Messenger RNA expression and methyiation of candidate tumor-suppressor genes and risk of ovarian cancer–a case-control analysis

Abstract

To investigate the association of expression and promoter methyiation of tumor-suppressor genes with risk of ovarian cancer, we conducted a case-control study of 102 patients with serous epithelial ovarian cancer and 100 patients without ovarian cancers. We measured mRNA expression levels (by real-time reverse transcription polymerase chain reaction) and methyiation status (by methylation-specific polymerase chain reaction) of five candidate genes (BRCA1, BRCA2, hMLHl, MGMT, and DNMT3B) in tumors from the cases and normal ovaries from the controls. We found that mRNA expression levels of the five genes were decreased in tumors than in normal ovaries with 0.39-fold for BRCA1, 0.25-fold for BRCA2, 0.42-fold for hMLHl, 0.45-fold for MGMT, and 0.87-fold for DNMT3B, calculated by the 2^-ΔΔCT method. Ovarian cancer risk (odds ratios, ORs) was associated with low expression of all genes (2.95 [95% confidence interval (CI), 1.51 - 5.78] for BRCA1, 3.65 (95% CI, 1.82 - 7.30) for BRCA2, 5.25 (95% CI, 2.52 - 10.96) for hMLHl, and 4.72 (95% CI, 2.32 - 9.62) for MGMT) but not DNMT3B. However, methyiation status was not associated with gene expression levels in the tumors, except for hMLHl whose mean (± SD) gene expression was significantly lower in methylated (13.0 ± 7.6) than in unmethylated (31.2 ± 44.8) tumors (P < 0.001). We concluded that low mRNA expression of these tumor-suppressor genes, likely due to molecular mechanisms in addition to the promoter methyiation in some instances, may be a biomarker for ovarian cancer risk in this study population. Larger studies are needed to validate our findings.

Introduction

Ovarian cancer is one of the most lethal malignancies in women worldwide [1, 2]. In the United States, ovarian cancer is the ninth most common malignancy and the fifth most common cause of death from female cancers. In 2009, the American Cancer Society estimated that 21,550 women will be diagnosed with ovarian cancer and that 14,600 women will lose their lives [3]. Because of the inability to detect ovarian cancer at its early stage that is highly treatable, more than two-third of patients are diagnosed with the advanced-stage disease, which leads to the survival rate essentially unchanged over the last decades. Although the molecular mechanisms leading to the development of ovarian cancer remain largely unknown, epigenetic alterations have been implicated. Therefore, further understanding epigenetic alterations underlying ovarian tumorigenesis may provide the basis for new tools for both identification of patients at risk and early diagnosis of ovarian cancer, which may ultimately reduce the incidence and mortality.

DNA methyiation at CpG sites in the promoter region of a gene can alter mRNA expression, which is one of the phenotypic characteristics of tumor development and progression [4-6]. The inactivation of tumor-suppressor genes due to aberrant methyiation of CpG islands has been implicated as one of the major pathways involved in the development of cancers, including ovarian cancer [6-8].

The importance of the role of aberrant methylation in the development of cancer has become increasingly apparent with the growing list of genes that has been shown to be susceptible to inactivation by promoter hypermethylation [9-15]. It has been observed that promoter methylation of specific genes in cancer occurs in both a tissue-specific and cell-specific manner, making the identification of methylation patterns a potentially useful tool for cancer diagnosis and management [9], particularly with the emerging high-throughput [16] and even genome-wide [13] technologies. It has been suggested that virtually all known cellular pathways contributing to carcino-genesis are more or less affected by epigenetic factors identified in cancer [13]. Because aberrant DNA methylation is frequently observed in early development of ovarian cancer, it has been predicted that such alterations can be detected in DNA circulating in the blood, potentially leading a non-invasive cancer detection test [17]. Specifically, frequent epigenetic inactivation of hMLHl, CDKN2A, and MG/W7were reported to be involved in ovarian carcinomas, using matched tumors and normal tissues from the same 18 patients [18], but another study showed a much less frequent methylation of hMLHl and MGMT in 13 ovarian cancer cell lines [19]. Such small studies often provide unstable estimates that are hard to replicate. In particular, the use of ovarian cancer cell lines without the control of normal ovaries from patients without ovarian cancer does not control for genetic effects on the carcinogenesis of normal ovaries.

