Methylation and Messenger RNA Expression of p15^INK4b but Not p16^INK4a Are Independent Risk Factors for Ovarian Cancer

Ovarian cancer remains a major threat to women's health in the United States. It was estimated that, in 2004, ∼25,580 women would be diagnosed with and 16,090 would die from ovarian cancer (1). Because of the inability to detect ovarian cancer at an early, more curable stage, its survival rate has been essentially unchanged over the past 20 years. Early detection and prevention, therefore, depend on the ability to identify genetic and epigenetic events that underlie the development and progression of ovarian cancer.

Epigenetic events, such as de novo cytosine-DNA methylation at CpG sites in the promoter region, can alter mRNA expression, one of the phenotypic characteristics of tumor development and progression (2). Mounting evidence suggests that aberrant methylation of CpG islands is one of the major pathways involved in the inactivation of tumor suppressor genes and the development of cancer (3). Indeed, several types of cancer, including ovarian cancer, exhibit a methylation phenotype (4, 5). In particular, certain types of tumors, including ovarian cancer, show aberrant methylation of CpG islands in the promoter regions of tumor suppressor genes, including the p15^INK4b (6) and p16^INK4a (7, 8). Both p15^INK4b and p16^INK4a proteins are inhibitors of cyclin-dependent kinases that prevent the cell from going through the G₁-S phase transition; therefore, inactivation of p15^INK4b and p16^INK4a is thought to be an important step in cancer development (6–9). p15^INK4b and p16^INK4a promoter methylation has been reported in several types of primary tumors and cancer cell lines, including acute myeloid leukemia (6, 10, 11) and cancer of the lung (4, 6, 12), breast (4, 6, 13), bladder (4, 14), head and neck (6, 15, 16), prostate (6), colon (6), liver (17), kidney (18), and stomach (19).

Several studies of the methylation status of p15^INK4b and p16^INK4a in ovarian cancer have been published (20–23). Whether either gene is aberrantly methylated in ovarian cancer is still controversial, particularly in the most common form of serous epithelial ovarian cancers. However, at least one study reported an association between methylation of p16^INK4a and disease progression, suggesting that an understanding of the regulation of methylation and the expression of p16^INK4a is critical to future management strategies for ovarian cancer (8). Thus, although it is clear that the expression of p15^INK4b or p16^INK4a is aberrant in ovarian cancer, it is not clear whether methylation contributes to the alterations in gene expression or whether these two events are independent in ovarian carcinogenesis. Therefore, the role of de novo DNA methylation in ovarian cancer remains uncertain.

We hypothesized that cytosine-DNA methylation differentially regulates the expression of p15^INK4b and p16^INK4a, and these differences may play a differential role in the processes contributing to ovarian tumorigenesis. Here, we report the methylation status and mRNA expression of both p15^INK4b and p16^INK4a in serous epithelial ovarian cancer tissue specimens from 52 patients compared with 40 apparently normal ovarian tissue samples from different individuals without ovarian cancer. Methylated CpG islands in the promoters of p15^INK4b and p16^INK4a were detected by methylation-specific PCR (MSP) and compared with levels of mRNA expression in the same individuals measured by real-time reverse transcription-PCR.

Tissue samples. Specimens from surgically resected ovarian tumors and from unaffected ovaries were used in this study. Briefly, between January 2000 and March 2003 at The University of Texas M.D. Anderson Cancer Center, tumor tissues were obtained from 52 patients with newly diagnosed primary serous ovarian carcinoma. Control specimens came from two sources: from apparently normal contralateral ovaries of 8 patients with unilateral ovarian cancer and from 40 other patients, of whom 27 underwent surgery for nonovarian cancer (9 with endometrial cancer, 11 with cervical cancer, 6 with uterine cancer, and 1 appendix mucinous tumor) and 13 for benign conditions (9 endometriosis, 2 cervical hyperplasia, 1 fallopian paratubal cysts, and 1 unknown). All samples were snap frozen after surgical removal, stored at −80°C, and tested after pathologic examination. DNA and RNA were extracted from ∼200 mg of fresh frozen tissue specimens. Informed consent was obtained from each patient, and the study was approved by M.D. Anderson Cancer Center institutional review board.

