|Home | About | Journals | Submit | Contact Us | Français|
Although the etiology of primary brain tumors is largely unknown, prior studies suggest that DNA repair polymorphisms may influence risk of glioma. Altered DNA repair is also likely to affect the risk of meningioma and acoustic neuroma, but these tumors have not been well studied. We estimated the risk of glioma (n = 362), meningioma (n = 134), and acoustic neuroma (n = 69) in non-Hispanic whites with respect to 36 single nucleotide polymorphisms from 26 genes involved in DNA repair in a hospital-based, case–control study conducted by the National Cancer Institute. We observed significantly increased risk of meningioma with the T variant of GLTSCR1 rs1035938 (ORCT/TT = 3.5; 95% confidence interval: 1.8–6.9; Ptrend .0006), which persisted after controlling for multiple comparisons (P = .019). Significantly increased meningioma risk was also observed for the minor allele variants of ERCC4 rs1800067 (Ptrend .01); MUTYH rs3219466 (Ptrend .02), and PCNA rs25406 (Ptrend .03). The NBN rs1805794 minor allele variant was associated with decreased meningioma risk (Ptrend .006). Risk of acoustic neuroma was increased for the ERCC2 rs1799793 (Ptrend .03) and ERCC5 rs17655 (Ptrend .05) variants and decreased for the PARP1 rs1136410 (Ptrend .03). Decreased glioma risk was observed with the XRCC1 rs1799782 variant (Ptrend .04). Our results suggest that common DNA repair variants may affect the risk of adult brain tumors, especially meningioma.
Tumors of the brain, meninges, and cranial nerves account for approximately 95% of all central nervous system tumors and include some of the most rapidly fatal cancer types.1,2 In normal cells, DNA damage is recognized by cellular mechanisms, and one of the chief responses to prevent the propagation of errors and subsequent initiation and growth of tumors is the repair of damaged DNA.3,4 Ionizing radiation, the only confirmed environmental risk factor for brain tumors in humans, produces several kinds of DNA damage, including oxidative damage to nucleotide bases, single- and double-strand breaks in DNA chains, and DNA–DNA or DNA–protein covalent cross links. Such damage also can be caused by other exogenous agents, as well as endogenous processes, such as oxidative metabolism and inflammation. The repair of this damage involves several molecular pathways of DNA repair, including base-excision repair, nucleotide excision repair, non-homologous end joining, and homologous recombination.5
Previous studies have noted that single nucleotide polymorphisms (SNPs) in DNA repair genes including CHAF1,6 LIG4, XRCC4,7 XRCC5, XRCC6,8 XRCC7,9 MGMT,10 XRCC1,11 XRCC3,11 ERCC1,12,13 ERCC2,14 and GLTSCR114 may modify glioma risk. Although fewer studies have examined meningioma, results suggest that genes in DNA repair pathways could play an important part in the etiology of this tumor,11,15,16 with increased meningioma risk for the BRIP1 rs4968551 minor allele variant being highly statistically significant.16 Previous studies have not examined meningioma risk with respect to variation in the GLTSCR1, MUTYH, or PCNA genes. To our knowledge, no previous studies have examined the risk of acoustic neuroma with respect to DNA repair polymorphisms. This study conducts an exploratory investigation of whether variation in common DNA repair genes is associated with acoustic neuroma. Using data from non-Hispanic whites in a hospital-based, case–control study conducted by the National Cancer Institute (NCI) between 1994 and 1998, we evaluated the risk of glioma (n = 362), meningioma (n = 134), and acoustic neuroma (n = 69) with respect to 36 SNPs from 26 genes involved in DNA repair. These genes and polymorphisms were selected based on the available data from the literature and SNP500 database (http://snp500cancer.nci.nih.gov/home_1.cfm) regarding relevance for brain tumors, common occurrence in the population, and potential functional relevance signaled by non-synonymous amino acid (AA) changes or occurrence in exonic or promoter regions of the gene (Table 1).