In this study, we used a case-control design to investigate the association between ovarian cancer risk and mRNA expression levels and methylation of five candidate tumor-suppressor genes involved in DNA repair.

Materials and methods

Study subjects

Ovarian tumor tissues were obtained from patients with primary serous epithelial ovarian cancer newly diagnosed between January 2000 and March 2005 at The University of Texas M. D. Anderson Cancer Center. Tissues from normal ovaries, used as the control, were obtained from patients who underwent surgery during the same time period for conditions other than ovarian cancer. Informed consent was obtained from each patient, and the study was approved by M. D. Anderson's institutional review board. All samples were snap-frozen after surgical removal and then stored at −80°C in the Gynecologic Cancer Tumor Bank at M. D. Anderson Cancer Center until pathologic examination and testing. For this case-control study, we obtained 102 surgically-resected ovarian tumors and 100 apparently normal ovarian tissues and DNA and RNA were extracted from about 200 mg of fresh-frozen tissue specimens.

Real-time reverse transcription polymerase chain reaction for gene expression

In this study, we measured five tumor-suppressor genes involved in DNA repair: BRCA1, BRCA2, hMLHl, MGMT, and DNMT3B using GAPDH as the internal control. 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 by 1% agarose gel electrophoresis for RNA degradation by visualizing the 18S and 28S RNA bands under ultraviolet light as shown previously with two clean bands [20]. The RNA concentration was determined with the Gene Quant Pro RNA/DNA Calculator (Amersham Pharmacia, Cambridge, England) before the detection of specific gene expression. The primers and probes for detecting mRNA levels of MGMT, hMLHl, and GAPDH were used as previously reported [20-22]. The cDNA sequences of BRCA1, BRCA2, and DNMT3B were referenced to design the primers and probes using express software from Applied Biosystems (Foster City, CA). All sequences of primers and probes are summarized in Table 1. Reverse transcription polymerase chain reaction (RT-PCR) was performed using TaqMan one-step RT-PCR Master Mix Reagents kit (Applied Biosystems) according to the manufacturer's protocol as previously described [20].

Table 1.

Oligonucleotide primer and probe sequences used in this study

Gene		Primer/probe sequence*	Position/PCR† product size
BRCA1	Forward primer	TTTCTATTTGGATCCCTTCGAGG	136 - 158
	Reverse primer	GTGAGCGCACTTCTGCCC	185 - 202/67 bp
	Probe	FAM-CCCCCGTGGCTGTGGAACCC-TAMRA	164 - 183
BRCA2	Forward primer	TGCTGCAAGCAACCTCCA	9587 - 9604
	Reverse primer	AGAAAAATCTCCAGCAAATAAAGTAAGAA	9631 - 9659/73 bp
	Probe	FAM-TGGCGACCAGAATCCAAATCAGGC-TAMRA	9606 - 9629
hMLH1	Forward primer	GTTCTCCGGGAGATGTTGCATA	1579 - 1600
	Reverse primer	TGGTGGTGTTGAGAAGGTATAACTTG	1661 - 1681/ 102 bp
	Probe	FAM-CCTCAGTGGGCCTTGGCACAGC-TAMRA	1627 - 1644
MGMT	Forward primer	CAATGAGAGGCAATCCTGTCC	494 - 514
	Reverse primer	CACGGCTCCGCTGCTG	546 - 561/ 68 bp
	Probe	VIC-CTCATCCCGTGCCACAGAGTGGTCT-TAMRA	520 - 544
DNMT3B	Forward primer	TCTCCTATCGAAAAGCCATGTA	1208 - 1229
	Reverse primer	GGGAAGGTCTTGCCAGC	1258 - 1274/ 67 bp
	Probe	FAM-CATGCTCTGGAGAAAGCTAGGGTGC-TAMRA	1231 - 1255
GAPDH	Forward primer	GAAGGTGAAGGTCGGAGTC	131 - 149
	Reverse primer	GAAGATGGTGATGGGATTTC	337 - 356/226 bp
	Probe	FAM-CAAGCTTCCCGTTCTCAGCC-TAMRA	308 - 327

^*VIC, FAM, and TAMRA are different dyes that were used to label two ends of the probes.

