Alterations of Cyclooxygenase-2 Methylation Levels Before and After Intervention Trial to Prevent Gastric Cancer in a Chinese Population Free

Gastric cancer is one of the most common cancers with high mortality rate (1). Helicobacter pylori (H. pylori), an important carcinogen for gastric cancer (2), was also found to be associated with increased risk for precancerous gastric lesions (3). Among the complicated and interplayed signal pathways in H. pylori–associated gastric carcinogenesis, COX-2 may play a crucial role (4–6).

COX, including COX-1 and COX-2 isoforms, is a rate-limiting enzyme for the production of prostaglandins. Rather than the constructive expression of COX-1 (7), COX-2 is normally absent in gastric mucosa, while induced by mitogens, cytokines, and carcinogens in precancerous lesions or gastric cancer (8, 9). It was reported to play important roles in promoting cell proliferation, inhibiting apoptosis and suppressing host immune response (10–12). Both genetic and epigenetic mechanisms may be involved in COX-2 regulation.

Evidences suggested that H. pylori infection might stimulate promoter hypermethylation to inactivate tumor suppressor genes, such as Cyclin-dependent kinase inhibitor 2A (CDKN2A), Runt-related transcription factor 3 (RUNX3), and Cadherin 1 (CDH1; refs. 13–15). Results from gastric cancer cells illustrated that the transcription of COX-2 could be repressed by hypermethylation of promoter and regained by demethylation (16, 17). However, relationships between COX-2 methylation and expression or H. pylori infection in the process of gastric lesion evolution are still unclear.

A few intervention trials indicated that anti-H. pylori or use of NSAIDs might reduce the risk of precancerous lesions or gastric cancer (18–23). From 2002 to 2006, we conducted an intervention trial in Linqu (Linqu Intervention Trial) using anti-H. pylori and/or selective COX-2 inhibitor (celecoxib; ref. 24), and found that COX-2 expression and its interaction with H. pylori infection were associated with the evolutions of gastric lesions (25). Whether the COX-2 methylation is one of the mechanisms for the effect of interventions still needs further investigation.

In the current study, we tested the dynamic COX-2 methylation changes in gastric mucosa before and after interventions by self-comparison, and evaluated the associations of COX-2 methylation alterations with interventions and gastric lesion evolutions.

In 2002, an initial screening program was conducted in Linqu County, a high-risk area of gastric cancer in China, with interview, carbon-13 urea breath test (¹³C-UBT), upper endoscopy and pathologic diagnosis (24). In 2004, a total of 1,024 eligible participants were invited to the intervention trial according to the selection and exclusion criteria described previously (24). Briefly, the trial included participants ages 35–64 years with H. pylori–positive and baseline histology of chronic atrophic gastritis, intestinal metaplasia, indefinite dysplasia, or dysplasia. Trial participants were randomly assigned to two interventions (anti-H. pylori and/or celecoxib treatments) or placebo in a 2×2 factorial design. Repeat upper endoscopy examinations and histologic diagnoses were conducted in 919 participants at the endpoint of the trial in 2006.

For the current study, a total of 809 participants were finally examined for COX-2 methylation levels at baseline and endpoint. For each subject, the gastric lesion evolution and COX-2 methylation alteration were both compared between the baseline biopsy site with the most severe diagnosis and the same stomach site at endpoint. A written informed consent was obtained from each participant and the study was approved by the Institutional Review Board of Peking University School of Oncology (Beijing, China).

Upper endoscopy examinations were conducted by four experienced gastroenterologists using fiber optic or video endoscopes (Olympus) at baseline and endpoint of our trial, respectively. The gastric mucosa was examined and five biopsies were taken from the standard sites of stomach according to Updated Sydney System (26). Each slide was reviewed by a panel of three senior pathologists and interpreted according to the Updated Sydney System (26) and Padova International Classification (27). Each biopsy was given a diagnosis based on the most severe histology, and each participant was assigned a global diagnosis based upon the most severe diagnosis.

