As summarized above, bladder cancer is primarily a disease induced by environmental carcinogens. However, there is also a clear genetic component to the aetiology of bladder cancer, as evidenced by the familial association of bladder cancer. Identification of low-penetrance genetic predisposition loci provides additional strong evidence for a genetic component of bladder cancer. Although the relative risks are modest, genetic polymorphisms might account for a large fraction of all bladder cancers because of their high prevalence in the population. Because there is a good understanding of the pathways and genes involved in carcinogen metabolism and host defence mechanisms, numerous studies evaluated associations between genetic polymorphisms in these pathways and bladder cancer risk. There are a few consistent associations, but most of the reported genetic associations have not been supported by replication in independent populations. We will not enumerate all the genetic variants that have been evaluated, but only summarize the most consistent associations and data from larger studies.
Carcinogen metabolism genes
In general, the metabolism of carcinogens consists of two phases, phase I and II. Phase I enzymes, mainly CYP450, typically activate carcinogens, whereas phase II enzymes generally detoxify carcinogens. The balance between phase I and II enzymes often determines the accumulation of toxic reactive intermediates.
There are about 60 human CYP genes arranged in 18 families and 42 subfamilies . Human CYP genes are highly polymorphic. For example, CYP2D6 has >100 variant alleles and CYP2C9 has 37 different variant alleles. Numerous studies have investigated various CYP gene polymorphisms and bladder cancer risk, including CYP1A1, CYP1A2, CYP1B1, CYP2C19, CYP2D6 and CYP2E1[71,72]; however, none of the genotypes showed consistent associations with bladder cancer risk. There are many reasons for this, including small sample sizes, weak correlations between genotypes and phenotypes, redundancy in enzyme functions and overlap in enzyme substrates, and variable frequencies of the polymorphic alleles in different ethnic populations.
NADPH quinine oxidoreductase-1 (NQO1) is a cytosolic enzyme that catalyses two- or four-electron reductions of quinoid compounds into less toxic hydroquinones. A non-synonymous single-nucleotide polymorphism (nsSNP), Pro187Ser, has been widely studied in many cancer types. Several case-control studies of bladder cancer with small sample sizes (<300 cases) have produced contradictory results [72,73], but a meta-analysis of six studies among Caucasians suggested a modestly increased risk for the variant genotypes (OR 1.20, 95% CI, 1.00–1.43) .
Glutathione S-transferases (GSTs) are a group of Phase II detoxification enzymes that catalyse the conjugation of glutathione to a wide variety of xenobiotics and carcinogenic compounds. At least seven mammalian GST families (α, µ, ο, π, σ, τ and ζ) have been identified and several functional polymorphisms in GST family genes have been closely assessed in cancer-association studies . The largest case-control study to date (1150 cases and 1149 controls) reported that the null genotype of GSTM1 was associated with a significantly increased bladder cancer risk . A meta-analysis of 28 studies (5072 cases and 6466 controls) produced a summary odds ratio of 1.5 (95% CI 1.3–1.6) . The Ile105Val variant of GSTP1 might have a modest effect on bladder cancer risk, as suggested in a recent meta-analysis . Other common polymorphisms, such as the GSTT1 null genotype and the Val224Ile variant of GSTM3, appear to have a minimal effect on bladder cancer risk .
NATs catalyse the metabolic activation of aromatic and heterocyclic amine carcinogens by acetylation. NAT1 and NAT2 are two distinct NAT isozymes that exist in humans. The NAT1 and NAT2 genes have extensive polymorphisms that stratify the population into rapid-, intermediate- and slow-acetylator phenotypes . The association between NAT2 slow-acetylator genotype and increased bladder cancer risk is among the most consistent and robust, and has become a classical example of gene–environment interaction in the aetiology of sporadic cancer. The two largest case-control studies and a meta-analysis of 31 studies (5091 cases and 6501 controls) showed that, compared with the NAT2 rapid- or intermediate-acetylators, NAT2 slow-acetylators had a 40% increased risk of bladder cancer [75,78]. In addition, there was an interaction between NAT2 genotype and smoking. The association of NAT1 slow-acetylator genotypes with bladder cancer is controversial and warrants further clarification [72,75,78]. Similarly, polymorphisms in many other phase II enzyme genes, including sulfotransferases, UDP-glucuronosyltransferases, myeloperoxidase, catechol-o-methyltransferase, manganese superoxide dismutase and glutathione peroxidase 1, are either contradictory or evaluated in only a few small studies . Well-designed large studies are needed to clarify their roles in bladder cancer susceptibility.
DNA repair genes
There are four major DNA repair systems in mammalian cells: nucleotide-excision, base-excision, double-strand break and mismatch repair. Tobacco and environmental carcinogens might cause varied DNA damage that requires distinct repair pathways. Polymorphisms in DNA repair pathways, particularly in the first three systems, might be important in the aetiology of bladder cancer. There are many small case-control studies suggesting such roles , but none of the commonly studied SNPs in DNA repair genes showed a consistent significant association with bladder cancer across large case-control studies published more recently [79–87]. These latter studies had large samples and took a pathway-based approach to genotype a panel of SNPs with potential functional impact (e.g. nsSNPs and SNPs in promoter and untranslated regions). Meta-analyses indicate that two XPD nsSNPs, Asp312Asn and Lys751Gln, might have modest effects on bladder cancer (García-Closas, personal communication). The same is true for the Thr214Met nsSNP in the XRCC3 gene (homozygous variant vs homozygous common genotype: OR 1.17, 95% CI, 1.00–1.36, from seven studies with 3086 cases and 3150 controls) .
