Regional hypermethylation and global hypomethylation are associated with altered chromatin conformation and histone acetylation in colorectal cancer
Article first published online: 19 JAN 2006
Copyright © 2006 Wiley-Liss, Inc.
International Journal of Cancer
Volume 118, Issue 12, pages 2999–3005, 15 June 2006
How to Cite
Deng, G., Nguyen, A., Tanaka, H., Matsuzaki, K., Bell, I., Mehta, K. R., Terdiman, J. P., Waldman, F. M., Kakar, S., Gum, J., Crawley, S., Sleisenger, M. H. and Kim, Y. S. (2006), Regional hypermethylation and global hypomethylation are associated with altered chromatin conformation and histone acetylation in colorectal cancer. Int. J. Cancer, 118: 2999–3005. doi: 10.1002/ijc.21740
- Issue published online: 27 MAR 2006
- Article first published online: 19 JAN 2006
- Manuscript Accepted: 3 NOV 2005
- Manuscript Received: 27 JUL 2005
- The Department of Veteran Affairs Medical Research Service
- Theodora Betz Foundation Fund
- gastrointestinal cancer;
- colorectal cancer;
- DNA methylation/epigenetics;
- microsatellite instability;
- loss of heterozygosity;
- chromosomal instability
Regional DNA hypermethylation and global DNA hypomethylation are 2 epigenetic alterations associated with colorectal cancers. However, their correlation with microsatellite instability (MSI) and chromosomal instability (CIN) in colorectal cancer, and their relationship with chromatin conformation and histone modification are not clear. In this study, we analyzed regional and global methylation in 16 cell lines and 64 primary colorectal cancers. We found that MSI and CIN are 2 alternative events in most cell lines and tumors. Furthermore, regional hypermethylation and global hypomethylation are also alternative events in most cases. We also observed a strong correlation between MSI and regional hypermethylation and between CIN and global hypomethylation. We further analyzed chromatin conformation and histone acetylation in cell lines with CIN or MSI. CIN cancers had open chromatin conformation and enriched histone acetylation in repetitive as well as in gene-specific regions. MSI cancers, on the other hand, had closed chromatin conformation and low levels of histone acetylation. After a MSI cell line was treated with 5-aza-2′-deoxycytidine or trichostatin A, the closed chromatin conformation became open, and histone acetylation was enriched. These observations support our hypothesis that in colorectal cancer, regional hypermethylation and global hypomethylation are associated with altered chromatin conformation and histone acetylation, which might have a causal correlation with MSI and CIN, respectively. © 2006 Wiley-Liss, Inc.
Since the discovery of hypomethylation in cancers,1 epigenetic alterations, including hypermethylation in region-specific CpG islands (regional hypermethylation or focal hypermethylation) and global hypomethylation, have been reported in many human cancers.2, 3, 4, 5 In colorectal cancer, hypermethylation is present in CpG islands in promoters of a variety of genes, such as hMLH1,4, 5APC6 and p16ink4A,7 leading to the silencing of these genes. In addition, in colorectal neoplasia, global DNA methylation levels have been reported to be lower than in normal mucosa.8 This global hypomethylation was suggested to contribute to chromosomal instability (CIN) in cancers,5 and this hypothesis was supported by studies of the tumors of mice carrying hypomorphic Dnmt1,9, 10 and also in several human cancers, such as cancers in breast, liver and prostate and Wilms tumor.11, 12, 13, 14, 15 Global hypomethylation was also detected in colorectal cancer.16 However, its correlation with CIN has not been reported. Although hypermethylation of gene promoters and global hypomethylation have frequently been reported in cancers, their relationship within a specific cancer is not clear.16, 17 In addition, chromatin conformation and histone modification in colorectal cancers with regional hypermethylation and global hypomethylation have not been reported. In this study of the phenotypes (microsatellite instability [MSI], and CIN) and epigenotypes (region-specific methylation and global methylation) in 16 cell lines and 64 primary colorectal cancers, we demonstrate that regional hypermethylation and global hypomethylation are 2 distinct epigenetic pathways, and these aberrations of regional and global methylation are correlated with changes in chromatin conformation and histone modification in the chromatin regions involved.
