Long Interspersed Nuclear Element 1 Hypomethylation Is a Marker of Poor Prognosis in Stage IA Non–Small Cell Lung Cancer

Despite recent advances in our understanding of the molecular mechanisms of carcinogenesis and in the use of multimodal cancer therapy, lung cancer remains one of the major causes of cancer-related deaths worldwide. Surgery is still the major treatment for non–small cell lung cancer (NSCLC). Even in patients with pathologic stage I disease, the cumulative 5-year survival rate is ∼60% (1, 2). A growing body of clinical evidence indicates that adjuvant chemotherapy after surgery confers a survival benefit for NSCLC patients (3, 4). Therefore, accurate prognostic markers are required to help select the optimal treatment modality for individual lung cancer patients, including the use of adjuvant chemotherapy.

Aberrant methylation of CpG dinucleotides is a commonly observed epigenetic modification in human cancer (5). The two phenomena of global genomic DNA hypomethylation and hypermethylation of gene promoter regions occur in parallel and are observed in a wide variety of cancer types (6, 7). Hypermethylation of gene promoters is often associated with transcriptional silencing of tumor suppressors. Numerous studies have suggested possible clinical uses of promoter hypermethylation as markers of early diagnosis (8–11) and as predictors of patient outcome (12–15). However, there is no consensus about the genes to be analyzed for specific clinical applications. We previously reported that p16 methylation was a candidate prognostic marker in NSCLC patients (16). A recent study also showed that APC, CDH13, and RASSF1 methylations were promising markers for predicting the early recurrence of lung cancer (17). On the other hand, the molecular mechanisms that underlie global DNA hypomethylation in tumorigenesis are poorly understood, although an involvement with genomic instability has long been suggested (18, 19). Only a few studies have analyzed global hypomethylation in primary cancers with the aim of exploring its clinical importance as a molecular marker (20–22).

Long interspersed nuclear element 1 (LINE-1) represents a family of non–long terminal repeat retroposons that are interspersed throughout genomic DNA and comprise ∼18% of the human genome (23, 24). Because of their high frequency in the genome, LINE-1 methylation serves as a useful surrogate marker of global methylation (25). LINE-1 is heavily methylated in normal human tissues; however, loss of methylation is consistently observed in human cancers (26, 27) and accounts for a substantial proportion of the genomic hypomethylation observed in this disease. Most LINE-1 sequences in the human genome are truncated in the 5′ region or are mutated, making transposition impotent (23). However, ∼100 copies of full-length LINE-1 sequence are present and have the ability to transpose (28, 29). Hypomethylation in the promoter region of potent LINE-1 sequence causes transcriptional activation of LINE-1, resulting in transposition of the retro-element and chromosomal alteration. LINE-1 methylation status may therefore be a key factor linking global hypomethylation with genomic instability (30, 31). The association between LINE-1 methylation and genomic instability also suggests that it may be a good prognostic marker in cancer because previous studies have reported associations between genomic instability and the outcome of cancer patients (32–34).

Following our earlier demonstration of prognostic value for p16 methylation in NSCLC (16), the aim of the present study was to evaluate other candidate methylation markers as predictors of patient outcome. Methylation levels of the APC, CDH13, and RASSF1 gene promoters and of LINE-1 were quantitatively assessed in a large series of unselected NSCLC and matching normal lung tissues. These were analyzed in relation to clinicopathologic features and to patient outcomes. In addition, we investigated the loss of heterozygosity (LOH) in a subset of tumor samples to correlate LINE-1 hypomethylation with genomic instability.

