Skip Navigation


Carcinogenesis Advance Access originally published online on March 26, 2007
Carcinogenesis 2007 28(9):1906-1913; doi:10.1093/carcin/bgm073
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
28/9/1906    most recent
bgm073v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (8)
Right arrowRequest Permissions
Google Scholar
Right arrow Articles by Liu, Y.
Right arrow Articles by Zhou, L.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Liu, Y.
Right arrow Articles by Zhou, L.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© The Author 2007. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org

Tagging SNPs in non-homologous end-joining pathway genes and risk of glioma

Yanhong Liu{dagger}, Haishi Zhang1,{dagger}, Keke Zhou1, Lina Chen2, Zhonghui Xu, Yu Zhong, Hongliang Liu, Rui Li, Yin Yao Shugart2,3, Qingyi Wei4, Li Jin, Fengping Huang1,*, Daru Lu* and Liangfu Zhou1

State Key Laboratory of Genetic Engineering and MOE Key Laboratory of Contemporary Anthropology, School of Life Sciences and Institutes for Biomedical Sciences, Fudan University, 220 Handan Road, Shanghai 200433, China
1 Neurosurgery Department of Huashan Hospital, Fudan University, 12 Wulumuqi Zhong Road, Shanghai 200040, China
2 Department of Social Medicine, University of Bristol, Bristol, UK
3 Department of Epidemiology, Johns Hopkins University, Baltimore, MD, USA
4 Department of Epidemiology, The University of Texas M. D. Anderson Cancer Center, Houston, TX, USA

* To whom correspondence should be addressed. Email: drlu{at}fudan.edu.cn;

Correspondence may also be addressed to F.Huang; Email: huangfengping_neuro{at}hotmail.com


   Abstract
Top
 Abstract
Introduction
 Materials and methods
 Results
 Discussion
 References
 
Ionizing radiation is known to cause DNA damage, including single-strand and double-strand DNA breaks (DSBs), and the unrepair of DNA damage, particularly DSBs, may cause chromosome aberrations. Although the etiology of gliomas remains unclear, exposure to ionizing radiation has been identified as the only established risk factor. We hypothesized that polymorphisms of candidate genes involved in the DSBs repair pathway may contribute to susceptibility to glioma. We used a haplotype-based approach to investigate the role of 22 tagging single-nucleotide polymorphisms (tSNPs) of XRCC5, XRCC6 and XRCC7 in 771 glioma patients and 752 healthy controls. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated by the unconditional logistic regression, haplotypes were inferred by the HAPLO.STAT program and gene–gene interactions were evaluated by the multifactor dimensionality reduction method. We found that, in the single-locus analysis, glioma risk was statistically significantly associated with three XRCC5 tSNPs (SNP1 rs828704, SNP6 rs3770502 and SNP7 rs9288516, P = 0.005, 0.042 and 0.003, respectively), one XRCC6 tSNP (SNP4 rs6519265, P = 0.044) but none of XPCC7 tSNPs. Haplotype-based association analysis revealed that gliomas risk was statistically significantly associated with one protective XRCC5 haplotype "CAGTT," accounting for a 40% reduction (OR = 0.60, 95% CI = 0.43–0.85) in glioma risk, and some positive gene–gene interactions were also evident. In conclusion, genetic variants of the genes involved in the DSB repair pathway may play a role in the etiology of glioma.

Abbreviations: CI, confidence interval; CVC, cross-validation consistency; DNA–PK, DNA-dependent protein kinase; DSB, double-strand DNA break; IR, ionizing radiation; LD, linkage disequilibrium; MDR, multifactor dimensionality reduction; NHEJ, non-homologous end joining; OR, odds ratio; tSNP, tagging single-nucleotide polymorphism


   Introduction
Top
 Abstract
 Introduction
Materials and methods
 Results
 Discussion
 References
 
Although the common use of computerized tomography and magnetic resonance imaging greatly improves the diagnosis of brain tumors for the last two decades, there is a tendency toward higher incidence rates in more developed, industrialized countries (1). Approximately 9% of human cancers are brain tumors, of which 90% are gliomas, and nearly 80% of the affected glioma patients die within the first year after diagnosis (2).

Gliomas are central nervous system neoplasms derived from glial cells that surround and support neurons. Although the etiology of gliomas remains unclear, exposure to ionizing radiation (IR) has been identified as the well-established risk factor (3–5). IR induces various types of DNA damage, including both single-strand breaks and double-strand breaks (DSBs). Among these, DSBs are arguably the most detrimental form of DNA damage, since they may lead to either chromosomal breakage or rearrangement, events that may result in carcinogenesis (6).

The repair of DSBs involves either non-homologous end joining (NHEJ) or homologous recombination (7–9). Homologous recombination repair is of high fidelity and error free, whereas NHEJ, the predominant pathway for repairing DSBs in mammalian cells, is potentially error prone. In the NHEJ repair process, a DSB is first recognized by the DNA-dependent protein kinase (DNA–PK) complex, which subsequently recruits the LIG4/XRCC4 complex, to perform the end-joining reaction (7–9).

