|Home | About | Journals | Submit | Contact Us | Français|
Methylation of promoter DNA contributes to transcriptional silencing of various tumor-suppressor genes in cancer. Transcriptional silencing of 15-lipoxygenase-1 (15-LOX-1) promotes tumorigenesis. Methylation of 15-LOX-1 promoter DNA occurs in some cancers, but its mechanistic role in 15-LOX-1 transcriptional silencing is unclear. We examined the mechanistic role of 15-LOX-1 promoter DNA methylation in 15-LOX-1 transcriptional regulation in human colorectal cancers. 15-LOX-1 promoter methylation occurred in colorectal cancer cells in vitro, in 36% of tumor tissue samples of colorectal cancer patients, and in virtually no normal colonic mucosa samples of 50 human subjects with no history of colorectal cancer or polyps. 15-LOX-1 promoter DNA methylation levels, however, did not correlate with 15-LOX-1 expression levels (Spearman’s r = .21; P = 0.38). We employed siRNA knockdown and genetic disruption models of DNA methyltransferases (DNMTs) to study the effects of this methylation on 15-LOX-1 expression in colon cancer cells. 15-LOX-1 promoter demethylation was insufficient to reestablish 15-LOX-1 expression. 15-LOX-1 transcription was activated by the histone deacetylase inhibitor suberoylanilide hydroxamic acid (SAHA) only after DNMT-1 dissociation from the 15-LOX-1 promoter and without altering 15-LOX-1 promoter DNA methylation. DNMT-1 protein hypomorphism impaired DNMT-1 recruitment to the 15-LOX-1 promoter, which allowed 15-LOX-1 transcription activation by SAHA. DNMT-1 has a direct suppressive role in 15-LOX-1 transcriptional silencing that is independent of 15-LOX-1 promoter DNA methylation.
Promoter DNA methylation is a well-recognized mechanism for inactivating various tumor suppressor genes in cancer (1). 15-Lipoxygenase-1 (15-LOX-1) is the rate limiting–step enzyme for the conversion of linoleic acid to 13-S-hydroxyoctadecadienoic acid (2, 3) and also is involved in the production of anti-inflammatory lipid signaling mediators from arachidonic acid (e.g., lipoxins (4)) and docosahexaenoic acid (e.g., protectin D1 (5)). Overexpression of 15-LOX-1 inhibits inflammation in animal models (4, 6). 15-LOX-1 is down-regulated in human colon (7–9), esophageal (10), breast (11), and pancreatic cancers (12). 15-LOX-1 re-expression via pharmaceutical agents (e.g., nonsteroidal anti-inflammatory drugs, histone deacetylase inhibitors [HDACIs]) or transfection with plasmid or adenoviral vectors induces apoptosis and inhibits tumorigenesis in cancer cells (8–10, 13–19).
The 15-LOX-1 promoter contains CpG islands (20, 21) and is methylated in lymphoma, lung, epidermoid, cervical, and prostate cancer cell lines (21, 22). Whether 15-LOX-1 promoter DNA methylation mechanistically contributes to 15-LOX-1 transcriptional suppression in human cancers, however, has remained controversial. Conflicting data have shown that the hypomethylating agent 5-aza-2-deoxycytidine (5-aza-dc) has both induced (23) and failed to induce (24) 15-LOX-1 expression in colon cancer cell lines and has both facilitated 15-LOX-1 transcriptional activation by interleukin-4 or HDACIs (in the L428 lymphoma cell line (21)) and inhibited 15-LOX-1 gene transcription (in prostatic cancer cell lines (22)). These conflicting data have clouded the role of 15-LOX-1 promoter methylation in 15-LOX-1 transcriptional silencing in cancer cells. In the present study, we tested whether 15-LOX-1 promoter DNA methylation is associated with 15-LOX-1 transcriptional suppression in patients with colorectal cancers, which are well known to have down-regulation of 15-LOX-1 (7–9); we also investigated the mechanistic contribution of 15-LOX-1 promoter methylation to the suppression of 15-LOX-1 transcription in colorectal cancer cells.
We collected colonic biopsy specimens during colorectal endoscopic procedures after obtaining written informed consent from participating patients. Study patients were selected from among patients seen at outpatient gastrointestinal clinics at The University of Texas M. D. Anderson Cancer Center and other hospitals within the Texas Medical Center (the Gastroenterology section at Baylor College of Medicine, an outpatient gastrointestinal endoscopy unit affiliated with St. Luke’s Hospital, and the Michael E. DeBakey VA Medical Center) for colorectal cancer screening and for the follow-up and management of colorectal cancers. This study was approved by the institutional review board at each participating institution.
