Polymorphisms in GLTSCR1 and ERCC2 are associated with the development of oligodendrogliomas

Authors


Abstract

BACKGROUND

Deletions of 19q have been associated with gliomas, especially oligodendrogliomas. In addition, cases with oligodendrogliomas with the 19q deletion have been observed to have a better survival compared with cases without the 19q deletion. The authors have previously described a 150-kilobase minimal deletion region in gliomas that maps to 19q13.33 and contains 3 novel candidate genes (GLTSCR1, EHD2, and GLTSCR2).

METHODS

The authors performed an association study using 141 cases with gliomas (61 cases with astrocytomas, 40 cases with oligodendrogliomas, 40 cases with mixed oligoastrocytomas) and 108 general controls. They evaluated 7 single nucleotide polymorphisms (SNPs) in 6 genes within and nearby the minimal 19q deletion region (ERCC2, RAI, ASE-1, ERCC1, GLTSCR1, and LIG1).

RESULTS

The prevalence of a germline GLTSCR1-exon-1 T allele (SNP rs1035938) was 40% in cases with oligodendrogliomas compared with 27% in controls (P = 0.029), and the prevalence of an ERCC2-exon-22 T allele (SNP rs1052555) was 35% in cases with oligodendrogliomas compared with 18% in controls (P = 0.043). One high-risk and 1 low-risk haplotype were associated with oligodendroglioma development (P = 0.003 and 0.026, respectively). Cases with oligodendrogliomas with the 19q deletion had a significantly higher frequency of the GLTSCR1-exon-1 T allele compared with cases without the 19q deletion (P = 0.01). It was noteworthy that cases with gliomas who were homozygous for the GLTSCR1-exon-1 T allele had a significantly better survival: 77% and 68% survival at 2 and 5 years compared with 56% and 34% for other genotypes (P = 0.02, log-rank test). Multivariable analysis identified grade, age, and the GLTSCR1-exon-1 and ERCC2-exon-22 genotypes as independent predictors for survival.

CONCLUSIONS

These results suggested that alterations in GLTSCR1 (or a closely linked gene) were associated with the development and progression of oligodendroglioma. Cancer 2005. © 2005 American Cancer Society.

Malignant gliomas are the most common primary central nervous system tumors affecting adults, with nearly 18,500 diagnosed annually in the U.S. and a mortality rate approaching 80% in the first year after diagnosis.1 Despite this poor prognosis, our understanding of glioma development and progression remains modest, and our abilities to effectively stratify and treat these malignancies continue to be limited.

Previous molecular analyses have demonstrated different genetic alterations associated with gliomas. Briefly, tumors of astrocytic lineage often demonstrate anomalies of chromosome arms 9p, 10p, 10q, 11p, 13q, 17p, 17q, 19q, and 22q, whereas oligodendrogliomas and mixed oligoastrocytomas (MOAs) commonly have alterations of 1p and 19q.2 A number of tumor suppressor genes have been cloned from these regions, including the CDKN2A/p16/p14ARF and CDKN2B/p15 gene cluster mapped to 9p, the p53 gene mapped to 17p, the RB gene mapped to 13q, and most recently, the PTEN3, 4 and DMBT5 genes mapped to 10q. However, the genes on 1p and 19q remain to be identified.

Chromosome 19 q-arm alterations are of particular interest because they have not been found to play a role in other common malignancies6 and because they are the only known genetic abnormalities shared by all 3 glioma subtypes.7–9 These observations, coupled with the high frequency of 19q alterations,10 strongly suggest that 19q harbors ≥ 1 gene important for the pathogenesis of gliomas, and further suggest that this gene may be of fundamental importance for glial development and growth regulation. Furthermore, a recent report by Cairncross, et al.11 suggested that combined loss of chromosome arms 1p and 19q in high-grade oligodendrogliomas was highly associated with chemotherapeutic response and longer disease recurrence-free survival after chemotherapy, whereas CDKN2A deletions conferred a worse prognosis. Recently, we have extended these observations to oligodendrogliomas of all grades. For example, tumors with both 1p and 19q deletions have a statistically significant prolonged survival compared with tumors without these alterations.12 We have also found that genetic alterations of 1p and 19q may predict chemotherapeutic response in a set of low-grade oligodendrogliomas and astrocytomas.13

We have previously described a 150-kilobase (kb) minimal deletion region in gliomas that maps to 19q13.33.14 This region contains 3 novel transcripts (EHD2, GLTSCR1, and GLTSCR2; GLTSCR stands for glioma tumor suppressor candidate region) and 2 known genes (SEPW1 and CRX).15 Although no tumor-specific alterations were observed in any of the putative transcripts, several recent observations suggest that genes in or near this region may be involved in tumor development. Polymorphisms in loci near this deletion region (ERCC1, ERCC2, RAI, ASE-1, and D19S246) have been associated with basal cell, breast, and lung carcinoma and MOAs.16–22 Specifically, individuals carrying the ERCC2 variant, R156R, have been shown to be at a higher risk for gliomas with the strongest association with MOAs.22 A polymorphism within GLTSCR1 has recently been associated with prostate carcinoma aggressiveness.23, 24 Based on this preliminary evidence, we conducted an association study in gliomas using these and other markers that map within and nearby the 150-kb minimally deleted region observed in gliomas.