^†The nucleotide (nt) position of the cDNAs with GenBank accession numbers XM_208274 for BRCA1, NM_000059.2 for BRCA2, U_07343 for hMLHl, M29971 for MGMT, NM_006892 for DNMT3B, and AK026525 for GAPDH.

Methylation-specific PCR

The methylation status of target genes were qualitatively analyzed as described previously [20]. Briefly, Genomic DNA samples were modified with sodium bisulfite. 1 μg of DNA was denatured by NaOH (50 μ1; final concentration, 0.2 M) for 10 min at 37°C, mixed with 30 μ1 of freshly prepared 10 mM hydroquinone (Sigma, St. Louis, MO) and 520 JLXI of 3 M, pH 5.0 sodium bisulfite (Sigma), and incubated under mineral oil at 55°C for 16 h. The DNA samples were desalted through Wizard columns (Promega, Madison, WI) and then desulphonated by NaOH treatment (final concentration, 0.3 M) for five minutes at room temperature followed by ethanol precipitation. DNA was resuspended in water and used shortly after reconstitution. For PCR amplification, the bisulfite-modified DNA (100 ng) was separately amplified using published primers specific for the methylated as well as the unmethylated sequences of genes including BRCA1 [23], MGMT [24], and hMLHl [25]. Since the BRCA2 gene is rarely methylated and there is no report on promoter methylation of the DNMT3B gene, we did not perform methylation-specific PCR assays for these two genes. CpGenome Universal Methylated DNA (Serologicals Corporation, Norcross, GA) 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 as previously described [20]. PCR products were analyzed on 2% agarose gels containing ethidium bromide (Figure 1A). Two researchers (JA and ZL) independently evaluated the results, and questionable assays were repeated to achieve complete agreement.

Figure 1.

A, Methylation-specific PCR analysis of the methylation status in BRCA1, hMLHl, and MGMT. Representative PCR products of the promoter region of these genes were amplified by the MSP method. P, positive control (CpGnome Universal Methylated DNA); Ca, ovarian cancer tissues; Cn, normal ovarian tissues; N, negative control (water blank); M, methylated; U, unmethylated. B, Relative mRNA expression levels of methylated and unmethylated BRCA1, hMLHl, and MGMT in the cases.

Statistical analysis

The 2^-ΔΔCT method was used to calculate changes in candidate gene expression levels in tumor tissues normalized to the internal control GAPDH and relative to the normal tissues as reported [26-28]. The Student's t-test was used to compare differences in the relative expression levels to the internal control GAPDH for the subgroups, which were analyzed as a continuous variable between groups. Two-sided X² test was used for the comparison of categorical variable distribution between two groups. For calculating odds ratios (ORs) and 95% confidence intervals (CIs), the median relative expression level of each gene in the controls was used as the cutoff point. Adjusted ORs were calculated by fitting logistic regression models with adjustment for age and ethnicity. All statistical analyses were performed with SAS software (version 9.1; SAS Institute Inc., Cary, NC).

Results

Demographic characteristics for the study population are summarized in Table 2. There was significant difference in age between the case and control groups. Controls (53%) were younger (<50 years) than patients (13%) (Table 2). The mean age of cases (62.3 ± 10.0 years [±SD]) was significantly higher than that of controls (50.7 ± 14.1 years) (P < 0.001), and ages ranged from 39 to 81 years for cases and from 23 to 85 years for controls (Table 3). About 78% of cases and 74% of controls were non-Hispanic whites. The other one-forth of subjects consisted of small numbers in minority groups including African-, Mexican-, and Asian-American and other ethnicities as shown in Table 2. All cases were diagnosed with high-grade serous ovarian tumors, which had 93%stage III or IV tumors (Table 2).

Table 2.