A detailed description of the study methods can be found elsewhere.17 Briefly, eligible patients were 18 years or older with a first intracranial intracranial glioma or neuroepitheliomatous tumor (ICD-O-2 codes 9380–9473 and 9490–9506), meningioma (ICD-O-2 codes 9530–9538), or acoustic neuroma (ICD-O-2 code 9560) diagnosed during 1994–1998 at 1 of 3 hospitals specializing in brain tumor treatment (in Boston, Phoenix, and Pittsburgh) within the 8 weeks preceding hospitalization. Ninety-two percent of the eligible brain tumor patients agreed to participate, and 489 patients with glioma, 197 with meningioma, and 96 patients with acoustic neuroma were enrolled, with all but 4% of acoustic neuromas being confirmed by microscopy.
Controls were admitted to the same hospitals for injuries (25%), circulatory system disorders (22%), musculoskeletal disorders (22%), digestive disorders (12%), or a variety of other non-neoplastic conditions, and were frequency-matched in a 1:1 ratio to all brain tumor patients based on age (18–29, 30–39, 40–49, 50–59, 60–69, 70–79, and 80–90 years); race/ethnicity (non-Hispanic white, Hispanic, African American, other), sex, hospital, and residential proximity to the hospital. Seven hundred ninety-nine control patients (86% of all contacted) were enrolled. The study protocol was approved by the Institutional Review Board of each participating institution, and written informed consent was obtained from each patient or proxy. This analysis was restricted to non-Hispanic whites (89% of all study participants) who provided blood samples. All study participants were alive at the time of interview and were therefore eligible and able to provide blood specimens. For non-Hispanic whites who had consented to provide blood samples, samples were genotyped for 362 patients with glioma (82% of all non-Hispanic whites), 134 patients with meningioma (82%), 69 patients with acoustic neuroma (78%), and 495 controls (69%). The main obstacle to obtaining blood samples was subject refusal, with non-participation in the blood draw being higher for controls (24%) than for cases (14%).
DNA repair polymorphisms were selected based on minor allele frequency >.05 according to SNP500, putative functional importance, and/or evidence of an association with cancer risk. DNA was extracted using a phenol-chloroform method, and genotyping was conducted using TaqMan assays. Primer and probe sequences as well as assay conditions can be found on the SNP500 website (http://snp500cancer.nci.hih.gov).18 Three hundred eighty-four-well plates were used, with each plate containing 368 study specimens and 16 controls (4 homozygous wild-type, 4 heterozygous, 4 homozygous variant positive controls, and 4 DNA-negative controls). Quality control (QC) specimens included 10–34 samples from 3 non-study participants and duplicates from 57 study subjects that were interspersed among all genotyping assays in a masked fashion. Three SNPs that were found to have replicate concordance of less than 95% were dropped from the analysis (MLH1 rs1799977, PRKDC rs7003908, and ERCC1 rs3212986). Results are reported for all other selected SNPs.
Statistically significant departure from the Hardy–Weinberg equilibrium for controls was assessed using the χ2 test. For each polymorphism, unconditional logistic regression was used to calculate odds ratios (ORs) and 95% confidence intervals (CI) for each major tumor type, adjusted for the study matching factors of age, sex, hospital, and residential proximity to hospital. Since controls were frequency matched to all tumor types, all controls were used in the models for each tumor type. Models were run under the assumption of co-dominant (AA vs Aa vs aa) and dominant (AA vs Aa or aa) inheritance. A score test of linear trend was conducted for each SNP using a 3-level ordinal variable. In order to evaluate possible bias introduced by using disease controls, regression models were repeated for each SNP, excluding one major subset of disease controls at a time. For each major tumor type (glioma, meningioma, acoustic neuroma), trend P values from the 3-level ordinal model (36 contrasts per tumor) were adjusted for multiple comparisons using the false discovery rate,19 with α = 0.05.
Percent agreement between the 3 non-study replicates ranged from 97% to 100% for all SNPs. Duplicate concordance was 93% for GLSTCR1 rs1035938, 95% for RAD2 rs11226, and ranged between 98% and 100% for all remaining SNPs. Hardy–Weinberg equilibrium in controls showed no significant deviation except for the LIG4 rs1805388 (P = .02) and XRCC3 rs861539 (P = .04) polymorphisms. Genotyped subjects, 1060 (75%) of the 1411 non-Hispanic white participants, were similar to all study subjects except for the lower proportion of those aged 70–90 and those with less education. Compared with controls, a larger proportion of glioma subjects were male, whereas subjects with meningioma and acoustic neuroma showed a female predominance and were, on average, older than controls (Table 2).
