Cyclooxygenases are rate-limiting enzymes in the synthesis of prostaglandins. Two isoforms, cyclooxygenase 1 (COX-1) and cyclooxyganese 2 (COX-2), are involved in this process,1 and COX-1 is the constitutive isoform expressed in various types of tissues, whereas COX-2 is an inducible enzyme. Overexpression of COX-2 has been reported in multiple types of gastrointestinal malignancies including colorectal, gastric, esophageal and pancreatic cancers.2–6 Studies that used animal models indicate that COX-2 plays important roles in cell adhesion, apoptosis and angiogenesis.7, 8 Numerous epidemiological studies suggest that the use of nonsteroidal anti-inframatory drugs (NSAIDs) decreases the incidence of gastrointestinal cancers and COX-2 is recognized as a major target of NSAIDs.9 Inhibition of COX-2 by NSAIDS or COX-2-specific inhibitors causes cell death in cancer cells,10, 11 indicating that COX-2 is an important molecular target for prevention and therapy in gastrointestinal cancers.
The precise mechanism of how COX-2 expression is controlled remains unclear. Recent studies indicate that both positive and negative regulators are involve in the regulation of COX-2 expression. Among these, a signal pathway related to RAS/MAP kinase is demonstrated to up-regulate COX-2.12 In addition, this pathway up-regulates COX-2 by increasing the stability of the transcript.13 On the other hand, p53 has been demonstrated to down-regulate COX-2.14 Therefore, activation of COX-2 can be partly explained by the alteration of several known oncogenes and tumor suppressor genes. Several lines of evidence suggest that COX-2 expression is not exclusive in all the tumors analyzed, although the mechanism of this phenomenon remains to be determined.15–17
A growing amount of evidence suggests a critical role of DNA methylation in the transcriptional silencing of the gene.18, 19 The subset of gastrointestinal cancers showed methylation of multiple genes, including a cyclin-dependent kinase inhibitor, p16INK4A and a mismatch repair gene, hMLH1.20 Recently, it was shown that methyl-CpG-binding proteins such as MeCP2 interact specifically with methylated DNA and mediate transcriptional repression through recruiting histone deacetylase.21 Therefore, both the methylation and acetylation status of the 5′ region of the gene is important in transcriptional regulation of the gene. In our study, we examined whether absence of COX-2 expression in gastric cancer is associated with aberrant methylation of 5′ region of the gene. We also examined the acetylation status of 5′ region of COX-2 using chromatin immunoprecipitation assay. The results shown in this study suggest that the absence of COX-2 expression in gastric cancer is closely associated with methylation of the 5′ region of the gene and histone deacetylation plays a role in methylation-dependent transcriptional silencing of the gene.
MATERIAL AND METHODS
Cell lines and specimens
Ninety-three primary gastric cancers used in this study were obtained from Sapporo Keiyukai Hospital. Eight gastric cancer cell lines were cultured in appropriate medium. DNA was extracted using Sepagene (Nippon Gene Co., Tokyo, Japan) and RNA was extracted using TRIZOL (GIBCO, Grand Island, NY) according to the manufacture's instruction. Cell lines were treated with 0.1, 1 or 10 μM of 5-deoxy-2′-azacytidine (5-aza-dC) (Sigma, St. Louis, MO) for 3 days. Trichostatin A (TSA) was added after 3 days treatment of 5-aza-dC and cells were harvested after 24 hr.
Two micrograms of DNA were treated with sodium bisulfite, essentially as described previously.22 After treatment, DNA was purified by Wizard PCR Preps (Promega, Madison, WI), precipitated with ethanol and resuspended in 20 μl of diluted water. Two microliters of solution was used for PCR reaction.
Combined Bisulfite Restriction Analysis (COBRA).
Combined Bisulfite Restriction Analysis (COBRA) was performed as described previously.23 PCR reactions were performed in a volume of 50 μl containing 1×PCR buffer (Takara, Tokyo, Japan), 0.25 mM of dNTP mixture, 0.5 μM of each primer and 2.5 U of Taq polymerase (Takara). Touchdown PCR was carried out as follows. After a host start, the cycling parameters were 94°C 30 sec, 53°C 30 sec, 72°C 30 sec for 3 cycles, followed by 94°C 30 sec, 51°C 30 sec, 72°C 30 sec for 4 cycles, 94°C 30 sec, 49°C 30 sec, 72°C 30 sec for 5 cycles and finally 94°C 30 sec, 47°C 30 sec, 72°C 30 sec for 26 cycles. Primers were designed based on the nucleotide sequences submitted to Genbank (AF044206, HSU44805 and D28235). Primers used for COBRA are COX-2F: 5′-GATTTGTAGTGAGYGTTAGGAGT-3′, COX-2R: 5′-RCCAAATACTCACCTATATAACTAAA-3′. After amplification, 20 μl of product were digested with restriction enzyme AfaI (Takara) or TaiI (Fermentas). After ethanol precipitation, DNA was electrophoresed in 3% agarose gel and stained with ethidium bromide.