The NHEJ mechanism depends strictly upon DNA–PK, one member of the phosphatidylinositol 3-kinase super-family of kinases (10). DNA–PK consists of KU heterodimers (i.e. KU70/KU80, encoded by the XRCC6/XRCC5 genes), a regulatory subunit, and DNA–PKcs (encoded by the XRCC7 gene), a large catalytic subunit. DNA–PK is not only required for the repair of IR-induced DNA damage but also for the V(D)J and immunoglobulin class-switch recombination in the immune system. Mice with inactivated components of DNA–PK show severe combined immunodeficiency as well as IR hypersensitivity (11,12). Furthermore, recent findings have implicated that DNA–PK has telomere functions, suggesting an association between DSBs repair and telomere maintenance (13).

Based on the fact that IR primarily causes DSBs that may increase risk of glioma, we hypothesized that genetic variants of the genes involved in the repair of DSBs may contribute to the etiology of glioma; alternatively, individuals with some adverse genotypes of these genes may be at high risk of developing glioma. Several studies have found associations between polymorphisms in DSBR genes and risk of cancers, including breast cancer and ovarian cancer (14–17). However, to date, few studies have investigated the role of polymorphisms in DSB repair genes in glioma risk. Hence, the primary aim of this study was to test the hypothesis that variants in XRCC5, XRCC6 and XRCC7 would modulate risk of glioma. Further, DNA–PK is a heterotrimer complex that is involved in protein–protein interactions and may have a synergistic effect on pathogenesis of glioma. The secondary aim therefore focused on the examination of the potential role of gene–gene interactions among XRCC5, XRCC6 and XRCC7 in the etiology of glioma.


   Materials and methods
Top
 Abstract
 Introduction
 Materials and methods
Results
 Discussion
 References
 
Study population
All subjects were genetically unrelated ethnic Han Chinese and were from Shanghai and the surrounding provinces (Zhejiang, Jiangsu and Anhui) in east China. Patients newly diagnosed with histopathologically confirmed glioma were consecutively recruited between October 2004 and May 2006 in the Department of Neurosurgery at Huashan Hospital of Fudan University (Shanghai, China) without the restrictions of age, gender and histology. The exclusion criteria included self-reported cancer history, previous radiotherapy and chemotherapy for unknown disease conditions. Among 900 eligible case patients, 771 agreed to participate in this study by providing an informed consent and the response rate was 85.7% (771/900). The cases enrolled in our study represented 91% (702/771) of the newly diagnosed glioma patients and 9% (69/771) of the recurrent glioma patients, mostly glioblastoma. The common reasons for refusal to participate included the patients being too sick or too upset to participate and time constraints. There were no significant demographic differences between individuals who agreed or refused to participate in the study.

The cancer-free control subjects consisted of trauma outpatients (20%) from the Emergency Medical Centre and hospital visitors (80%) who came to the health examination clinic for an annual check-up at the same hospital (Huashan Hospital) during the same study period. The exclusion criteria for the control subjects included central nervous system-related diseases, self-reported history of any cancer and previous radiotherapy and chemotherapy for unknown disease conditions. All the control subjects were frequency matched to the cases on age (±5 years), gender and residence area (urban or rural). The trauma outpatients did not differ from annual check-up subjects by demographic data (data not shown). Of 950 eligible control subjects we contacted for recruitment, 752 agreed to participate in this study by providing an informed consent, and the response rate was 79.2% (752/950). The common reasons for refusal to participate included the subjects being reluctant to donate blood and time constraints or other inconvenient situations. The non-respondents did not differ appreciably from the respondents with regard to age, sex and residence area (data not shown).

A structured questionnaire developed by the Department of Epidemiology at M. D. Anderson Cancer Center (provided by Dr Melissa Bondy) served as a model to develop our own questionnaire in China, which is shorter than the original M. D. Anderson Cancer Center brain tumor study questionnaire. Each participant was scheduled for a face-to-face interview with the structured questionnaire that detailed information on demographic factors, occupational radiation exposure history, family history of cancer and health characteristics. For childhood glioma patients under age 18, baseline data were collected from their parents. Family history of cancer was defined as any self-reported cancer in the first-degree relatives. Those who had smoked less than one cigarette per day and less than 1 year in their lifetime were defined as non-smokers, otherwise they were considered smokers. Occupational radiation exposure history included work as a pilot, flight attendant, astronaut, unanium miner, industrial and nuclear power plant worker or X-ray medical worker, radiology technologist, specialist (a dentist and a hygienist) physicians and other workers exposed to IR in the workplace. Since estimating levels of actual IR exposure was difficult, we used proxy measures of IR exposure based on radiological work-years obtained from the questionnaire.

After interview, a one-time sample of approximately 3–5 ml venous blood was collected from each participant. Genomic DNA was extracted from white blood cell fractions using the Qiagen Blood Kit (Qiagen, Chatsworth, CA). The research protocol was reviewed and approved by the Fudan University Ethics Committee for Human Subject Research.

Selection of the tagging single-nucleotide polymorphisms
Because haplotype-based designs of choosing tagging single-nucleotide polymorphisms (tSNPs) for association studies are more powerful than the single-allele approach and the consideration of haplotype diversity and linkage disequilibrium (LD) structure is the key to the success in association studies (18), we utilized a haplotype tSNPs approach in this study. Several methods have been developed for selecting tSNPs on the basis of published or observed genotype or haplotype data (19–21). Therefore, we selected tSNPs by using genotype data obtained from the International HapMap Project (http://www.hapmap.org) data (released # 19/PhaseII Oct 05). The HapMap samples consist of 90 Yoruba individuals from Ibadan, Nigeria (YRI), 90 individuals of European descent from Utah (CEU), 45 Han Chinese individuals from Beijing (CHB) and 45 Japanese individuals from Tokyo (JPT).

