Chronic nonsteroidal anti‐inflammatory drug (NSAID) use suppresses multiple CpG islands hyper methylation (CIHM) of tumor suppressor genes in the human gastric mucosa

Abstract

There have been reports showing a protective role of nonsteroidal anti‐inflammatory drugs (NSAIDs) against gastrointestinal cancers. CpG island hyper methylation (CIHM) of tumor suppressor genes is a major event in carcinogenesis. We investigated the CIHM status of non‐cancerous gastric mucosa in chronic NSAID users and non‐users and assessed the effect of NSAIDs on CIHM. Gastric mucosa samples were obtained from 51 chronic NSAID users and 180 non‐users. CIHM of p14(ARF), p16(INK4a), death‐associated protein kinase (DAP‐kinase), and E‐cadherin (CDH1) genes were determined by methylation‐specific PCR. CIHM high was defined as two or more CpG islands methylated. CIHM of p14, p16, CDH1, and CIHM high were lower in chronic NSAID users than in non‐users (p14: non‐users vs users = 32.2%vs 9.8%, P = 0.003; p16: non‐users vs users = 35.0%vs 15.7%, P = 0.02; CDH1: non‐users vs users = 36.1%vs 9.8%, P = 0.0009; CIHM high: non‐users vs users = 44.4%vs 17.6%, P = 0.0009). NSAID use was also associated with decreased number of CIHM by anova (R = –0.32, P < 0.0001). Multivariate logistic regression analysis with adjustment for sex, age, Helicobacter pylori infection, and NSAID use revealed that NSAID use was inversely correlated with all four CIHM and CIHM high as an independent factor (p14: odds ratio [OR] = 0.17, 95% confidence interval [CI] = 0.06–0.48; p16: OR = 0.32, 95% CI = 0.14–0.75; DAP‐kinase: OR = 0.45, 95% CI = 0.22–0.92; CDH1: OR = 0.18, 95% CI = 0.06–0.48; CIHM high: OR = 0.21, 95% CI = 0.09–0.49). No association was found between CIHM status and the duration or dose of NSAIDs. Chronic NSAID use suppresses CIHM in human gastric mucosa. NSAIDs may have a suppressive role against methylation‐related gastric carcinogenesis. (Cancer Sci 2009; 100: 1192–1197)

Abbreviations: 

CIHM

CpG island hyper methylation

H. pylori

Helicobacter pylori

MSP

methylation‐specifc PCR

NSAID

nonsteroidal anti‐inflammatory drug

NSAIDs are widely used for prevention of ischemic heart or cerebral diseases or for the treatment of chronic inflammation and various musculoskeletal disorders. Their usage, however, remains limited because of their potential risk for gastrointestinal damage.^{( 1} , ² , ³ , ^{4 )}

However, the hypothesis that NSAIDs prevents gastrointestinal cancers is also gaining support. Mechanisms involving the inhibitory effect of aspirin and other NSAIDs on the cyclooxygenase‐2 (COX‐2) enzyme's production of various prostaglandins, resulting in modulation of inflammation and immunoresponse, induction of cell apoptosis, and inhibition of angiogenesis, have been suggested^{( 5 )} as the antitumorgenetic effects.

Epidemiologic studies have also addressed association between NSAID use and reduced risk of cancers in the colorectal^{( 6} , ⁷ , ⁸ , ^{9 )} esophagus^{( 10} , ¹¹ , ¹² , ¹³ , ^{14 )} and stomach.^{( 10} , ¹¹ , ^{12 )} Recent study has shown that chronic users of celecoxib, a selective COX‐2 inhibitor, showed more significant regression of gastric intestinal metaplasia than non‐users after H. pylori eradication therapy,^{( 15 )} suggesting that the use of NSAIDs may suppress the early processes of carcinogenesis.

CIHM is shown to be an important mechanism in gene silencing. In the stomach, CIHM of several tumor suppressor genes have been documented in gastric cancer tissues, precancerous lesions, and H. pylori‐infected non‐neoplastic mucosa.^{( 16} , ¹⁷ , ¹⁸ , ¹⁹ , ^{20 )} Among several genes, p14(ARF), p16(INK4a), death‐associated protein kinase DAP‐kinase, and E‐cadherin (CDH1) have been frequently methylated in non‐neoplastic gastric mucosa, in relation to age,^{( 20 )} H. pylori infection^{( 18} , ²¹ , ^{22 )} and histological degree of gastritis,^{( 22 )} and gastric cancer occurrence.^{( 23} , ²⁴ , ^{25 )} Therefore, epigenetic silencing of these genes may be involved in early steps of carcinogenesis.