Bisulfite-SSCP analysis was performed as described previously.24 Briefly, PCR was carried out with Cy5-labeled sense primer. One microliter of PCR products was diluted with 9 μl formamide dye, heat denatured for 5 min at 95°C and loaded onto 5% polyacrylamide non-denaturing gel (99:1, acrylamide to bisacrylamide) containing 5% glycerol in an ALF automated DNA sequencer (Pharmacia Biotech, Uppsala, Sweden). Electrophoresis was performed for 400 min and during electrophoresis, the gel was kept at 25°C with a circulator instrument. Obtained data were analyzed by using the Fragment Manager Software (Pharmacia Biotech).
Bisulfite sequencing analysis
Sodium bisulfite modified genomic DNA were amplified by PCR by using primer pairs, COX-2F and COX-2R. Amplified PCR products were cloned into pCR2.1 vector by using the TOPO-TA Cloning Kit for Sequencing (Invitrogen, La Jolla, CA), according to manufacture's protocol. Plasmid DNA were purified with QIAprep spin Mini prep Kit (Qiagen, Chatsworth, CA). Sequence reaction was carried out with Big Dye Terminator Cycle Sequencing Ready Reaction Kits (Applied Biosystems, Foster City, CA). Samples were sequenced by ABI PRISM 310 Genetic Analyzer (Applied Biosystems). In total, 145 alleles from 4 cell lines, 7 gastric mucosa adjacent to tumor tissues and 7 gastric cancer tissues were examined.
Reverse transcription was performed by using Thermoscript Reverse Transciptase (GIBCO). Five micrograms of total RNA was transcribed in a volume of 50 μl and 1 μl of synthesized cDNA was used for PCR reaction. RT-PCR was performed as follows: 94°C 3 min, 94°C 30 sec, 53°C 30 sec, 72°C 30 sec for 35 cycles. Primers used were RT-F: 5′-CAAAAGC TGGGAAGCCTTCTCTAACC-3′ and RT-R: 5′ -GCCCAGCCCGTTGGTGAAAG-3′. The integrity of RNA was evaluated by amplification of human GAPDH mRNA. Primer sequences for GAPDH were 5′-CAGCCGAGCCACATCG-3′ (sense), 5′-TGAGGCTG TTGTCATACTTCTC-3′ (antisense).
Chromatin immunoprecipitation analysis
To cross link DNA with chromatin, 1 × 106 cells were incubated with 1% folmaldehyde at 37°C for 10 min. Cells were harvested, washed with phosphate-buffered saline (PBS) and resuspended in lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.0, with protease inhibitor). DNA with chromatin was fragmented to 200–1,000 bp by sonication. Immunoprecipitation was performed using anti-acetylated histone H3 antibody (Upstate Biotechnologies, Lake Placid, NY) for 16 hr at 4°C. After immunoprecipitation, immune complexes were collected using protein A-agarose. DNA was purified by phenol/chloroform extraction, precipitated with ethanol and resuspended in diluted water. About 1/100 of the immunoprecipitated DNA was used for PCR reaction; 1/100 of the solution before adding antibody was used as internal control for the amount of DNA. PCR was performed in a solution containing 1× PCR buffer (Takara), 1 μM of primers 0.25 mM of dNTP mixture and 2.5 U of Taq polymerase (Takara). Primer sequences for PCR reaction are as follows: CH-F: 5′-AAGGGGAGAGGAGGGAAAAATTTGTG-3′; CH-R: 5′-GAGGCGCTGCTGA GGAGTTCCTG-3′. PCR reactions were performed for 35 cycles of 95°C 30 sec, 62°C for 30sec and 72°C for 1 min. As a positive control, 5′ region of GAPDH were also amplified using the primers as follows: F: 5′-TCGGTGCG TGCCCAGTTGAACC-3′; R: 5′-ATGCGGCTGACTGTCGAACAGGAG-3′. PCR products were analyzed with agarose gel electrophoresis.