We aimed at defining a set of tSNPs that have an estimated r2 > 0.8 with those untyped SNPs. However, in the case of XRCC5 tSNP selection, because of the extensive haplotype diversity (there are 160 available SNPs and 9 LD blocks from the HapMap CHB data) and weak LD across the XRCC5 gene, we set an r2 threshold of 0.6 following Carlson et al. (22). Using the HAPLOVIEW program (http://www.broad.mit.edu/mpg), within each gene and the flanking 10 kb on either side, we selected the SNPs that have a minor allele frequency >0.05 in CHB. For XRCC5 (2q35), which spans ~97 kb and contains 21 exons, we selected eight tSNPs at a resolution of one SNP per 12.1 kb by using r2 threshold 0.6. For XRCC6 (22q13.2-q13.31), which spans ~43 kb and contains 12 exons, three HapMap-based tSNPs were selected by using r2 threshold 0.8 by forcing HAPLOVIEW to include two additional SNPs (rs2267437 and rs132793) that were reported in literature (15,16). Thus, five SNPs were selected with an SNP resolution of one SNP per 8.6 kb. For XRCC7 (on chromosome 8q11), which spans ~186 kb and contains 86 exons, nine tSNPs were selected with an SNP resolution of 1 SNP per 20.7 kb by using r2 threshold 0.8. Therefore, a total of 22 SNPs were chosen for our study, of which the vast majority was intronic (Table I).


View this table:
[in this window]
[in a new window]

  		Table I. Information about 22 genotyped SNPs of XRCC5, XRCC6 and XRCC7 genes

 
Genotyping assays
We genotyped all samples for the selected 22 tSNPs using the fluorogenic 5' nuclease TaqMan assay (Applied Biosystems, Foster City, CA). The TaqMan primers and FAM- or VIC-labeled probes were designed using the Primer Express Oligo Design software v2.0 (ABI PRISM) and available upon request. We scanned the completed polymerase chain reaction plates with ABI PRISM 7900HT Sequence Detector in the end point mode using the Allelic Discrimination Sequence Detector Software. The polymerase chain reaction assays for both case patients and control subjects were arrayed together in 384-well plates, with eight no-template controls and eight duplicated samples in each 384-well format as quality controls. On average, 97% of genotypes were successfully determined for all SNPs (Table I).

Statistical methods
Goodness-of-fit to the Hardy–Weinberg equilibrium expectation in control subjects was assessed by {chi}2 test for each SNP. The Akaike's information criterion was used to determine the genetic model for each SNP (23). Genotype frequencies of case patients and control subjects were compared using {chi}2 tests. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated by unconditional logistic regression analysis with adjustment for age and gender. We used 18 years as age cut points and divided the subjects into two groups of children (≤18) and adults (>18). In addition, we stratified the case patients into three subgroups of glioblastoma, astrocytomas except for glioblastoma (including diffuse astrocytomas, anaplastic astrocytomas or other astrocytomas), and other gliomas (including oligodendrogliomas, enpendymomas or mixed glioma). For each gene, the Bonferroni correction was made for P value for the results of any tSNP by multiplying the number tSNPs tested for the gene. All statistical tests were two sided.

LD and D' for each pair of SNPs were performed by the Arlequin 2.0 Software (24). Haplotypes of the tSNPs were inferred using the HAPLO.STAT program (http://www.mayo.edu/hsr/Sfunc.html) (25) and adjusted for possible confounding variables (i.e. age and gender).

For the evaluation of gene–gene interactions, we used the multifactor dimensionality reduction (MDR) method (26–28). This method includes a combined cross-validation/permutation-testing procedure that minimizes false-positive results by multiple examinations of the data. Cross-validation divides the data into a training set and a testing set. With 10-fold cross-validation, the data are divided into 10 equal parts, and the model is developed on 9/10 of the data (the training set) and then tested on 1/10 of the remaining data (the testing set). This is repeated for each possible 9/10 and 1/10 of the data, and the resulting 10 testing accuracies are averaged. In addition to the testing accuracy, we also report the cross-validation consistency (CVC), a measure of how many times out of 10 divisions of the data that MDR found the same best model. Models that are true-positives are likely to be generalized to independent data sets and will have estimated testing accuracies of >0.5. The permutation test was performed by shuffling the case–control stats 1000 times and repeating the MDR analysis on each randomized data set. The MDR analysis was performed by using the version 0.5.1 of the open-source MDR software package that is freely available online (http://www.epistasis.org/software.html).


   Results
Top
 Abstract
 Introduction
 Materials and methods
 Results
Discussion
 References
 
Characteristics of the study population
Our final analysis included 771 glioma cases and 752 cancer-free controls. The characteristics of case patients and control subjects are summarized in Table II. The case patients and control subjects appeared to be adequately matched on age and sex (P = 0.935 and 0.780, respectively). Mean age (age at diagnosis for case patients and age at inclusion for control subjects) for the case patients and control subjects was 44.10 and 42.63 years, respectively. For occupational radiation exposure histories, only 21 glioma cases and 9 cancer-free controls persons reported a history of occupational exposure to IR (3.64% versus 1.70%; P = 0.047). Also, case patients were significantly more likely than the controls to report a family history of cancer (19.94% versus 14.68%; P = 0.01) in their first-degree relatives, and a family history of cancer appeared to account for ~45% increased glioma risk (OR = 1.45, 95% CI = 1.08–1.94) in this study population. Among 771 case patients, 295 (38.26%) had glioblastoma, 242 (31.39%) had astrocytomas except for glioblastoma, 234 (30.35%) had other gliomas, including oligodendrogliomas (8.17%), enpendymomas (9.47%), or mixed glioma (12.71%).