MATERIALS AND METHODS

Case–Control Populations

Blood specimens were collected from 249 study subjects: 141 cases with glioma and 108 general controls. The 141 gliomas were surgically resected at Mayo Clinic (Rochester, MN) between December 1987 and December 2001. This existing resource was initiated in 1987 and stored tissue samples of every patient whose surgically resected glioma was large enough to spare for research purposes. Tissue samples were obtained for 35% of all surgical cases. For the cases for whom we had tissue samples, we actively sought patients' consent to donate a blood sample, and the time from surgery to blood draw was within 30 days for > 95% of the patients.

The controls were selected from a pool of Olmsted County residents enrolled as general controls in the Mayo Clinic Cancer Center between 1997 and 2001, who had a blood sample remaining after their clinical tests were performed.25, 26 This design was chosen because the Mayo Clinic is a major primary care provider for the local population, and > 90% of Olmsted County residents visit the Mayo Clinic ≥ 1 time in any 3-year period.27 Eligible control subjects had no current or previously diagnosed malignancy (except nonmelanoma skin carcinoma) as of the date of phlebotomy. Controls received a self-administered questionnaire and a request for permission to use their blood samples.

In accordance with the institutional review board of Mayo Clinic, clinical information was collected and is regularly updated for the patients with glioma through follow-up and questionnaires. These data include date of birth, gender, date of diagnosis of primary tumor, date of surgical resection or biopsy and pathology of the specimen, treatment with chemotherapy and/or radiotherapy for primary and/or recurrent lesions, date of last follow-up, and status of patient (living/deceased) at the time of last follow-up.

Tumors were classified morphologically and graded according to the World Health Organization system28 by three independent neuropathologists. Consensus morphology and grade were established for each glioma, as previously described.10 Table 1 summarizes the morphology of the cases and other basic characteristics of cases and controls. There were 61 cases with astrocytomas, 40 cases with oligodendrogliomas, and 40 cases with mixed oligodendrogliomas. The 19q deletion status of the gliomas was determined by fluorescence in situ hybridization analysis of paraffin-embedded material using previously described methods.12

Table 1. Basic Description of 141 Glioma Cases and 108 Controls
CharacteristicsGeneral controls (n = 108) (%)All gliomas (n = 141) (%)Astrocytomas (n = 61) (%)Mixed oligoastrocytomas (n = 40) (%)Oligodendrogliomas (n = 40) (%)
  • a

    P < 0.05 determined by the Pearson's chi-square test, comparing with the general control group.

  • b

    P < 0.05 determined by the Pearson's chi-square test, comparison among three tumor subgroups.

  • c

    Includes: Iowa, Illinois, Indiana, Michigan, Minnesota, North Dakota, South Dakota, and Wisconsin.

  • d

    Includes: Brazil, Greece, Hungary, Mexico, Peru, Saudi Arabia, Spain, and Yugoslavia.

Gender     
 Men 67 (62) 94 (67)42 (69)28 (70)24 (60)
 Women 41 (36) 47 (33)19 (31)12 (30)16 (40)
Race/ethnic background     
 Non-Hispanic white103 (95)128 (94)57 (97)36 (92)35 (92)
 Hispanic white  1 (1)  1 (1)0 —0 — 1 (3)
 Non-U.S. white  4 (4)  7 (5) 2 (3) 3 (8) 2 (5)
 Unknown0 —5 —2 —1 —2 —
Age group (yrs)a     
 < 50 10 (9) 91 (64)26 (43)29 (72)36 (90)
 ≥ 50 97 (91) 50 (36)35 (58)11 (28) 4 (10)
Geographic region of residence     
 Midwest statesc108 (100) 97 (70)42 (69)34 (85)21 (55)
 Non-Midwest states 32 (23)12 (20) 5 (13)15 (40)
 Outside U.S.d 10 (7) 7 (11) 1 (2) 2 (5)
Tumor gradeb     
 II 42 (30) 4 (7)13 (32)25 (62)
 III 39 (28) 5 (8)21 (52)13 (33)
 IV 60 (42)52 (85) 6 (15) 2 (5)
Primary vs. recurrent brain tumorb     
 Primary110 (78)54 (89)26 (65)30 (75)
 Recurrent 31 (22) 7 (11)14 (35)10 (25)
Initial diagnosisb     
 Mayo Clinic 96 (71)49 (83)22 (56)25 (66)
 Elsewhere 40 (29)10 (17)17 (44)13 (34)
 Unknown5 —2 —1 —2 —