Recently, we have reported that chronic aspirin users show lower prevalence of CDH1 methylation in non‐neoplastic gastric mucosa than non‐users.^{( 26 )} Since DNA methylation play a important role in the early steps of gastric carcinogenesis, it is speculated that the chronic use of NSAIDs might influence methylation status.

This is a consequent study of this recent preliminary report. To clarify whether the chronic use of various NSAIDs might also prevent methylation‐related gastric carcinogenesis through suppression of CIHM of multiple CpG promoters, we examined the prevalence of CIHM in four candidate promoters (p14, p16, DAP‐kinase, and CDH1) in the non‐neoplastic gastric mucosa of chronic users of various NSAIDs and non‐users and assessed its influence on DNA methylation.

Materials and Methods

Chronic NSAID users and non‐users.  Enrolled were 231 subjects who were attending the Endoscopy Center of Fujita Health University Hospital from January 2005 to May 2008. All subjects underwent upper gastroscopy for various indications. The reasons for gastroscopy included health check, secondary complete check‐up of stomach cancer following to barium X‐ray examination, follow‐up examination of ulcer diseases, or for the complaint of abdominal discomfort. Among 231 subjects, 51 chronic NSAID users were identified as having chronic use of NSAIDs for at least 3 months or more. The other 180 subjects had no history of chronic use of NSAIDs in the past, or occasional use of NSAIDs within 12 months. This cohort was partly recruited from recent study investigating the association between the promoter CIHM and severity of chronic gastritis,^{( 22 )} host genetic factors,^{( 27 )} and chronic aspirin use.^{( 26 )} Those who had a history of continuous or occasional use of NSAIDs in a period of less than 3 months were excluded from this study. Patients who had severe systemic disease or malignancy in the stomach or other organ were also excluded. The Ethics Committee of the Fujita Health University School of Medicine approved the protocol, and prior, written informed consent was obtained from all participating subjects.

Tissue sample, DNA extraction, and detection of H. pylori infection.  Biopsy specimens were taken from the antrum as well as the upper corpus along the greater curvature from grossly non‐pathological mucosa in all patients. The specimens taken from the antrum were cut into two pieces, and one part of specimens was immediately frozen and stored at –80°C until use. Genomic DNA was extracted directly from these frozen specimens using the standard phenol/chloroform method. Other specimens were fixed in 10% buffered formalin and embedded in paraffin for microscopic histological examination. H. pylori infection status was assessed by histological analysis of both the biopsy specimens from the greater curvature of the gastric antrum and upper corpus using antibodies to H. pylori to avoid false‐negative results for H. pylori infection as much as possible.^{( 28} , ^{29 )} NSAID users were ordered to stop using their drugs 1 week before and after gastroscopy to avoid gastrointestinal bleeding.