View this table:
[in this window]
[in a new window]

  		Table II. Frequency distribution of selected characteristics of study subjects by the case–control status

 
Association between individual SNP and risk of glioma
The genotype distributions of 22 selected tSNPs of case patients and control subjects are summarized in Table III. All genotype distributions of control subjects were consistent with those expected from the Hardy–Weinberg equilibrium (all P > 0.05) (Table I). In the single-locus analyses, we observed statistically significant differences between case patients and control subjects in genotype distributions of three XRCC5 tSNPs [SNP1 rs828704 A/C, SNP6 rs3770502 G/A and SNP7 rs9288516 T/A and P = 0.005, 0.042 and 0.003, respectively, and the significance remained for SNP1 rs828704 A/C (P = 0.040) and SNP7 rs9288516 T/A (P = 0.024) after the Bonferroni correction], one XRCC6 tSNP (SNP4 rs6519265 A/G, P = 0.044; no more significant after the Bonferroni correction), but none of tSNPs in XRCC7.


View this table:
[in this window]
[in a new window]

  		Table III. Genotype frequencies of 22 SNPs of XRCC5, XRCC6 and XRCC7 genes among cases and controls and their associations with risk of glioma

 
Further logistic regression analyses revealed that in the co-dominant effect model selected by the Akaike's information criterion, significant protection against glioma risk was associated with the variant genotypes of XRCC5 SNP1 rs828704 A/C [adjusted OR = 0.78 (95% CI = 0.54–0.92) for AC genotypes, compared with the AA genotype], whereas significantly increased risk was associated with variant genotypes of XRCC5 SNP6 rs3770502 G/A [adjusted OR = 1.29 (95% CI = 1.02–1.65) for GA genotypes, compared with the GG genotype], XRCC5 SNP7 rs9288516 T/A [adjusted OR = 1.52 (95% CI = 1.19–1.93) for TA genotypes, compared with the TT genotype] and XRCC6 SNP4 rs6519265 G/A [adjusted OR = 1.51 (95% CI = 1.11–1.96) for GA genotypes, compared with the GG genotype] (Table III).

Furthermore, we performed stratified analyses by age, IR exposure history and glioma histological type on the two XRCC5 SNPs (SNP1 rs828704 A/C and SNP7 rs9288516 T/A) that were significantly associated with glioma risk (Table IV). Compared with the common homozygous, the protective effect of SNP1 rs828704 A/C variant were more evident in the adults [adjusted OR = 0.72 (95% CI = 0.54–0.95) for AC], subjects without IR exposure histories [adjusted OR = 0.71 (95% CI = 0.52–0.98) for AC] and anaplastic astrocytomas subgroups [adjusted OR = 0.61 (95% CI = 0.40–0.93) for AC], whereas the increased risk associated with the SNP7 rs9288516 T/A variant were more pronounced in the adults [adjusted OR = 1.45 (95% CI = 1.17–1.88) for TA, and 1.56 (95% CI = 1.16–2.14) for AA], subjects without IR exposure histories [adjusted OR = 1.48 (95% CI = 1.11–1.97) for TA, and 1.43 (95% CI = 1.00–2.04) for AA] and patients with anaplastic astrocytomas [adjusted OR = 1.56 (95% CI = 1.13–2.16) for TA, and 1.54 (95% CI = 1.04–2.29) for AA], and others glioma patients subgroups [adjusted OR = 1.48 (95% CI = 1.03–2.12) for TA, and 1.56 (95% CI = 1.01–2.41) for AA].


View this table:
[in this window]
[in a new window]

  		Table IV. Logistic regression analysis of XRCC5 SNP1 rs828704 and SNP7 rs9288516 genotypes stratified by age, IR exposure history and histological type of glioma

 
Association between haplotypes and risk of gliomas
LD between each pair of SNPs in XRCC5, XRCC6 and XRCC7 are shown in Table V. Strong LD was observed between each pair of the tSNPs of XRCC6 and XRCC7. In XRCC5, except that five tSNPs (SNP 4–8) were in strong LD with each other (D' > 0.7) and therefore formed a haplotype block, LD among other tSNPs and between any of these tSNPs and any of the five-SNP set was weak. On the basis of the LD pattern of XRCC5, we conducted haplotype-based risk assessment for the five-tSNP set.


View this table:
[in this window]
[in a new window]

  		Table V. Linkage disequilibrium (D') between pairs of SNPs in XRCC5, XRCC6 and XRCC7 genes

 
Table VI summarizes the associations between frequencies of the haplotypes and risk of glioma. Haplotypes with a frequency >0.03 were pooled into a mixed group. For XRCC5, one risk haplotype "AAATC" (adjusted OR = 1.33, 95% CI = 1.06–1.70) and two protective haplotypes "CGGTT" (adjusted OR = 0.66, 95% CI = 0.52–0.81) and "CAGTT" (adjusted OR = 0.60, 95% CI = 0.43–0.84) were found, compared with the most common haplotype CGGAT. The protective haplotype "CAGTT" of XRCC5 had the lowest OR, indicating the greatest protective effect, accounting for a 40% reduction in risk of developing glioma. The risk haplotype "CGAAA" in XRCC6 had a 2.36-fold increased risk of gliomas (adjusted OR = 2.36, 95% CI = 1.59–3.54), but no such a difference was observed for XRCC7.