Tumor samples were obtained from a consecutive series of 379 NSCLC patients who underwent surgery at Kanazawa University Hospital. Corresponding normal lung tissues were available for 333 of these patients. The patients comprised 248 males and 131 females and ranged in age from 13 to 83 y (mean, 64.3 y). Smoking history was obtained from the health interview questionnaire. Current and former smokers were classified as “smoker” and never smokers as “nonsmoker.” All tissue samples were fixed in formalin and embedded in paraffin followed by histologic diagnosis with H&E staining. Tissues for DNA isolation were dissected manually from formalin-fixed and paraffin-embedded tissue sections (10 μm thickness). After deparaffinization using xylene and ethanol, genomic DNA was isolated using the QIAamp DNA mini kit (Qiagen) according to the manufacturer's protocol. Approval for this project was obtained from the Kanazawa University Medical Ethics Committee.

DNA samples were subjected to bisulfite treatment using a CpGenome DNA Modification kit (Chemicon) according to the manufacturer's protocol. LINE-1 methylation was measured using a methylation-specific real-time PCR assay as previously described (27). Real-time reactions for unmethylated and methylated LINE-1 sequences were done simultaneously in a 96-well plate. The percentage of methylated LINE-1 was calculated using the formula: 100 × methylated reaction/(unmethylated reaction + methylated reaction). The level of promoter methylation for APC, CDH13, and RASSF1 was measured with a MethyLight assay as previously described (35, 36). The amount of Alu element, a primate-specific short interspersed nuclear element (ALU) product measured by methylation-independent reaction was used for normalization. Methylation values were calculated using CpG methylase (M.SssI)–treated genomic DNA as the constant reference sample and were expressed as a percentage of the methylated reference (PMR). Oligonucleotide sequences for primers and probes were as previously described (37). Real-time PCR was conducted using the ABI-PRISM 7900 Sequence Detection System (Applied Biosystems) and the Premix Ex Taq (TaKaRa Bio) following the protocol provided by the manufacturer.

LOH status was investigated by screening three microsatellite loci that flank the APC, P16, and P53 loci. DNA from tumor and matching normal tissues of 51 patients selected randomly was amplified with fluorescence-labeled primers. The PCR conditions were 35 cycles of 98°C for 20 s and 60°C for 60 s. The sets of primers used for specific amplification of the microsatellite sequences were as follows:

• D5S346 forward primer, FAM-ACTCACTCTAGTGATAAATCGGG

• D5S346 reverse primer, AGCAGATAAGACAGTATTACTAGTT

• D9S942 forward primer, FAM-GCAAGATTCCAAACAGTA

• D9S942 reverse primer, CTCATCCTGCGGAAACCATT

• TP53 forward primer, FAM-TGCCCCATTCCCCTTTCCCT

• TP53 reverse primer, GATACTATTCAGCCCGAGTT

LOH analysis was conducted by capillary electrophoresis using the ABI-PRISM 310 Sequence Detection System and the GeneMapper software version 4.0 following the protocol provided by the manufacturer (Applied Biosystems). Allelic imbalance was calculated using the formula: (peak of allele 1 in tumor sample/peak of allele 2 in tumor sample)/(peak of allele 1 in normal sample/peak of allele 2 in normal sample). Allelic imbalance values that were >1.35 or <0.67 were considered to represent LOH.

Associations between gene methylation levels and clinicopathologic variables were analyzed by the Mann-Whitney U test or the Kruskal-Wallis test. The statistical significance of methylation status as a prognostic factor was evaluated using the Cox proportional hazard regression model. The cumulative survival rate was calculated by the Kaplan-Meier method and statistical significance was analyzed by log-rank test. All statistical analyses were carried out using the R software package version 2.8.0 (38).

Methylation levels were analyzed in an initial cohort of 246 matched normal and tumor tissues. DNA extraction was unsuccessful in 31 normal tissues, resulting in a total of 215 normal and 246 tumor tissues that were analyzed for methylation. CDH13 and RASSF1 methylation levels were significantly higher in tumor compared with normal tissue, whereas LINE-1 methylation was significantly lower (Fig. 1). APC methylation levels were not significantly different between tumor and normal tissues.