Bisulfite modification and MSP.  To detect CIHM, we chose four candidate promoter CpG islands, which have been shown to be frequently methylated in gastric mucosa in relation to age, gender,^{( 20 )} and H. pylori infection:^{( 18} , ²¹ , ^{22 )} p14, p16, DAP‐kinase, and CDH1. For the examination of DNA methylation, genomic DNA was treated with sodium bisulfite using the BislFast DNA Modification Kit for Methylated DNA Detection (Toyobo, Osaka, Japan). MSP was carried out with the following primers: p14 methylated forward (p14 MF): 5′‐gtgttaaagggcggcgtagc‐3′, p14 methylated reverse (p14 MR): 5′‐aaaaccctcactcgcgacga‐3′, which amplified 122‐bp product; p14 unmethylated forward (p14 UF): 5′‐tttttggtgttaaagggtggtgtagt‐3′, p14 unmethylated reverse (p14 UR): 5′‐cacaaaaaccctcactcacaacaa‐3′, which amplified 132‐bp product;^{( 30 )} p16 methylated forward (p16 MF): 5′‐ttattagagggtggggcggatcgc‐3′, p16 methylated reverse (p16 MR): 5′‐gaccccgaaccgcgaccgtaa‐3′, which amplified 150‐bp product; p16 unmethylated forward (p16 UF): 5′‐ttattagagggtggggtggattgt‐3′, p16 unmethylated reverse: (p16 UR): 5′‐caaccccaaaccacaaccataa‐3′, which amplified 151‐bp product;^{( 31 )} DAP‐kinase methylated forward (DAP‐kinase MF): 5′‐ggatagtcggatcgagttaacgtc‐3′, DAP‐kinase methylated reverse (DAP‐kinase MR): 5′‐ccctcccaaacgccga‐3′, which amplified 98‐bp product; DAP‐kinase unmethylated forward (DAP‐kinase UF): 5′‐ggaggatagttggattgagttaatgtt‐3′, DAP‐kinase unmethylated reverse (DAP‐kinase UR): 5′‐caaatccctcccaaacaccaa‐3′, which amplified 106‐bp product;^{( 32 )} CDH1 methylated forward (CDH1 MF): 5′‐ ttaggttagagggttatcgcgt‐3′, CDH1 methylated reverse (CDH1 MR): 5′‐taactaaaaattcacctaccgac‐3′, which amplified 115‐bp product; CDH1 unmethylated forward (CDH1 UF): 5′‐taattttaggttagagggttattgt‐3′, CDH1 unmethylated reverse (CDH1 UR): 5′‐cacaaccaatcaacaacaca‐3′, which amplified 97‐bp product.^{( 31 )} An annealing temperature and times were determined using DNA from the peripheral blood of a young individual without H. pylori infection and DNA methylated with SssI methylase (New England BioLabs, Beverly, MA, USA). MSP was carried out in a volume of 20 µL containing 0.1 µg of bislufite‐modificated DNA. The DNA was denatured at 95°C for 5 min, followed by 33–35 cycles at 95°C for 30 s and 57–69°C according to primers for 1 min, and 72°C for 1 min with a final extension at 72 for 5 min. MSP reactions were done using EX Taq HS (Takara Bio, Shiga, Japan). PCR products (10 µL) were separated by electrophoresis in 2.5% agarose gels, and visualized by UV illumination using ethdium bromide staining. CIHM was defined as the presence of a positive methylation band, showing signals approximately equivalent to or greater than that of the size marker (10 ng/µL: 100 bp DNA Ladder; Takara Bio), irrespective of the presence of unmethylated bands. Samples giving faint positive signals were analyzed a further two times and only those samples with consistent positive methylation bands were considered as positive. In addition, we measured the fluorescence intensities of methylated bands for 50 randomly selected CIHM samples using a digital densitometer (Lane Analyzer; ATTO, Tokyo, Japan), and confirmed that the fluorescence intensities of all 50 methylated bands were approximately equivalent to or greater than that of the size marker (data not shown). CIHM high was also defined as two or more CpG islands methylated.

Statistical analysis.  Statistical analysis was using the χ²‐test for the comparison of CIHM between NSAID users and non‐users. Association between the number of methylated CpG islands and NSAID use was assessed by anova. The strength of association between CIHM and various clinicopathological factors was assessed by calculating the odds ratio (OR) and 95% confidence interval (CI). A probability value of less than 0.05 was considered statistically significant in all analyses.

Results

Characteristics of subjects.  Age and sex distributions, H. pylori infection status, and prevalence of active peptic ulcer disease among chronic NSAID users and non‐users are shown in Table 1. Age and prevalence of active peptic ulcer disease were significantly higher among aspirin users, while sex distribution and H. pylori infection status were not different among these two groups. Clinicopathological characteristics of chronic NSAID users are shown in Table 2. Most NSAID users consisted of users of aspirin (33 subjects, 64.7%) and loxoprofen sodium (12 subjects, 23.5%). NSAID use was occasional in four (7.8%) loxoprofen sodium users, and for all others, usage was continuous. The reasons for NSAID use were prevention of ischemic heart or cerebral diseases in 31 (60.8%) subjects (they were all 100 mg or 200 mg/day aspirin users), and for control of muscular pain or arthralgia in 18 (35.3%) subjects. In 39 (76.5%) subjects, detailed information on previous periods of chronic NSAIDs use was available.

Table 1.

Age and sex distribution, H. pylori infection status, and prevalence of active peptic ulcers in NSAID users and non‐users

Variables	NSAID non‐users	NSAID users	P‐values
Number of subjects	180	51
Age (mean ± SD) †	59.9 ± 13.0	68.0 ± 11.0	< 0.0001
Male : female ‡	105:75	29:22	0.85
H. pylori positive
Subjects, n (%) ‡	99 (55.0)	29 (56.9)	0.81
Active peptic ulcer, n (%) ‡	19 (10.6)	18 (35.3)	P = 0.0001

^† Student's t‐test,

^‡ χ²‐test.