H. pylori infection was determined by ¹³C-UBT as described previously (28). Briefly, the baseline breath sample was collected from each participant after an overnight fast. Then, participants were required to take 75 mg ¹³C-urea, and the second breath samples were collected after 30 minutes. The content of ¹³CO₂ in the two vials of breath samples for each participant was analyzed by a gas isotopic ratio mass spectrometer. The subject was considered H. pylori–positive if the concentration of ¹³CO₂ increased >4 parts per 1,000 when comparing the second breath sample with the baseline one.

High molecular weight genomic DNA was isolated and treated with bisulfite as reported previously (29). After being deparaffinized and rehydrated, biopsy tissues were digested by lysis buffer containing proteinase K at 50°C overnight and then modified with sodium bisulfite. Bisulfite-modified DNA was then purified with a genomic DNA purification kit (Promega).

Universal primers used to amplify the promoter of COX-2 (D28235) were 5′- TTGGAGAGGAAGTTAAGTGTTT-3′ and 5′-ATTAGATAGGAGAGTGGGGAT-3′ (188 bp), respectively (30). The 20-μL PCR reaction mixture contained 1× reaction buffer, 100 ng of bisulfite-modified DNA, 5 pmol of each primers, and 0.5 U of Taq DNA polymerase (QIAGEN GmbH). PCR was accomplished by an initial denaturation of 95°C for 15 minutes, followed by 40 cycles of 95°C for 30 seconds, 54°C for 45 seconds, and 72°C for 30 seconds, with a final extension at 72°C for 10 minutes.

The hyper- and hypomethylated PCR products were separated using a DNASep analytical column (Transgenomic) at the partial denaturing temperature (55°C; ref. 30). The completely methylated and unmethylated DNA from human promyelocytic leukemia cells (HL-60) and gastric cancer cells (MGC803) were kindly provided by Dr. Da-Jun Deng at Peking University Cancer Hospital & Institute (Beijing, China) as positive and negative control. The proportion of hypermethylated copies of COX-2 was calculated by dividing the methylation-peak height (M) by total peak heights of methylation and unmethylation (M+U) for each sample (30). For quality control, all the detections were performed blind by one technician according to the standard protocol with interventions and baseline lesions randomly mixed, and about 10% samples were randomly selected for duplicate detection.

To determine the evolution of gastric lesions before and after intervention by self-comparison, each subject was assigned histologic severity scores at baseline (A) according to the most severe histology in the five standard stomach sites and at endpoint (B) in the same stomach site, respectively. The histologic severity scores were defined as follow: 0 for superficial gastritis/normal; 1 for mild or moderate chronic atrophic gastritis; 2 for severe chronic atrophic gastritis; 3 for superficial intestinal metaplasia; 4 for deep intestinal metaplasia; 5 for indefinite dysplasia; 6 for low-grade dysplasia; 7 for high-grade dysplasia; and 8 for gastric cancer. The evolution status was identified as regression (B < A), progression (B > A), and no change (B = A), respectively.

Similarly, COX-2 methylation alterations before and after interventions were determined using the same pairs of biopsy samples as those for lesion evolution assessment. The COX-2 methylation level alteration for each subject was represented as the methylation level difference between 2006 and 2002 by self-comparison. The methylation alteration trends were classified as decrease, no change, or increase, when the COX-2 methylation level alterations were less than, equal to, or more than 0, respectively.

Because of the few dysplasia subjects in each intervention group, we combined dysplasia with indefinite dysplasia as indefinite dysplasia/dysplasia group for our later analysis. The Pearson χ² test was used to examine the overall differences among four intervention groups in gender, smoking, drinking status, and baseline pathology. The age variable was compared among treatment groups by one-way ANOVA. The baseline methylation levels and methylation level alterations were compared by non-parametric test among various interventions, lesion evolution, or expression score groups. With the methylation no decrease group as reference, we utilized unconditional logistic regression to calculate the ORs and 95% confidence intervals (CI) for associations of methylation alterations with interventions or lesion evolutions, adjusting for age, sex, smoking, drinking status, or baseline histology. Two-sided P value of less than 0.05 was considered statistically significant. These analyses were performed with Statistical Analysis System Software (version 6.12; SAS Institute).

