Nonsmall cell lung cancer (NSCLC) frequently exhibits genomic instability, such as high fractional allelic loss (FAL). Genomic instability may result from unrepaired or misrepaired double-strand breaks (DSBs). The authors of this report postulated that polymorphisms in genes of the nonhomologous end-joining (NHEJ) pathway, which is the major DSB repair pathway in mammalian cells, may modulate lung cancer susceptibility and prognosis.
Patients with NSCLC (n = 152) and a group of appropriate age-matched and sex-matched controls (n = 162) were subjected to genotype analysis of the NHEJ pathway genes x-ray repair complementing defective repair in Chinese hamster cells 6 (Ku70) (reference single nucleotide polymorphism number [rs] 2267437), x-ray repair complementing defective repair in Chinese hamster cells 5 (Ku80) (rs3835), x-ray repair complementing defective repair in Chinese hamster cells 4 (XRCC4) (rs1805377), and DNA ligase IV (LIG4) (rs1805388). The gene-gene interaction (joint effect), genotype-environmental (ie, smoking) correlation, and genotype-phenotype (ie, FAL) correlation were examined. The Kaplan-Meier method and log-rank tests were used to assess the prognostic effect.
There was a significant association between the XRCC4 and LIG4 genotypes with NSCLC risk in an analysis of individual polymorphism associations, and the risk of NSCLC increased further in a combined analysis of multiple polymorphisms (adjusted odds ratio [OR], 8.74). The patients who had a homozygous variant guanine/guanine genotype of the XRCC4 gene had a poorer prognosis compared with other patients (P = .015). There was a significant difference between the patient smokers and controls for XRCC4 (adjusted OR, 2.67) and LIG4 (adjusted OR, 2.04). In addition, polymorphisms in XRCC4 and LIG4 were linked significantly with patients who had high FAL (adjusted OR, 2.03-3.84).
Lung cancer is among the most common malignancies in the world and is the leading cause of cancer deaths in industrial countries.1 The major risk factor for lung cancer is an excessive exposure to tobacco smoke. However, only approximately 11% of tobacco smokers ultimately develop lung cancer,2 suggesting that genetic factors may influence the risk of lung cancer among those who are exposed to carcinogens. Therefore, it is rational to speculate that certain common genetic variants or polymorphisms in genes involved in the repair of DNA damage may have an impact on lung cancer risk.3-7 DNA double-strand breaks (DSBs) are 1 of the most severe types of DNA damage and can result from a variety of factors, including ionizing radiation, free radicals, replication errors, and telomere dysfunction.8, 9 Unrepaired or misrepaired DSBs can lead to cell death, genomic instability, and oncogenic transformation.8-11 Our previous microsatellite genotyping study indicated that patients with lung cancer frequently exhibit markers of chromosome instability, such as high fractional allelic loss (FAL).12, 13 Chromosomal instability with FAL has been proposed as the pathogenic basis of tumorigenesis caused by the functional inactivation of DSB repair genes.9, 11, 14 Therefore, we postulated that polymorphisms in the DSB repair genes are involved in lung cancer susceptibility and prognosis.
In mammalian cells, there are 2 major DSB repair pathways, ie, homologous recombination (HR) and nonhomologous end-joining (NHEJ). NHEJ plays a predominant role under most conditions in mammalian cells.15 This is mainly because searching for homologous sequences for HR does not seem to be efficient due to the complexity of mammalian genomes. In addition, most of the mammalian genome consists of noncoding sequences.16 The NHEJ pathway provides a relatively minor cost of efficient, though imprecise, DSB repair to avoid lethality caused by unrepaired damage. Four major NHEJ components include the x-ray repair complementing defective repair in Chinese hamster cells 6 (Ku70) and x-ray repair complementing defective repair in Chinese hamster cells 5 (Ku80) proteins, which function as a DNA end-binding complex, and the x-ray repair complementing defective repair in Chinese hamster cells 4 (XRCC4) and DNA ligase IV (LIG4) proteins, which function as a complex in end ligation.18, 19 In this context, we decided to investigate the role of the polymorphisms in these 4 major NHEJ proteins in the susceptibility to and the prognostic value of nonsmall cell lung cancer (NSCLC) using a nested case-control study.