Methylation-specific PCR. DNA was prepared by overnight digestion with 200 μg/mL proteinase K (Life Technologies, Inc., Rockville, MD) at 42°C followed by phenol/chloroform (1:1) extraction and ethanol precipitation. We used MSP to examine the promoter methylation status of the p15^INK4b and p16^INK4a genes. Briefly, genomic DNA (1 μg) was modified with sodium bisulfite by using a CpGenome DNA Modification kit according to the manufacturer's instructions (Serologicals Corp., Norcross, GA). The bisulfite-modified DNA (100 ng) was separately amplified by using primers specific for methylated p15^INK4b (forward GCGTTCGTATTTTGCGGTT and reverse CGTACAATAACCGAACGACCGA) and for unmethylated p15^INK4b (forward TGTGATGTGTTTGTATTTTGTGGTT and reverse CCATACAATAACCAAACAACCAA; Genbank accession no. S75756; ref. 24) as well as primers specific for methylated p16^INK4a (forward TTATTAGAGGGTGGGGCGGATCGC and reverse GACCCCGAACCGCGACCGTAA) and for unmethylated p16^INK4a (forward TTATTAGAGGGTGGGGTGGATTGT and reverse CAACCCCAAACCACAACCATAA; Genbank accession no. X94154; ref. 24). For methylation detection, CpGenome Universal Methylated DNA (Serologicals) was used as the positive control for amplification of methylated alleles, and water blanks without added DNA were included as the negative PCR controls in each assay. DNA amplification was carried out by using reagents supplied in a HotStarTaq DNA Polymerase kit (Qiagen, Inc., Valencia, CA). The PCR mixture contained PCR buffer [16.6 mmol/L ammonium sulfate (pH 8.8), 67 mmol/L Tris (pH 8.8), 1.5 mmol/L MgCl₂, 10 mmol/L 2-mercaptoethanol], deoxynucleoside triphosphates (each at 1.25 mmol/L), and primers (300 ng each per reaction) in a final volume of 50 μL. Reactions were hot started at 95°C for 15 minutes, and amplification was carried out in a PTC-200 Peltier thermal cycler (MJ Research, Inc. Waltham, MA) for 40 cycles (30 seconds at 95°C, 1 minute at the annealing temperature 60-65°C, and 1 minute at 72°C) followed by a final 10-minute extension at 72°C. PCR products were analyzed on 2% agarose gels containing ethidium bromide (Fig. 1). The results were evaluated independently by two researchers, and samples with questionable results were retested to achieve agreement between observers.