Results for dominant and co-dominant models for SNP main effects are presented in Table 3. The GLTSCR rs103598 T variant was associated with significantly increased risk of meningioma (ORCT = 3.5, 95% CI: 1.8–7.0; ORTT = 3.6, 95% CI: 1.2–11.2, Ptrend = .0006), which persisted after controlling for multiple comparisons (adjusted P value = .019). Risk of meningioma was also significantly increased for the minor allele variants of ERCC4 rs1800067 (ORAG/AA = 2.1, 95% CI: 1.2–3.6; Ptrend .01); MUTYH rs3219466 (ORCT/TT = 2.1, 95% CI: 1.0–4.3; Ptrend .02), and PCNA rs25406 (ORCT/TT = 2.1, 95% CI: 1.3–3.5; Ptrend .03). The NBN rs1805794 minor allele variant was associated with decreased meningioma risk (ORCG/CC = 0.5, 95% CI: 0.3–0.8; Ptrend .006). Decreased glioma risk was observed with the XRCC1 rs1799782 variant (ORCT/TT = 0.7, 95% CI: 0.4–1.0; Ptrend .04). The small sample size for acoustic neuroma limited our ability to assess risk by genotype, particularly for homozygous variant genotypes. Nonetheless, the risk of acoustic neuroma was statistically significantly increased for the ERCC2 rs1799793 (ORAG/AA = 2.1, 95% CI: 1.2–3.7; Ptrend .03) and ERCC5 rs17655 (ORCG/CC = 1.8, 95% CI 1.0–3.1; Ptrend .05) variants and decreased for the PARP1 rs1136410 (ORAG/AA = 0.5, 95% CI: 0.3–1.0; Ptrend .03). We observed no statistically significant associations at a P-level of < .05 between genotype and risk of meningioma, glioma, or acoustic neuroma for the remaining polymorphisms. Results remained very similar when major groups of disease controls were excluded from the analysis, one at a time.
Prior studies of brain tumors, mainly glioma, have indicated that common genetic variation in DNA repair genes might affect brain tumor risk.6–16 Although very few prior studies have examined the risk of meningioma with DNA repair genes, there is some indication that meningioma may be more susceptible to common changes in DNA repair genes than glioma. In a previous large multicenter study that examined the risk of glioma and meningioma in 1127 polymorphisms of DNA repair genes, the top finding for meningioma, BRIP1 rs4968451, was associated at a level of P = 8.95 × 10−6 (OR = 1.6, 95% CI: 1.3–1.9),16 whereas the top finding for glioma, CHAF1A rs243356, was associated at a level of P = 2 × 10−4 (OR = 1.3, 95% CI: 1.1–1.5) despite having a larger sample size and therefore more power to detect an association.6 Prior studies have not examined meningioma risk with respect to variation in the GLTSCR1, MUTYH, or PCNA genes. To our knowledge, the risk of acoustic neuroma with respect to DNA repair polymorphisms has not been examined in prior epidemiological studies.