Recent evidence suggests that several single nucleotide polymorphisms (SNPs) in the NHEJ genes may be relevant to modify the risk of multiple myeloma,20 glioma,21 breast cancer,22-24 and bladder cancer25, 26 and in patients with pterygium.27 However, inconsistent results were reported, perhaps because of the inappropriate case and control selection and failure to evaluate the effect of multiple pathophysiologic-related genes, which are especially important for the application to association studies of complex diseases such as cancers. Therefore, for this study, we designed a nested case-control series by recruiting 152 patients with NSCLC and a control group of 162 cancer-free individuals matched by age and sex with homogenous ethnic background. We tested the hypothesis that SNPs in the NHEJ genes may modulate the risk and prognosis of NSCLC first by performing individual SNP association analysis and a combined analysis of multiple SNPs to study whether there were gene-gene interactions. Second, because smoking is the predominant risk factor for lung cancer, we also stratified our analysis by smoking status to determine whether there were gene-environment interactions. Finally, to gain insight into the potential mechanism for increased cancer risk, such as genomic instability, conferred by these different SNPs in the NHEJ pathway, we determined genotype-phenotype correlations in patients with cancer who had markers of chromosome instability (ie, high FAL). To our knowledge, this is the first study to demonstrate an association between SNPs in the NHEJ genes and cancer risk and prognosis in NSCLC.
MATERIALS AND METHODS
In total, 152 patients with NSCLC who were admitted to China Medical University Hospital, Taichung, and Veterans General Hospital, Taipei, Taiwan, between 2000 and 2006 were included in this study after obtaining appropriate institutional review board permission and informed consent from the patients. Among the 152 patients, 83 had adenocarcinomas, and 65 had squamous cell carcinomas. The histologic determinations, including tumor type and disease stage, were performed according to the World Health Organization classification and the TNM classification system, respectively. Information on smoking history for patients with lung cancer was obtained from hospital records. Patients were categorized as nonsmokers (never smokers) and smokers, including regular smokers (if they smoked 1 cigarette per day for 6 months or longer) and occasional smokers. The follow-up of 152 postoperative patients without chemotherapy was performed at 2-month intervals in the first year after surgery and at 3-month intervals thereafter in the outpatient clinics or by routine telephone calls. The end of follow-up was defined as February 2008 for all patients. The mean follow-up for all patients was 18 months (range, 0.5-59 months). For the 96 patients who remained alive at the end of follow-up (censored patients), the mean follow-up was 20 months. For the 56 patients who died during follow-up, the mean follow-up was 13 months.
In total, 162 unrelated, age-matched (±3 years) and sex-matched, healthy individuals who had no known medical illness or hereditary disorders and who were not taking any medications comprised the control group. The control group was recruited from individuals who visited the same hospital for a health check-up. Before its commencement, this study was approved by the Research Ethics Committee of China Medical University, and informed consent was obtained from each participant.
Genomic DNA was extracted from peripheral blood lymphocytes by proteinase K digestion and phenol/chloroform extraction. We selected 1 SNP for each of the 4 major NHEJ genes: Ku70 (reference SNP number [rs] 2267437), Ku80 (rs3835), XRCC4 (rs1805377), and LIG4 (rs1805388). These SNPs were selected either because some of them had been associated with the risk of other cancers19, 21, 23-25 or because the percentage of the minor allele was >5% in the general Taiwanese population21 to increase statistical power for detecting their association with cancer risk and prognosis. The genotypes were determined by a polymerase chain reaction (PCR) restriction fragment length polymorphism assay (RFLP). The PCR primers and the PCR conditions are listed in Table 1. The representative figures of PCR-RFLP are shown in Figure 1. For quality control, the genotyping analysis was performed with the investigators blinded to individual case-control status and was repeated twice for all participants. The results of genotyping were 100% concordant. To confirm the genotyping results, 30 randomly selected, PCR-amplified DNA samples for each gene were examined by DNA sequencing, and the results also were 100% concordant.
Table 1. List of Primer Sequences and Restriction Enzymes Used in the Single Nucleotide Polymorphism Study
SNP indicates single nucleotide polymorphism; PCR, polymerase chain reaction; bp, base pairs; Temp, temperature; RFLP, restriction fragment length polymorphism; Ku70, x-ray repair complementing defective repair in Chinese hamster cells 6; rs, reference single nucleotide polymorphism number; C, cytosine; G, guanine; A, adenine; T, thymine; Ku80, x-ray repair complementing defective repair in Chinese hamster cells 5; IVS, intervening sequence; XRCC4, x-ray repair complementing defective repair in Chinese hamster cells 4; LIG4, DNA ligase IV; Thr, threonine; Ile, isoleucine.