DNA Extraction, Polymerase Chain Reaction Amplification, and Genotype Determination

Table 2 summarizes the 7 single nucletotide polymorphisms (SNPs) in 6 genes that were analyzed. These 7 SNPs previously had been shown to be associated with basal cell carcinomas (BCC), breast carcinomas, or MOAs.16–22 The ASE-1-exon-3 SNP was previously identified as an ERCC1-3′ untranslated region (UTR) SNP.21 The July 2003 assembly of the University of California Santa Cruz's (Santa Cruz, CA) human genome database (available from URL: http://genome.ucsc.edu/) shows that this SNP lies outside of ERCC1 and within the coding region of ASE-1.

Table 2. SNPs Analyzed
RS no.aGeneaLocationaChromosome 19 positionbFunction of changeAlleles (frequencies)c
  1. a

    SNP: single nucleotide polymorphism; RS no.: accession number in the National Center for Biotechnology Information (NCB1) SNP database.

  2. a Gene and location are from the NCB1 SNP database., Available from URL: http://www.ncbi.nlm.nih.gov/SNP/ [accessed].

  3. b In Mb, from July 2003 freeze of the University of California Santa Cruz (UCSC) human genome database.; Available from URL: http://genome.ucsc.edu/ [accessed].

  4. c Found among general controls in our pilot study.

rs1052555ERCC2Exon 2250547364D711DC(0.69), T(0.31)
rs238406ERCC2Exon 650560149R156RC(0.53), A(0.47)
rs6966RAIExon 650574802Untranslated regionA(0.83), T(0.17)
rs3212986ASE-1Exon 350604576K504QC(0.76), A(0.24)
rs3177700ERCC1Exon 450615493N118NT(0.63), C(0.37)
rs1035938GLTSCR1Exon 152875583S397SC(0.75), T(0.25)
rs20580Lig1Exon 653346365A170AC(0.52), A(0.48)

Genomic DNA was extracted from blood specimens using the phenol-chloroform method. Polymerase chain reaction (PCR) primers (Table 3) were purchased from Integrated DNA Technologies (Coralville, IA). PCR reactions were performed in a 25-μL reaction volume containing 2.50 μL GeneAmp10X PCR Gold Buffer (Applied Biosystems, Foster City, CA), 1.5 mM MgCl2, 200 μM each deoxynucleotide triphosphate, 7 pmol of each PCR primer, 0.625 U of AmpliTaq Gold (Applied Biosystems), and 25 ng of genomic DNA. The PCR reactions to amplify the GLTSCR1-exon-1 fragment included 4% dimethylsulfoxide and an additional 0.125 U of AmpliTaq Gold. PCR amplification was performed using a PTC-225 Peltier thermal cycler (MJ Research, Waltham, MA). Cycling conditions for all sequences except GLTSCR1-exon-1 were 50 cycles of denaturation at 95 °C for 30 seconds, annealing for 30 seconds, and extension at 72 °C for 30seconds followed by a final extension at 72 °C for 10 minutes. Annealing was performed at 61.5 °C for ERCC2-exon-6, RAI-exon-6, ASE-1-exon-3, and Lig1-exon-6. ERCC2-exon-22 was annealed at 53.8 °C, and ERCC1-exon-4 was annealed at 59.0 °C. GLTSCR1-exon-1 was amplified using the following cycling conditions: initial denaturation at 95 °C for 10 minutes, 40 cycles of denaturation at 95 °C for 30 seconds, annealing at 60 °C for 30 seconds, and extension at 72 °C for 30 seconds followed by a final extension at 72 °C for 10 minutes.

Table 3. Primers Used for PCR and Pyrosequencing
LocusPrimers
5′-Biotin labeled for PCRUnlabeled for PCRPyrosequencing
  1. PCR: polymerase chain reaction.

ERCC2-exon-225′-GAGCACCTCACAGATGCCA5′-CCATCTGCCGCAGGAAGTA5′-CCACCTGGACACCCT
ERCC2-exon-65′-CACAGCCTCACAGCCTCCTA5′-CCATCAAATTCCTGGGACAAGAGT5′-CCAGTAACCTCATAG
RAI-exon-65′-CATAACCACAATGATGAGCATGTA5′-CTCTCCCCAATTAAGTGCCT5′-ATTAAGTGCCTTCACACAG
ASE-1-exon-35′-CTCTGGGGAGGGATTCTGG5′-AGGGCCACTGAATTCAGAGTC5′-GGACAAGAAGCGGAA
ERCC1-exon-45′-CAGAGCTCACCTGAGGAACA5′-CATCCCTATTGATGGCTTCTGC5′-GTACTGAAGTTCGTGCG
GLTSCR1-exon-15′-CAAGCGAACGTCTTCAAGCA5′-GGCAGTAGGAACTGGTTCTGG5′-TTGGTTCAGGAGGTGGA
Lig1-exon-65′-CGGAGGAGAGGACCAGAAACT5′-CCATCTGACCGTTCTGTCTATCGT5′-ACAGAGGCTGAAGTGG