Real-time reverse transcription-PCR for mRNA expression. Total RNA was extracted with Tri-Reagent according to the manufacturer's protocol (Molecular Research Center, Cincinnati, OH). We assessed the quality of the extracted total RNA on 1% agarose gels after electrophoresis by visualizing the 18S and 28S RNA bands under UV light (Fig. 2), and RNA concentration was determined with the GeneQuant Pro RNA/DNA Calculator (Amersham Pharmacia Biotech, Cambridge, United Kingdom) before it was used for assessment of mRNA levels. The primers and probes for detecting p15^INK4b, p16^INK4a, and GAPDH cDNA sequences (Genbank accession nos. XM_027626, XM_027621, and AK026525, respectively) were designed by using Primers Express software (Perkin-Elmer Applied Biosystems, Foster City, CA). They were forward CGTGGGAAAGAAGGGAAGAGT, reverse CCCCAGACGCGCAGC, and probe FAM-CGGCCAACGGTGGATTATCCGG-TAMRA for p15^INK4b; forward CTCACGCCCTAAGCGCA, reverse AGTCGACAGCTTCCGGAGG, and probe FAM-TCATGTGGGCATTTCTTGCGAGCC-TAMRA for p16^INK4a; and forward GAAGGTGAAGGTCGGAGTC, reverse GAAGATGGTGATGGGATTTC, and probe FAM-CAAGCTTCCCGTTCTCAGCC-TAMRA for GAPDH (used as an internal control for relative quantification). The fluorogenic probes contained a reporter dye [FAM (6-carboxyfluorescein)] covalently attached at the 5′ end and a quencher dye [TAMRA (6-carboxytetramethylrhodamine)] covalently attached at the 3′ end. Total RNA from each sample (∼50 ng) was used to quantify the cDNA copy number of p15^INK4b, p16^INK4a, and GAPDH. Reverse transcription-PCR was done by using the TaqMan One-Step Reverse Transcriptase-PCR Master Mix Reagents with an ABI PRISM 7700 Sequence Detection System according to the protocol of the manufacturer (Perkin-Elmer Applied Biosystems). The system detects the fluorescence emitted from the fluorogenic oligonucleotide probes, and the signal is directly proportional to the number of template molecules in the reaction mixture after they have crossed a fluorescence detection threshold (25). The ABI PRISM 7700 Sequence Detector is a thermal cycler designed to monitor multiple fluorescent signals in a 96-well format. Samples were amplified in 96-well optical PCR trays and caps, and the data were directly collected during the cycling procedure. After activation of the AmpliTaq Gold at 48°C for 30 minutes and 95°C for 10 minutes, the samples were subjected to 40 cycles, each consisting of 95°C for 15 seconds and 60°C for 1 minute. The GAPDH signal was used as the internal control. The expression level of the target genes was calculated based on a standard curve established with a standard human RNA sample that contained a large number of copies to ensure the detection of GAPDH mRNA (∼1 × 10⁵ copies/ng, Perkin-Elmer Applied Biosystems). The standard curve was constructed with a continuation of six data points equivalent to 50,000, 5,000, 500, 50, 5, and 0.5 pg of human RNA. Each sample was measured in triplicate and the means of the three values were calculated for statistical analysis. About 10% of the samples were retested using the same RNA samples, and the measurements were found to be consistent between repeats.

Statistical analysis. Student's t test was used to compare the differences in the relative mRNA expression levels, which were analyzed as a continuous variable between groups. The χ² test was used to test associations between methylation status and risk for ovarian cancer, and the Pearson correlation coefficients were calculated for correlation analysis. For calculation of crude odds ratios (OR) and 95% confidence intervals (95% CI), the median relative expression level of p15^INK4b and p16^INK4a mRNA in the controls was used as the cutoff point for defining lower versus higher expression level. Adjusted ORs were calculated by fitting logistic regression models with adjustment for patient age and ethnicity and other variables of interest. All statistical analyses were done with SAS software version 8.0e (SAS Institute, Inc., Cary, NC).

p15^INK4b and p16^INK4a methylation and expression in ovarian tumor and noncancerous ovarian tissues. Raw data on p15^INK4b and p16^INK4a promoter methylation and relative mRNA expression levels in ovarian tumor specimens (the cases) are presented in Table 1 and those in normal ovary specimens (the controls) in Table 2. The promoter methylation results are presented as either “U” for unmethylated or “M” for methylated. Overall, cases were older than controls (61.5 ± 9.4 versus 50 ± 11.5 years; P = 0.001), but no difference was found in the distribution of ethnic groups between cases (45 whites, 3 non-White Hispanics, 2 African Americans, and 2 Asians) and controls (36 whites, 6 non-white Hispanics, 5 African Americans, and 1 Asian; P = 0.256). The methylation status of p15^INK4b and that of p16^INK4a were not associated among either the 52 cases (χ² = 1.9259; P = 0.165) or the 48 controls (χ² = 0.140; P = 0.708), nor was methylation status associated with the relative mRNA expression levels (P = 0.172 for p15^INK4b and P = 0.365 for p16^INK4a among the cases; P = 0.561 for p15^INK4b and P = 0.716 for p16^INK4a among the controls). However, the relative mRNA expression levels of p15^INK4b and p16^INK4a were highly correlated with each other (r = 0.5 and P < 0.001 for the cases and r = 0.7 and P < 0.001 for the controls). When the 48 controls were stratified by their primary cancer status, the mean age was 49.3 ± 11.9 years for those with ovarian cancer (n = 8), 52.2 ± 12.5 years for those with cancers other than ovarian cancer (n = 27), and 46.1 ± 8.4 years for those without any cancer (n = 13; Table 3).