A total of 809 subjects (362 males and 447 females) finally finished detections for COX-2 methylation levels before and after interventions, including 206 in placebo, 196 in anti-H. pylori, 213 in celecoxib, and 194 in anti-H. pylori followed by celecoxib treatment groups. Although the frequency of alcohol drinking was higher in the placebo and celecoxib treatment group (P = 0.03), no significant difference in age, sex, smoking, or baseline pathology was found among four intervention groups (all P > 0.05; Table 1).

The average COX-2 methylation levels at baseline were 9.90% ± 12.22% in placebo, 12.04% ± 14.59% in anti-H. pylori, 11.17% ± 11.67% in celecoxib, and 14.28% ± 14.46% in anti-H. pylori followed by celecoxib groups (P = 0.002), respectively. After interventions, COX-2 methylation levels were increased slightly in placebo group (0.15%), while decreased in anti-H. pylori (−6.04%), celecoxib (−2.57%), and anti-H. pylori followed by celecoxib (−3.71%) groups, respectively (P < 0.001).

With few methylation increasing subjects in active intervention groups, we divided COX-2 methylation alterations into decrease and no decrease groups for further multivariate analysis. Compared with placebo group, logistic regression analysis found higher frequencies of COX-2 methylation decrease in anti-H. pylori (OR, 3.30; 95% CI, 2.16–5.02), celecoxib (OR, 2.04; 95% CI, 1.36–3.07), and anti-H. pylori followed by celecoxib (OR, 2.10; 95% CI, 1.38–3.17) groups, respectively (Table 2). While no additive effect on COX-2 methylation was found for anti-H. pylori followed by celecoxib than two treatments alone.

We also assessed the association between COX-2 methylation alterations and interventions after stratification by baseline histology. Significant differences were found in indefinite dysplasia/dysplasia subjects with COX-2 methylation levels increased by placebo (2.97%), while decreased by anti-H. pylori (−3.40%), celecoxib (−1.83%), and anti-H. pylori followed by celecoxib (−4.16%), respectively (P = 0.001). Compared with the placebo group, COX-2 methylation levels were decreased significantly in indefinite dysplasia/dysplasia subjects by anti-H. pylori (OR, 3.65; 95% CI, 1.96–6.80) and celecoxib (OR, 2.43; 95% CI, 1.34–4.39), as well as anti-H. pylori followed by celecoxib treatment (OR, 2.80; 95% CI, 1.52–5.15) without additional effect than two treatments alone. On the other hand, with placebo groups as reference, COX-2 methylation levels were only significantly decreased by celecoxib treatment (methylation level decreased −7.84%; OR, 2.86; 95% CI, 1.23–6.68) in chronic atrophic gastritis subjects and by anti-H. pylori treatment (methylation level decreased −10.99%; OR, 6.25; 95% CI, 2.38–16.40) in intestinal metaplasia subjects (Table 3).

After the intervention trial, we have reported the regression and progression status of the participants according to the global severity scores at baseline and endpoint (24). Different from the previous report, the evolution of gastric lesions for each subject was currently compared between the most severe histology biopsy site at baseline and the same stomach site at endpoint. Similar with our previous results, we found higher proportions of regression and lower proportions of progression by the three active arms (anti-H. pylori: 73.3%, 11.0%; celecoxib: 68.8%, 11.5%, anti-H. pylori followed by celecoxib: 66.7%, 6.9%, respectively) than by placebo (62.8%, 11.7%), P = 0.078. Because the case numbers of progression subjects in four intervention groups were limited (13–24 cases respectively), we combined the progression and no change groups as no regression group for the further analysis.