Pyrosequencing was performed on a PSQ 96 (Biotage AB, Uppsala, Sweden) using streptavidin-labeled Dynabeads (Dynal Biotech, Oslo, Norway) according to the protocol provided.29

Statistical Analysis

The frequency distribution at each SNP locus was tested against the Hardy–Weinberg equilibrium under the allele Mendelian biallelic expectation using the chi-square test. No significant distortion was detected.

Data for both primary and recurrent tumors were used to evaluate the association between the seven candidate SNP markers and the risk of glioma in a case–control design. Pooled cases as well as each morphologic subgroup were compared with the control group. Univariable associations of allele (each chromosome as a unit) and genotype (a person as a unit) with disease were evaluated using contingency table methods in SAS software, version 8.2 (SAS Institute, Inc., Cary, NC). These associations were assessed using the Pearson chi-square and Fisher exact tests for dichotomous categorical variables, and the Cochran–Armitage trend tests for genotype associations. Genotype-specific odds ratios (ORs), as an assessment of relative risk and potential confounding effects from other loci, were estimated and adjusted, respectively, using logistic regression for the significant markers identified from the univariate analysis. The multiple SNP marker-disease association by estimated haplotype was evaluated using haplo.score (a Mayo-developed software), which accounts for ambiguous linkage phase.30 Linkage disequilibrium (LD) was assessed using the Graphical Overview of Linkage Disequilibrium (GOLD) software package (Center for Statistical Genetics, University of Michigan, Ann Arbor, MI).31

Survival analyses were performed using cases with glioma, with survival defined as the time from surgical diagnosis of the primary tumor to death or time of last follow-up. Cases who were alive at their last follow-up were censored. Survival distributions were estimated with Kaplan–Meier curves32 and compared among patient subsets with log-rank tests.33 Univariate associations of patient characteristics, tumor characteristics, and SNP markers with survival were ascertained with a Cox proportional hazards model. Multivariate analyses were performed with classification and regression tree (CART) models,34, 35 as well as with multivariate Cox proportional hazards models. The multivariate analyses were done on the pooled cases. Potential prognostic variables included tumor type (i.e., astrocytoma, oligodendroglioma, or MOA), the other variables in Table 1 (treating age as a continuous variable), and 7 SNP markers. The statistical tests were 2-sided tests and P values ≤ 0.05 were statistically significant.

RESULTS

Characteristics of Cases and Controls

Gender and ethnic background distributions were similar between pooled cases and controls and among tumor types (cases). However, because of the study design (i.e., both groups were selected from existing resources), there was a significant difference between cases and controls with regard to age at blood sample collection and geographic region of residence. Tumor grade, the proportion of primary tumors, and the proportion of patients with an initial diagnosis at Mayo Clinic differed significantly among the tumor subgroups. These differences are not unexpected and reflect the nature of the tumors. Patients with astrocytomas tend to present with a higher grade than patients with oligodendrogliomas and mixed oligodendrogliomas.

An Association Study Using Candidate SNPs

We compared allele frequencies of each of the 7 SNPs between general controls and glioma cases by morphologic subtypes (Table 4). The presence of a germline GLTSCR1-exon-1 T allele was significantly associated with the development of oligodendroglioma (P = 0.029). The association between the ERCC2-exon-22 C allele and oligodendroglioma also achieved significance (P = 0.043). Similar results were observed when the data were analyzed using genotype-based methods. For example, GLTSCR1-exon-1 genotypes TT, TC, and CC were observed in 18%, 45%, and 37% of cases with oligodendrogliomas compared with 7%, 39%, and 54% of controls, respectively (P = 0.032). ERCC2-exon-22 genotypes TT, TC, and CC were observed in 0%, 37%, and 63% of cases with oligodendrogliomas, respectively, compared with 10%, 42%, and 49% of controls, respectively (P = 0.04). It is noteworthy that of the 13 cases with glioma with the GLTSCR1-exon-1 TT genotype, 7 (53%), 2 (15%), and 4 (31%) developed oligodendrogliomas, MOAs, and astrocytomas, respectively. No other significant associations between SNP genotypes and oligodendrogliomas or other glioma subtypes were observed in our study.

Table 4. Allele-Based Analysis of Association of Selected 19q SNPs with Glioma Development
LocusGeneral controls (n = 108)All gliomas (n = 141)Astrocytomas (n = 61)Mixed oligoastrocytomas (n = 40)Oligodendrogliomas (n = 40)
(%)No. (%)P valueaNo. (%)P valueaNo. (%)P valueaNo. (%)P valuea
  • SNP: single nucleotide polymorphism; Chrs: chromosome.