Material and methods
Cell lines and primary tumors
Colorectal cancer cell lines Caco2, Colo320, H498, HCT8, HT29, LS123, SW48, SW1116 and SW1463 were obtained from American Type Culture Collection (Manassas, VA). Cell lines VACO5, VACO6 and VACO10P were kindly provided by Dr. Sanford D. Markowitz. Cell line C1a was derived from 5583s, provided by Dr. Fred T. Bosman. RKO and C cells were from Dr. Michael Brattain. RW2982 cell line was from Dr. Lance M. Tibbetts. Cells were grown in DMEM supplemented with 10% fetal bovine serum at 37°C in 5% CO2 atmosphere. Primary tumors were obtained at San Francisco Veteran Affairs Medical Center, University of Chicago Hospitals and Minneapolis Veteran Affairs Medical Center. All tumors were from patients with no known family history of colorectal cancer, meeting the criteria of familial adenomatous polyposis (FAP) or hereditary non-polyposis colorectal cancer (HNPCC). Tumors were microdissected from formalin-fixed, paraffin-embedded histological sections stained with haematoxylin–eosin. Normal mucosa of the same patient was also microdissected from the histological sections of normal blocks taken at least 5 cm away from the tumors.
5-Aza-2′-deoxycytidine (5-aza-dC) and trichostatin A (TSA) treatment
Cells were treated with 1 μM 5-aza-dC (Sigma, St. Louis, MO) for 24 hr, or with 300 nM TSA (Sigma) for 24 hr, as described.18
MSI status of primary tumors were determined by comparing the polymerase chain reaction (PCR) patterns of tumors with their normal counterparts at the polymorphic loci BAT26, D3S2420, APC, D11S1999 and D18S877. Samples showing differences in PCR patterns between tumor and normal mucosa in 2 or more loci were scored as MSI. Those showing no difference or difference in only 1 locus were scored as microsatellite stable (MSS). MSI status of cell lines was determined by PCR using the BAT26 locus. Since BAT26 patterns are essentially monomorphic within the human population, any difference reflects MSI. Therefore, no normal DNA control from each case is required when cell lines are investigated.19 Samples exhibiting band 3 or more base pairs (bp) shorter than blood cell DNA were scored as MSI.
Fractional genomic alteration
Fractional genomic alteration was detected by array-based comparative genomic hybridization (CGH).20, 21, 22 Human 1.14 arrays were obtained from the Cancer Center Array Core of University of California San Francisco. The arrays consisted of 2,463 bacterial artificial chromosome clones that cover the human genome at a 1.5 megabase resolution. Five hundred nanograms of tumor and reference DNA was fluorescently labeled with Cy3- and Cy5-dUTP by random priming. Hybridization was performed for 48 hr in a formamide-based buffer in the presence of Cot-1 DNA (Invitrogen, Carlsbad, CA). Three single-color fluorescence intensity images (DAPI, cy3 and cy5) were collected from each array using a charge coupled device camera (Sensys, Photometric, equipped with a Kodak KAF 1400 chip) with a 1× optical system. Image data were analyzed by Spot and Sproc software as previously described.20, 21, 22 A series of 10 combined reference DNA samples versus individual reference DNA hybridizations were performed to define the normal variation of the test to reference log2 intensity ratio for each target clone. The log2 ratios for each case were median centered to zero. The fraction of the genome altered including the fraction of the genome gained and lost for each case was calculated as the sum of genomic distances represented by each clone.
Loss of heterozygosity analysis
Loss of heterozygosity (LOH) in cell lines and tumors were analyzed as described previously23, 24 at polymorphic loci D5S1461, D5S1453, D5S1466, D5S1468, D5S1478 (in chromosome 5q21 region), D8S1130, D8S1106, D8S1463, D8S1125, D8S1121, D8S255, D8S1098 (chromosome 8p12-22), D17S1298, D17S1537, S17S1541, D17S1303 (chromosome 17p13), D18S877, D18S536, D18S846, D18S851 and D18S858 (chromosome 18q21). After normalizing the ratio of two alleles from the normal mucosa, tumors with the ratio of 2 alleles <0.5 or >2.0 were scored as LOH. Cell lines with single band at all consecutive loci within a chromosome region were scored as LOH.24
Regional methylation analysis
DNA methylation status in the promoters of hMLH1, RASSF1A, RASSF2, HIC1, p16ink4A genes and MINT1 and MINT31 loci were determined by methylation specific PCR (MSP) as described previously.25, 26 Equal amounts of NaHSO3-treated DNAs were amplified by PCR separately, using either methylation-specific primers or unmethylation-specific primers. PCR products from methylation- and unmethylation-specific primers were run side by side on a 2% agarose gel. For each sample, the ratio of methylation product over the total methylation and unmethylation products indicated the percent of methylation at this locus. The samples with >50% methylation were scored as methylation positive. The regional methylation level for each sample was calculated as the percent of these 7 genes or loci showing methylation positivity.