Immunohistochemical staining for COX-2 protein
Antibody binding was detected by using the peroxidase-conjugated streptavidin biotin complex method using HISTOFINE SAB-PO(M) Kit (NICHIREI, Tokyo, Japan). After incubation of primary antibody at 4°C for 16 hr, slides were rinsed with PBS and the biotylinated secondary immunogloblin G antibody was applied for 5 min. The chromogen 3, 3′ diaminobenzidine (Merck, Darmstadt, Germany) was subsequently added and color reaction was observed at light microscopy. The primary monoclonal antibody (MAb) against human COX-2 (Cayman Chemical, Ann Arbor, MI) recognizes a 19 amino acid sequences at the carboxy terminus that is absent in COX-1.25
Methylation status of 5′ region of COX-2 in gastric cancer cell lines was examined by a semi-quantitative methylation assay, combined bisulfite restriction analysis (COBRA), using the primers which amplify the 5′ region of the gene (Fig. 1a). Out of 8 cell lines examined, methylation was detected in 2 cell lines (MKN28 and KATOIII) (Fig. 1b). To determine whether methylation of COX-2 is also detected in primary gastric cancers, we examined a series of gastric cancers by COBRA. Out of 93 cases examined, 11 (12%) cases showed aberrant methylation (Fig. 1c). Examination of clinicopathological features revealed no statistical difference in age, gender, tumor size, depth of invasion and lymphnode metastasis in the cases with or without methylation of COX-2. In 18 cases, methylation status of COX-2 in adjacent gastric mucosa was examined. No significant methylation (>10%) was detected in any of the samples examined (Fig. 1c).
To examine the methylation status of COX-2 in detail, bisulfite-SSCP, which is useful to examine the density of methylation,24 was performed (Fig. 2a). The results were consistent with those of COBRA, indicating not only the restriction sites examined but that the entire region of bisulfite-PCR products was densely methylated. Next, methylation status of specific CpG sites in COX-2 CpG island was examined by bisulfite-sequencing by using the same primers used for COBRA and bisulfite-SSCP (Fig. 2b). Representative results are shown in Figure 2b. In total, 16 CpG sites around transcriptional start site were examined. In MKN28 and KatoIII that showed dense methylation by COBRA and bisulfite-SSCP, 108 out of 112 CpGs from 7 alleles (96%) and 153 out of 160 CpG sites from 10 alleles (96%) were methylated, respectively. MKN7 and MKN45 showed very little methylation. In addition, 7 primary gastric cancers and 7 gastric mucosa adjacent to the tumors were examined. The 3 cases that showed aberrant methylation by COBRA and bisulfite-SSCP had dense methylation of entire region examined. In contrast, 4 unmethylated tumors and all gastric mucosa adjacent to the tumors had no or slight amount of methylation.
Next, we examined whether methylation of COX-2 is correlated with absence of expression in gastric cancer cell lines (Fig. 3a). Expression of COX-2 was detected in all cell lines (6 out of 6) without methylation by RT-PCR. No expression was detected in MKN28 and KATOIII that are methylated for COX-2. Recently, deacetylation of histone has been demonstrated to be involved in the gene silencing of associated with methylation.21 Therefore, we have examined the acetylation status of histone in cell lines with or without COX-2 methylation. Chromatin immunoprecipitation (ChIP) analysis was performed using the anti-acetylated histone H3 antibody (Fig. 3b,c). In all cell lines that express COX-2 (MKN45, MKN74, AZ521, JRST, MKN7 and NUGC3), distinct PCR products from COX-2 promoter was detected, indicating that histone of 5′ region of COX-2 is hyperacetylated. In contrast, MKN28 and KatoIII, which do not express COX-2, showed no distinct band, indicating that histone is deacetylated in this region. Therefore, histone acetylation of COX-2 promoter was closely correlated with gene expression and inversely correlated with methylation.
To further examine the role of methylation and histone deacetylation in silencing of COX-2, we have treated MKN28 with 5-aza-dC, a methyltransferase inhibitor and/or with TSA, a histone deacetylase inhibitor (Fig. 4a). Re-expression of COX-2 was not detected by low-dose treatment with 5-aza-dC treatment or treatment with TSA alone. High dose 5-aza-dC restored COX-2 and combining treatment of low dose 5-aza-dC and TSA restored the expression of synergistically, indicating the role of DNA methylation and histone deacetylation in silencing of the gene. We also examined whether inhibition of methyltransferase and/or histone deacetylase restores the acetylation of histone; ChIP analysis was performed using the immunoprecipitated DNA from MKN28 cells treated with 5-aza-dC, TSA and 5-aza-dC with TSA (Fig. 4b,c). Treatment with TSA, 5-aza-dC and TSA significantly restored the histone acetylation of COX-2. In contrast, histone acetylation was restored only slightly in MKN28 cells treated with 5-aza-dC alone.
To examine whether aberrant methylation of 5′ region of COX-2 is correlated with loss of expression in primary gastric cancers, immunohistochestry was performed (Fig. 5). We have examined 27 cancers that were determined for the methylation status of COX-2. Of these, 6 cases were positive for COX-2 methylation and 21 cases were negative. Normal gastric mucosa did not express COX-2 (Fig. 5a). Of the 27 cases, 19 cases expressed of COX-2 in various level (Fig. 5c,d) and 8 cases were negative for COX-2 (Fig. 5b). All 19 cases that expressed COX-2 were unmethylated tumors and all 6 cases with methylation did not express the gene. There is significant correlation between methylation and absence of COX-2 protein (p<0.0001, Fisher's Exact Test).