Distribution of demographic characteristics of patients with ovarian cancer (cases) and with normal ovarian tissues (controls)

	Cases (n = 102)		Controls (n = 100)
Variable	No.	%	No.	%	P value*
Age (years)					<0.001
<50	13	(12.8)	53	(53.0)
≥50	89	(87.2)	47	(47.0)
Ethnicity					0.007
Non-Hispanic white	80	(78.4)	74	(74.0)
African American	5	(4.9)	7	(7.0)
Mexican American	9	(8.8)	19	(19.0)
Others†	8	(7.9)	0	(0.0)
Tumor grade
High	102	(100.0)
Tumor stage
I + II	7	(6.9)
III	71	(69.6)
IV	24	(23.5)

^*Two-sided χ² tests.

^†Others included five Asian cases and three cases with unknown ethnicity.

Table 3.

mRNA Expression of candidate genes in ovarian tumors and normal ovarian tissues

Variable	Controls		Cases		P vlaue†	Stage I + II		Stage III		Stage IV		P value‡
	No.	Mean ± SD*	No.	Mean ± SD*		No.	Mean ± SD*	No.	Mean ± SD*	No.	Mean ± SD*
Age, range (years)	100	50.7 ± 14.1, 23-85	102	62.3 ± 10.0, 39-81	<0.001	7	64.0 ± 10.7, 48-78	71	63.5 ± 10.1, 39 - 81	24	58.4 ± 8.9 45-74	0.087
mRNA Expression
BRCA1	97	25.4 ± 31.5	98	17.0 ± 54.6	0.190	7	12.6 ± 28.6	69	17.9 ± 61.8	22	15.6 ± 34.3	0.963
Methylated	50	29.0 ± 37.1	45	20.7 ± 76.3	0.511	3	2.2 ± 0.6	27	29.0 ± 79.5	15	9.4 ± 34.3	0.674
Unmethylated	47	21.6 ± 24.1	53	13.8 ± 24.7	0.118	4	20.4 ± 38.1	42	10.7 ± 21.1	7	28.8 ± 52.7	0.173
		0.246		0.567			0.409		0.341		0.380
BRCA2	99	12.9 ± 15.5	91	6.2 ± 9.6	<0.001	7	11.7 ± 25.9	63	5.3 ± 5.5	21	7.1 ± 10.4	0.224
hMLH1	97	47.4 ± 25.4	102	27.1 ± 40.3	<0.001	7	12.8 ± 8.1	71	29.2 ± 43.3	24	25.0 ± 36.4	0.570
Methylated	33	43.7 ± 23.7	23	13.0 ± 7.6	<0.001	2	12.1 ± 8.5	13	12.1 ± 8.2	8	14.6 ± 7.1	0.769
Unmethylated	64	49.3 ± 26.2	79	31.2 ± 44.8	0.003	5	13.1 ± 8.9	58	33.1 ± 47.0	16	30.3 ± 43.9	0.636
		0.308		<0.001			0.901		0.002		0.183
MGMT	100	94.1 ± 91.2	101	66.9 ± 123.2	0.057	7	23.4 ± 12.8	70	77.0 ± 144.0	24	41.6 ± 46.5	0.314
Methylated	14	84.3 ± 95.3	33	48.4 ± 47.9	0.199	3	26.6 ± 16.4	24	57.2 ± 52.4	6	23.9 ± 22.8	0.227
Unmethylated	86	95.7 ± 91.0	68	72.9 ± 146.1	0.262	4	21.1 ± 11.5	46	87.4 ± 173.4	18	47.4 ± 51.3	0.479
		0.667		0.214			0.622		0.282		0.294
DNMT3B	97	6.2 ± 17.7	93	5.9 ± 6.6	0.895	7	4.6 ± 4.1	66	5.9 ± 7.0	20	6.5 ± 6.2	0.817

^*mRNA expression is the expression level relative to that of the GAPDH gene / 10.

The sample size in each gene is less than the total number because the assay failed or the expression values are out of the 90% confidence interval.

^†Two-sided Student's t-tests for the difference s in the means between cases and controls.

^‡Analysis of variance tests for the differences among the stages within cases.