Our study involved a total of 100 patients divided into two groups of 50 patients each. The first group included patients with colorectal cancers in whom samples were obtained from the colorectal cancers and from normal-appearing mucosa at least 10 cm from the tumor. The second group included subjects who had a normal complete colonoscopic examination at the time of sample procurement and no personal history of colorectal cancer or polyps. In this normal-colon group, two sets of samples of the colonic mucosa were obtained—one from the left and one from the right colon.
Subjects in both groups were required to be between 45 and 85 yr old, to have no history of hereditary colon cancer (familial colorectal polyposis syndrome, hereditary nonpolyposis colon cancer syndrome, or family history of one or more first-degree relatives with colon cancer), and to be U.S. citizens or permanent residents (to reduce the potential for large variability in risk factors such as dietary habits if international patients were included (25, 26)). Patients were excluded if they had a history of inflammatory bowel disease, had received chemotherapy within 4 wk prior to the colonoscopy, had participated in a chemopreventive study during the month prior to the colonoscopy, had a history of bleeding diathesis, had a history of another active cancer within 5 yr prior to enrollment (except for non-melanoma skin cancer), were taking warfarin, or were taking anti-inflammatory medications (e.g., nonsteroidal agents, aspirin, sulfasalazine) within 1 wk of the colonoscopies. Samples were collected between 2001 and 2006. All tissue samples were fresh frozen and stored at −80°C until the time of laboratory analyses.
We obtained Caco-2 and SW480, LOVO, SW620, HT29, and DLD1 human colon carcinoma cell lines from the American Type Culture Collection (Manassas, VA, USA) and RKO human rectal carcinoma cells from Dr. Michael Brattain (The University of Texas, San Antonio, TX). The HCT-116 parental colorectal cell line, DNMT-1 HM [HCT-116 cells with DNA methyltransferase 1 genetic disruption resulting in DNMT-1 hypomorph protein (DNMT-1 HM)], DNMT-3B KO (HCT-116 cells with DNMT-3B knockout), and DKO (HCT-116 cells with both DNMT-1 HM and DNMT-3B KO) cells were gifts from Dr. Bert Vogelstein (The Johns Hopkins University School of Medicine, Baltimore, MD).
We purchased anti-human DNMT-1 and anti-human DNMT-3B antibodies from Upstate Cell Signaling Solutions (Lake Placid, NY, USA); siGENOME SMARTpool and ON-TARGET plus siRNA for DNMT-1, DNMT-3A, and DNMT-3B and a nonspecific control siRNA (siGLO RISC-Free siRNA) were obtained from Dharmacon, Inc. (Lafayette, CO, USA). Normal human-derived tracheobronchial epithelial cells cultured in a multilayered, pseudostratified, and highly differentiated in vitro model that closely resembles the epithelial tissue of the respiratory tract were obtained from MatTek Corporation (Ashland, MA, USA). Rabbit polyclonal antiserum to recombinant human 15-LOX-1 was obtained as described previously (14); 5-aza-dc was purchased from Sigma-Aldrich (St. Louis, MO, USA), and suberoylanilide hydroxamic acid (SAHA) was provided by the National Cancer Institute’s Chemical Carcinogen Reference Standards Repository. Other reagents, molecular-grade solvents, and chemicals were obtained from commercial manufacturers or as specified.
The HCT-116 parental colorectal cell line and DNMT-1 HM, DNMT-3B KO, and DKO cells were grown in McCoy’s 5A medium; SW480, LOVO, SW620, HT29, RKO, and DLD1 cells were grown in RPMI 1640 medium; and Caco2 cells were grown in MEM medium (Cambrex, East Rutherford, NJ, USA) with added L-glutamine in a humidified atmosphere containing 5% CO2 at 37°C. Media contained 10% fetal bovine serum and were supplemented with 1% penicillin–streptomycin as described previously (14).
HCT-116 and DNMT-1 HM cells were cultured to 40%–50% confluence and then transfected with 100 nmol/L of a pooled mixture of either four SMART-selected or ON-TARGET plus siRNA duplexes for the targeted gene (Dharmacon) or with a nonspecific control siRNA (siGLO RISC-Free siRNA; Dharmacon) using Lipofectamine 2000 (Invitrogen Corp., Carlsbad, CA, USA).