Associations between methylation and clinicopathologic features are shown in Table 1 and in Supplementary Table S1. APC methylation was significantly lower in squamous cell carcinoma compared with other histologic types. RASSF1 methylation was significantly higher in tumors from older patients. Striking associations were observed between LINE-1 methylation and all clinicopathologic features examined. The observed associations of LINE-1 hypomethylation with male gender, smoking, and squamous cell histology are likely due to the close associations between these clinicopathologic variables (gender and smoking history, P < 0.0001; gender and tumor histology, P < 0.0001; smoking history and tumor histology, P < 0.0001; χ² test).

Survival information was available for 235 patients from the initial cohort (median follow up time, 45.5 months; range, 2-149 months). In an exploratory analysis, a variety of cutoff values were set for methylation and the patients were classified accordingly into hypermethylated or hypomethylated groups for each gene. The survival of these patient groups was compared using the Kaplan-Meier method and the log-rank test. The relationships between methylation cutoff values and P values (log-rank test) are shown Fig. 2 (left) for each methylation marker. CDH13 methylation levels in tumor tissue did not show prognostic significance with any of the cutoff values used (Fig. 2B, left). In contrast, APC, RASSF1, and LINE-1 methylation levels were significantly associated with patient survival [Fig. 2A, C, and D (left), respectively]. The strongest associations were observed using cutoff values of 15 PMR for APC, 65 PMR for RASSF1, and 90% methylation for LINE-1. These cutoff values were used for Kaplan-Meier survival analysis (Fig. 2, right) and for multivariate analyses. Hypermethylation of APC and RASSF1 in tumor tissue was a marker of poor prognosis [Fig. 2A and C (right), respectively], whereas the hypomethylation of LINE-1 was associated with poor prognosis (Fig. 2D, right).

Multivariate analysis was used to determine whether APC, RASSF1, and LINE-1 methylation were associated with patient prognosis independently of other clinicopathologic variables. The analysis included the variables of age, gender, smoking history, histologic type, pathologic stage, and methylation status defined by the abovementioned cutoff values. RASSF1 methylation status was not a significant prognostic factor in multivariate analysis, whereas APC and LINE-1 methylation remained significant together with the factors of age, gender, and pathologic stage (Table 2).

Although APC and LINE-1 methylation levels were independently associated with the patients' prognosis in the multivariate analysis, the most significant prognostic factor was pathologic stages (P < 0.0001; Table 2). Therefore, these methylation markers have less clinical value in predicting the prognosis of the NSCLC patients with all pathologic stages. To clarify the clinical value of APC and LINE-1 methylation levels, the results were reanalyzed according to different stage subgroups. APC methylation stratified according to a cutoff value of 15 PMR showed no prognostic significance in any of the stages (IA, P = 0.22; IB, P = 0.97; II, P = 0.20; III, P = 0.22). Although not reaching statistical significance, trends for prognostic value were observed for LINE-1 methylation in stage IA (P = 0.058) and IB (P = 0.053) patients. These results suggest that APC and LINE-1 methylation may be novel prognostic factors for NSCLC. However, it is unclear whether they are of clinical value complementing pathologic stage system and whether LINE-1 methylation is a significant prognostic factor in early-stage NSCLC.

To examine whether LINE-1 methylation is a significant prognostic factor in stage subgroups, 133 additional tumors were evaluated for this marker, thus increasing the statistical power for analysis of individual stages. As shown in Supplementary Table S2, there were no significant differences in the profile of clinicopathologic features between the initial cohort of 246 cases and the additional 133 cases. The distribution of LINE-1 methylation was also identical between the two tumor cohorts (Supplementary Fig. S1). Combination of the two cohorts caused 379 cases for analysis of the prognostic significance of LINE-1 methylation (stage IA, n = 128; stage IB, n = 76; stage II, n = 34; stage III, n = 129; stage IV, n = 12). Survival information was available for 364 patients and the median follow-up time was 44.5 months (range, 2-158 months). Using a LINE-1 methylation cutoff value of 90% to stratify patients, a significant difference in prognosis was observed (P < 0.001; Supplementary Fig. S2).

