Introduction

Urinary bladder cancer (UBC) is one of the most common cancers of the urinary system and is much more common in males worldwide, including in Japan. UBC is classified by the stage and grade of the tumor, which are highly correlated with recurrence, progression and patient survival rates. Tumor stage is determined by the degree of invasiveness and metastasis, whereas tumor grade classification is based on the degree of differentiation. Non-invasive UBCs are designated as stage Ta, whereas stages T1, T2, T3 and T4 refer to invasion into the subepithelial connective tissue, muscle, perivesical tissue and adjacent organs, respectively. With regard to tumor grades, the degrees of differentiation are based on tumor histology: well differentiated (G1), moderately differentiated (G2) and poorly differentiated (G3) cancer.1, 2 Most of the UBC cases clinically fall into one of the two distinct tumor entities at diagnosis related to the risk of patients.3, 4 The first comprises non-invasive low-grade carcinoma of relatively well-differentiated histology (G1 or G2) with stage Ta, which rarely progresses to a higher stage and pose a minimal risk for the patients. On the other hand, the second refers to aggressive tumors corresponding to stages T1 to T4 with higher grades (G2 to G3), showing high recurrence and progression rates, and therefore pose a serious risk for the patients.

Carcinogen exposure is one of the major risk factors contributing to UBC incidence.5 Genes involved in DNA repair mechanisms, DNA replication, transcription, DNA damage signaling, cell cycle and metabolic processes influence the development and progression of UBC. A number of genetic polymorphisms in these types of genes were reported to be associated with genetic susceptibility to UBC.6, 7, 8 However, genetic association analyses focusing on the invasiveness of UBC are yet to be performed.

In this study, we focused on 121 genes acting on the four pathways of DNA repair (base excision repair, nucleotide excision repair, double strand break repair and mismatch repair), as well as those related to DNA synthesis, cell cycle and metabolism. We designed a panel of 1536 single nucleotide polymorphism (SNP) markers from these genes for extensive genotyping study, of which 802 SNPs corresponding to 114 genes that passed quality control process were statistically analyzed for their association with UBC invasiveness in Japanese patients.

Materials and methods

Study population

DNA samples were collected at the Department of Urology, Kyoto University, Kyoto, Japan. The stage and the grade of UBC were judged by the histological examinations at diagnosis. Non-invasive group (group-N) corresponds to stage Ta patients with grades G1 or G2, whereas invasive group (group-I) refers to those of stages T1 to T4 with grades G2 or G3. The characteristics and the detailed clinical diagnosis of the study populations were summarized in Table 1 and Supplementary Table 1, respectively. All the patients provided their written informed consents according to the protocols approved by the ethical review board of Kyoto University Graduate School of Medicine.

Table 1 Characteristics of UBC patients used for statistical analysis

Selection of candidate genes and construction of SNP genotyping panel

Identification of SNPs was performed by resequencing exons and flanking regions of 121 genes (summarized in Supplementary Table 2), using 32 each of DNA samples of Japanese, French, Congolese and Thai populations as reported.9 Alignment and genotyping were performed using the Genalys software (Centre National de Génotypage, Institut Génomique, Commissariat à l'Énergie Atomique (CEA), Evry, France, http://software.cng.fr/).10 As for the selection of SNP markers, SNPs located in linkage disequilibrium (LD) blocks encompassing the 121 genes and essentially covering all the major haplotypes with frequencies greater than 5% were initially chosen from the International HapMap Project data of Japanese, Caucasians and Africans.11 Additional SNPs identified in at least one of the four populations by the SNP discovery were included in the selection. Finally, we generated a panel of 1536 SNPs, which were chosen as tag SNP markers for genetic analyses of multiple populations using GoldenGate technology (Illumina, San Diego, CA, USA).

SNP genotyping

Genomic DNAs were prepared from peripheral blood leukocytes using Qiagen DNA Extraction kits (Qiagen, Hilden, Germany). The first screening was performed using GoldenGate assay (Illumina). The validation analysis of marker rs17650301 was performed with additional samples by sequencing. The forward primer (5′-AGGCTGGTAGGGTCCAAAGT-3′) and the reverse primer (5′-AGGGTTAGGTTGAGCATCCC-3′) were used for the PCR and an internal primer (5′-GAAGTTTTCACCGTGTTGCC-3′) was used for the sequencing with BigDye Terminator v3.1 Cycle Sequencing kit (Life Technologies, Foster City, CA, USA).