To examine DNA methylation patterns in the CpG island of the 15-LOX-1 gene promoter, we extracted genomic DNA from patient samples and the cell lines using a standard phenol–chloroform method. In a subset of patients in whom both 15-LOX-1 promoter DNA methylation and 15-LOX-1 RNA expression were examined (19 colon cancer patients), both total RNA and genomic DNA were extracted from the same sample by using TRI Reagent (Molecular Research Center, Inc., Cincinnati, OH, USA). Each biopsy specimen was homogenized in 0.5 ml of TRI Reagent. The homogenates were separated into aqueous and organic phases by adding bromochloropropane and then by centrifugation. Total RNA was precipitated from the aqueous phase, and the DNA was precipitated from the phenol and interphase. The isolated genomic DNA was treated with sodium bisulfite (Sigma, St. Louis, MO, USA) as described previously (27). Bisulfite-treated DNA was stored at −20°C until it was used for bisulfite cloning sequencing or pyrosequencing assays.
Bisulfite-treated DNA was amplified using a pyrosequencing method. PCR amplified 150 bp of the 15-LOX-1 promoter region (Fig. 1A) that contains three CpG sites (located between +13 and +22 bp) using 10 pM of forward primer (GGGTTTTAGGTTGGGTTATTTA), 1 pmol of reverse primer with universal overhang (GGGACACCGCTGATCGTTTAACCAACCACAACTACACCTAATTAT), and 10 pmol of universal biotinylated primer (GGGACACCGCTGATCGTTTA). The PCR conditions were as follows: 95°C for 5 min, followed by 50 cycles of 95°C for 30 s, 50°C for 45 s, and 72°C for 45 s, and a final incubation at 72°C for 4 min. Biotin-labeled PCR product was isolated with streptavidin sepharose HP (Amersham Biosciences, Uppsala, Sweden) and subjected to pyrosequencing as described previously (27). The methylation levels at three different CpG sites within the 15-LOX-1 promoter sequence were averaged to represent the degree of methylation in the 15-LOX-1 promoter.
A region of the 15-LOX-1 promoter that contains 10 CpG sites (Fig. 1A) was PCR amplified from bisulfite-treated genomic DNA of four colorectal cancer cell lines and sequenced to confirm the pyrosequencing results. The PCR step amplified the targeted 244-bp region of the 15-LOX-1 promoter using 2 μl of bisulfite-treated DNA and the following oligonucleotide primers: 5′-GATAGTGGTTTTTATTTTTTGTTTT-3′ (sense primer) and 5′-AACCCATCTTACTCAAAAATATTTC-3′ (antisense primer). The PCR conditions were 94°C for 3 min and then 94°C for 20 s, 60°C for 30 s, and 71.5°C for 70 s for 35 cycles. The PCR products were purified and subcloned into pCRII-TOPO vector (Invitrogen). Plasmid DNA was extracted from individual clones using a QIAprep Spin Miniprep Kit (QIAGEN, Inc., Valencia, CA, USA) and sequenced at the SeqWright Company (Houston, TX, USA). Five clones were sequenced for each cell line.
Total RNA was extracted from cells using TRI Reagent. One microgram of total RNA from each sample was reverse transcribed in a 20-μl reaction using the iScript cDNA synthesis kit (Bio-Rad Laboratories, Hercules, CA, USA). Real-time PCR was carried out in triplicate with 25 μl of each reaction containing 1 μl of cDNA (50 ng/μl), 12.5 μl of 2×TaqMan Universal PCR Master Mix (Applied Biosystems, Inc., Foster City, CA, USA), and 1.25 μl of primer and probe mixture (Applied Biosystems) using a 7300 real-time PCR system (Applied Biosystems) with the following program: 50°C for 2 min, 95°C for 10 min, 40 cycles at 95°C for 15 s and at 60°C for 1 min. A sequence-detection program calculated a threshold cycle number (Ct) at which the probe cleavage–generated fluorescence exceeded the background signal (28). We calculated the relative RNA expression level using a comparative threshold cycle method (ddCt), as previously described (18).
Western blot analyses for 15-LOX-1 protein expression were performed as described previously (14).