Multivariate analysis showed that LINE-1 methylation remained significant as a prognostic factor (P = 0.016) together with age (P = 0.001), gender (P = 0.001), and pathologic stage (P < 0.0001). In a subgroup analysis of stage IA cases, survival was significantly worse for patients with LINE-1 hypomethylation (P = 0.018; Fig. 3A). Of the 126 stage IA patients, 91 were treated with surgery alone and 35 received adjuvant treatment that comprised VP16 + OK432 (n = 23), OK432 (n = 1), NK421 (n = 5), or UFT (n = 6). Patients who did or did not receive adjuvant chemotherapy showed no difference in LINE-1 methylation.

Multivariate analysis of stage IA patients that included the variables of age, gender, smoking history, tumor size, histologic type, and postoperative therapy (with or without adjuvant treatment) revealed that LINE-1 methylation was the only significant prognostic factor (P = 0.026; Table 3). Subgroup analysis failed to show the prognostic value for LINE-1 methylation in all other NSCLC disease stages (Fig. 3B-D).

The prognostic significance of LINE-1 methylation in the initial 235 cases and in the additional 129 cases is shown separately and for each disease stage in Supplementary Fig. S3. The survival curves for the two cohorts were similar for stage IA and III patients (Supplementary Fig. S3B and E) but not for stage IB and II patients (Supplementary Fig. S3C and D), probably due to the small number of patients in each group.

To investigate the molecular basis for the association between LINE-1 methylation and patient prognosis, LOH status was analyzed in 51 randomly selected tumors. Microsatellite analysis of the matching normal tissues showed that 30, 45, and 46 of the 51 cases were heterozygous for the D5S346, D9S942, and TP53 markers, respectively and, therefore, suitable for LOH analysis at these loci. All heterozygous cases were suitable for the evaluation of tumor LOH status except three that showed unstable allelic peaks at D9S942, indicating the presence of the microsatellite instability phenotype. LOH was observed in 13 of 30 (43.3%), 22 of 42 (52.4%) and 26 of 46 (56.5%) tumors at D5S346, D9S942, and TP53, respectively. Representative results of the LOH analysis are shown in Supplementary Fig. S4A. Tumors showing LOH at one or more loci were considered to be LOH positive. Using this criterion, 36 were LOH positive and 15 were LOH negative. The median LINE-1 methylation level in LOH-positive tumors was significantly lower (61.7%; range 20.8-96.6%) than in LOH- tumors (84.9%, range 47.4-97.4; P = 0.004; Supplementary Fig. S4B). This result supports the hypothesis that LINE-1 hypomethylation causes chromosomal instability through the activation of its transposition, resulting in the accumulation of genetic abnormalities and hence poor prognosis in these patients.

In this study, we explored the prognostic significance of gene promoter and global methylation in tumor DNA from NSCLC patients. APC, CDH13, and RASSF1 were analyzed for promoter methylation and LINE-1 methylation was assessed as a surrogate marker of global methylation. Both univariate and multivariate analyses of the initial cohort of 234 cases suggested that APC and LINE-1 methylation were promising prognostic factors in NSCLC. Because APC and LINE-1 methylation were also associated with pathologic stage (Table 1), it was unclear from the study of this initial cohort whether they were independent prognostic factors. APC methylation showed no prognostic significance in the subgroup analysis of each stage, whereas LINE-1 methylation showed marginal significance in stage IA and IB cases only. These initial results prompted us to further investigate LINE-1 methylation as a candidate prognostic factor in a larger series of tumors. Subgroup analysis of a larger cohort of 379 cases showed that LINE-1 methylation was an independent prognostic factor in stage IA NSCLC (Table 3; Fig. 3A). Even for patients with early-stage disease, the prognosis of NSCLC is relatively poor. Therefore, accurate prediction of the likely outcome of stage IA patients is very important for their postoperative management, including decisions on the use of adjuvant chemotherapy and the frequency of follow-up examination. The current results on the prognostic significance of LINE-1 methylation should be validated in prospective, large-scale clinical studies of NSCLC.