The p15^INK4b methylation rate was higher among the 52 cases (31%) than among all 48 controls (6.3%; P = 0.004) and also higher than any of the control subgroups [8 with primary ovarian cancer (12.5%; P = 0.510), 27 with primary cancer other than ovarian (3.7%; P = 0.012), and 13 without cancer (7.8%; P = 0.178)]. In contrast, the p16^INK4a methylation rate seemed to be lower among the cases (25%) than among the controls (37.5% for all 48 controls, 37.5% for the 8 with unilateral ovarian cancer, 29.6% for the 27 with other cancers, and 53.9% for the 13 with no cancer), but none of the apparent differences was statistically significant (Table 3). Because the 8 controls with unilateral primary ovarian cancer had a higher p15^INK4b methylation rate (12.5%) than the other controls, they were excluded from the control group. The final control group used for further case-control comparisons consisted of 40 subjects. Compared with the case group, the 40 subjects in this final control group were younger (49 years; P = 0.01), had a lower p15^INK4b methylation rate (5.0%; P = 0.05), and had a slightly higher p16^INK4a methylation rate (37.5%; P = 0.288; Table 3).

Consistent with the methylation data, the mean ± SD relative p15^INK4b mRNA expression levels were lower among the 52 cases (12.9 ± 41.2%) than among all 48 controls (40.4 ± 56.9%; P = 0.05) and among the control subgroups [primary ovarian cancer (n = 8), 33.8 ± 35.1%; P = 0.180, primary cancer other than ovarian cancer (n = 27), 32.2 ± 47.4%; P = 0.065, and those without cancer (n = 13), 61.6 ± 80.2%; P = 0.003]. Similarly, the p16^INK4a expression levels were generally lower among the cases (27.0 ± 114%) than among the controls (31.9 ± 67.0% for all 48 controls, 27.6 ± 23.5% for the 8 with unilateral ovarian cancer, 20.7 ± 32.3% for the 27 with other cancer, and 57.9 ± 118.2% for the 13 with no cancer), but none of the apparent differences was statistically significant (Table 3). The final control group (n = 40) had higher levels of p15^INK4b expression (41.7 ± 61.1%; P = 0.008) and slightly higher levels of p16^INK4a expression (32.8 ± 27.0%; P = 0.754) than did those with ovarian cancer (Table 3).

Association between risk for ovarian cancer and p15^INK4b and p16^INK4a methylation status and mRNA expression levels. We then performed logistic regression analysis that included known risk factors for ovarian cancer, such as age (in years) and ethnicity, and used methylation status and dichotomized mRNA expression levels [using the median relative mRNA expression levels of the final controls (n = 40) as the cutoff points (23.6% for p15^INK4b and 14.7% for p16^INK4a)]. We found that ovarian tumor tissues were more likely to have methylated p15^INK4b than normal ovarian tissues (age- and ethnicity-adjusted OR, 5.68; 95% CI, 1.14-28.2), but no such association was evident for p16^INK4a (adjusted OR, 0.47; 95% CI, 0.16-1.37; Table 4). Furthermore, low mRNA expression levels of p15^INK4b were associated with a >9-fold increased risk of ovarian cancer (adjusted OR, 9.04; 95% CI, 2.51-32.5); and, low mRNA expression levels of p16^INK4a were associated with ∼3-fold increase in risk of ovarian cancer (adjusted OR, 2.81; 95% CI, 1.04-7.62; Table 4). Although the 52 cases were substantially older than the 40 controls (a 10-year difference in the mean age), the ORs with and without adjustment for age and ethnicity (Table 4) did not change the results substantially, suggesting that age may not have a major effect on either methylation or mRNA expression levels in this study population.

Finally, to assess whether the risk for ovarian cancer associated with low mRNA expression of both p15^INK4b and p16^INK4a was independent, we fit all variables (age, ethnicity, and methylation status and expression levels of both p15^INK4b and p16^INK4a) in one logistic regression model. We found that the risk for ovarian cancer was associated with both high methylation status (adjusted OR, 5.67; 95% CI, 0.85-37.9) and low mRNA expression levels (adjusted OR, 8.98; 95% CI, 1.58-50.9) of p15^INK4b but not of p16^INK4a (adjusted OR, 0.30; 95% CI, 0.09-1.07 for high methylation status and OR, 0.94; 95% CI, 0.21-4.29 for low mRNA expression levels), suggesting that p15^INK4b may play a major role in ovarian carcinogenesis but that other unknown factors may have caused changes in both genes.