H. pylori, Helicobacter pylori; NSAID, nonsterioidal anti‐inflammatory drug.

Table 2.

Characteristics of 51 NSAID users

Variables (n)	Number of subjects
Aspirin (n = 33)
80 mg/day	1
100 mg/day	29
200 mg/day	2
330 mg/day	1
Loxoprofen sodium (n= 12)
60 mg/day	6
120 mg/day	2
180 mg/day	2
Unknown dose	2
Lornoxicam (n= 1)
8 mg/day	1
Dicrofenac sodium (n= 2)
Unknown dose	2
Etodolac (n= 1)
400 mg/day	1
Other NSAIDs (n= 2)
Unknown dose	2
Purpose of NSAID use
Prevention of ischemic heart or cerebral diseases	31
Control of muscular pain or arthralgia	18
Unknown	2
Past periods of NSAID use
3 months~ <1 year	12
1 year~ <2 years	10
2 year~	17
Unknown	12

Note: NSAID use was occasional in only four of the 12 Loxoprofen sodium users, and all others were continuous users.

NSAID, nonsterioidal anti‐inflammatory drug.

CIHM in chronic NSAID users and non‐users.  All 231 gastric mucosa samples were available for MSP analysis. In all subjects, CIHM was found in 63 (27.2%) for p14, 71 (30.7%) for p16, 91 (39.4%) for DAP‐kinase, and 70 subjects (30.3%) for CDH1. CIHM high was also found in 89 subjects (38.5%).

In the comparison of frequency of CIHM between chronic NSAID users and non‐users, we found that rates of CIHM of p14, p16, CDH1, and CIHM high were significantly lower in chronic NSAID users than in non‐users (p14: non‐users vs users = 32.2%vs 9.8%, P = 0.003; p16: non‐users vs users = 35.0%vs 15.7%, P = 0.02; CDH1: non‐users vs users = 36.1%vs 9.8%, P = 0.0009; CIHM high: non‐users vs users = 44.4%vs 17.6%, P = 0.0009; Fig. 1). We also investigated the association between the number of methylated CpG islands NSAID use. In all subjects, none, one, two, three, or all four CIHM were found in 75 (32.5%), 67 (29.0%), 49 (21.2%), and 40 subjects (17.3%), respectively, while the prevalence of NSAID users in the four groups were 31 (41.3%) for none, 11 (16.4%) for one, six (10.2%) for two, and one subject (2.5%) for three or all four CIHM. This association was significant by anova (R = –0.32, P < 0.0001; Fig. 2).

Figure 1.

Rates of CpG island hyper methylation (CIHM) of four CpG islands and CIHM high in the gastric mucosa of nonsteroidal anti‐inflammatory drug (NSAID) users and non‐users. CIHM of p14, p16, E‐cadherin (CDH1), and CIHM high were significantly lower in NSAID users than in non‐users. CIHM high was defined as two or more CpG islands methylated. DAP‐kinase, death‐associated protein kinase.

Figure 2.

Association between the number of methylated CpG islands and nonsteroidal anti‐inflammatory drug (NSAID) use. The number of methylated CpG islands was inversely associated with NSAID use. Statistical analysis was performed by anova.

Univariate and multivariate logistic regression analysis for relationship between CIHM and clinicopathological factors.  Since CIHM in non‐neoplastic gastric mucosa would be influenced by clinicopathological factors such as age, sex,^{( 20 )} and H. pylori infection,^{( 18} , ²¹ , ^{22 )} univariate (Table 3) and multivariate logistic regression analysis (Table 4) were conducted. Univariate analysis showed that H. pylori infection was closely associated with CIHM of p16 (OR = 3.69, 95% CI = 1.97–6.91, P < 0.0001), CDH1 (OR = 3.58, 95% CI = 1.91–6.70, P < 0.0001), and CIHM high (OR = 3.12, 95% CI = 1.77–5.51, P < 0.0001), while NSAID use was inversely correlated with CIHM of p14 (OR = 0.23, 95% CI = 0.09–0.61, P = 0.003), p16 (OR = 0.35, 95% CI = 0.15–0.78, P = 0.02), CDH1 (OR = 0.19, 95% CI = 0.07–0.51, P = 0.0009), and CIHM high (OR = 0.27, 95% CI = 0.12–0.58, P = 0.0009) (Table 3).