HCT-116, Caco2, HT29, and SW480 cells were treated with 5 μM SAHA or DMSO for 4 h and then cross-linked by adding formaldehyde to the culture medium to a final concentration of 1% for 10 min at 37°C. Chromatin immunoprecipitation (ChIP) assays were carried out using a commercial assay kit according to the manufacturer’s protocol (Upstate Cell Signaling Solutions). Chromatin was immunoprecipitated using anti-human DNMT-1 and anti-human DNMT-3B (Upstate Cell Signaling Solutions). Real-time PCR was used to assess ChIP DNA of the 15-LOX-1 promoter region between −150 nt and −300 nt relative to the ATG site. The PCR reaction consisted of 1 μl of immunoprecipitated DNA (20 ng/μl), 12.5 μl of 2×TaqMan Universal PCR Master Mix (Applied Biosystems), and 1.25 μl specific primer and probe mixture (Applied Biosystems). The PCR reaction was performed using the following program: 50°C for 2 min, 95°C for 10 min, 40 cycles at 95°C for 15 s and at 60°C for 1 min in a 7300 real-time PCR system (Applied Biosystems). For each experiment, the threshold bar was set within the linear range of PCR amplication. To assess the relative binding of a protein (e.g., DNMT-1) to the 15-LOX-1 promoter region, we used the comparative Ct method to calculate the relative sequence abundance of the ChIP DNA fraction to the input sample (% of input) according to the following formula: % input = [2Ct(input)−Ct(ChIP)] × 100. The Ct value of the ChIP DNA sample is donated by Ct(ChIP) and that of the input by Ct(input) (29, 30).
We used Pearson’s correlation coefficient for correlating the methylation levels from the pyrosequencing and cloning sequencing methods. We used the t test for two-group comparisons (e.g., methylation levels between cancerous and normal mucosa in patients with colorectal cancer). Chi-square and Kruskal-Wallis tests were used to compare clinical characteristics between the colon cancer group subsets (i.e., highly methylated versus nonmethylated cancer tissue subsets). Spearman’s correlation method was used to correlate the methylation levels with 15-LOX-1 mRNA levels. For analyses involving the simultaneous consideration of two factors, we performed two-way analysis of variance (ANOVA).
Our analyses proceeded as follows: We first tested the interaction effect, and if it was significant, we subsequently performed specific comparisons to investigate which differences were driving this effect, using the Bonferroni correction to adjust for the multiple testing problem. This means that if we performed k comparisons, an individual comparison would not be considered significant unless its P value was less than 0.05/k. If the interaction effect was not significant, we tested the individual main effects. Then, if those were significant, we determined which pairwise comparisons were significant, again adjusting for multiplicities using the Bonferroni correction. For analyses involving single factors, we performed a one-way ANOVA. If the overall ANOVA test was significant, then we performed pairwise comparisons, adjusting for multiplicities using the Bonferroni correction. All tests were two sided and conducted at the P ≤ 0.05 level. All quantitative analyses (sample size < 35) were done on the log-transformed data because we found that the log transformation decoupled the relationship between the mean and variance and accommodated the normal-distributional assumptions underlying the methods. Data were analyzed using SAS software (SAS Institute, Cary, NC, USA).
The results for the 15-LOX-1 promoter methylation levels of the cloning sequencing assay highly correlated with the ones from the pyrosequencing assay (Pearson’s correlation coefficient, r = .999, P = 0.0006). HCT-116 cells with both DNMT-1 and DNMT-3B KO (i.e., DKO cells) had very low methylation density (2%) (Fig. 1B). A high methylation level was defined as a methylation density of >10% (i.e., two times higher than the bisulfite/pyroseqencing assay background) (27). All eight tested colorectal cancer cell lines had high methylation levels (Fig. 1C). Methylation densities were >50% in seven of the eight colorectal cancer cell lines (Fig. 1C), and only Caco2 cells had a level <50% (14%).
The demographic and clinical characterestics of subjects enrolled in the study are shown in Table 1. The anatomic distribution for the colorectal cancer was as follows: cecum 2%, ascending colon 2%, transverse colon 4%, descending colon 4%, sigmoid colon 22%, and rectum 66%.
The cancerous mucosa had high methylation levels on average, but the levels were highly variable (14% ± 13.75%, mean ± SD). High methylation levels (>10% methylation density) were detected in a subset of patients (36%) (methylation levels 28% ± 14%) (Fig. 1D). The remaining portion of the cancer patients had methylation levels in the colonic mucosa of 6% ± 1.86%. The cancer patients with high methylation levels were statistically significantly younger than those with normal methylation levels (Table 2). The two cancer subsets had no significant differences in sex, tumor stage, or location distribution (Table 2).
The paired normal tissues from cancer patients had low methylation levels (6% ± 2%). Within this group, only two of 50 adjacent normal tissues had levels >10%; in both cases, the levels were 11%. In specimens obtained from colonic biopsies from patients with normal colons, the methylation levels were low and similar for left and right sides (6% ± 2%; P = 0.22). Only three of 100 mucosal samples from patients with normal colons had levels >10%. The levels in those three samples were 11% (right side), 13% (left side), and 17% (left side). The matched samples from the other side of these subjects’ colons were <10% (6% [left side], 6% [right side], and 7% [right side], respectively). Methylation levels were statistically higher in the cancer tissues than they were in the paired normal tissues from the same patients and in the samples from patients with normal colons (P < 0.0001). Methylation levels were not statistically different between normal colonic mucosa from colorectal cancer patients and normal mucosa from patients with normal colons (P = 0.36).