We found here that promoter hypermethylation and low expression of p15^INK4b but not p16^INK4a were more common in ovarian cancer tissues than in normal ovarian tissues and that both changes were associated with increased risk of ovarian cancer. Although no association was found between the methylation status and mRNA expression levels for either p15^INK4b or p16^INK4a, their expression levels were highly correlated. In multivariate analysis, however, both hypermethylation rate and low expression of p15^INK4b but not p16^INK4a were independent risk factors for ovarian cancer. To the best of our knowledge, no reported studies have simultaneously investigated the association between promoter methylation status and mRNA levels of p15^INK4b and p16^INK4a and risk of ovarian cancer. Our results of p16^INK4a are consistent with other ovarian cancer studies, but our findings on p15^INK4b are novel.

The p15^INK4b and p16^INK4a genes are colocalized on chromosome 9p21. The p15^INK4b protein binds to one or more cyclin-dependent kinases and inhibits its functions in vitro, and ectopic expression of p15^INK4b inhibits cell growth in vitro (26). Although p15^INK4b and p16^INK4a have similar biochemical characteristics, their proteins have distinct functions in vivo. The expression of p15^INK4b but not that of p16^INK4a can be induced by transforming growth factor-β (27). The prevalence of point mutations in p16^INK4a vary in different tumor lineages, but point mutations are extremely rare in p15^INK4b (26), supporting the notion that other mechanisms may differentially regulate the expression of these two genes. Nevertheless, aberrant cytosine-DNA methylation in the promoter region of these genes that could disrupt the function of p15^INK4b and p16^INK4a has been found in numerous tumors, including ovarian cancer (4, 6, 9, 20, 28–31).

Although hypermethylation of p16^INK4a is common in breast, renal cell, colon, and prostate carcinomas (4, 28), the reported rates of p16^INK4a methylation in ovarian cancer tissues have ranged from 0% to 40% (21, 23, 32, 33). One study indicated that methylation of p16^INK4a was present in 40% of ovarian cancers and was associated with disease progression (8), whereas other studies showed methylation of p16^INK4a to be a rare event (21, 32–34). The discrepancies among these results may reflect different experimental designs, inclusion of ovarian cancer of different histologic types, and different methylation detection methods. By including only serous tumors and using a well-tested and sensitive MSP method (35), we found p16^INK4a to be methylated in 13 of 52 cases of serous epithelial ovarian cancers (25%), consistent with the most recent published findings (8). Interestingly, we also detected methylation of p16^INK4a in 15 of 40 normal ovarian tissues (37.5%), a finding that has not been reported elsewhere, and we found that methylation status did not differ substantially between normal and tumor tissues, suggesting that methylation of p16^INK4a may not play a major role in ovarian carcinogenesis.

However, previous studies of p16^INK4a expression in ovarian cancer are less discrepant. One study showed that 26% of 42 ovarian cancer samples did not express p16^INK4a protein, and this finding was unrelated to DNA methylation (36). In another study, low p16 protein expression was found in 22 of 60 ovarian epithelial tumors (37%) and correlated significantly with low p16 mRNA expression but was unrelated to gene deletion or point mutation (20). However, two later studies of different histologic subtypes of primary ovarian carcinoma showed that loss of p16 protein expression could be caused by hypermethylation of regions other than that of the promoter (37, 38). Although we did not measure p16^INK4a protein expression, we found that serous epithelial ovarian cancer tissue expression had similar levels of p16^INK4a mRNA to those of normal ovarian tissues, that mRNA expression of p16^INK4a was not associated with methylation of p16^INK4a, and that low mRNA expression of p16^INK4a was not an independent risk factor for ovarian cancer. However, the expression of p16^INK4a correlated highly with the expression of p15^INK4b, suggesting that p15^INK4b may have a role in ovarian carcinogenesis.