Table 3.

Univariate analysis for relationship between CIHM and clinicopathological factors

Genes	p14		Univariate		p16		Univariate		DAP‐kinase		Univariate		CDH1		Univariate		CIHM high		Univariate
CIHM	UM	M	OR (95% CI)	P‐ values	UM	M	OR (95% CI)	P‐ values	UM	M	OR (95% CI)	P‐ values	UM	M	OR (95% CI)	P‐ values	(–)	(+)	OR (95% CI)	P
Variables
Gender (n)
Female	70	27	Reference		66	31	Reference		58	39	Reference		67	30	Reference		56	41	Reference
Male	98	36	0.95 (0.53–1.71)	0.87	94	40	0.91 (0.52–1.59)	0.73	82	52	0.94 (0.55–1.69)	0.83	94	40	0.95 (0.54–1.68)	0.86	86	48	0.76 (0.45–1.30)	0.32
Age (Mean ± SD)	60.9 ±  13.4	63.4 ±  11.8	1.02 (0.99–1.04)	0.11	62.3 ±  13.6	60.4 ±  11.7	0.99 (0.97–1.01)	0.31	60.5 ±  13.0	63.6 ±  13.0	1.02 (0.99–1.04)	0.08	62.5 ±  13.3	60.0 ±  12.3	0.99 (0.97–1.01)	0.19	61.7 ±  13.2	61.7 ±  12.9	1.00 (0.98–1.02)	0.99
H. pylori infection (n)
Negative	78	25	Reference		86	17	Reference		69	34	Reference		86	17	Reference		78	25	Reference
Positive	90	38	1.32 (0.73–2.37)	0.36	74	54	3.69 (1.97–6.91)	<0.0001	71	57	1.63 (0.95–2.79)	0.08	75	53	3.58 (1.91–6.70)	<0.0001	64	64	3.12 (1.77–5.51)	<0.0001
NSAID use (n)
Non‐users	122	58	Reference		117	63	Reference		104	76	Reference	115	65		Reference		100	80	Reference
Users	46	5	0.23 (0.09–0.61)	0.003	43	8	0.35 (0.15–0.78)	0.02	36	15	0.57 (0.29–1.12)	0.10	46	5	0.19 (0.07–0.51)	0.0009	42	9	0.27 (0.12–0.58)	0.0009

CIHM high was defined as two or more CpG islands methylated.

CDH1, E‐cadherin; CI, confidence interval; CIHM, CpG island hyper methylation; DAP‐kinase, death‐associated protein kinase; M, methylated; NSAID, nonsteroidal anti‐inflammatory drug; OR, odds ratio; UM, unmethylated.

Table 4.

Multivariate logistic regression analysis for relationship between CIHM and clinicopathological factors

CIHM	p14		p16		DAP‐kinase		CDH1		CIHM high
	Multivariate OR (95% CI)	P‐values	Multivariate OR (95% CI)	P‐values	Multivariate OR (95% CI)	P‐values	Multivariate OR (95% CI)	P‐values	Multivariate OR (95% CI)	P‐values
Variables
Gender
Female	Reference		Reference		Reference		Reference		Reference
Male	0.85 (0.46–1.58)	0.61	0.77 (0.42–1.41)	0.39	0.85 (0.49–1.47)	0.56	0.80 (0.43–1.49)	0.49	0.61 (0.34–1.10)	0.10
Age	1.03 (1.01–1.06)	0.01	0.99 (0.97–1.02)	0.64	1.03 (1.00–1.05)	0.02	0.99 (0.97–1.02)	0.55	1.01 (0.99–1.03)	0.35
H. pylori infection
Negative	Reference		Reference		Reference		Reference		Reference
Positive	1.36 (0.73–2.51)	0.34	4.07 (2.13–7.77)	<0.0001	1.69 (0.97–2.93)	0.06	4.09 (2.12–7.89)	<0.0001	3.62 (1.99–6.61)	<0.0001
NSAID use
Non‐users	Reference		Reference		Reference		Reference		Reference
Users	0.17 (0.06–0.48)	0.0006	0.32 (0.14–0.75)	0.009	0.45 (0.22–0.92)	0.03	0.18 (0.06–0.48)	0.0007	0.21 (0.09–0.49)	0.0003

Note: CIHM high was defined as two or more CpG islands methylated.