View this table:
[in this window]
[in a new window]

  		Table VI. Haplotype frequencies in cases and controls in our hypothesis generating data seta

 
Furthermore, global score test also showed statistically significant differences in haplotype profile between case patients and control subjects for both XRCC5 and XRCC6 (global stat = 104.77 and 39.41, respectively; both P < 0.000). Unlike XRCC5, however, the significant effect of the XRCC6 tSNPs was due almost entirely to the difference in the frequency of the rare haplotypes (Table VI).

Locus–locus and gene–gene interactions
In the NHEJ pathway, the Ku complex binds to the ends of DNA first and then recruits DNA–PKcs to form a DNA–PK complex. Therefore, separate MDR analyses were conducted for the Ku complex (heterodimer of Ku70 and Ku80) and DNA–PK complex (heterotrimer of DNA–PKcs and Ku70/Ku80). Table VII summarizes the best interaction models obtained in the MDR analysis. The model with the lowest prediction error and the highest CVC was selected and further evaluated using the permutation test.


View this table:
[in this window]
[in a new window]

  		Table VII. Summary of MDR results for gene–gene interactions on glioma risk

 
Firstly, in the Ku complex, one 4-locus (i.e. SNP4 rs668844 and SNP5 rs207916 of XRCC5 and, SNP2 rs132770 and SNP5 rs132793 of XRCC6) model was selected. It had a minimum prediction error of 40.62% and a perfect CVC of 10.0 that was statistically significant (P = 0.001), as determined empirically by the permutation testing. The four loci in the 4-locus model were located within the two different subunit genes, which may describe possible interactions between XRCC5 and XRCC6. Secondly, in the DNA–PK complex, one 5-locus (i.e. XRCC5 SNP4 rs668844 and SNP5 rs207916, XRCC6 SNP2 rs132770 and SNP5 rs132779, and XRCC7 SNP1 rs7830743) had a significant probability value (P = 0.001). This 5-locus model consisted of those four SNPs in the above 4-locus model in the Ku complex and an additional XRCC7 SNP1 rs7830743. However, this 5-locus model had a lower CVC of 6 and a higher prediction error of 41.3% than that in the 4-locus model (Table VII).


   Discussion
Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
References
 
In this haplotype-based case–control study, we found, for the first time, that four tSNPs (SNP1 rs828704, SNP6 rs3770502 and SNP7 rs9288516 of XRCC5 and SNP4 rs6519265 of XRCC6) were significantly associated with glioma risk in this Chinese study population. We observed a strong protective effect of haplotype "CAGTT" of XRCC5 that was associated with a 40% reduction in risk of developing glioma. Furthermore, the MDR analysis suggested that the association was even stronger when gene–gene interactions within the DNA–PK complex were considered.

It is of importance to note that four tSNPs were associated with glioma risk in the present study, of which three (SNP1 rs828704, SNP6 rs3770502 and SNP7rs9288516) were from XRCC5. Moreover, two XRCC5 tSNPs, SNP6 rs3770502 and SNP7 rs9288516, were both located within intron 16 and in strong LD with each other (D' = 0.921), suggesting that either one of these SNPs may influence the splicing sites or be in linkage with a risk variant somewhere in this gene, most likely downstream of intron 16. Because the intronic SNP7 rs9288516 had a strongest association with risk of glioma, it may have a tight linkage with other untyped functional SNPs. Therefore, the exact location and biological functions of the real causal SNPs in XRCC5 is of great interest and warrant further investigation.

Although the functional relevance of the XRCC5 tSNPs is unknown, several lines of evidence suggest that these findings are biologically plausible. Studies have suggested that the radiation-resistant phenotype of a repair-proficient hybrid may co-segregate with the human 2q35 chromosome fragment (29). As a major gene involved in NHEJ, XRCC5 is located on chromosome 2q35 and encodes the 80 kD subunit of the Ku autoantigen, a heterodimer that binds to DSBs, facilitating the NHEJ repair. We propose that diverse haplotypes and weak LD across XRCC5 could perhaps explain the failure of previous studies to detect associations with polymorphisms in this gene, highlighting the importance of fully characterizing the LD and haplotype profiles of any genes before embarking on their association studies.

Several previous studies have assessed SNPs in NHEJ genes for their associations with risk of cancers, of which one reported a positive association of XRCC6 SNP1 rs2267437 variant with risk of breast cancer (15); however, subsequent studies failed to confirm this finding (16). In the present study, we found no evidence of an association of this XRCC6 variant with risk of glioma. The SNP4 rs7003908 of XRCC7 had previously been reported to be associated with risk of glioma in a US population (30), but this could not be replicated in our Chinese population. A similar study also found no evidence of an association between risk of renal cell carcinoma and this SNP of XRCC7 (31).

Our study has several strengths. One is that we adopted a haplotype-based approach. To the best of our knowledge, this study is the first haplotype-based study that describes tSNPs in the NHEJ repair pathway and their potential associations with cancer risk. Moreover, none of the tSNPs in our study have been previously examined in glioma association studies, except for the XRCC7 SNP4 rs7003908 polymorphism (30). Previous studies were focusing on only one or two variants in NHEJ genes, which might not be sufficient to capture the full effects of susceptibility genes. By using a haplotype-based association approach, an increasingly accepted approach to genetic association studies, we were able to provide reasonably strong evidence that variation in NHEJ genes may contribute to susceptibility to glioma.