Global methylation analysis
Global methylation status was determined using the combined bisulfite-restriction assay (COBRA)27 with primers for the long interspersed element 1 (LINE1) repetitive sequence.28 The NaHSO3-treated DNA was amplified by PCR using LINE1 primers (5′ GYGTAAGGGGTTAGGGAGTTTTT and 5′ AACRTAAAACCCTCCRAACCAAATATAAA). PCR was performed by 30 cycles with annealing at 50°C for analyzing cell lines, and 35 cycles with annealing at 50°C for analyzing primary tumors. The PCR product was digested with restriction enzyme Taq I, and separated on a 2% agarose gel. The ratio of intensity of the digested band over the sum of digested and undigested indicated the percent of global methylation.
Chromatin conformation analysis
Conformation of heterochomatin (represented mainly by repetitive DNA sequences) was analyzed by nuclease accessibility assay.18, 29 The nuclei of cell lines were prepared, and digested with restriction enzymes Hinf I and Alu I separately. DNA was then extracted from the digested nuclei with proteinase K/phenol procedure. DNA from Hinf I digested nuclei was amplified by PCR with LINE1 primers (5′ AGGCATTGCCTCACCTGGGA [upstream] and 5′ CTGCTTTGTTTACCTAAGCAAG [downstream]), and electrophoresed on a 2% agarose gel (Fig. 2a). The absence or presence of a 300 bp fragment indicated that Hinf I is or is not accessible to the chromatin, corresponding to LINE1 repetitive sequence, respectively. To assess the input amount of chromatin, another upstream primer (5′ TCGCTGATTGCTAGCACAG) was mixed in the PCR. Since there is no Hinf I site between this primer and downstream primer, their PCR product (105 bp) represented the amount of input chromatin regardless of digestion. Thus, the ratio of the 300 bp:105 bp bands indicates the amount of Hinf I undigested chromatin corresponding to LINE1 repeat sequence (Fig. 2a). Similarly, DNA from Alu I digested nuclei was also amplified with Alu primers (5′ CGTCTCTACTAAAAATACAAAAATTAGC [upstream], 5′ ACTCGGGAGGCTGAGGCA [upstream] and 5′ GCCCAGGCTGGAGTGCAGTG [downstream]). The absence or presence of a 152 bp product (compared with a control of 94 bp) indicated that Alu I is or is not accessible to the chromatin corresponding to Alu repeat sequence, respectively (Fig. 2a). When the ratio of top band over bottom band (152:94 in Alu repetitive sequence or 300:105 in LINE1 sequence) for the undigested nuclei was normalized as 1, the ratio for the digested nuclei indicated the percent of the tightly bound global chromatin conformation in this sample (Fig. 2b).
Chromatin immunoprecipitation assay
Chromatin immunoprecipitation (ChIP) assay was performed using ChIP assay kit (Upstate Biotechnology, Charlottesville, VA) following the manufacturer's protocol with some modifications. Cells were treated with 1% formaldehyde at 37°C for 10 min for the crosslinking of histone to DNA, and washed twice with PBS containing protease inhibitors. After cells were lysed in SDS lysis buffer containing protease inhibitors, the crosslinked chromatin was sheared by sonication on ice. The sonication conditions were optimized for different cell lines to achieve DNA size between 200 and 1000 bp. The sheared chromatin was diluted in ChIP dilution buffer, and precipitated with 5 μl anti-acetylated histone H3 antibody (Upstate Biotechnology, Charlottesville, VA) at 4°C with stirring overnight. The antibody/histone/DNA complex was absorbed with salmon sperm DNA/protein A agarose slurry. The agarose slurry was then washed with low salt immune complex wash buffer, high salt immune complex wash buffer, LiCl immune complex wash buffer and twice with 10 mM Tris-HCl pH 8.0, 1 mM EDTA (1× TE buffer). The antibody/histone/DNA complex was then eluted with a freshly-made solution containing 1% SDS and 0.1 M NaHCO3. After histone-DNA cross-linking was reversed by heating at 65°C for 4 hr in 0.2 M NaCl, DNA was recovered by proteinase K digestion, phenol extraction and ethanol precipitation, and dissolved in 1× TE buffer. DNA was amplified by PCR separately with primers of β-actin (5′ TCACCAACTGGGACGACATG and 5′ ACCGGAGTCCATCACGATG), Alu repeat (see chromatin conformation analysis), LINE1 repeat (see chromatin conformation analysis) and hMLH1 gene (5′ GAAGAGACCCAGCAACCCA, and 5′ TTTGGCGCCAGAAGAGCCA). The PCR products were separated on a 2% agarose gel. The density of each band was measured with a densitometer. The ratio of density from antibody-precipitated chromatin DNA over that from input control (chromatin DNA collected before antibody-precipitation) indicates the level of the acetylated H3 in the region amplified by PCR.