CDH1, E‐cadherin; CIHM, CpG island hyper methylation; CI, confidence interval; DAP‐kinase, death‐associated protein kinase; H. pylori, Helicobacter pylori; NSAID, nonsteroidal anti‐inflammatory drug; OR, odds ratio.

After multivariate logistic regression analysis with adjustment for sex, age, H. pylori infection, and NSAID use, it was revealed that NSAID use was inversely correlated with all four CIHM and CIHM high as a independent factor (p14: OR = 0.17, 95% CI = 0.06–0.48, P = 0.0006; p16: OR = 0.32, 95% CI = 0.14–0.75, P = 0.009; DAP‐kinase: OR = 0.45, 95% CI = 0.22–0.92, P = 0.03; CDH1: OR = 0.18, 95% CI = 0.06–0.48, P = 0.0007; CIHM high: OR = 0.21, 95% CI = 0.09–0.49, P = 0.0003; Table 5).

Table 5.

Association between CHIM and past duration of NSAID use in 39 chronic NSAID users

CIHM (n)	Past duration of NSAIDs(Mean ± SD): years	Multivariate analysis
		OR (95% CI)	P‐values
p14
UM (34)	1.46 ± 1.04	Reference
M (5)	3.09 ± 3.36	1.68 (0.77–3.66)	0.20
p16
UM (34)	1.49 ± 0.91	Reference
M (5)	2.88 ± 3.77	1.46 (0.83–2.58)	0.19
DAP‐kinase
UM (27)	1.51 ± 0.91	Reference
M (12)	2.03 ± 2.49	1.22 (0.76–1.95)	0.41
CDH1
UM (35)	1.49 ± 1.01	Reference
M (4)	3.28 ± 3.92	1.65 (0.84–3.25)	0.14
CIHM high
(–) (33)	1.42 ± 0.93	Reference
(+) (6)	3.17 ± 3.19	1.72 (0.88–3.38)	0.12

Note: All data were adjusted for age, sex, and Helicobacter pylori infection status. CIHM high was defined as two or more CpG islands methylated.

CDH1, E‐cadherin; CI, confidence interval; CIHM, CpG island hyper methylation; DAP‐kinase, death‐associated protein kinase; M, methylated; NSAID, nonsteroidal anti‐inflammatory drug; OR, odds ratio; UM, unmethylated.

Multivariate logistic regression analysis for relationship between CHIM and past duration of NSAID use in chronic NSAID users.  In 39 subjects for whom information on previous NSAID use was available, we investigated the association between CHIM and past duration of NSAID use. Although past duration of NSAID use was relatively longer in methylated subjects for all CIHM and CHIM high, differences were not significant by multivariate logistic regression analysis with adjustment for sex, age, and H. pylori infection (Table 5). We also investigated whether CIHM status would be affected by different NSAIDs as well as dosage, but no association was found between CIHM status and such factors (data not shown).

Discussion

This is a consequent study of our recent preliminary report showing a suppressive role of chronic aspirin use against CIHM of CDH1 promoter in gastric mucosa.^{( 26 )} We investigated the prevalence of CIHM of four candidate promoters (p14, p16, DAP‐kinase, and CDH1) in the non‐neoplastic gastric mucosa of chronic users of various NSAIDs and non‐users.

Chronic use of various NSAIDs is associated with lower frequency of CIHM in human gastric mucosa compared to non‐users. In particular, rates of CIHM of p14, p16, CDH1, and CIHM high were significantly lower in chronic NSAID users than in non‐users. NSAID use was also associated with a decreased number of CIHM by anova. Furthermore, multivariate logistic regression analysis also revealed that NSAID use was inversely correlated with all four CIHM and CIHM high as an independent factor.