The strongest association observed in our candidate gene study of 36 SNPs in the DNA repair pathway was the 3.5-fold increased risk of meningioma with the GLTSCR1 rs1035938 T allele. This association remained statistically significant even after controlling for multiple comparisons. Although previous studies have not examined the risk of meningioma with the GLTSCR1 rs1035938 polymorphism, one prior study has reported a 3-fold increased risk of oligodendroglioma with the T allele of GLTSCR1 rs1035938 polymorphism.14 Consistent with our results for glioma, the same study observed no association between the GLTSCR1 rs1035938 polymorphism and risk of all glioma. Interestingly, individuals with gliomas and oligodendrogliomas with the TT genotype had better survival than CT/CC individuals.14 GLTSCR1, which stands for glioma tumor suppressor candidate region 1, is highly conserved among humans, chimps, mice, and rats, suggesting that it may contain key genetic information. Data from MTN (multiple tissue Northern blot) hybridization suggest that the gene transcript is approximately 6.5 kb and exhibits moderate expression in the brain.20 The GLTSCR1 rs1035938 polymorphism is known to alter a CpG within the CpG island flanking the 5′ region of the gene. This may affect the transcription of GLTSCR1 and other candidate genes in the region.14
The minor allele variants for NBN rs1805794, ERCC4 rs1800067, MUTYH rs3219466, and PCNA rs25406 were also statistically significantly associated with meningioma in our study (P < .05). Associations between meningioma and gene variants in GLTSCR1, MUTYH, and PCNA have not been previously reported. However, our observation of increased risk with ERCC4 rs1800067 is consistent with a statistically significant association observed in a multicenter European study of DNA repair genes and meningioma (P = .02), and several SNPs in the NBN gene were found to be of borderline significance in that same study.16
Only one SNP was associated with glioma at P < .05, XRCC1 rs1799782. Although previous studies have not reported on this particular polymorphism, 2 other SNPs in the XRCC1 gene (rs3213266 and rs2854496) were associated with glioma risk in a multicenter study of glioma and DNA repair.6 No other SNPs in XRCC1 were associated with glioma risk in this or a previous study.9 Our study results did not replicate previously reported findings of statistically significant or borderline associations for ERCC2 rs13181 and glioma13 or XRCC1 rs25487 and glioma.11
To our knowledge, this is the first study to examine the risk of acoustic neuroma with respect to common variants in DNA repair genes. Given the relatively small sample size for acoustic neuroma, we underscore the exploratory nature of the findings for this tumor. We found an elevated risk of acoustic neuroma with the ERCC2 rs1799793 and ERCC5 rs17655 variants and decreased risk for the PARP1 rs1136410 minor allele variant. Although these variants have not been examined previously in the context of acoustic neuroma, ERCC2 has been reported to be down-regulated in astrocytoma compared with normal brain tissue,21 and other polymorphisms in ERCC2 have been associated with meningioma,15 oligodendroglioma,14 and certain subsets of glioma.13 The C allele of ERCC5 rs17655, although not associated with glioma or meningioma risk in a previous study,6,16 has been associated with a borderline main effect association with breast cancer risk in previous studies,22–24 with some suggestion that individuals carrying the C allele variant of this SNP are more susceptible to the effects of ionizing radiation.24 PARP1 rs1136410 was not associated with glioma or meningioma risk in one previous study,6,16 but a small study in French patients reported that 21 rare genetic variants of PARP1, including rs1136410, were detected in 11% of patients with breast cancer.25
Our study had adequate statistical power to detect moderate to strong main effects (OR ≥ 1.5) of common genetic polymorphisms for glioma and meningioma. Other strengths include standardized genotyping, high reproducibility of the genotyping results in most of the QC samples, and controls in Hardy–Weinberg equilibrium for all but 2 polymorphisms. Given that deviation from Hardy–Weinberg equilibrium was not extreme (P < .01) for either of these polymorphisms and that we observed no significant associations for the 2 SNPs in question, this is unlikely to affect our results. Rapid ascertainment of brain tumor cases and blood collection close to the date of diagnosis reduced the possibility that survival bias affected our results. Although there is the possibility of bias introduced by the use of hospital controls, results of the analyses were very similar after excluding major groups of disease controls, one at a time, suggesting that bias is unlikely to completely explain the findings.
Nevertheless, we underscore the need for replication of our findings given the false-positive reports frequently generated in genetic association studies, the possibility that the notable SNPs are actually in linkage disequilibrium with other causally relevant polymorphisms, and the relatively low duplicate concordance rate for GLSTCR rs1035938. While non-participation in the blood draw was higher among controls than cases, we believe that this is unlikely to be related to genotype, and thus unlikely to bias our results. Our results for acoustic neuroma are limited by the small sample size for this tumor.
Our findings suggest that GLTSCR1, NBN, and ERCC4 are especially promising candidate genes to examine in terms of meningioma risk. Other genes of interest are MUTYH and PCNA in terms of meningioma risk, XRCC1 for glioma, and ERCC2, ERCC5 rs17655 and PARP1 for acoustic neuroma. Replication of these results in large consortial studies of brain tumors is needed.
This research was supported by intramural funds from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services. This project has been funded in whole or in part with federal funds from the National Cancer Institute, National Institutes of Health, under contract N01-CO-12400.
The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.
Conflict of interest statement. None declared.