MSI and CIN are 2 alternative genetic aberrationsin colorectal cancer
MSI analysis was performed in 16 colorectal cancer cell lines using PCR at BAT26 locus. Seven of 16 cell lines examined in this study were found to be MSI positive (Table I). CIN status was analyzed by array-based CGH and LOH techniques (Table I). Fraction of the genome altered in the 7 MSI cell lines was significantly lower (16.1 ± 6.0 [mean ± standard deviation]) than in the 9 MSS cell lines (46.7 ± 10.3, p < 0.001). When CIN was defined as >40% of fraction of the genome altered, 8/9 (89%) of cell lines with MSS and none of the cell lines with MSI showed CIN. LOH at chromosomal regions 5q21, 8p12-22, 17p13 and 18q21 was also performed in these cell lines. There was, as expected, a close correlation between FGA and number of regions with LOH. In 8 cell lines with less than 30% fraction of genome altered, no LOH or LOH at only 1 region was observed. In the other 8 cell lines with more than 40% fraction of genome altered, LOH at 2 or more regions was present. Since the assessment of CIN by array-based CGH and LOH analysis yielded similar results, and the method of analysis of LOH was simpler, we used LOH analysis to assess CIN status in primary colorectal cancers. MSI and LOH status were determined in 64 primary tumors. When tumor showing LOH at 1 or more of the 4 chromosome regions was scored as LOH positive, 44 (98%) of 45 tumors with MSS showed LOH positive, while only 5 (26%) of 19 tumors with MSI showed LOH positive (p < 0.001, Table II). The significant correlation between MSI and absence of CIN, or between MSS and CIN in cell lines and primary tumors indicate that the phenotypes of MSI and CIN represent 2 alternative pathways of colorectal carcinogenesis. We also observed that MSI and CIN coexisted in 5 tumors and neither MSI nor CIN existed in 1 cell line and 1 tumor (Tables I and II). This suggests that other pathways may also play the role in the development of colorectal cancer.
|Fraction of genome altered (%)||Regional methylation (%)||Global methylation (%)|
|Cell lines with MSI|
|Total||16.1 ± 6.01||80 ± 122||49 ± 143|
|Cell lines with MSS|
|Total||46.7 ± 10.01||40 ± 222||22 ± 103|
|Fraction of genome altered (%)||Case number with LOH (%)||Regional hypermethylation (%)1||Global hypermethylation (%)2|
|MSI||16.1 ± 6.03||7/7 (100)4||1/7 (14)5|
|MSS||46.7 ± 10.03||2/9 (22)4||8/9 (89)5|
|MSI||5/19 (26)6||7/12 (58)7||3/11 (27)8|
|MSS||44/45 (98)6||6/40 (15)7||32/34 (94)8|
Regional hypermethylation and global hypomethylation in colorectal cancers with MSI and CIN
Regional methylation and global methylation levels were measured in colorectal cancer cell lines and primary tumors. When plotting global methylation level against regional methylation level of the same cell line, a general trend was observed in which cell lines with low levels of regional methylation showed low levels of global methylation, while cell lines with high levels of regional methylation showed high levels of global methylation (Fig. 1). The correlation of global methylation levels (y) and regional methylation levels (x) could be demonstrated by the equation: y = 0.53x + 3.34, (R2 = 0.64). In MSI cell lines, the regional methylation levels were significantly higher (80% ± 12%) than in MSS cell lines (40% ± 22%, p < 0.001, Table I), or in 4 samples of normal mucosa (0%, p < 0.001). Using greater than 50% of regional methylation level as the criteria for defining hypermethylation, 7/7 (100%) cell lines with MSI and only 2/9 (22%) MSS cell lines showed regional hypermethylation (p < 0.01, Table II and Fig. 1). Similarly, 7/12 (58%) MSI tumors and 6/40 (15%) MSS tumors showed regional hypermethylation (p < 0.01, Table II). In summary, MSI or MSS were associated with the existence or absence of the regional hypermethylation respectively in 14/16 (88%) cell lines and 41/52 (79%) tumors. The levels of global methylation in MSI cell lines (49% ± 14%) were similar to those in 12 samples of normal mucosa (55% ± 14%), but the global methylation level was significantly lower in MSS cell lines (22% ± 10%, p < 0.001, Table I and Fig. 1). Only 1/7 (14%) MSI cell lines showed global hypomethylation, defined as less than 40% of global methylation level (Fig. 1). In contrast, 8/9 (89%) of MSS cell lines exhibited global hypomethylation (p < 0.01, Table II and Fig. 1). Similarly, 3/11 (27%) MSI tumors and 32/34 (94%) MSS tumors showed global