We were also interested to test whether COX-2 methylation alterations were associated with gastric lesion evolutions after the two-year interventions. In total participants, COX-2 methylation levels were decreased in regression (−2.71%) and no regression (−3.84%) groups without statistical significance (P = 0.907). When stratified by baseline histology (Table 4), COX-2 methylation levels were increased slightly in no regressed indefinite dysplasia/dysplasia subjects (0.20%), while decreased significantly in regression group (−2.02%), P = 0.039. Compared with subjects without methylation reduction, higher opportunity for regression of gastric lesions was found in subjects with decreased COX-2 methylation levels, especially for indefinite dysplasia/dysplasia subjects (OR, 1.92; 95% CI, 1.03–3.60). Similarly, we used the data of previously detected COX-2 expression changes in gastric mucosa before and after interventions (25), and found an inverse correlation between COX-2 expression increase and regression of indefinite dysplasia/dysplasia subjects (OR, 0.56; 95% CI, 0.38–0.81; Supplementary Table S1).

As COX-2 methylation and expression may play important roles in regression of indefinite dysplasia/dysplasia subjects, we further compared the methylation and expression alterations between indefinite dysplasia and dysplasia subjects who regressed to superficial gastritis, chronic atrophic gastritis, or intestinal metaplasia. We found greater decreasing trends of COX-2 methylation levels in regressed dysplasia (2.10% ± 11.51%) than in regressed indefinite dysplasia subjects (1.81% ± 16.68%), although the P value (0.961) showed no significance. COX-2 expression changes also showed no significant difference between regressed dysplasia and indefinite dysplasia subjects, and the frequencies of expression decrease were 43.5% and 39.8%, respectively (P = 0.876).

Finally, we evaluated the association between COX-2 methylation and expression levels at baseline. COX-2 expression levels were graded semiquantitatively from score 0 to 3. The average methylation levels were decreased gradually from 14.28% at score 0 to 11.90% at score 1, 10.28% at score 2, and 9.31% at score 3, P = 0.066 (Table 5). Compared with the negative expression subjects, COX-2 methylation levels were decreased significantly in strongly expressed subjects (OR, 0.53; 95% CI, 0.29–0.94 at score 3; Table 5). While after the interventions, COX-2 methylation levels showed no relationship with expression scores and remained at low levels as 8.39%, 9.06%, 8.93%, and 7.82% from score 0 to 3, respectively.

On the basis of an intervention trial in a high-risk area of gastric cancer (24), we evaluated COX-2 methylation alterations before and after interventions and their associations with gastric lesion evolutions. We found that COX-2 methylation levels were decreased markedly by anti-H. pylori or celecoxib treatment alone, which was associated with the regression of precancerous lesions, such as indefinite dysplasia/dysplasia.

Mounting evidences have suggested that cytokines, reactive oxygen species, and nitric oxide generated by complex inflammatory and immune reaction may activate DNA methyltransferase and hypermethylate specific genes after the exposure of gastric epithelial cells to H. pylori (31, 32). It is highly reasonable that eradication of H. pylori, an initiator during the complex H. pylori–induced carcinogenesis sequence, could reverse aberrant COX-2 methylation levels in various gastric lesions from chronic atrophic gastritis to intestinal metaplasia and dysplasia in our current study and previous report (33).

In this study, we also assessed the association between COX-2 methylation alterations and interventions stratified by baseline histology. The findings of the placebo groups in various histopathology subgroups are of specific interest, because COX-2 methylation levels were decreased in chronic atrophic gastritis (−5.43%), while increased in intestinal metaplasia (0.59%) and indefinite dysplasia/dysplasia (2.97%) subjects, suggesting a natural withdrawal of inflammatory reaction and subsequent decreased COX-2 methylation levels in early gastric lesions such as chronic atrophic gastritis, while continuous increasing methylation levels in intestinal metaplasia and indefinite dysplasia/dysplasia without interventions. This result indicated that persistent H. pylori–induced inflammation and other risk factors might play important roles in the progression of gastric lesions. However, after interventions (especially anti-H. pylori treatment), more significant decreasing trend of COX-2 methylation was found in intestinal metaplasia or indefinite dysplasia/dysplasia, and showed a favorable impact on the regression of indefinite dysplasia/dysplasia subjects. These findings provided evidence that COX-2 methylation might be one of the mechanisms of precancerous lesion (indefinite dysplasia/dysplasia) regression and might serve as a potential biomarker for chemoprevention efficacy.