Quality control and association analysis

Among the 1536 SNPs genotyped, 20 SNPs on X-chromosome and one triallelic SNP were initially removed before quality control process. During the quality control, we first examined the presence of DNA samples showing high degrees of kinship by plink (Center for Human Genetic Research, Massachusetts General Hospital, Boston, MA, USA, http://pngu.mgh.harvard.edu/~purcell/plink/)12 and found there were no such samples. Three DNA samples with call rates smaller than 0.90 were removed. Regarding the SNP markers, 104 SNPs with calling rates smaller than 0.95 and 609 SNPs with minor allele frequencies smaller than 0.01 were also excluded. After these steps, the remaining 802 SNP markers corresponding to 114 genes were used for statistical analyses. The association was examined by trend exact test for genotype distribution. Odds ratios and 95% confidence intervals were estimated with additive model. The heterogeneity of the effect size between the first screening and the validation analysis was evaluated by Breslow–Day test.13 The LD structure of the POLG2 locus was established by using Japanese data from the International HapMap Project.11

Expression quantitative trait loci analysis

Allele-specific effect on the expression of rs17650301 was performed using Japanese data set GSE6536 of expression profile Gene Expression Omnibus (National Center for Biotechnology Information, National Institutes of Health, Bethesda, MD, USA, http://www.ncbi.nlm.nih.gov/geo/).14 The association P-values were obtained by Jonckheere-Terpstra method using R software (Department of Statistics, The University of Auckland, Auckland, New Zealand, http://www.r-project.org/) or SPSS (version 18, IBM, Armonk, NY, USA, http://www-01.ibm.com/software/data/spss/products/statistics/statistics-desktop/).

Results

First screening

Among 1536 SNPs corresponding to 121 candidate genes in the genotyping panel, a total of 802 SNPs derived from 114 candidate genes were examined for their association with UBC invasiveness, using 106 group-N samples and 93 group-I samples (for details, see Materials and Methods). Although no SNPs showed significant association (P=6.23 × 10−5) after Bonferroni's correction for multiple testing, we identified 55 SNPs with nominal P-values smaller than 0.05, of which nine SNPs showed P-values smaller than 0.01 (Table 2). The strongest P-value was obtained for rs17650301 (P=3.40 × 10−3) located in intron 5 of DNA polymerase subunit γ2 gene (POLG2) at 17q24.1. Five SNPs in the POLL gene, rs3730476, rs1055364, rs3730465, rs3730472 and rs3095795, were in complete LD with each other.

Table 2 Results of association analysis in the first screening

We then performed a subgroup analysis by dividing the UBC patients according to their gender. In the male study, six SNPs showed P-values smaller than 0.01, of which the strongest association was again obtained for rs17650301 (P=6.93 × 10−4) (Table 2). However, rs17650301 did not show association in the female study. In addition, rs2518968, which showed the second lowest P-value in the joint study (P=4.50 × 10−3), showed association in the male study (P=4.81 × 10−3), although it was much weaker than that of rs17650301.

Validation analysis

We focused on rs17650301 and performed a validation analysis by direct sequencing, using DNA samples of an independent set of UBC patients consisting of 65 group-N samples and 114 group-I samples. Although the association was not observed in the male–female joint study (P=0.432), we obtained a marginal P-value (P=0.0396) in the male study (Table 3). The combined analysis by pooling the genotypes of the two studies returned a P-value of 9.45 × 10−3 and 1.67 × 10−4 in the joint study and in the male study, respectively (Table 3). The P-value of heterogeneity did not show evidence of differentiation in effect size (P=0.17 for the joint study and P=0.47 for the male study). A logistic regression analysis adjusted for age and sex returned a P-value of 5.28 × 10−3 for the joint study. Also, the P-value did not change (P=1.67 × 10−4) after adjusting for age in the male study.

Table 3 Summary of the association analysis of rs17650301 with UBC invasiveness

Linkage disequilibrium block

The LD block of the DNA region encompassing rs17650301 in the Japanese population was established using genotype data of the International Hapmap Project.11 In the LD block encompassing a 34-kb DNA between rs8068311 and rs6504238, there were no SNP markers showing strong LD (r20.8) with rs17650301, except rs1991401 (r2=1). Rs1991401 is located in the 5′ untranslated region of the DDX5 (DEAD box polypeptide 5) gene adjacent to POLG2 (Figure 1).

Figure 1
figure 1

LD structure of the 34-kb region spanning the POLG2 and DDX5 genes. The structure and transcriptional polarity of POLG2 and DDX5 are shown according to NCBI Reference Sequence Build 36. Exons are shown by filled and open rectangles representing the coding and untranslated regions respectively. Pairwise LD estimation scores between SNPs within the region were converted into colors according to the color scheme of Haploview. The SNPs included in the genotyping panel for the first screening are indicated by an asterisk.