Numbers in parentheses represent the position (measured in bp) from the transcription site, and letters represent nucleotide changes. The positions of the amino acids are given for SNPs located in exons.
rs1805377 (IVS7-1A→G, splice site)
rs1805388 (exon 2+54C→T, Thr9Ile)
Genetic Instability Index (FAL)
In total, 269 informative microsatellite polymorphic markers spanning the 39 nonacrocentric autosomal arms with an average interval of approximately 20 cM were used for genome-wide loss of heterozygosity (LOH) allelotyping. All fluorescently labeled dinucleotide, trinucleotide, and tetranucleotide markers were based on the Marshfield Map and were purchased from PE Applied Biosystems (Foster City, Calif). The representative LOH values are shown in Figure 2.
Surgically resected tumor samples were immediately snap-frozen and subsequently stored in liquid nitrogen. For LOH assays, 5-μm serial sections of formalin-fixed, paraffin-embedded primary tumors and nearby normal lung tissues were used for the preparation of genomic DNA using proteinase K digestion and phenol-chloroform extraction. Twenty nanograms of genomic DNA from microdissected tumor cells and adjacent, noncancerous lung tissues were used for PCR analyses. Multiplex-PCR reactions and subsequent electrophoresis were done as described previously.11, 12 FAL was calculated for each tumor as the number of markers that demonstrated LOH divided by the number of informative markers. Patients with genomic instability were indicated by high FAL, which we defined as higher than the mean FAL for all 152 patients who were analyzed.
Differences in clinical data between the group of patients with NSCLC and the control group were evaluated with the Mann-Whitney U test. Hardy-Weinberg equilibrium was assessed using a goodness-of-fit chi-square test for biallelic markers. The Pearson chi-square test was used to compare genotype distributions among different ethnic groups and between various clinical factors (eg, tumor stages and smoking habits) and genetic alteration factors (eg, FAL) in the patients with NSCLC. Statistical modeling with logistic regression was used to calculate the relative risk (odds ratio [OR]) of genotypes for the case-control study. ORs were expressed together with 95% confidence intervals (95% CIs). Multivariate logistic regression analysis was adjusted for age and sex. Gene-gene interactions, such as the joint effect of individual SNPs in the NHEJ pathway to increase the risk of lung cancer, were examined by using conventional logistic regression analysis, and P values for trend were determined. Type III censoring was performed on individuals who remained alive at the end of the study. The Kaplan-Meier method was used to estimate the probability of survival as a function of time and median survival. The log-rank test was used to assess the significance of the risk of genotypes among pairs of survival probabilities. P values <.05 were considered significant.
Characteristics of Patients and SNPs
Genomic DNA samples from 152 patients with NSCLC and 162 cancer-free control individuals were analyzed to determine the distribution of SNPs in 4 major genes of the NHEJ pathway. Individuals in the control group were matched for sex and age (±3 years) to the patient population (cases). We selected 1 SNP for each NHEJ gene: Ku70 (rs2267437), Ku80 (rs3835), XRCC4 (rs1805377), and LIG4 (rs1805388). These SNPs were selected to increase statistical power for detecting their association with cancer risk and prognosis, either because some of them have been associated with risk of other cancers20, 22, 24-26 or because the percentage of the minor allele is >5% in the general Taiwanese population.22 None of the genotype distributions for the controls differed significantly from those expected under Hardy-Weinberg equilibrium. In our primary analysis, the minor-variant allele at each gene was treated as the “risk” allele. However, after a cross-validation study to evaluate the main effect on risk status, we reassigned the risk genotype for each locus as the homozygous wild-type cytosine/cytosine (C/C) of Ku70, heterozygous and homozygous variant genotypes guanine/adenine (G/A) + A/A of Ku80, homozygous variant genotype G/G of XRCC4, and heterozygous homozygous variant genotypes G/A + A/A of LIG4 on the basis of the cross-validation study. The assignment for risk genotype also was confirmed in another study that was conducted in Taiwan.22
Risk Associated With Individual SNPs
Table 2 shows the distribution of the 4 NHEJ polymorphisms by case-control status in 162 noncancer controls and 152 lung cancer patients along with the patients' clinicopathologic characteristics. Overall, there was a significant difference in genotype distribution of the XRCC4 and LIG4 genes between all noncancer controls and all lung cancer patients. The XRCC4 polymorphism intervening sequence 7-1 (IVS7-1) A→G (splice-site) apparently had a recessive association with lung cancer risk. The homozygous variant G/G genotype indicated a significantly increased risk of developing lung cancer compared with other genotypes (crude OR: 2.29; 95% CI, 1.03-5.48 [P = .045] and adjusted OR: 2.38; 95% CI, 1.05-5.76 [P = .043], using the logistic regression model). In addition, the LIG4 exon 2 (Ex2) + 54 cytosine to tyrosine (C→T) (threonine-9-isoleucine [Thr9Ile]) SNP had a dominant effect on lung cancer susceptibility. The combined G/A and A/A variant genotypes were associated with a significantly increased risk of lung cancer compared with the wild-type G/G genotype (adjusted OR: 1.64; 95% CI, 1.03-2.62 [P = .038]). However, there was no association between cancer risk and polymorphisms in the Ku70 (−1310C→G) and Ku80 (IVS19 + 6929G→A) genes.