Another strength is that we used the MDR method to detect interactions among genes involved in the DNA–PK complex and to test them globally as genetic risk factors for glioma. It is well known that epistasis or gene–gene interactions play an important role in the etiology of common human diseases. However, such interactions are often difficult to detect by traditional parametric statistical methods, such as the logistic regression analysis, because of the sparseness of the data in high dimensions. To address this problem, we adopted the MDR method for collapsing high-dimensional genetic data into a single dimension, thus permitting interactions to be detected in studies with relatively small sample sizes (26–28). As our results show, the interaction of the five tSNPs (i.e. XRCC5 SNP4 rs668844 and SNP5 rs207916, XRCC6 SNP2 rs132770 and SNP5 rs132779, and XRCC7 SNP1 rs7830743) in XRCC5, XRCC6 and XRCC7 genes, which are known to form a trimeric complex, DNA–PK, is the best model for predicting risk of glioma.

Interestingly, the majority of tSNPs kept in the optimal MDR model were not associated with glioma risk in early single-locus association analysis. This serves as evidence of epistasis, that is, the effect of one gene may be too week to be detected, when the effect of another gene is not accounted for. Furthermore, the loci included in these models were located within different genes or even on different chromosomes, which indicates that the interaction crossed chromosomal boundaries between the NHEJ genes may exist. Taken together, these data suggest that SNPs in XRCC5, XRCC6 and XRCC7 may act together to contribute to the risk of glioma. More importantly, our results suggest that the risk of glioma depends not only on the effect of individual NHEJ genes but also on the interaction between SNPs of these NHEJ genes.

Finally, our sample size (771 glioma patients and 752 control subjects) is relatively large among glioma association studies published to date. Several studies investigating putative associations between various SNPs and glioma have been published (32–36), but the number of glioma patients in these studies has rarely exceeded 500. Our relatively large sample size plus the haplotype-based study design and the MDR analysis ensured a sufficient power to detect risk of glioma.

Some limitations are inherent in this type of case–control study and must be noted. Although we used a statistical correction to adjust for multiple testing for a given gene, current epidemiologic and statistical literature is not unanimously clear on when and how to make such corrections. Although being frequently used, both Bonferroni correction and Bayesian techniques (37) are problematic in correcting multiple comparisons. Some authors believe that corrections were not needed when different associations in a study are of interest on a purely one-at-a-time basis (38,39).

Another potential concern was population admixture, which is a known confounding factor for association analysis and may also result in inflated type-I error (false positive). In this study, we used two types of controls, "annual check-up subjects" and "trauma outpatients" in the same hospital, to avoid possibility that one may have more pronounced selection bias. However, this bias is unlikely to be of significance, because they did not differ in the distributions of demographic variable and genotype frequencies. Therefore, we believe that the use of two control groups greatly reduced possible selection bias if any. Moreover, the observed allelic frequencies of the genotyped SNPs showed consistent results with the International HapMap Project data, we achieved a relatively high response rate in both cases and controls (85.7% versus 79.2%), and there is no substantial population admixture in Chinese populations.

Finally, though exposure to IR is a well-established risk factor for glioma, only a small percentage of the participants reported previous exposure to IR. Specifically, <4% of the glioma patients had IR exposure histories. One possibility of the low reported exposure to IR is because the subjects were not fully aware of IR they had exposed to or they had been exposed to unknown sources of IR. This hypothesis has to be validated by future larger prospective studies. Nevertheless, in the absence of reliable quantitative dose assessment of lifetime IR exposure on a daily basis and appropriateness of linear extrapolation from high to low dose, proving or disproving such gene–radiation interactions remains a challenge in molecular epidemiology studies. Several previous studies examined the association between DSBR gene polymorphisms and cancer risk, but none examined interactions of candidate genes with IR exposures, except for studies on people exposed specifically to medical IR (40,41). Future studies are needed to address whether polymorphisms in DSBs repair genes interact with IR exposure and thus modify the genetic susceptibility to glioma.

In summary, this study is, to the best of our knowledge, the first and largest haplotype-based study that describes tSNPs in the NHEJ gene and their associations with glioma risk. Our study demonstrated that some representative tSNPs of XRCC5 and XRCC6 genes might modulate risk of glioma. In particular, the association was stronger for gene–gene interactions involving the DNA–PK complex than for a single gene. Given that glioma is a highly fatal malignancy, our findings may have prevention implications through identifying at-risk population, once these findings are replicated by other studies on a larger scale or prospective studies.


   Footnotes
 
{dagger} These authors contribute equally to this work. Back


   Acknowledgments
 
We thank Drs Jian Yu, Mei Chong and Xilan Mei for their help with sample collection and archiving and Yin Wang and Wenting Wu for preparing DNA samples. We also thank Dr Melissa Bondy for providing the M. D. Anderson brain tumor questionnaire. Particular thanks go to Aiping Zhang and Dandan Zhang for their helpful comments and discussion. We are most indebted to the participants and staff at the Huashan Hospital of Fudan University who helped make this study possible. This work was partially supported by the China National Key Basic Research Program Grants 2002CB512902 (to D.L.) and National "211" Environmental Genomics Grant (to D.L.).