hypomethylation (p < 0.001, Table II). In summary, the regional methylation levels in MSI cancers were significantly increased compared with normal mucosa and MSS cancers. The global methylation levels in MSS cancers, which showed CIN in 52/54 cell lines and tumors, were significantly decreased compared with normal mucosa and MSI cancers. These observations suggest a strong association between regional hypermethylation and global hypomethylation and MSI and CIN phenotypes, respectively. Since MSI and CIN appear to be alternative genetic events in most colorectal cancers, and regional hypermethylation and global hypomethylation are closely associated with MSI and CIN respectively, we compared the relationship between regional hypermethylation and global hypomethylation within a specific cancer cell line or tumor. In 60 cancers (including 16 cell lines and 44 tumors) where information on both regional and global methylation was available, 47 cancers (78%) showed only one kind of epigenetic aberrations, including 34 cancers (7 cell lines and 27 tumors) with global hypomethylation and 13 cancers (7 cell lines and 6 tumors) with regional hypermethylation. However, there were 4 tumors (7%) showing no changes in regional and global methylation levels, indicating that these cancers may develop from other pathways. Finally, in 9 cancers (15%, 2 cell lines and 7 tumors) both regional hypermethylation and global hypomethylation were observed. Interestingly, all these 9 cancers are from mucinous cancer patients, indicating that a subset of colorectal cancer may develop through these 2 sequential epigenetic pathways.
Chromatin conformation in colorectal cancer cells with different epigenetic status
In a previous study using nuclease accessibility assay, we demonstrated that chromatin conformation in the region of a specific gene promoter (such as APC gene promoter) was open in cells expressing this gene. In cells which did not express this gene, the chromatin conformation was closed, which could be changed to open conformation after the cells were treated with methylation inhibition agent, 5-aza-dC.18 To understand the epigenetic alterations that occur in cancers with CIN, we used a similar nuclease accessibility assay technique to determine global chromatin conformation. In this assay, 2 primer sets were used, 1 for Alu repetitive sequence and 1 for LINE1 repetitive sequence. The global chromatin conformation was determined in Alu repetitive sequence in RKO, HT29, Caco2 and LS123 cell lines (Fig. 2). For RKO cells, which showed a high global methylation level (61%, Table I), the ratio of 152:94 was not reduced compared with no digestion control, indicating that the chromatin was not accessible to the restriction enzyme Alu I, and the global chromatin conformation is closed. However, in HT29, Caco2 and LS123 cells with global hypomethylation (21, 21 and 15% respectively, Table I), the ratios of 152:94 were clearly reduced by Alu I digestion, indicating that Alu I could reach the recognition site in the chromatin, and the global chromatin conformation was open. The closed chromatin conformation in RKO cells and open structure in HT29, Caco2 and LS123 cells were also observed using LINE1 primers (data not shown). RKO cells were treated with demethylation agent 5-aza-dC repetitively for 4 passages. After each treatment, cells were harvested for the analysis of global methylation and chromatin conformation (P1–P4, Fig. 3). After the drug was removed, cells were passaged 6 more times and collected (P5–P10, Fig. 3). Before RKO cells were treated, the global methylation level was 61%, and the global chromatin conformation was 100% closed. After 4 treatments with 5-aza-dC, global methylation levels decreased to 15% (P1–P4). The levels recovered to 49% gradually after the removal of the drug (P5–P10). The chromatin conformation gradually opened to 33% after the consecutive treatments (P1–P4), and began to close again to 45% after 5-aza-dC was removed (P5–P7). We also observed the similar effect when we used a histone deacetylase inhibitor TSA to treat RKO cells; chromatin structure began to open simultaneously with decreasing global methylation levels. However, the changes caused by TSA treatment were slower than the treatment with 5-aza-dC. After 6 TSA treatments, the levels of global methylation and chromatin conformation reached the similar levels as the cells treated with 5-aza-dC twice.