Expression of COX-2 is regulated by positive and negative regulatory factors. For example, overexpression of COX-2 has been shown to contribute to tumorigenicity of ras-transformed intestinal epithelial cells.26 It has also been demonstrated that oncogenes such as RAS and c-MYB positively regulate COX-2.10, 13, 27 In contrast, tumor suppressor genes such as p53 negatively regulate COX-2.14 In addition to the transcriptional regulation of the gene, several experimental results suggested that epigenetic changes such as DNA methylation and histone acetylation affect gene expression by modifying the chromatin structure.18, 19, 21 Here, we have shown that methylation status of 5′ region of COX-2 is closely associated with loss of expression of COX-2. In cancer cells which do not express COX-2, dense methylation of the 5′ region of the gene was detected, indicating that methylation play a role in the regulation of the gene. To examine the methylation status of 5′ region of COX-2 in detail, we have performed several different techniques. Using COBRA, bisulfite-PCR followed by restriction digestion, methylation status of 2 CpG sites, AfaI and TaiI sites, were examined in a semi-quantitative manner. Aberrant methylation of these sites was closely associated with absence of expression of COX-2 in gastric cancer cell lines. The results obtained by COBRA were concordant with the results of bisulfite-SSCP, which allowed us to examine the density of methylation in the entire region amplified. Bisulfite-sequencing of the same region revealed the dense methylation of entire region examined, indicating that density of methylation in this region is critical for the silencing of the gene. Our data is consistent with the results that demonstrated transcriptional repression by methyl binding domain proteins such as MeCP2 and MBD1 is dependent on the density of methylation.28, 29
Because treatment of 5′-aza-dC readily induced expression of COX-2, aberrant methylation of promoter can suppress COX-2 even in the presence of the transcription factors. Our data also demonstrated that deacetylation of histone play a role in methylation dependent silencing of COX-2. Chromatin immunoprecipitation analysis revealed that histone around COX-2 promoter was deacetylated in the gastric cancer cell lines with methylation. Although treatment of histone deacetylase inhibitor, TSA, could restore acetylation of histone in a methylated cell line, expression of COX-2 was not restored unless methylation was inhibited. These results indicated that methylation of cytosine can repress gene transcription histone deacetylation independent manner. However, combined treatment of methylation inhibitor and histone deacetylase inhibitor reactivate gene transcription synergistically; histone deacetylation does play a part of methylation dependent gene repression. These results are consistent with the previous findings that DNA methylation can suppress gene dominant by histone acetylation.30 The precise mechanism of how methylation can suppress gene in the presence of acetylated histone remains to be determined. One possible mechanism is the methylation-dependent exclusion of the transcription factors by occupying the binding sites.31 Obviously, further study is necessary to clarify which methyl CpG binding proteins interact with methylated COX-2 promoter.
It has been reported that a subset of colorectal and gastric cancers showed methylation of multiple CpG islands, indicating that these tumors have methylator phenotype.20, 32 More recently, aberrant methylation of COX-2 is specifically found in the colorectal cancer with methylator phenotype.33 These results suggested that methylation of COX-2 is not a random event followed by selective advantage of cells which do not express COX-2. Methylation of COX-2 may occur concordant with other methylation alterations and could be a part of the genome wide methylation defect in these cancers. Because expression status of any of DNA methyltransferase is correlated with methylation of endogenous genes,34 aberrant methylation may not simply be caused by over expression of methyltransferases. Myohanen et al. reported that RNA helicase (RHA) interact with 5′ region of p16INK4A only in cell lines that are unmethylated for the gene.35 RHA has ATP-dependent DNA winding activity and it could be possible that RHA prevents the access of methylation machinery by remodeling the chromatin. Identification of such a cis or trans-acting factor that binds differentially to the methylated and unmethylated promoter may facilitate understanding of the mechanisms of aberrant methylation in cancer.
In conclusion, we have shown that aberrant methylation of COX-2 plays a role in regulation of the gene expression. Two gastric cancer cell lines, which are methylated and do not express COX-2, may be good experimental models to study COX-2 independent mechanisms of NSAIDs induced apoptosis. The results also indicate that gastric cancers with COX-2 methylation may not be good indications for the treatment of NSAIDs or COX-2 specific inhibitor.
The authors thank Dr. W.F. Goldman for editing the manuscript. This study is supported by the grant-in-aid from the Ministry of Education, Science, Sports and Culture. (F.I., M.T. and K.I). T.K. is a research fellow from the Japanese Society for the Promotion of Science. H.S. is a postdoctoral fellow from the Japanese Society for the Promotion of Science.