Rs17650301 and expression of POLG2

Although rs17650301 is located in an intronic region, it may have functional impact in the transcription of POLG2. We performed an expression quantitative trait loci (eQTL) analysis by examining the expression profiles of 44 lymphoblastoid cells of Japanese origin in Gene Expression Omnibus database14 and found an increase of POLG2 transcription by the risk allele (C) of rs17650301 (P=0.004, Figure 2). However, a search for cis-regulatory elements using Jaspar database (The Bioinformatics Centre, Department of Biology, University of Copenhagen, Copenhagen, Denmark, http://jaspar.genereg.net/)15 failed to identify any known motifs covering rs17650301.

Figure 2
figure 2

Allelic expression of POLG2 with the genotypes of rs17650301. The levels of POLG2 transcription in cell lines carrying the three genotypes are indicated by a square. ‘C’ and ‘A’ represent the risk and the alternative alleles, respectively.

Discussion

In this study, we conducted a case-control association analysis of UBC invasiveness by taking a candidate gene approach of 1536 SNPs in 121 genes related to DNA repair mechanisms, DNA synthesis and metabolic processes. To our knowledge, this is the first genetic study focusing on the invasiveness of UBC. Statistical analysis was performed for 802 SNPs in 114 genes that passed quality control. The majority of the SNP markers (627 out of 734) excluded from the statistical test showed minor allele frequencies smaller than 0.01 in the Japanese population. This is due to the design of the genotyping panel for trans-ethnic study by covering SNPs identified in at least one of the four populations (Japanese, French, Thai and Congolese). Indeed, there were as many as 934 SNPs whose minor allele frequencies in the SNP discovery were smaller than 0.01 in Japanese and greater than 0.01 in at least one of the three other populations.

There was no SNP showing a significant association after Bonferroni's correction for multiple testing (P=6.23 × 10−5). The Bonferroni method is a very stringent test; although the possibilities of false positives are decreased, it comes together with the risk of losing potential candidates with true associations. In order to overcome this drawback, we took a strategy to screen SNP markers for potential associations in two stages. The most important aspect of our results is that the association of rs17650301 with UBC invasiveness was reproducibly obtained in two independent sample collections, without evidence of differentiation in effect size. However, the detection of significant association will require replication analyses using other sample collections.

The genotyping of 802 SNP markers followed by the validation analysis identified rs17650301 located in intron 5 of the POLG2 gene as the strongest candidate for the invasiveness of UBC in Japanese male patients. POLG2 encodes the 55-kDa accessory subunit of mitochondrial DNA polymerase. This subunit, together with the 140-kDa catalytic subunit (POLG), stimulates the polymerase and exonuclease activities in the replication process of mitochondrial DNA.16, 17 Although the association between POLG2 polymorphisms and bladder cancer has not yet been established, its association with the risk of head and neck cancer was demonstrated in the French population.18 There is no strong biological evidence to support the functional importance of POLG2 to the invasiveness of UBC. However, the increase of transcription level with the risk allele of rs17650301 may be a reason for the association. Because there are no other known SNPs in strong LD (r20.8) with rs17650301 in the POLG2 locus, rs17650301 is a good candidate marker for the invasiveness of UBC in Japanese males.

The strong association of rs17650301 with the UBC invasiveness was observed only in males. As such, the reasons for the observed higher UBC incidence in males than females remain uncertain. Different non-genetic risk factors including sex hormones, life style and environment may contribute to the disease onset, resulting in the sex-specific association. For instance, involvement of androgen and its receptor in bladder cancer is demonstrated in a mouse study.19 Epidemiologically, postmenopausal women have a higher risk of development and progression of UBC than premenopausal women.20 Also, smoking is one of the major risks of urinary tract cancer and considered to increase the cancer risk by approximately threefold,21, 22 and the percentage of smokers is much higher in males than in females in Japan (36.8 versus 9.1%, as of 2008 in the National Survey by the Ministry of Health, Labour and Welfare of Japan). If such environmental and life style-related factors contribute to the predisposition of bladder cancer together with POLG2, this may explain the observed effect of POLG2 with UBC invasiveness in males only. However, further functional characterization of POLG2 for its involvement in carcinogenesis of indolent and aggressive tumors are required for the elucidation of the molecular mechanism underlying the prognosis of UBC.