We correlated the levels of 15-LOX-1 expression and 15-LOX-1 promoter DNA methylation to examine the relevance of 15-LOX-1 promoter DNA methylation to the loss of 15-LOX-1 expression in cancer cells. 15-LOX-1 mRNA expression levels in colorectal cancer cell lines had a weak inverse correlation with 15-LOX-1 promoter DNA methylation levels that failed to reach statistical significance (n = 8; Spearman’s r = −.5; P = 0.21; Fig. 2A). 15-LOX-1 RNA expression levels were lower in the cancerous than in the paired normal mucosa (P =0.006; data not shown). In the subset of 19 colon cancer patients whose DNA and RNA were simultaneously extracted from the same samples, the correlation between the 15-LOX-1 mRNA and DNA promoter methylation level was, however, nonsignificant (n = 19; Spearman’s r = .21; P = 0.38; Fig. 2B).
We further studied the mechanistic importance of 15-LOX-1 promoter DNA methylation to 15-LOX-1 transcriptional silencing by examining the effects of reversing 15-LOX-1 promoter methylation on 15-LOX-1 transcriptional suppression in cells that had high DNA methylation of the 15-LOX-1 promoter. We used the HCT-116 colon cancer cell line, which we found to have a high level of 15-LOX-1 methylation (Fig. 1B, C) and which has been used extensively in the past to study DNA methylation. We used three approaches to demethylate the 15-LOX-1 promoter: (a) siRNA to rapidly down-regulate the expression of DNMT-1, -3A, and -3B, (b) treatment with 5-aza-dc, a well-known DNA demethylating agent, and (c) combined DNMT-1 and -3B genetic disruption in HCT-116 cells.
DNMT-1 siRNA reduced DNMT-1 mRNA levels by 78%–80% in cells transfected with DNMT-1 siRNA compared with those transfected with nonspecific siRNA (P < 0.0001; Fig. 3A). DNMT-3A siRNA transfection reduced DNMT-3A mRNA levels by 41% in cells transfected with DNMT-3A siRNA compared with those transfected with nonspecific siRNA (P < 0.0001; Fig. 3B). DNMT-3B siRNA transfection reduced DNMT-3B mRNA levels by 60%–64% in cells transfected with DNMT-3B siRNA compared with those transfected with nonspecific siRNA (P < 0.0001; Fig. 3C). 15-LOX-1 promoter methylation levels were reduced in cells transfected with DNMT-1 and -3B siRNA or with DNMT-1, -3A, and -3B siRNA by 42% and 37%, respectively; however these differences failed to reach statistical significance (P = 0.264; Fig. 3D).
5-Aza-dc treatment reduced 15-LOX-1 promoter methylation levels for the two concentrations tested, 1 μM and 2.5 μM, by 51% and 38%, respectively (P = 0.0545; Fig. 3D). DNMT-1 and -3B siRNA transfection with and without -3A siRNA increased 15-LOX-1 mRNA levels by onefold compared with nonspecific siRNA (P < 0.0001; Fig. 3E). 5-Aza-dc treatment also increased 15-LOX-1 mRNA by 30% (P = 0.09) for the 1 μM concentration and 79% (P = 0.0126) for the 2.5 μM concentration (Fig. 3E). Neither DNMT-1, -3A, or -3B siRNA down-regulation nor 5-aza-dc treatment was able to induce protein expression of 15-LOX-1 in HCT-116 cells (Fig. 3F). Combined DNMT-1 and -3B genetic disruption in HCT-116 cells (DKO) resulted in significant reduction of the average methylation scores, from 91% to 2.5% (P < 0.0001; Fig. 3G). 15-LOX-1 expression levels were, however, lower in DKO cells than they were in the parental HCT-116 cells (P = 0.0002; Fig. 3H).