Table 2. Comparison of Genotype Distribution According to Clinical Parameters in Patients With Lung Cancer
OR indicates odds ratio; 95% CI, 95% confidence interval; XRCC4, x-ray repair complementing defective repair in Chinese hamster cells 4; rs, reference single nucleotide polymorphism number; A, adenine; G, guanine; Ref, reference group; FAL, fractional allelic loss; LIG4, DNA ligase IV; C, cytosine; T, thymine; Ku70, x-ray repair complementing defective repair in Chinese hamster cells 6; Ku80, x-ray repair complementing defective repair in Chinese hamster cells 5.
The genotypes G/G of rs1805377, C/T+T/T of rs1805388, C/C for rs2267437, and G/A+A/A for rs3835 were considered putative high-risk genotypes.
The total number is not the same for some genes or clinical subgroups, because some genotype analyses were not successful.
ORs were calculated to measure the association of the high-risk and nonhigh-risk genotypes for lung cancer.
P values were estimated by using chi-square tests.
Adjusted by sex and age.
P values were estimated by using a multivariate logistic regression model.
Combined Analysis of Multiple SNPs
To test our hypothesis that multiple SNPs in the same pathway may have a joint effect on lung cancer risk, we estimated the combined effect of these SNPs (Table 3). When all 4 genes were considered together, no apparent increase in lung cancer risk was observed as the number of risk genotypes increased (data not shown). Therefore, we separated the joint effect based on the 2 different steps of NHEJ (ie, Ku70 and Ku80 for DNA end-binding complex and XRCC4 and LIG4 for end-ligation complex). Table 3 shows that the risk of lung cancer increased significantly with the number of putative risk genotypes (adjusted OR: 4.65; 95% CI, 1.19-24.08 [P = .040] for the Ku70 and Ku80 pathway and adjusted OR: 8.75; 95% CI, 2.27-57.77 [P = .006] for the XRCC4 and LIG4 pathway, because patients had 2 risk genotypes).
Table 3. Adjusted Odds Ratios of Developing Lung Cancer Associated With the Number of Putative High-risk Genotypes of Nonhomologous End Joining Pathway Genes
OR indicates odds ratio; 95% CI, 95% confidence interval; Ref, reference group. Ku70, x-ray repair complementing defective repair in Chinese hamster cells 6; Ku80, x-ray repair complementing defective repair in Chinese hamster cells 5; rs, reference single nucleotide polymorphism number; Ref, reference group; XRCC4, x-ray repair complementing defective repair in Chinese hamster cells 4; LIG4, DNA ligase IV.
The genotypes cytosine/cytosine (C/C) of rs2267437, guanine/adenine (G/A)+A/A of rs3835, G/G of rs1805377, and cytosine/thymine (C/T)+T/T of rs1805388 were considered putative high-risk genotypes.
P values were estimated by using chi-square tests.
Adjusted by sex and age.
P values were estimated by using a multivariate logistic regression model.
