non-CpG site ensured successful bisulfite conversion. The amount of C relative to the sum of the amounts of C and T at each CpG site was calculated as the percentage (i.e., 0–100). The average of the relative amounts of C in the 4 CpG sites was used as the overall LINE-1 methylation level in a given tumor (Fig. 1). In published literature, we have validated our LINE-1 methylation pyrosequencing assay; we have performed bisulfite conversion on five different DNA specimen aliquots and repeated PCR-pyrosequencing five times using four macro-dissected cancers. Bisulfite-to-bisulfite (between-bisulfite treatment) standard deviation (SD) ranged from 1.4 to 2.9 (median, 2.3), and run-to-run (between-PCR pyrosequencing run) SD ranged from 0.6 to 3.3 (median, 1.2) [21]. In this study, we used “LINE-1 methylation level” for LINE-1 methylation as a continuous variable and “LINE-1 hypomethylation” for LINE-1 methylation as a categorical variable (i.e., hypomethylation vs. hypermethylation).

For the statistical analyses, we used the JMP (Version 9; SAS Institute, Cary, NC, USA) and the SAS software programs (Version 9.1; SAS Institute). All p values were two sided. To compare the means, we performed the t test assuming unequal variances. For the survival analysis, the Kaplan–Meier method was used to assess the survival time distribution, and the log-rank test was used. To assess the independent effect of the LINE-1 methylation level on mortality, the tumor stage (I, II, III + IV) was used as a stratifying (matching) variable in Cox models using the “strata” option in the SAS “procphreg” command to avoid residual confounding and overfitting. We constructed a multivariate, stage-stratified Cox proportional hazard model to compute a hazard ratio (HR) according to LINE-1 methylation status, containing sex (male vs. female), age at surgery (continuous variable), tumor location (lower vs. middle or upper), and histological type (intestinal vs. diffuse). A backward stepwise elimination with a threshold of p = 0.20 was used to select variables in the final model. We initially performed the Cox regression analysis with LINE-1 methylation as a continuous variable and then performed the Cox regression analysis with LINE-1 methylation as a categorical variable. An interaction was assessed by including the cross product of the LINE-1 variable and another variable of interest in a multivariate Cox model; thereafter, the Wald test was performed.

We first examined LINE-1 methylation level in 74 gastric cancer tissues and matched noncancerous mucosa samples. The cancer tissues exhibited significantly lower levels of LINE-1 methylation [median 74.9, mean 72.3, SD 10.1 (all in 0–100 scale)] than matched noncancerous mucosa (median 79.4, mean 79.2, SD 5.6) (p < 0.0001 by the paired t test) (Fig. 2a).

Next, we quantified the LINE-1 methylation in 206 cancer specimens and obtained valid results in 203 (99 %) of cases. LINE-1 methylation levels in the 203 cancers (Fig. 2b) were approximately normally distributed: mean 71.4, median 74.4, SD 12.9, range 11.6–97.5; inter-tertile range 70.0–77.4 (all in 0–100 scale). The LINE-1 methylation level was then divided into tertiles [Ter1 (77.4–97.5, n = 68), Ter2 (70.1–77.3, n = 66), Ter3 (11.6–70.0, n = 69)] for further analyses. We found that the LINE-1 methylation level was associated with tumor stage (p = 0.039; Table 1). However, in the analysis with LINE-1 methylation as a continuous variable, there was no significant relationship between LINE-1 methylation level and tumor stage (p = 0.64, Fig. 3). LINE-1 methylation was not significantly associated with other clinical or pathological variables.

During the follow-up of the 203 patients, there were a total of 56 deaths. The median follow-up time for censused patients was 2.9 years. The primary statistical survival analysis was the Cox regression test with LINE-1 methylation as a continuous variable. LINE-1 hypomethylation was associated with a statistically significant increase in overall survival rate (univariate analysis p = 0.014). The univariate hazard ratio for overall survival rate associated with a 20 % decrease in LINE-1 methylation was 1.96 [95 % confidence interval (CI) = 1.33–2.87]. We also performed analyses using categorical variables (i.e., tertile). In a univariate Cox regression analysis, compared to first tertile (Ter1) cases, the third tertile (Ter3) cases experienced a significantly lower overall survival rate (p = 0.017, HR 2.24, 95 % CI 1.15–4.63). The second tertile (Ter2) cases experienced a slightly, but not significantly, lower overall survival rate compared to Ter1 cases (p = 0.12, HR 1.76, 95 %CI 0.85–3.74) (Table 2; Fig. 4). Based on these results, we made a dichotomous LINE-1 methylation variable (i.e., hypomethylation vs. hypermethylation), defining Ter1 as the “hypermethylated group” and combining Ter2 and Ter3 into the “hypomethylated group.” Thus, in this study, “LINE-1 hypomethylation” was defined as “≤77.3 %” and “LINE-1 hypermethylation” was defined as “≥77.4 %.”

In the Kaplan–Meier analysis, LINE-1 hypomethylators (i.e., Ter2 and Ter3 cases) experienced significantly shorter overall survival (log rank p = 0.029) than those with hypermethylation (Fig. 4). In the univariate Cox regression analysis, compared to LINE-1 hypermethylated cases, LINE-1 hypomethylators experienced a significantly lower overall survival rate (HR 2.01, 95 % CI 1.09–3.99, p = 0.023) (Table 2). In the multivariate Cox model adjusted for the clinical and pathological features, LINE-1 hypomethylation was found to be associated with a significantly lower overall survival rate (multivariate HR 1.98, 95 % CI 1.04–4.04, p = 0.036). Another independent prognostic factor was histological diffuse type (multivariate HR 1.91, 95 % CI 1.09–3.29, p = 0.023), whereas neither sex, age, nor tumor location was significantly associated with overall survival rate.

We also examined whether the influence of LINE-1 hypomethylation on the overall survival was modified by any of the clinical and pathological variables. We did not observe a significant effect of modification by any of the covariates in survival analysis (all p interaction >0.25). Notably, there was no significant interaction between LINE-1 methylation and tumor stage (p interaction = 0.68 for stage I, II vs. III, IV; p interaction = 0.97 for stage I vs. II–IV).