  • a

    P value = Pearson's chi-square test (in comparison to the general control group).

ERCC2-exon-22210 Chrs274 Chrs0.694122 Chrs0.89876 Chrs0.41876 Chrs0.043
 Allele T64 (30.5)79 (28.8) 38 (31.1) 27 (35.5) 14 (18.4) 
 Allele C146 (69.5)195 (71.2) 84 (68.9) 49 (64.5) 62 (81.6) 
ERCC2-exon-6208 Chrs274 Chrs0.518122 Chrs0.17074 Chrs0.66978 Chrs0.884
 Allele C110 (52.9)153 (55.8) 74 (60.7) 37 (50.0) 42 (53.9) 
 Allele A98 (47.1)121 (44.2) 48 (39.3) 37 (50.0) 36 (46.1) 
RAI-exon-6216 Chrs276 Chrs0.216118 Chrs0.17380 Chrs0.74278 Chrs0.118
 Allele A183 (84.7)222 (80.4) 93 (78.8) 69 (86.2) 60 (76.9) 
 Allele T33 (15.3)54 (19.6) 25 (21.2) 11 (13.8) 18 (23.1) 
ASE-1-exon-3210 Chrs282 Chrs0.548122 Chrs0.91580 Chrs0.31480 Chrs0.749
 Allele A51 (24.3)62 (22.0) 29 (23.8) 15 (18.8) 18 (22.5) 
 Allele C159 (75.7)220 (78.0) 93 (76.2) 65 (81.2) 62 (77.5) 
ERCC1-exon-4214 Chrs280 Chrs0.996122 Chrs0.59878 Chrs0.36780 Chrs0.867
 Allele C78 (36.5)102 (36.4) 48 (39.3) 24 (30.8) 30 (37.5) 
 Allele T136 (63.5)178 (63.6) 74 (60.7) 54 (69.2) 50 (62.5) 
GLTSCR1-exon-1216 Chrs280 Chrs0.392120 Chrs0.77080 Chrs0.58980 Chrs0.029
 Allele T58 (26.9)85 (30.4) 34 (28.3) 19 (23.8) 32 (40.0) 
 Allele C158 (73.1)195 (69.6) 86 (71.7) 61 (76.2) 48 (60.0) 
LIG1-exon-6216 Chrs282 Chrs0.907122 Chrs0.90580 Chrs0.58580 Chrs0.323
 Allele A103 (47.7)133 (47.2) 59 (48.4) 41 (51.3) 33 (41.2) 
 Allele C113 (52.3)149 (52.8) 63 (51.6) 39 (48.7) 47 (58.7) 

Table 5 summarizes the haplotype-based analysis of oligodendrogliomas. Twenty-five haplotypes were identified and only haplotypes with a frequency of ≥ 0.03 in either cases or controls are shown. One high-risk (TCACCTA) and 1 low-risk (TCAACCA) haplotype were identified (simulated P = 0.014 and 0.023, respectively). These 2 haplotypes differ only by the presence of an ASE-1-exon-3 allele (C or A) or a GLTSCR1-exon-1 allele (T or C). The high-risk haplotype was computationally identified by the DNA markers program in two cases with oligodendrogliomas and none with another glioma or in a control. No haplotypes were significantly associated with astrocytoma or MOA development in the current study.

Table 5. Haplotype-Based Analysis of Association of Selected 19q SNPs with Oligodendroglioma Development
ERCC2-exon-22ERCC2-exon-6RAI-exon-6ASE-1-exon-3ERCC1-exon-4GLTSCR1-exon-1LIG1-exon-6Empirical P valueSimulated P valueHaplotype frequency
OverallControlsCases
  1. SNP: single nucleotide polymorphism.

TCAACTC0.350.350.0200.0160.046
TCAACCA0.0260.0230.0680.0960.004
CAAACTC0.280.290.0260.0260.057
CAAACCC0.970.970.0340.0420.000
TCACCTA0.0030.0140.0160.0000.054
CCACCCC0.610.620.0270.0340.000
TCACTCA0.170.180.0540.0670.000
TCACTCC0.190.170.0390.0440.021
CAAACCA0.510.500.0290.0000.056
CCTCTTC0.060.070.0320.0210.048
CCTCTCA0.590.600.0420.0370.050
CCTCTCC0.130.140.0610.0390.102
CAACTTA0.600.610.0710.0280.097
CAACTTC0.590.600.0660.0760.039
CAACTCA0.600.590.0730.1120.042
CAACTCC0.790.790.1280.1300.170

LD analysis was performed using both D' (close to 1) and R2 (> 0.3), to detect pairwise LD among all 7 loci. Significant LD was found between ASE-1-exon-3 and ERCC1-exon-4 in controls, astrocytomas, and MOAs. Borderline significant LD was found in oligodendrogliomas between ERCC2-exon-22 and ERCC2-exon-6 (D' = 1.00 and R2 = 0.20) and between ERCC2-exon-6 and RAI-exon-6 (D' = 1.00 and R2 = 0.28). No LD was detected between GLTSCR1-exon-1 and any of the other six 19q SNPs in the study populations.