Celecoxib, a selective inhibitor of COX-2, was widely used for rheumatoid arthritis, and has been demonstrated the potency to reduce colorectal adenoma (34, 35) and precancerous gastric lesions (24). Our current study revealed that COX-2 methylation was decreased in H. pylori–positive subjects after two-year low-dose (200 mg twice daily) intake of celecoxib. Although the mechanism of COX-2 methylation reduction by celecoxib still remains unclear, it may primarily involve the down-regulation of inflammatory reaction by direct COX-2 activity inhibition and the subsequent reduction of catalyzed prostaglandins or cytokines even in subjects with H. pylori remained in the gastric epithelial (36). A more significant decreasing trend of COX-2 methylation by celecoxib in chronic atrophic gastritis rather than in intestinal metaplasia might be resulted from the more significant prevalence of chronic inflammation in chronic atrophic gastritis than in intestinal metaplasia lesions.

We also found that COX-2 methylation levels were decreased by anti-H. pylori followed by celecoxib, but not more effectively than by anti-H. pylori or celecoxib treatment alone. This tendency is consistent with the lesion evolution outcome in our intervention trial, which did not show additional beneficial effects for anti-H. pylori followed by celecoxib on the regression of precancerous gastric lesions (24). Our findings further confirmed that the decrease of COX-2 methylation level and the subsequent precancerous gastric lesion regression might be mainly induced by anti-H. pylori or celecoxib treatment alone.

The association between COX-2 methylation and expression is very complicated. In baseline H. pylori–infected gastric mucosa, COX-2 methylation showed a repressive effect on expression, which was consistent with our previous findings in a case–control study between H. pylori–positive and negative subjects (30). After removing of inflammatory stimulators by interventions, COX-2 methylation was decreased to low level and showed no relationship with expression. Our results, together with the findings of constitutive activation of the COX-2 promoter without concomitant increase in protein (37), supported that methylation might be one of the mechanisms involved in COX-2 expression regulation, and the relationship between expression and methylation after interventions might be affected by various factors, such as H. pylori status, different interventions and gastric lesions evolution.

The strength of our study was that the intervention trial using anti-H. pylori and/or celecoxib provided us a unique opportunity to assess COX-2 methylation levels dynamically before and after interventions, and the associations with interventions or lesion evolutions. Our self-comparison design before and after interventions at the same stomach site for each subject can control the corresponding confounders, such as the difference in baseline methylation levels among various groups. The gastric lesions selected from this population-based study included various precancerous lesions from chronic atrophic gastritis to dysplasia. While because of the limited number of normal and dysplasia subjects at baseline and few gastric cancer cases in the follow-up period, we could not analyze the evolutions of normal, dysplasia, or gastric cancer subjects.

In conclusion, depending on an intervention trial, our study suggested that anti-H. pylori or celecoxib treatment alone could significantly decrease COX-2 methylation levels in gastric mucosa. COX-2 methylation alteration may be associated with regression of precancerous gastric lesions, such as indefinite dysplasia/dysplasia, and may serve as a potential biomarker for chemoprevention efficacy.

No potential conflicts of interest were disclosed.

Conception and design: H.-M. Zeng, K.-F. Pan, W. You

Development of methodology: Y. Zhang, H.-M. Zeng, X.-R. Nie

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Zhang, L. Zhang, J.-L. Ma, J.-Y. Li

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Zhang, H.-M. Zeng, X.-R. Nie

Writing, review, and/or revision of the manuscript: Y. Zhang, K.-F. Pan, W. You

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.-M. Zeng, J.-Y. Li

Study supervision: K.-F. Pan, W. You

The authors thank the residents, field staff, and the government of Linqu County for supporting the intervention trial.

This work was supported by National Natural Science Foundation of China (81171989, to K.-F. Pan; 30801346, to Y. Zhang), and National Basic Research Program of China (973 Program: 2010CB529303; to W.-C. You).

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.