Ku70 (rs2267437) and Ku80 (rs3835)
XRCC4 (rs1805377) and LIG4 (rs1805388)
Interaction Between Smoking and Genetic Factors
Because smoking is the predominant risk factor for lung cancer, we also stratified our analyses by smoking status of lung cancer patients. There was no evidence of a statistically significant increased risk for never-smoker patients with all 4 SNPs even in the adjusted logistic regression model (data not shown). However, there was a significant difference between the patients who smoked and the controls for XRCC4 (adjusted OR: 2.67; 95% CI, 1.07-7.16 [P = .041]) and LIG4 (adjusted OR: 2.04; 95% CI, 1.15-3.64 [P = .015]) (Table 2).
Association of Polymorphisms in the NHEJ Pathway With Genomic Instability
To examine the association of high chromosomal instability with SNPs in the NHEJ pathway, we analyzed fractional allelic loss FAL status, an indication of genomic instability, in 152 patients with NSCLC using LOH genotyping with 269 microsatellite markers (Fig. 2). FAL was calculated for each patient as the number of markers that demonstrated LOH divided by the number of informative markers. Patients who had genomic instability were categorized with high FAL, which was defined as FAL higher than the mean FAL of all patients analyzed. Our data indicated that patients with a risk allele in the XRCC4 and LIG4 genes had significantly high FAL (adjusted OR: 3.84; 95% CI, 1.59-9.75 [P = .003] for XRCC4; and adjusted OR: 2.03; 95% CI, 1.15-3.66 [P = .016] for LIG4) (Table 2).
Prognostic Significance of the Polymorphisms in NHEJ Genes in Nonsmall Cell Lung Cancer
All 152 patients without chemotherapy were followed after surgery. After several training tests, we observed that there was a trend toward shorter survival in those patients who had 3 putative risk alleles in the NHEJ pathway according to the combined analysis (P = .053) (Fig. 3A). It is noteworthy that the homozygous variant G/G genotype in XRCC4 had the strongest prognostic effect (P = .015; log-rank test) (Fig. 3B). The estimated median survival for patients who had the G/G genotype and other genotypes was 13 months and 42 months, respectively.
Although the potential involvement of defects in the NHEJ repair system and of chromosomal instability is conceivable, to date, no direct clinical evidence has been obtained in patients with lung cancer. Previous studies have demonstrated that polymorphic alleles of NHEJ genes would predispose carriers to a high risk of developing cancer.20-27 Therefore, we hypothesized that polymorphisms in the NHEJ genes were associated with cancer risk and prognosis in lung cancer. In addition, we postulated that some risk SNPs may exhibit gene-environment (smoking) and gene-phenotype (high FAL) correlations in lung tumorigenesis. It also is conceivable that genetic factors for lung cancer may be mutifactorial. Therefore, we tested the gene-gene interaction (joint effect) among SNPs of the NHEJ pathway for lung cancer risk and prognosis. The data generated in this study provide the first evidence to our knowledge that genetic polymorphisms of the NHEJ pathway are involved and exhibit a joint effect in lung cancer susceptibility and a poor prognosis. The risk effect of these polymorphisms is apparent in smoking patients and/or patients with lung cancer who have tumors with high genomic instability.
In the individual SNP analyses, the allele frequencies of XRCC4 (rs1805377, IVS7-1A→G, splice-site) and LIG4 (rs1805388, Ex2 + 54C→T, Thr9Ile) differed significantly between patients and controls (Table 2). In addition, the XRCC4 homozygous variant G/G genotype was associated significantly with poor survival (Fig. 3). The XRCC4 IVS7-1A→G polymorphism involves a substitution of G→A in the intron 7/exon 8 junction region of XRCC4. This SNP potentially abolishes an acceptor splice site at exon 8.28 Although little is known regarding this SNP in terms of the potential impact on repair capacity of XRCC4 and the consequent risk of carcinogenesis, our results study indicate that it is associated strongly with lung cancer risk and prognosis, suggesting that this SNP may alter XRCC4 expression or protein function. The LIG4 Ex2 + 54C→T polymorphism is a nonsynonymous SNP at the N-terminal of the LIG4 protein, which is essential for its activity,29 and is predicted to increase the hydrophobicity of that region.30 Although LIG4 interacts with XRCC4 through the region located between the 2 C-terminal breast cancer 1 (BRCA1) C-terminus (BRCT) domain,31 our gene-gene interaction analysis revealed that SNPs in the XRCC4 (IVS7-1A→G; splice-site) and LIG4 (Ex2 + 54C→T; Thr9Ile) genes interacted to modulate the risk of lung cancer as a joint effect (adjusted OR, 8.75) (Table 3). Biologically, LIG4 forms a heterodimer with XRCC4 to execute the final rejoining step of NHEJ.32 It has been demonstrated that XRCC4 interacts, stabilizes, and stimulates the activity of LIG4.31 In addition, XRCC4 acts as a bridge that links LIG4 to other components of the NHEJ apparatus.33 Our combined analyses also indicated that there was a joint effect between SNPs in the Ku70 and Ku80 genes (adjusted OR, 4.65) (Table 3), although their effect on cancer risk in individual SNP analyses was not significant. The data provided indicate that genes involved in the same pathway can modify the tumorigenic effects of the other partners. Further functional assays will provide the ultimate answer concerning the expressional influence of these polymorphisms.