To estimate the risk of developing oligodendrogliomas associated with the GLTSCR1-exon-1 genotype, we calculated the ORs of CT and TT using CC as the referent genotype (Table 6). The unadjusted OR for individuals with CT was 1.7 (95% confidence interval [95% CI], 0.7–3.7), and for those with the TT genotype, it was 3.4 (95% CI, 1.1–10.8). There was an indication that genotype CT may also increase the risk of oligodendrogliomas, but the OR was not statistically greater than unity with our current sample size.

Table 6. OR for GLTSCR1 Genotypes Using Multiple Logistic Models
GLTSCR1-exon-1 genotypeNormal controls (n = 108)Oligodendroglioma cases (n = 40)OR (95% CI)ORa (95% CI)ORb (95% CI)
  • OR: odds ratio; 95% CI: 95%; confidence interval.

  • a

    GLTSCR1-exon-1 OR from a model adjusted for ERCC1-exon-1.

  • b

    GLTSCR1-exon-1 OR from a model adjusted for ERCC2-exon-22.

CC (reference)5815
CT42181.65 (0.75, 3.65)1.63 (0.73, 3.62)1.44 (0.64, 3.24)
TT873.38 (1.06, 10.82)3.25 (1.01, 10.41)2.17 (0.61, 7.69)
TT/CT50251.93 (0.92, 4.07)1.89 (0.89, 4.00)1.57 (0.73, 3.39)

We also attempted to estimate the risk of developing oligodendrogliomas associated with the ERCC2-exon-22 genotype. Based on our allele-based analysis (shown in Table 4), individuals with the CC and TT genotypes are expected to be at the highest and the lowest risk, respectively, for developing oligodendrogliomas. Because there were no cases with oligodendrogliomas in our study sample that had the TT genotype at the ERCC2-exon-22 locus, we used the CT and TT genotypes combined as the reference group to estimate the risk associated with the CC genotype. There was an increased OR (1.8, 95% CI, 0.9–3.9) for the development of oligodendrogliomas among cases with the CC genotype, but this did not quite achieve statistical significance.

We also performed a two-locus analysis as an attempt to control potential confounding effects from other loci on the observed association between GLTSCR1 and oligodendroglioma risk. The risk for developing oligodendrogliomas among cases who were TT at the GLTSCR1 locus did not change significantly after adjusting for the other 6 SNP genotypes, as demonstrated by adjusting for ERCC1-exon-1 and ERCC2-exon-22 in Table 6.

Because the controls were selected from an existing resource (i.e., the Olmsted County normal control pool) and were not matched to the cases with glioma (Table 1), we conducted stratified analyses by age groups, gender, ethnic background, geographic region of residence, place of initial tumor diagnosis, tumor grade, and primary versus recurrent tumor (data not shown). The changes in allele and genotype frequencies did not substantially change our reported associations.

Correlations between Germline SNPs with Glioma 19q Deletion Status

We performed a stratified analysis to determine whether there were associations between the 7 SNPs and glioma 19q deletion status. The GLTSCR1-exon-1 T allele is associated with glioma 19q deletion status. GLTSCR1-exon-1 genotypes TT, TC, and CC were observed in 9%, 56%, and 34%, respectively, of cases whose glioma had 19q deletion, compared with 5%, 34%, and 63%, respectively, of cases whose glioma lacked 19q deletion (P = 0.060). GLTSCR1-exon-1 genotypes TT, TC, and CC were observed in 15%, 60%, and 25% of cases whose oligodendroglioma had 19q deletion compared with 11%, 11%, and 78%, respectively, of cases whose oligodendroglioma lacked 19q deletion (P = 0.019). No other significant associations were observed between SNP genotypes and cases whose tumors had 19q deletion.

Haplotype analysis was performed for cases with oligodendrogliomas, 20 with and 9 without 19q deletion. One new high-risk haplotype was observed (CCTCTTC, simulated P = 0.01) in the group with the 19q deletion, and none were observed none in the group without deletion. This haplotype is similar to the high-risk haplotype identified for the oligodendrogliomas as a whole and shares the ASE-1-exon-3 C and the GLTSCR1-exon-1 T alleles. Also this haplotype contains the high-risk ERCC2-exon-22 C allele. The number of mixed oligodendrogliomas and MOAs with 19q loss was too small for stratified analysis.