Histone modification in colorectal cancer cell lines with different epigenetic status
By ChIP analysis, histone acetylation levels have been shown to be enriched in chromatin regions corresponding to active genes, such as hMLH1, indicating that histone acetylation plays an important role in keeping chromatin in an active state.30 To investigate the difference in the histone modification status between cancer cells with normal global methylation levels and those with hypomethylation, we used ChIP assay to determine the histone acetylation levels in the regions of repetitive sequences in the cells with different global methylation levels, including Caco2 (21%), SW48 (51%) and RKO (61%). Chromatin from different cell lines were crosslinked with formaldehyde, sheared by sonication, and precipitated with anti-acetylated histone H3 antibody. After purification, DNA from the immunoprecipitated chromatin was analyzed by PCR with primers for Alu repeats, LINE1 repeats, and for hMLH1 and β-actin gene promoters. The ratio of the intensity of the PCR band of DNA from the anti-acetylated H3-precipitated chromatin over the intensity of PCR band from input control is indicative of the level of the acetylated H3 in the region amplified by PCR (Fig. 4). The ratios in the β-actin region were similar among cell lines Caco2, SW48 and RKO (0.68, 0.58 and 0.55, respectively), indicating that they have similar levels of histone acetylation in the chromatin of β-actin gene (Fig. 4). However, the intensity ratios of SW48 and RKO were lower than Caco2 in the regions of hMLH1 promoter (0.12, 0.20 and 0.52, respectively), Alu repeat (0.05, 0.05 and 1.20, respectively) and LINE1 repeat (0.12, 0.21 and 1.05, respectively). These comparisons suggest that the chromatin regions of hMLH1 promoter, Alu repeat and LINE1 repeat contain lower levels of acetylated histone in SW48 and RKO cells than in Caco2 cells. This result is consistent with our previous observations that in SW48 and RKO cells, hMLH1 expression is silenced,19 chromatin conformation in hMLH1 region and global region is tightly bound (data not shown), and DNA methylation levels in hMLH1 promoter region and Alu and LINE1 repeat regions are higher (Table I), as compared to Caco2 cells. This suggests that in cells with global hypomethylation, global chromatin conformation is open, and histone in chromatin is frequently acetylated. To investigate the mechanisms involved in chromatin modification in cancer cells, we further analyzed the histone acetylation levels in cells treated with 5-aza-dC and TSA. RKO cells treated with 5-aza-dC 2 times and RKO cells treated with TSA 6 times were used in this study. Figure 4 showed that after 5-aza-dC or TSA treatment, the intensity ratios increased in hMLH1 promoter (from 0.20 to 1.55 and 1.35 respectively), Alu repeat (from 0.05 to 1.52 and 1.05 respectively) and LINE1 repeat (from 0.21 to 1.22 and 1.24 respectively), indicating that the histone acetylation levels were increased in all regions.