We examined whether 15-LOX-1 promoter DNA methylation might play a secondary role in 15-LOX-1 transcriptional silencing by investigating the effects of reversing DNA methylation on 15-LOX-1 transcriptional activation by the HDACI, SAHA. Treatment of HCT-116 cells with the hypomethylating agent 5-aza-dc resulted in a small increase in 15-LOX-1 mRNA expression (0.6-fold); SAHA treatment resulted in a higher increase of 15-LOX-1 mRNA expression (3.22-fold; P < 0.0001; Fig. 4A). The combination of 5-aza-dc and SAHA resulted in a markedly enhanced increase in 15-LOX-1 mRNA (8.8-fold; P < 0.0045; Fig. 4A). As was the case with 5-aza-dc, DMNT-1 and -3B combined genetic disruption in HCT-116 cells (i.e., in DKO cells) markedly increased 15-LOX-1 expression by SAHA from 3.22- to 11.74-fold compared with that in parental HCT-116 cells (P < 0.0001; Fig. 4B). DKO cells were generated from HCT-116 cell line that had homozygous genetic disruption of DNMT-3B (DNMT-3B KO) and DNMT-1 (DNMT-1 HM) (31). DNMT-1 genetic disruption by deleting exons 3, 4, and 5 of the DNMT-1 protein produced a truncated hypomorph protein (DNMT-1 HM) that retained the c-terminal catalytic domain for DNA methyltransferase activity (32). SAHA induced 15-LOX-1 mRNA to a level similar to that in the DKO cells (Fig. 4C) in DNMT-1 HM but not DNMT-3B KO cells (DKO versus DNMT-1 HM [P = 0.12], DKO versus DNMT-3B [P < 0.0001]). DNMT-1 HM and DKO cells but not DNMT-3B KO cells had 15-LOX-1 protein expression induction by SAHA (Fig. 4D). DNMT-1 siRNA transfection significantly reduced the expression of DNMT-1 in HCT-116 cells compared with that in mock-transfected or nonspecific siRNA–transfected cells (P < 0.0001; Fig. 4E). DNMT-1 siRNA down-regulation of DNMT-1 by SAHA treatment resulted in a marked increase in 15-LOX-1 expression levels in HCT-116 cells compared with those in cells treated identically with SAHA but transfected with nonspecific siRNA (P = 0.03; Fig. 4F). The increase in 15-LOX-1 expression with the combination of DNMT-1 down-regulation (by siRNA) and SAHA treatment in the parental HCT-116 cells was similar to that achieved in DNMT-1 HM cells treated with SAHA and transfected with nonspecific siRNA (P = 0.06; Fig. 4F).
Although 15-LOX-1 promoter methylation levels were significantly lower in DKO cells than they were in parental HCT-116 cells (P < 0.0001), the methylation levels were not significantly different between parental HCT-116, DNMT-1 HM, and DNMT-3B KO cells; SAHA had no significant effect on methylation levels in all cell lines (P = 0.662; Fig. 5A). SAHA treatment markedly increased 15-LOX-1 mRNA expression levels in three of the colon cancer cell lines (Caco-2, HT-29, and SW-480) compared with that in the HCT-116 cell line (P < 0.0001; Fig. 5B), although it failed to change the 15-LOX-1 promoter DNA methylation levels significantly in the HCT-116, Caco-2, SW-480, or HT-29 cell lines (P = 0.724; Fig. 5C). SAHA markedly reduced DNMT-1 recruitment to the 15-LOX-1 promoter at 4 h in DNMT-1 HM, Caco-2, HT-29, and SW-480 (P < 0.0001 for all four cell lines) but not the HCT-116 colon cancer cells (P = 0.09; Fig. 5D). SAHA failed to reduce DNMT-3B binding to the 15-LOX-1 promoter in Caco-2, SW-480, or HT-29 cell lines (P = 0.08; data not shown).
The major findings of this study demonstrate that DNMT-1 has a direct repressive role in 15-LOX-1 transcriptional suppression independently of 15-LOX-1 promoter DNA methylation. We found that despite its being limited to colorectal cancer cells, 15-LOX-1 promoter DNA methylation had no correlation with 15-LOX-1 mRNA expression levels, and its demethylation failed to reestablish 15-LOX-1 expression. In contrast, the dissociation of DNMT-1 from the 15-LOX-1 promoter was necessary to activate 15-LOX-1 transcription in colon cancer cells.
The evidence for the specific and direct contribution of DNMT-1 to the suppression of 15-LOX-1 transcription included 1) the genetic disruption of DNMT-1 but not DNMT-3B was necessary for SAHA activation of 15-LOX-1 transcription in HCT-116 cells lines, 2) the down-regulation of DNMT-1 by siRNA in HCT-116 cells allowed SAHA to induce 15-LOX-1 transcription to levels similar to those of DNMT-1 HM (HCT-116 cells with the genetic disruption of DNMT-1), and 3) SAHA activation of 15-LOX-1 transcription in various colon cancer cell lines (HCT-116, Caco-2, SW-480, and HT-29) was associated with a reduction in DNMT-1 but not DNMT-3B recruitment to the 15-LOX-1 promoter. These new findings demonstrate the importance of the effects of HDACIs on DNMT-1 to activate gene transcription in cancer cells, as in the case of 15-LOX-1. This study focused on examining this mechanism in relation to 15-LOX-1 transcription activation, which contributes to the apoptotic effects of the HDACIs (16). This mechanism is likely, however, to contribute to the transcription activation of some other important genes in cancer cells. A recent report showed that HDACIs reduce DNMT-1 binding to the estrogen receptor promoter during estrogen receptor transcription activation in the MDA-MB-231 human breast cancer cell line (33). With the use of quantitative ChIP/real-time PCR and various models of DNMT gene disruption and down-regulation, our current findings demonstrate how HDACIs can activate a gene transcription by inhibiting DNMT-1 recruitment to the gene promoter in cancer cells.