Conflict of Interest Statement: None declared.


   References
Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 

  1. Parkin DM, et al. Cancer Incidence in Five Continents (2002) Lyon: IARC Press.
  2. Kleihues P, et al. Pathology and Genetics of Tumors of the Nervous System (2000) Lyon: IARC Press.
  3. Little MP, et al. Risks of brain tumour following treatment for cancer in childhood: modification by genetic factors, radiotherapy and chemotherapy. Int. J. Cancer (1998) 78:269–275.[CrossRef][Web of Science][Medline]
  4. Bondy ML, et al. Gamma-radiation sensitivity and risk of glioma. J. Natl. Cancer Inst. (2001) 93:1553–1557.[Abstract/Free Full Text]
  5. Neglia JP, et al. New primary neoplasms of the central nervous system in survivors of childhood cancer: a report from the childhood cancer survivor study. J. Natl. Cancer Inst. (2006) 98:1528–1537.[Abstract/Free Full Text]
  6. Morgan WF, et al. DNA double-strand breaks, chromosomal rearrangements, and genomic instability. Mutat. Res. (1998) 404:125–128.[Web of Science][Medline]
  7. Khanna KK, et al. DNA double-strand breaks: signaling, repair and the cancer connection. Nat. Genet. (2001) 27:247–254.[CrossRef][Web of Science][Medline]
  8. Lieber MR, et al. Mechanism and regulation of human non-homologous DNA end-joining. Nat. Rev. Mol. Cell Biol. (2003) 4:712–720.[CrossRef][Web of Science][Medline]
  9. Thacker J. The role of homologous recombination processes in the repair of severe forms of DNA damage in mammalian cells. Biochimie (1999) 81:77–85.[Medline]
  10. Hartley KO, et al. DNA-dependent protein kinase catalytic subunit: a relative of phosphatidylinositol 3-kinase and the ataxia telangiectasia gene product. Cell (1995) 82:849–856.[CrossRef][Web of Science][Medline]
  11. Ferguson DO, et al. The nonhomologous end-joining pathway of DNA repair is required for genomic stability and the suppression of translocations. Proc. Natl. Acad. Sci. USA (2000) 97:6630–6633.[Abstract/Free Full Text]
  12. Singleton BK, et al. Molecular and biochemical characterisation of xrs mutants defective in Ku80. Mol. Cell. Biol. (1997) 17:1264–1273.[Abstract/Free Full Text]
  13. Hsu HL, et al. Ku acts in a unique way at the mammalian telomere to prevent end joining. Genes Dev. (2000) 14:2807–2812.[Abstract/Free Full Text]
  14. Auranen A, et al. Polymorphisms in DNA repair genes and epithelial ovarian cancer risk. Int. J. Cancer (2005) 117:611–618.[CrossRef][Web of Science][Medline]
  15. Fu Y-P, et al. Breast cancer risk associated with genotypic polymorphism of the nonhomologous end-joining genes: a multigenic study on cancer susceptibility. Cancer Res. (2003) 63:2440–2446.[Abstract/Free Full Text]
  16. Lee K-M, et al. Genetic polymorphisms of selected DNA repair genes, estrogen and progesterone receptor status, and breast cancer risk. Clin. Cancer Res. (2005) 11:4620–4626.[Abstract/Free Full Text]
  17. Kuschel B, et al. Variants in DNA double-stranded break repair genes and breast cancer susceptibility. Hum. Mol. Genet. (2002) 11:1399–1407.[Abstract/Free Full Text]
  18. Johnson GCL, et al. Haplotype tagging for the identification of common disease genes. Nat. Genet. (2001) 29:233–237.[CrossRef][Web of Science][Medline]
  19. Daly MJ, et al. High-resolution haplotype structure in the human genome. Nat. Genet. (2001) 29:229–232.[CrossRef][Web of Science][Medline]
  20. Gabriel SB, et al. The structure of haplotype blocks in the human genome. Science (2002) 296:2225–2229.[Abstract/Free Full Text]
  21. Zhang K, et al. A dynamic programming algorithm for haplotype block partitioning. Proc. Natl. Acad. Sci. USA (2002) 99:7335–7339.[Abstract/Free Full Text]
  22. Carlson CS, et al. Selecting a maximally informative set of single-nucleotide polymorphisms for association analyses using linkage disequilibrium. Am. J. Hum. Genet. (2004) 74:106–120.[CrossRef][Web of Science][Medline]
  23. Akaike H. A new look at the statistical model identification. IEEE Trans. Automat. Contr. (1974) 19:716–723.[CrossRef]
  24. Schneider S, et al. Arlequin Ver. 2.000: A Software for Population Genetics Data Analysis. Genetics and Biometry Laboratory (2000) Geneva: University of Geneva.
  25. Schaid DJ, et al. Score tests for association between traits and haplotypes when linkage phase is ambiguous. Am. J. Hum. Genet. (2002) 70:425–434.[CrossRef][Web of Science][Medline]
  26. Ritchie MD, et al. Multifactor-dimensionality reduction reveals high-order interactions among estrogen-metabolism genes in sporadic breast cancer. Am. J. Hum. Genet. (2001) 69:138–147.[CrossRef][Web of Science][Medline]
  27. Moore JH, et al. New strategies for identifying gene-gene interactions in hypertension. Ann. Med. (2002) 34:88–95.[CrossRef][Web of Science][Medline]
  28. Hahn LW, et al. Multifactor dimensionality reduction software for detecting gene-gene and gene-environment interactions. Bioinformatics (2003) 19:376–382.[Abstract/Free Full Text]
  29. Chen DJ, et al. Regional assignment of a human DNA repair gene (XRCC5) to 2q35 by x-ray hybrid mapping. Genomics (1994) 21:423–427.[CrossRef][Web of Science][Medline]
  30. Wang L-E, et al. Polymorphisms of DNA repair genes and risk of glioma. Cancer Res. (2004) 64:5560–5563.[Abstract/Free Full Text]
  31. Hirata H, et al. Polymorphisms of DNA repair genes are associated with renal cell carcinoma. Biochem. Biophys. Res. Commun. (2006) 342:1058–1062.[CrossRef][Web of Science][Medline]
  32. Chen P, et al. Association of an ERCC1 polymorphism with adult-onset glioma. Cancer Epidemiol. Biomarkers Prev. (2000) 9:843–847.[Abstract/Free Full Text]
  33. Parhar P, et al. Possible association of p53 codon 72 polymorphism with susceptibility to adult and pediatric high-grade astrocytomas. Mol. Brain Res. (2005) 137:98–103.[Medline]
  34. Yang P, et al. Polymorphisms in GLTSCR1 and ERCC2 are associated with the development of oligodendrogliomas. Cancer (2005) 103:2363–2372.[CrossRef][Web of Science][Medline]
  35. Peters ES, et al. NAT2 and NQO1 polymorphisms are not associated with adult glioma. Cancer Epidemiol. Biomarkers Prev. (2001) 10:151–152.[Free Full Text]
  36. Inoue R, et al. A genotype of the polymorphic DNA repair gene MGMT is associated with de novo glioblastoma. Neurol. Res. (2003) 25:875–879.[CrossRef][Web of Science][Medline]
  37. Benjamini Y, et al. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. Roy. Stat. Soc. B (1995) 57:289–300.
  38. Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiology (1990) 1:43–46.[Medline]
  39. Perneger TV. What's wrong with Bonferroni adjustments. Br. Med. J. (1998) 316:1236–1238.[Free Full Text]
  40. Millikan RC, et al. Polymorphisms in DNA repair genes, medical exposure to ionizing radiation, and breast cancer risk. Cancer Epidemiol. Biomarkers Prev. (2005) 14:2326–2334.[Abstract/Free Full Text]
  41. Sadetzki S, et al. Genotyping of patients with sporadic and radiation-associated meningiomas. Cancer Epidemiol. Biomarkers Prev. (2005) 14:969–976.[Abstract/Free Full Text]
Received January 31, 2007; revised March 15, 2007; accepted March 21, 2007.


Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?


This article has been cited by other articles:


Home page
 Neuro OncologyHome page
P. Rajaraman, A. Hutchinson, S. Wichner, P. M. Black, H. A. Fine, J. S. Loeffler, R. G. Selker, W. R. Shapiro, N. Rothman, M. S. Linet, et al.
DNA repair gene polymorphisms and risk of adult meningioma, glioma, and acoustic neuroma
Neuro Oncology, January 1, 2010; 12(1): 37 - 48.
[Abstract] [Full Text] [PDF]

Home page
 Neuro OncologyHome page
A. P. Kyritsis, M. L. Bondy, J. S. Rao, and C. Sioka
Inherited predisposition to glioma
Neuro Oncology, January 1, 2010; 12(1): 104 - 113.
[Abstract] [Full Text] [PDF]

Home page
 Cancer Epidemiol. Biomarkers Prev.Home page
R. McKean-Cowdin, J. Barnholtz-Sloan, P. D. Inskip, A. M. Ruder, M. Butler, P. Rajaraman, P. Razavi, J. Patoka, J. K. Wiencke, M. L. Bondy, et al.
Associations between Polymorphisms in DNA Repair Genes and Glioblastoma
Cancer Epidemiol. Biomarkers Prev., April 1, 2009; 18(4): 1118 - 1126.
[Abstract] [Full Text] [PDF]

Home page
 Cancer Epidemiol. Biomarkers Prev.Home page
Y. Liu, M. E. Scheurer, R. El-Zein, Y. Cao, K.-A. Do, M. Gilbert, K. D. Aldape, Q. Wei, C. Etzel, and M. L. Bondy
Association and Interactions between DNA Repair Gene Polymorphisms and Adult Glioma
Cancer Epidemiol. Biomarkers Prev., January 1, 2009; 18(1): 204 - 214.
[Abstract] [Full Text] [PDF]

Home page
 Clin. Cancer Res.Home page
Z. Hu, H. Liu, H. Wang, R. Miao, W. Sun, G. Jin, Y. Wang, H. Ma, L. Jin, Q. Wei, et al.
Tagging Single Nucleotide Polymorphisms in Phosphoinositide-3-Kinase-Related Protein Kinase Genes Involved in DNA Damage "Checkpoints" and Lung Cancer Susceptibility
Clin. Cancer Res., May 1, 2008; 14(9): 2887 - 2891.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
28/9/1906    most recent
bgm073v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (8)
Right arrowRequest Permissions
Google Scholar
Right arrow Articles by Liu, Y.
Right arrow Articles by Zhou, L.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Liu, Y.
Right arrow Articles by Zhou, L.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?