It has been established that tumorigenesis can be prompted by selective environmental factors and genetic factors. We observed an increased risk of SNPs in Ku70, XRCC4, and LIG4 especially in smoking patients (Table 2). The data reaffirmed that smoking is a risk factor in lung cancer and indicated a biologically meaningful interaction between NHEJ and smoking. In addition, SNPs in XRCC4 and LIG4 are linked significantly with high FAL, an indicator of genomic instability (Table 2). Chromosomal instability with high FAL has been proposed as the pathogenic basis of tumorigenesis caused by the functional inactivation of DSB repair genes, BRCA1 or XRCC5 genes, and p53 damage-response genes in animal models.14, 34, 35 It is a reasonable hypothesis to correlate the variability in DSB repair with lung cancer susceptibility and prognosis. A strength of the current study was the relatively large panel of markers (269 microsatellites) used to define the genomic stability status of each patient who was analyzed. Although the potential involvement of defects in the DSB repair system and of chromosomal instability is conceivable, to our knowledge, no direct evidence has been obtained to date in clinical patients with lung cancer. Given the high frequency of chromosomal abnormalities and perhaps mutations in these patients who had high FAL verified by intensive microsatellite genotyping, our study provided the first evidence that defective NHEJ genes by germ-line mutation may lead causally to an increased likelihood of multiple genetic alterations and, ultimately, carcinogenesis.
In the current study, only patients who underwent surgery without chemotherapy were enrolled. These patients had accurate pathologic staging, and their prognosis analysis would not be complicated by chemotherapy. Our results indicate a significant prognostic effect of the XRCC4 IVS7-1A→G, splice-site G/G variant genotype in lung cancer (Fig. 3). We did not perform a stratification analysis by tumor stage, because there was no difference in XRCC4 IVS7-1A→G G/G variant genotype distribution within various tumor stages of lung cancer (data not shown). The mechanism responsible for the association between the XRCC4 IVS7-1A→G splice-site polymorphism and NSCLC remains to be elucidated. The homozygous variant G/G genotype may encode the lowest XRCC4 protein level, thereby leading to the worst prognosis for patients. Further functional assays for the expressional influence of this polymorphism will be needed.
Several previous studies have assessed SNPs in individual gene or combinable genes in the NHEJ pathway for their associations with risk of different cancers; however, their results were mixed. This may be because of the inclusion of different diseases and ethnic groups or geographic areas. Alternatively, cancer is a complex disease in which the study of a single gene is likely to reveal only a partial component of the genetic etiology. In addition, gene-environment and genotype-phenotype correlations should be taken into account during the study of a complex disease such as cancer. To our knowledge, this is the first nested case-control study of polymorphisms in the NHEJ pathway in relation to lung cancer susceptibility and prognosis. This also is the first study to address the issues of gene-gene interaction, genotype-environment correlation, and genotype-phenotype correlation in the NHEJ pathway in lung cancer. However, given the number of comparisons and the sample size of the current study, the conclusions should be confirmed in larger samples and/or in other ethnic groups. Nevertheless, the use of the polymorphisms in the NHEJ pathway may supplement for the current clinical evaluation methods of risk assessment in population studies and perhaps for disease monitoring of lung cancer in the future. Our study also suggests a future direction for association studies.
We are grateful to professor Yuh-Shan Jou for providing us with the microsatellite primers used in this study.
Conflict of Interest Disclosures
Supported by grants DOH95-TD-G-111-004 and NSC96-2628-B-006-048-MY3 from the National Science Council, Executive Yuan, Republic of China.