We conducted RT-PCR assays to assess the relative mRNA expression levels of BRCA1, BRCA2, hMLHl, MGMT, and DNMT3B in cases and controls (Table 3). We performed an analysis of variance for differences in the relative mRNA expression levels of these genes among subgroups of age and ethnicity in both cases and controls but did not find statistically significant differences. Therefore, we combined all ethnicity groups together in the following analysis. Overall, mean mRNA expression levels of BRCA2 and hMLHl were significantly lower in ovarian tumors than in normal ovaries (P < 0.001 for both genes), whereas the difference in expression levels of MGMT between cases and controls was approaching significant (P = 0.057). There were no statistically significant differences in the mean mRNA expression levels of BRCA1 and DNMT3B between cases and controls.

We then evaluated the association between the risk of ovarian cancer and mRNA expression levels of the five tumor-suppressor genes and found that the risk was associated with low levels of mRNA expression in all genes but DNMT3B. Specifically, using the control median as the cutoff value, low expression levels were associated with a 2.95-fold increased risk (95% CI, 1.51 - 5.78) for BRCA1, a 3.65-fold increase (95% CI, 1.82 -7.30) for BRCA2, a 5.25-fold increase (95% CI, 2.52 - 10.96) for hMLHl, and a 4.72-fold increase (95% CI, 2.32 - 9.62) for MGMT after adjustment for age and ethnicity. In contrast, no increased risk was associated with the mRNA expression level of DNMT3B (adjusted OR, 0.59; 95% CI, 0.31 - 1.12) (Table 4).

Table 4.

Crude and adjusted odds rations (ORs) and 95% confidence intervals (CIs) for the relative gene expression levels in ovarian tumors and normal ovarian tissues

Expression level*	No. (%) of cases(N = 102)	No. (%) of controls (N = 100)	P value†	Crude OR (95% CI)	Adjusted OR‡ (95% CI)
BRCA1
High	25 (25.5)	49 (50.5)	0.0003	1.00	1.00
Low	73(74.5)	48 (49.5)		2.98(1.63-5.45)	2.95(1.51-5.78)
BRCA2
High	23 (25.3)	50 (50.5)	0.0004	1.00	1.00
Low	68(74.7)	49 (49.5)		3.02(1.63-5.58)	3.65(1.82-7.30)
hMLHl
High	17 (16.7)	49 (50.5)	<0.001	1.00	1.00
Low	85 (83.3)	48 (49.5)		5.10(2.65-9.83)	5.25 (2.52 -10.96)
MGMT
High	20 (19.8)	50 (50.0)	<0.001	1.00	1.00
Low	81 (80.2)	50 (50.0)		4.05(2.16-7.58)	4.72(2.32-9.62)
DNMT3B
High	59 (63.4)	48 (49.5)	0.053	1.00	1.00
Low	34 (36.6)	49 (50.5)		0.57(0.32-1.01)	0.59(0.31-1.12)

^*The median relative mRNA expression level in the controls was used as the cutoff point for each gene. The sar in each gene is less than the total number because the assay failed or the expression values are out of the 90% confidence interval.

^†Two-sided χ²-test

^‡Adjusted for age (in years) and ethnicity (non-Hispanic whites versus others) in a logistic regression model.

Finally, we assessed whether low mRNA expression by these tumor-suppressor genes in ovarian tumors was due to altered promoter methylation status. The expression levels by methylation status in BRCA1, hMLHl, and MGMT in the cases are summarized in Figure 1B. We found that only methylated MGMT was significantly higher in the cases than in the controls (32.7% vs. 14.0%; P = 0.002) (data not shown). The stratification of mRNA expression levels by methylation status in ovarian tumors and normal ovarian tissues is presented in Table 3. Although in cases and controls, methylated MGMT and methylated hMLHl showed lower gene expression levels than their unmethylated counterparts did, the difference was statistically significant only for hMLHl in cases: the mean expression 13.0 ± 7.6 in 23 methylated tumors and 31.2 ± 44.8 in 79 unmethylated tumors (P < 0.001).