Glioma Survival and Candidate SNP Markers

Using the above case cohort, we have also compared the association of ASE-1-exon-3, ERCC2-exon-22, and GLTSCR1-exon-1 polymorphisms with case survival and other clinical variables. Importantly, the cases with gliomas with the GLTSCR1-exon-1 TT genotype had a better survival: 77% and 68% survival at 2 and 5 years for the TT genotype compared with 56% and 34% at 2 and 5 years for the CT/CC genotype (P = 0.02, log-rank test) (Fig. 1). This difference in survival was also observed for the cases with oligodendrogliomas alone but did not reach statistical significance (P = 0.08). The ASE-1 and ERCC2 polymorphisms were not significantly associated with glioma 19q deletion status or with the morphologic grade of glioma. In addition, the ASE-1 and ERCC2 polymorphisms were not associated with survival after univariate analysis.

Figure 1.

Kaplan–Meier curve comparing the survival of cases with glioma with the GLTSCR1-exon-1 TT genotype with cases with glioma with GLTSCR1-exon-1 CT or CC genotypes. Straight line: CT/CC; dashed line: TT.

To identify subgroups with the longest and shortest survival, we used CART34, 35 modeling to determine clinical and genetic variables that were independently associated with survival. The CART model selected tumor grade, age at diagnosis, GLTSCR1-exon-1 genotype, and ERCC2-exon-22 genotype as the informative variables for generating survival subgroups. Morphology type was not identified as an important predictive factor. Six subgroups of cases with glioma with different survivals were generated (Table 7). The survival reference group consisted of 34 cases who were < 57 years old at the time of a Grade < 4 glioma diagnosis and who were CT/CC at GLTSCR1-exon-1 and CC at ERCC2-exon-22. Eight cases with Grade < 4 glioma, who were < 57 years old at the time of diagnosis and who carried the TT genotype at GLTSCR1-exon-1, had the best survival (hazard ratio = 0.11; 95% CI, 0.02–0.85). The 40 Grade 4 cases with glioma, who were > 47 years old at diagnosis, had the worst survival (hazard ratio = 8.67; 95% CI, 4.51–16.68).

Table 7. Relative Survival Time of Glioma Case Sub-groups Partitioned by CART Analysis
CART groupNo. in group (no. dead)HR (95% CI)
  1. CART: classification and regression tree; HR: hazard ratio; 95% CI: 95%; confidence interval.

Grade < 4  
 Age < 57 yrs  
  GLTSCR1-exon-1: TT8 (1)0.11 (0.02–0.85)
  GLTSCR1-exon-1: TC or CC  
   ERCC2-exon-22: TT or CT31 (10)0.45 (0.19–1.05)
   ERCC2-exon-22: CC34 (16)1.00 (reference group)
 Age ≥ 57 yrs8 (7)2.21 (0.90–5.42)
Grade = 4  
 Age < 4820 (14)1.95 (0.94–4.04)
 Age ≥ 4840 (35)8.67 (4.51–16.68)

DISCUSSION

The current study data suggest that alterations in GLTSCR1 (or a closely linked gene) and ERCC2 are associated with the development of oligodendrogliomas and with glioma survival. Specifically, our results indicate a threefold increased risk of oligodendrogliomas among cases with a TT genotype compared with cases with a CC genotype at a GLTSCR1-exon-1 polymorphic locus. There is limited, but suggestive, evidence in the published literature that some individuals are predisposed to develop gliomas. For example, an inherited susceptibility to gliomas has been reported to be present in 5–7% of families.36–39 Ten percent of these families (or approximately 1% of all patients with gliomas) carry germline high-penetrance mutations in the p53, NF1, NF2, hMLH1, hPMS2, or p16 genes).36 The genetic basis for the remainder of the families is unknown. Linkage studies have been limited mainly due to lack of informative pedigrees. One recent study reported a low-penetrance susceptibility locus for familial glioma at 15q23-q26.3.37

The association of low-penetrance genetic alterations, particularly DNA sequence polymorphisms with glioma development, only has been assessed by a few studies. Candidate genes evaluated include polymorphisms in DNA repair pathways such as base-excision repair, nucleotide-excision repair, and double-strand-break repair genes,38 and in metabolic pathways of carcinogens such as N-Nitroso compounds, pesticides, and polycyclic aromatic hydrocarbons.39 Our finding of an ERCC2 variant as a potential risk genotype is consistent with a previous study,22 although the precise nature of the variation needs to be further investigated.