In contrast to p16^INK4a, methylation and mRNA expression of p15^INK4b in ovarian cancer have not been extensively investigated. Inactivation of p15^INK4b by CpG island hypermethylation has been reported to occur selectively in leukemias and gliomas but not in colon, breast, or lung carcinomas (4, 28). In one study, a p15^INK4b mutation occurred in only 1 of 70 ovarian tumors and homozygous deletion of p15^INK4b was observed in only 1 additional case (20), leading the authors to conclude that p15^INK4b did not have an important role in ovarian tumorigenesis (20). However, we found more hypermethylation and less mRNA expression of p15^INK4b in ovarian cancer than in normal ovarian tissues, and both were independent risk factors for ovarian cancer, further suggesting that inactivation of the p15^INK4b gene through epigenetic regulation could be one of the major events during ovarian carcinogenesis.

In the present study, we found that only 5% of normal ovarian tissues (2 of 40) had p15^INK4b methylation, similar to that 8% rate in healthy nonsmokers/nondrinkers found in a previous study (16). However, we found methylated p15^INK4b in 31% of ovarian cancers (16 of 52), a rate comparable with the 33% (15 of 45 cases) in a Japanese study (39). Studies of other cancers showed rates that were similar, such as 29% of 271 patients with acute lymphoblastic leukemia (40), or higher, such as 65% of 20 patients with head and neck squamous cell carcinoma (16) and 49% of 51 patients with hepatocellular carcinoma (41); however, in another study, no methylated p15^INK4b was detected in 44 medulloblastomas (42). Thus, methylation of p15^INK4b may be tumor specific and may be affected by environmental factors involved in the etiology of each type of tumor.

The fresh human tumor samples that we used may have contained both normal and tumor tissue, making detection of tumor-specific changes difficult. However, the sensitivity of MSP makes it useful for evaluating primary tumors, because it allows aberrantly methylated alleles to be detected even if they contribute relatively little to the overall DNA in a sample (24), thereby overcoming false-negative results due to contamination of normal tissues. The tissue samples obtained from unaffected ovaries in the controls in our study should not have had such problems, and it is unlikely that they would have yielded false-positive results. Although we cannot rule out the possibility that some cross-contamination might have occurred during experimentation, the consistency of our results with published data makes this possibility unlikely.

Overall, in the present study, abnormal methylation of both p15^INK4b and p16^INK4a in ovarian cancer and normal ovarian tissues did not predict mRNA expression levels, suggesting that other molecular mechanisms may have caused the changes in p15^INK4b and p16^INK4a mRNA expression. However, when methylation status and mRNA expression of both p15^INK4b and p16^INK4a were considered in the same multivariate logistic regressional model, both methylation status and mRNA expression of p15^INK4b but not p16^INK4a remained independent risk factors for ovarian cancer, suggesting that inactivation of the p15^INK4b gene through epigenetic regulation could be a key event during ovarian carcinogenesis. However, only ∼25% of epithelial ovarian cancers in general have methylated p15^INK4b, suggesting that other mechanisms, such as aberrations in the p14^ARF-MDM2-p53 pathway (43) or in the hMLH and PTEN (44) or SOCS genes (45), may contribute to the development of epithelial ovarian cancer in addition to aberrant p15^INK4b. Whether these epigenetic events play a role in the prognosis of epithelial ovarian cancer warrants further investigations (4, 25).

In summary, we found in this case-control analysis that both methylation and mRNA levels of p15^INK4b were independent risk factors for ovarian carcinogenesis and that, compared with p15^INK4b methylation, p15^INK4b mRNA levels were a strong predictor for the risk of ovarian carcinogenesis. However, no such associations were observed for p16^INK4a. These results support the hypothesis that cytosine-DNA methylation may differentially regulate the expression of p15^INK4b in ovarian cancer. It is likely that other regulating mechanisms, in addition to hypermethylation of the p15^INK4b promoter, may cause lower levels of p15^INK4b mRNA in patients with epithelial ovarian cancer. However, these findings need to be verified in larger studies with unaffected ovaries from cancer-free individuals.

We thank Ely J. Celestino for obtaining and processing samples and clinical data management, Betty Jean Larson and Joanne Sider for article preparation, and Rachel Williams (Department of Scientific Publications) for scientific editing.