Our results indicate that the suppressive effect of DNMT-1 on 15-LOX-1 transcription was independent from 15-LOX-1 promoter methylation. First, the responsiveness of colon cancer to SAHA activation of 15-LOX-1 transcription occurred independently of 15-LOX-1 promoter methylation. For example, methylation levels were similar among HCT-116, DNMT-1 HM, and DNMT-3B KO cells; SAHA, however, activated 15-LOX-1 transcriptional suppression in DNMT-1 HM but not HCT-116 or DNMT-3B KO cells. Second, SAHA activated 15-LOX-1 transcription in the responsive colon cancer cell lines (DNMT-1 HM, DKO, Caco-2, SW-480, and HT-29) without altering the 15-LOX-1 promoter DNA methylation levels. DNMT-1 methyltransferase function is essential for maintenance of normal chromosomal stability and cell survival (34, 35). However, DNMT-1 has other active domains, besides its methyltransferase domain, that interact with various transcriptional repressors (e.g., HDAC1, HCAC2, the histone methyltransferase SUV39H1, and the polycomb group protein EZH2) (36–40) and, thus, it has been speculated that DNMT-1 acts as a corepressor independently of its methyltransferase function (41). The significance of this direct DNMT-1 corepressor function was questioned in a recent study showing that the expression of DNMT-1 protein with an inactivated methyltransferase domain (via a specific point mutation) fails to rescue DNMT-1 KO mouse embryonic stem cells from the lethal effects of DNMT-1 gene loss (42). Dissecting the significance of DNMT-1 repressor function from its methyltransferase function is difficult, however, when totally depriving the cell of the DNMT-1 methylation function, which is essential for cell survival. In our current study, we evaluated these other DNMT-1 repressive functions using DNMT-1 HM cells, which have a genetically truncated DNMT-1 protein with retained methylation function (31, 32). We found that this truncated DNMT-1 protein had a markedly reduced ability to be recruited to the 15-LOX-1 promoter and to be susceptible to the effects of SAHA in dissociating it from the 15-LOX-1 promoter, thus leading to the activation of 15-LOX-1 transcription. SAHA, as expected for an HDACI, failed to alter 15-LOX-1 promoter methylation levels during these events, indicating that the transcriptional activation that occurred by truncating the DNMT-1 protein was independent of 15-LOX-1 methylation status.
15-LOX-1 promoter DNA methylation was specific to colorectal cancer cells: it was observed in all the colorectal cancer cells tested in vitro and in a substantial subset of colorectal cancer patients (18 of 50; 36%). In contrast, 15-LOX-1 promoter DNA methylation levels in the normal colorectal mucosal specimens were below the established limits for a high methylation level (>10%), apart from very few questionable exceptions. These questionable exceptions included two normal mucosa samples from the same 50 cancer patients (both had a borderline level, 11%) and three of 100 samples from 50 subjects with normal colons (all three levels were only moderately >10%, and the matched normal mucosa from the contralateral side of the same colon had methylation levels <10%). These new findings show that 15-LOX-1 DNA methylation was thus limited to the cancerous mucosa. We believe that this investigation is the first to include sufficiently large numbers of cancer patients and control subjects for meaningful comparisons of 15-LOX-1 promoter DNA methylation levels. The only previous study to make such comparisons involved five cancer-free control subjects and 43 prostate cancer patients (22). Our study, however, included 50 individuals who were carefully matched prospectively to the cancer patients to address potential confounding factors (history of prior colon cancer or polyps, hereditary or familial colorectal cancers, or use of chemotherapeutic or chemopreventive agents) and thus to provide an accurate comparison of 15-LOX-1 DNA promoter methylation between normal and cancerous mucosa.