Of specific relevance to our current study are association studies of 19q alleles with cancer (and glioma) development. Recent linkage studies have associated markers at 19q12 and 19q13.2 with aggressive prostate carcinoma.23, 24 Polymorphisms in the ERCC2 and the RAI genes have been associated with the development of BCC and postmenopausal breast carcinoma.16–19 A genome-wide survey recently reported the association of D19S246 (mapped to 19q13.3) with the development of adenocarcinoma of the lung.20

Two recent studies have suggested that polymorphisms in the ASE-1 and ERCC2 genes are associated with the development of MOAs.21, 22 In these reports, oligodendrogliomas were grouped with the MOAs. Importantly, the ERCC2, RAI, and ASE-1 genes map within 3 Mb of the 150-kb minimally deleted 19q region observed in primary gliomas.14GLTSCR1 is also mapped within this region. Nexo et al.19 have reported that a polymorphism in GLTSCR1 is in LD with the RAI and ERCC2 alleles associated with BCC. Furthermore, the marker with the peak LOD (Log-Odd) score for the association of aggressive prostate carcinoma (D19S902) lies within the GLTSCR1 gene.24

GLTSCR1 is highly conserved among humans, chimps, mice, and rats. The 5′ flanking and UTR regions of GLTSCR1 are GC rich. The GLTSCR1 polymorphism we tested alters a CpG within this 5′ CpG island. It is possible that germline alterations of CpGs may affect the transcription of GLTSCR1 and other candidate genes in the region. Two recent reports have implicated 2 other 19q13.33 genes in the development of oligodendrogliomas. Both are located within 3 Mb of the GLTSCR1-exon-1 polymorphism. In a platelet-derived growth factor/retrovirus mouse Model 1 of glioma, vector sequences were integrated with p190RhoGAP (also know as GRLF1) in a small proportion of tumors.40 In addition, methylation of a CpG island 5′ to the gene ZNF342 was observed in a large proportion of oligodendrogliomas with the 19q deletion.41

Two loci in LD in our relatively small sample could have occurred by chance but could also indicate a tight linkage of a causal variant within or close to the markers being study. We did not observe LD between GLTSCR1-exon-1 and any of the other six 19q SNPs in the study populations, suggesting that this synonymous SNP is tightly linked to a causal variant that was not included in this study, or that the T allele at this locus may be causally related to oligodendrogliomas through a to-be-revealed mechanism. Conversely, we detected LD between ASE-1 and ERCC1 in our controls and cases with astrocytomas or MOAs, but not in cases with oligodendrogliomas. In addition, a borderline significant LD was detected in a 3-locus cluster of ERCC2-exon-2, ERCC2-exon-6, and RAI-exon-6 only among cases with oligodendrogliomas.

The associations of the GLTSCR1-exon-1 T allele with glioma development and survival are provocative. The data suggest that it is not simply the 1p and 19q deletion status that determines the prognosis of patients with gliomas. Rather, our data suggest that the 19q genotype, and putatively an associated alteration in a target gene, is an important biologic and clinical variable.

Two notable limitations of our study are potential referral bias of the cases and the choice of the control group. Bias due to referral cases is considered minimal because seeking care at our institution was not predetermined by genotype, and the allele distributions of the seven markers in our study are in agreement with data reported in the literature. Our choice of a control group was not ideal because it has been always difficult to find proper controls in a tertiary referral clinic where a large number of new cases can be rapidly enrolled. Because our cases with glioma are mainly referral based, ideal controls should be matched with cases by age, gender, race, geographic referral area, and duration of care at Mayo Clinic (i.e., equally referred). However, finding such ideal controls is not feasible mainly because approximately 60% of our cases are from outside of the tristate area (Minnesota, Wisconsin, and Iowa) in which a limited number of eligible controls can be found and enrolled. Moreover, this generates more concerns because these patients were typically referred for other disorders, which may be related to the conditions being study. Recognizing the imperfection of our population-based control group, we were able to generate intriguing preliminary results, justifying further rigorous investigations with a proper design. Our control group could also provide estimates of the expected allele frequencies of the candidate markers in a defined population.

Our results require confirmation using a relatively large series of patients with gliomas and controls. Although GLTSCR1 is a promising candidate gene underlying oligodendroglioma development, the data with this SNP in exon-1 need to be reproduced with independent study samples with closely matched cases and controls. In addition, other SNPs within GLTSCR1 (or mutations in another gene(s) in LD with GLTSCR1) may underlie the associations we observed. Thus, additional SNPs at 19q13.33 must be investigated. Finally, 19q deletions have been associated with response to therapy as well as survival. It must be determined if 19q SNPs are associated with the response of gliomas to radiotherapy and/or chemotherapy.

Acknowledgements

The consensus pathologic diagnosis of the cases was determined by Caterina Giannini, Bernd Scheithauer, Peter Burger, Allan Yates, Andrew Bollen, and Kenneth Aldape—neuropathologists who are active participants in the program project Grant CA85779. The authors thank them for their careful review of the cases.

Secretarial assistance was provided by Ms. Heidi Bollum and Miss Susan Ernst.

Ancillary