CIHM has been reported in many kinds of cancers as an important mechanism of gene silencing. Issa et al.^{( 33 )} first reported that CIHM of the estrogen receptor (ER) promoter increased as a function of age and suggested that ER was a tumor suppressor gene in the colon. Subsequently, age‐related CIHM of other tumor suppressor genes has been reported in the colon^{( 34} , ³⁵ , ³⁶ , ^{37 )} and stomach.^{( 16} , ¹⁷ , ¹⁸ , ¹⁹ , ^{20 )} CIHM in non‐cancerous gastric mucosa has been shown to be associated with degree of H. pylori‐related gastritis and gastric cancer occurrence;^{( 22} , ²³ , ²⁴ , ^{25 )} thus, this epigenetic change has been considered to be a early step in gastric carcinogenesis. Reversal or prevention of the CIHM of tumor suppressor genes in the stomach, resulting in their increased expression, could contribute to a delay and decrease in the risk of cancer. Our current data support a series of epidemiologic studies reporting the suppressive role of NSAIDs in carcinogenesis in the digestive tract,^{( 6} , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ^{12 )} and indicate the novel evidence that use of various NSAIDs may reduce the risk of methylation‐related carcinogenesis in human gastric mucosa.

The theory of methylation reduction by NSAIDs is not a new one. Pereira et al. first reported that administration of celecoxib, and celecoxib and α‐difluoromethylornithine (DFMO), significantly reduced the number of methylated sites in the ER‐α gene at both 7 and 28 days, resulting in an increase of ER‐α gene mRNA expression in colon tumors induced in rats by azoxymethane.^{( 38 )}

Although we did not investigate the mRNA expression of tumor suppressor genes in human gastric mucosa, it is possible that NSAIDs may uphold mRNA levels by suppressing methylation, and that they may reduce the risk of gastric cancer in the long term.

Concerning the possible mechanisms of CIHM, inflammation and cell proliferation have considered to be promoting factors. Cell proliferation itself has been reported to accelerate de novo DNA methylation.^{( 37} , ^{39 )} Inflammation has also been known to promote de novo methylation through repressing the expression of many genes.^{( 39} , ⁴⁰ , ⁴¹ , ^{42 )} In our study, CIHM of p16, CDH1, and CIHM high were associated with H. pylori infection by both univariate and multivariate analysis, which is in agreement with previous studies and our recent study.^{( 18} , ²¹ , ^{22 )} H. pylori‐related inflammation in methylation induction seems to be clearly evident in the stomach. On the other hand, NSAIDs blocks COX‐2 and H. pylori‐associated enhanced epithelial prostaglandin production and associated hyperproliferation.^{( 43} , ^{44 )} Since COX‐2 expression has been documented in gastric cancer^{( 45} , ^{46 )} and H. pylori‐infected chronic gastritis and premalignant lesions,^{( 47 )} this inhibitory effect against cell proliferation may be a main, possible explanation of our results.

Recently, a novel animal study disclosed that celecoxib could significantly decrease gastric tumor volume, but had no significant effect on COX‐2 and prostaglandin signals,^{( 48 )} implying that NSAIDs also have an antitumorigenic effect which is involved in the COX‐2‐independent pathway, such as nuclear factor–kappa B (NF‐κB), and I‐kappa B (IκB).^{( 48} , ⁴⁹ , ⁵⁰ , ^{51 )} Further study would be necessary to trace the exact mechanism of aspirin in counteracting gastric carcinogenesis, especially in a stage as early as the precancerous stage.

We also hypothesized that CIHM status would be influenced by past duration of NSAID use and the different agents as well as their dosage. In this study, however, contrary to our expectation, multivariate logistic regression analysis revealed that no association was found between CIHM status and past duration. Also, no association was found between CIHM status and different agents and their doses. This result may be due to the small number of subjects for whom information on past use of NSAIDs was available. Future study will be needed of a larger sample to resolve this issue.

In conclusion, we have shown that chronic NSAID use is significantly associated with a reduced risk of CHIM in human gastric mucosa. The suppressive role of various NSAIDs in CHIM of multiple promoters in human gastric mucosa provided a novel finding. However, our result is not meant to imply that such alterations in CHIM status in human gastric mucosa lead to direct manifestation of the NSAIDs as chemopreventive agents, because the detailed mechanism by which NSAIDs reduce the CHIM of multiple promoters is unclear. Moreover, whether this epigenetic alteration would actually reduce the development of gastric cancer is unclear. In addition, since our result has been obtained by retrospective observational study, evaluation of the effect of NSAIDs on CHIM by comparison of the methylation status before and after NSAID use would be also needed as part of longitudinal research.

Furthermore, there has been strong evidence that use of NSAIDs is accompanied by risk for gastrointestinal damage and bleeding.^{( 1} , ² , ³ , ^{4 )} Careful consideration of the clinical significance of NSAIDs as chemopreventive agents is required in regards to the stomach.