15-LOX-1 promoter DNA methylation is unlikely to play a primary role in the transcriptional suppression of 15-LOX-1 in cancer cells. This conclusion is based on our following findings: 1) 15-LOX-1 promoter methylation levels had no significant correlation with 15-LOX-1 mRNA expression levels in either the colorectal cancer cell lines or in the patients’ colorectal tumor specimens, 2) 15-LOX-1 DNA promoter hypomethylation by 5-aza-dc or by siRNA down-regulation of either DNMT-1 and -3B or DNMT-1, -3A, and -3B failed to significantly increase 15-LOX-1 transcription, and 3) complete 15-LOX-1 promoter demethylation by genetic disruption of both DNMT-1 and -3B failed to activate 15-LOX-1 transcription. These data clarify the mechanistic role of 15-LOX-1 DNA methylation in the transcriptional silencing of 15-LOX-1 in cancer cells. Prior investigators have examined this mechanistic role using only a pharmacologic approach via 5-aza-dc and have reported conflicting results, ranging from 5-aza-dc inducing 15-LOX-1 expression by itself (23), facilitating transcriptional activation by IL-4 or HDACIs (21), having no effects (24), to inhibiting 15-LOX-1 transcriptional activation (22). However, we used a more specific and direct approach via siRNA and genetic disruption of DNMTs and were able to achieve complete 15-LOX-1 promoter demethylation instead of the partial reversal of methylation achieved with 5-aza-dc. Nevertheless, 15-LOX-1 promoter demethylation was insufficient to activate 15-LOX-1 transcription. Promoter DNA methylation is a well-recognized mechanism that significantly contributes to tumorigenesis by being a primary event in the silencing of various tumor suppressor genes (1). Our data, however, provide evidence that DNA methylation might not always be the primary event for the transcriptional repression maintenance of some genes during tumorigenesis, as in the case of 15-LOX-1.
The mechanistic link between inflammation and cancer is increasingly being supported by emerging data (43), especially in the case of colonic tumorigenesis (44). This link between cancer and inflammation could suggest that the 15-LOX-1 expression loss during colonic tumorigenesis [current manuscript, (7–9)] might potentially contribute to tumorigenesis through the loss of the anti-inflammatory products of 15-LOX-1, including lipoxins (4) and the other recently identified potent anti-inflammatory products of the omega-3 polyunsaturated fatty acid docosahexaenoic acid, resolvins, and protectins (45). Conversely, inflammation might influence 15-LOX-1 transcription, as the inflammatory cytokinase IL-6 up-regulates DNMT1 expression (46, 47), thus suggesting that inflammation suppresses 15-LOX-1 transcription through DNMT-1 modulation during tumorigenesis, leading to decreased production of the 15-LOX-1 pro-apoptotic product, 13-S-hydroxyoctadecadienoic acid (18, 48). Further studies are warranted to further explore these potential roles for 15-LOX-1 transcriptional silencing in the emerging link between inflammation and cancer. The new knowledge gained from studying the mechanisms for silencing 15-LOX-1 transcription through DNMT-1 in tumorigenesis might also provide insight into the mechanisms behind the effects of 15-LOX-1 in various other noncancerous diseases (e.g., atherosclerosis, glomerulonephritis) through its anti-inflammatory role (49, 50).
Our new findings thus indicate that in addition to its essential function as a methyltransferase, DNMT-1 also contributes to transcriptional suppression in cancer cells via mechanisms that are independent of DNA methylation. Identification of these mechanisms is important for better understanding the role of DNMT-1 in the transcriptional repression of tumor suppressor genes. It will also be important for developing novel molecular targeting studies in which we can selectively target the non-methyltransferase suppressor functions of DNMT-1 in cancer cells without compromising its methyltransferase function, which is essential to the survival of normal cells.
We are indebted to Drs. Madhukar Kaw, Harish Gagneja, Sandeep Lahoti, Patrick Lynch, William Ross, Norio Fukami, F. Lyone Hochman, Robert S. Bresalier, Gulchin Ergun, Rhonda Cole, and David Graham for assistance with clinical sample accrual. We also thank Karen Phillips from the Department of Scientific Publications at The University of Texas M. D. Anderson Cancer Center for editing the manuscript. In addition, we acknowledge the technical assistance of Xiu L. Yang. This project was performed in part using a compound (SAHA) provided by the National Cancer Institute’s Chemical Carcinogen Reference Standards Repository operated under contract by Midwest Research Institute, No. N02-CB-07008. I. Shureiqi was supported by the National Cancer Institute, NIH, Department of Health and Human Services K07 grant CA86970 and R01 grant CA106577, by the American Cancer Society Scholar Award RSG-04-020-01-CNE, by the Caroline Wiess Law Endowment for Cancer Prevention, and by the National Colorectal Cancer Research Alliance.