We also compared mRNA expression levels among tumor stages I and II, III, and IV but found no statistical differences or trends. However, methylated hMLHl in 13 stage III tumors had significantly lower expression_than unmethylated hMLHl did in 58 stage III tumors (P = 0.002) (Table 3).

Discussion

In this case-control study, we found that low levels of the relative mRNA expression of BRCA1, BRCA2, hMLHl, and MGMT, but not of DNMT3B, were associated with a significantly increased risk of ovarian cancer. However, except for hMLHl, the methylation status of the genes did not appear to explain the observed lower expression levels.

Inactivation of tumor-suppression genes BRCA1 and BRCA2 in ovarian tumors has been reported by other investigators. For example, promoter hypermethylation of the BRCA1 gene was found to be between 5% and 36% of tumors in primary ovarian carcinomas, a molecular event that has been proposed as a potential cause of the gene inactivation [8, 10, 23, 29-33]. We found in our study that the relative mRNA expression levels of BRCA1 were significantly lower in ovarian tumors than in normal ovaries of subjects without ovarian cancer; however, this difference was not attributable to the promoter methylation status in BRCA1. We detected a much higher methylation status of BRCA1 (46.6%) than previous reports did [8, 10, 23, 29-33], and we even found that BRCA1 methylation commonly existed in unaffected ovaries (51.5%) of the subjects with conditions other than ovarian cancer. It is known that BRCA1 and BRCA2 genes are involved not only with DNA repair but also with hormone regulation; therefore, the ovarian cell type and status may need to be strictly defined in such methylation studies, and the best controls may be the normal ovaries of subjects without hormonally related conditions or cancers other than ovarian cancer.

A correlation between hMLHl hypermethylation, loss of expression, and microsatellite instability has been demonstrated in colorectal, gastric, endometrial [25, 34-36], and ovarian cancers [31, 37, 38]. The frequencies of hMLHl promoter methylation have been reported to range from 9% to 39% [32, 33, 39, 40]. In our study, we observed lower hMLHl expression that was associated with increased risk of ovarian cancer, and among the 102 cases, hMLHl expression was significantly lower in 23 methylated tumors than that in 79 unmethylated tumors.

We also observed that MGMT mRNA expression was lower in ovarian tumors than in normal ovaries, with 31.1% of the promoters methylated in cases, a finding consistent with a recently published study that had a much smaller number of study subjects (only 18 ovarian carcinomas) [18] and reported a high frequency (48%) of methylation status of MGMT. MGMT hypermethylation was less frequently observed in one study using ovarian cancer cell lines (in only 23% of 13 cell lines) [19] and in another study of ovarian granulosa cell tumors (in 33% of 43 subjects) [41]; however, MGMT hypermethylation was not detected in a recent study of 120 patients with endometrial cancer [42]. Our results indicate that the incidence rate of MGMT promoter methylation was significantly higher in cases than in controls. However, abnormal promoter methylation of MGMT in serous ovarian tumors and normal ovaries did not predict mRNA expression levels in our study, suggesting that molecular mechanisms other than methylation may contribute to the altered MGMT mRNA expression observed in this study.

The strength in the present study is the use of fresh-frozen tissues in a relatively large number of study subjects and a population of cases with homogenous high-grade serous epithelial ovarian cancer. Also, we compared the relative mRNA expression levels and methylation status of candidate tumor suppressor genes of both serous ovarian tumors from cancer patients and unaffected ovaries from subjects without ovarian cancer. In the present study we did not observe an association between DNMT3B gene expression and ovarian cancer risk, which has not been reported to date. The results of DNMT3B gene expression may serve as an internal control in the present study, which suggest no systematic errors in the measurements of gene expression that may have occurred in the experiments of other genes. Therefore, the observed risk in the present study could not be biased by systematic errors in the assays, by sample collection method, or by possible experimental error. However, because ovarian cancer likely arises from the ovarian surface epithelium and can be contaminated with normal tissues, microdissection of ovarian tumor tissues may be required for future ovarian cancer association studies. Further detection of genetic mutations in ovarian tumor tissues may help explain the underlying mechanisms of reduced gene expression, with quantization of CpG methylation to provide a more accurate estimation of methylation status and correlation with gene expression levels.