Although LINE-1 has been used to assess global methylation in several cancer types (39–41), to our knowledge, only one study of LINE-1 methylation in lung cancer has thus far been reported (42). LINE-1 methylation was lower in NSCLC tissues compared with adjacent normal tissues, consistent with observations in other malignancies such as colorectal cancer (27), leukemia (43), and ovarian cancer (44). These results indicate that LINE-1 hypomethylation is a common event in a variety of cancer types and reflects global hypomethylation of tumor DNA. Similar to the present study of NSCLC, previous workers have reported that LINE-1 hypomethylation was associated with poor prognosis in colorectal cancer (22), leukemia (20), and ovarian cancer (21) patients. This is despite the use of different methods to measure methylation level. Preliminary work from our group using the same analytic method as in the current study also found that LINE-1 hypomethylation was a marker of poor prognosis in colorectal cancer (45). Together, these results suggest that LINE-1 hypomethylation may have clinical application as a prognostic factor in a variety of malignancies.

The mechanism by which LINE-1 methylation is associated with patient prognosis may be linked to the function of LINE-1 sequence as a retroposon. The LINE-1 sequence is 6 kb in length and contains a 5′ untranslated region, two open reading frames, and a 3′ untranslated region (46). The 5′ untranslated region has internal promoter activity, whereas the second open reading frame encodes domains of nuclease and reverse transcriptase activities that are necessary for transposition (47, 48). Increased expression of LINE-1 following hypomethylation may be associated with chromosomal breaks through an increase in nuclease activity. This could result in chromosomal instability and lead to a variety of alterations such as deletion, amplification, and translocation. Chromosomal instability is a characteristic phenotype of more aggressive cancers, suggesting that LINE-1 hypomethylation and subsequent expression are associated with more aggressive tumors and worse patient prognosis. In support of this hypothesis, LINE-1 hypomethylation in human primary cancer has been linked to genomic instability as observed by frequent LOH (42, 49). Our study also found that tumors showing frequent LOH at the D5S346, D9S942, and TP53 loci have significantly lower LINE-1 methylation compared with tumors without LOH. Investigation of the mechanisms that underlie LINE-1 expression and chromosomal breaks may be of great importance in controlling the progression of tumors with LINE-1 hypomethylation.

We previously reported that p16 methylation was a prognostic factor in 246 NSCLC cases studied here as the initial cohort (16). However, in the present study, CDH13 methylation showed no prognostic significance with any of the cutoff values, whereas the prognostic significance of RASSF1 methylation was lost in multivariate analysis. APC methylation remained significant in multivariate analysis but showed no clear association with patient outcome in any of the pathologic stage subgroups. p16 methylation on the other hand was associated with patient prognosis in stage IA disease (16). These results suggest that p16 is the most promising candidate among the promoter methylation markers. A previous study reported that a combination of p16 and CDH13 methylation gave promising results for the prediction of outcome in stage I NSCLC patients (17). Combinations of different gene promoters in the present study may also have resulted in stronger prognostic value than individual methylation markers. However, multiple comparisons can lead to false-positive results by chance, and therefore, we did not explore combinations of markers. This should be analyzed prospectively using fewer markers to avoid the chance of false-positive results. In this regard, the study of p16 and LINE-1 methylation could be of great interest.

In conclusion, we have shown that LINE-1 methylation is significantly associated with patient prognosis in stage IA NSCLC. Lung cancer is a leading cause of cancer-related death and has poor prognosis even at early stages of disease. Hence, the ability to accurately predict the prognosis of patients with stage IA disease should improve strategies for deciding upon postoperative treatments and follow-up examinations. Further validation of the clinical significance of LINE-1 methylation as a prognostic marker in early-stage NSCLC would appear warranted.

No potential conflicts of interest were disclosed.

We thank Dr. Barry Iacopetta (School of Surgery, University of Western Australia) for the critical reading of the manuscript.

Grant Support: Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science.

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.