In this study, we showed that in most colorectal cancers with MSI, there is no evidence of chromosomal instability (the values of fraction of genome altered are low, and LOH is rare), regional and global methylation levels are high, chromatin conformations (regional and global) are closed, and the acetylated histones in chromatin (regional and global) are less abundant. On the other hand, in colorectal cancers with CIN (the values of fraction of genome altered are high, and LOH is frequently present), microsatellite instability is not seen, regional and global methylation levels are low, chromatin conformations (regional and global) are open and acetylated histones in chromatin (regional and global) are enriched. These 2 alternative genetic and epigenetic events appear to represent 2 different pathways in the development of colorectal cancers. However, we also observed that in 5 of 64 tumors (8%) both MSI and LOH co-existed, while in 1 of 64 tumors (2%) and 1 of 16 cell lines (6%) neither MSI nor LOH existed. These relatively rare cases might represent alternative pathways of colorectal carcinogenesis and suggest the heterogeneity of colorectal cancers. We also noticed that some colorectal cancers with MSS showed extensive methylation (8/49, 16%), while some MSI cancers showed no extensive methylation (5/19, 26%). These observations suggest that in some cancers with extensive methylation, hMLH1 gene is not the target of methylation, and the inactivation of mismatch repair genes (e.g., gene mutations) could happen in cancers without extensive methylation. Finally, we observed that even though the regional hypermethylation and global hypomethylation are present separately in most cancers (47/60, 78%), they do co-exist in 2 mucinous cancer cell lines and 7 mucinsou cancers. Our results are consistent with the reports from other laboratories showing lack of relationship between the occurrence of global hypomethylation and regional hypermethylation in colorectal16 and gastric17 neoplasms, and suggest that regional hypermethylation and global hypomethylation may represent 2 independent processes in colorectal carcinogenesis. Either 1 of these 2 events may contribute to the development of cancer. However, in our study, overlap in both processes occurred in some cancers (9/60, 15%), indicating that other mechanisms may also be involved. In ovarian cancer, the relationship between 2 processes was largely random,31 while they appeared to occur sequentially in prostate cancer.32 By analyzing methylation levels of LINE1 sequence with methylation sensitive enzyme/Southern analysis, Suter et al. observed that all 8 colorectal cancers that they studied had global hypomethylation, and suggested that gradual loss of global methylation might finally contribute to a paradox of hypermethylation of tumor suppressor genes during the process of tumor progression.33 However, their conclusion that global hypomethylation is always the first event (since all tumor showed global hypomethylation) differs from our observation that not all cancers showed global hypomethylation, and that either regional hypermethylation or global hypomethylation could be present in cancer independently. This difference may in part be due to the different methods in analyzing global methylation, or to differences in the sample size. In summary, our observations of genetic and epigenetic events in cancers with MSI and LOH indicate that colorectal cancer is a multistep, multipathway disease. Apart from 2 main pathways (CIN and MSI), other pathways of genetic and epigenetic events might also be involved.
In normal cells, absence of DNA methylation in the core regions of promoters in active chromatin domains (open chromatin conformation with enriched acetylated histone) are important for maintaining appropriate levels of gene activity. In addition, heavily methylated CpG sites in repetitive sequences and compact chromatin conformations with under-acetylated histones in heterochromatin (consisting mainly of repetitive DNA sequences) are also required to maintain the integrity of chromatin in cells.34, 35, 36 In cancer cells, global hypomethylation has been proposed to contribute to CIN through DNA double-strand break, isochromosomes, unbalanced juxtacentromeric translocations and whole arm deletions.13 In our study, we demonstrated that global hypomethylation is closely correlated with CIN. Global hypomethylation was present in 8/9 (89%) cell lines and 32/34 (94%) tumors with CIN. When RKO cells were treated with 5-aza-dC, we observed that chromatin conformation was opened, that acetylated histones became enriched in heterochromatin and that global methylation levels were decreased. The integrity of chromosome structure is maintained by its compact conformation, especially in heterochromatin, where hypermethylation of CpG sites exists. When DNA methyltransferase was inhibited by the treatment with 5-aza-dC, global methylation levels were decreased, leading to decreased binding of DNA methyl binding proteins and histone deacetylases, and to increase in acetylated histones in heterochromatin and to open global chromatin conformation. Under these conditions, DNA is likely to become more susceptible to mutagenesis, leading to DNA break or other types of chromosomal damage.
5-aza-dC is now being used as a chemotherapeutic agent for leukemia and myelodysplastic syndrome. The rationale for its use as therapeutic agent is that it induces DNA damage and cell growth arrest, as well as demethylation of silenced genes. The hypothesis that 5-aza-dC may reduce global methylation levels leading to chromosomal instability and DNA damage has been supported by the recent work, in which 5-aza-dC induced DNA damage and cell growth arrest via the p53/p21 WAF1 pathway.37, 38 Histone deacetylase inhibitors, such as TSA, are also candidates of chemotherapeutic agents. In this study, we showed that TSA treatment also induced global hypomethylation as well as open chromatin conformation and histone acetylation. As a histone deacetylase inhibitor, TSA can increase the levels of histone acetylation, leading to an open chromatin conformation. The reduction of DNA methylation levels by treatments that open chromatin conformation supports the notion of a gene silencing loop, whereby gene expression modulates DNA methylation, and vice versa.30, 39
We thank Drs. Shih-fan Kuan, Samuel B. Ho and Carolyn K. Montgomery for kindly providing primary colorectal cancers.