Other pathways and candidate genes
In addition to carcinogen metabolism and DNA repair, there are numerous individual reports on candidate SNPs in cell cycle control , apoptosis , inflammatory response , cell-adhesion molecules [89,90], tumour microenvironment , folate metabolism [92,93], G proteins , angiogenesis , and other pathways in relation to bladder cancer risk , but the individual results need replication in independent populations. In addition, the rapid advance of high-throughput genotyping technology has allowed large scale genotyping. García-Closas et al. genotyped 1433 SNPs in 386 genes in a large case-control study, and found that a SNP in VEGF was the most significant one. They further analysed 29 additional SNPs in VEGF and found a few additional SNPs and two haplotype blocks encompassing the promoter and 5′ UTR associated with bladder cancer risk . We (X.W.) recently used the iSelect platform (Illumina, San Diego, CA, USA) and genotyped ≈10 000 SNPs in ≈1000 cancer-related genes in the Texas bladder cancer case-control study (data unpublished), and we will validate the top 10 promising loci in an independent population.
Gene-gene and gene-environment interaction
The candidate-gene approach has given a few examples of true associations (e.g. GSTM1 and NAT2), but also produces numerous inconsistent results. The trend in cancer-association studies is to move beyond a limited candidate approach, and apply a pathway-based genotyping and analytical approach, and ultimately genome-wide association (GWA) studies. Several recent bladder cancer case-control studies examined the combined effect of multiple genes, as well as gene-gene and gene-environment interactions, using nonparametric data-mining tools, e.g. classification and regression tree (CART) and multifactor dimensionality reduction [79–83]. Evaluating 44 SNPs in DNA repair and cell-cycle genes, Wu et al. found a significant gene-dosage effect for increasingly elevated risks of bladder cancer with increasing numbers of high-risk alleles. CART analysis revealed potential higher-order gene-gene and gene-smoking interactions, and categorized a few higher-risk subgroups for bladder cancer based on distinct genotype and smoking combinations. Moreover, subgroups with a higher cancer risk also had higher levels of induced genetic damage. It is apparent that gene-environment interaction is important in bladder cancer aetiology, particularly NER genes and smoking interactions. However, from current reports, the interactions identified from various posthoc data-mining tools in different populations do not overlap [79–83]. Unfortunately, validation of interactions is more demanding than that of individual associations and requires large sample sizes.
GWA study of bladder cancer
Some cancer susceptibility loci cannot be identified by a candidate-gene approach, as recent GWA studies of other cancers suggest. High-resolution GWA studies, with extensive replications of positive findings in other case and controls series, can map such susceptibility loci. For bladder cancer, there are at least three independent GWA studies ongoing, all using Illumina’s high-density SNP chips. We expect that there will be exciting results in the coming year or two.
Phenotypic markers and bladder cancer risk
Although the vast majority of modern reports focus on genotypes, because of clear advantages in terms of easy sample collection, invariable genotype data and robust technology, phenotypic assays remain important. For most known genotypes, the functional impact is either not clear or not strong enough to be biologically relevant, contributing to the inconsistent results of genotypic data. Phenotypic assays measure the combined effects of multiple genes as well as environmental factors. Phenotypic markers are generally more consistent if the technical reproducibility is established, and require many fewer samples.
Most of the phenotypic assays measure DNA damage level and DNA repair capacity. Schabath et al. applied the Comet assay to measure baseline, benzo[a]pyrene diol epoxide (BPDE)- and γ-radiation-induced DNA damage in peripheral blood lymphocytes (PBLs) of bladder cancer cases and controls. They found that higher levels of DNA damage at baseline (OR 1.84, 95% CI 1.07–3.15), after γ-radiation (1.81, 1.04–3.14) and after BPDE treatment (1.69, 0.98–2.93) were all associated with increased risks of bladder cancer. Mutagen sensitivity is an established genetic predisposing factor for cancer . A recent large case-control study showed that high BPDE and γ-radiation sensitivity was associated with significantly increased bladder cancer risks (OR 1.48, 1.33–1.64 and 1.92, 1.71–2.15, respectively) . Lin et al. developed a modified host-cell reactivation assay to measure DNA repair capacity for DNA damage induced by 4-aminobiphenyl, an aromatic amine and known bladder carcinogen, and found that poor DNA repair capacity was associated with a 3.42-fold increased bladder cancer risk.
Another consistent phenotypic marker is telomere length in PBLs. Three independent studies showed that constitutive telomere shortening is a cancer-susceptibility factor for bladder cancer. Wu et al. provided the first epidemiological evidence that telomeres in PBLs of cases were significantly shorter than those of controls for four different types of cancer, including lung, head and neck, bladder, and kidney cancer. Dichotomized at the 75th percentile value of telomere length in controls, individuals with shorter telomeres were associated with a significantly increased risk of these cancers (OR 4.51, 2.31–8.81). Broberg et al. showed that the risk of bladder cancer was significantly increased with decreasing telomere length, with an odds ratio of 4.5 for the shortest quartile compared to the longest quartile of telomere length. In a recent prospective case-control study nested within the Health Professionals Follow-up Study and the Nurses’ Health Study, McGrath et al. showed that the OR for bladder cancer was 1.88 (1.05–3.36) for the shortest quartile compared to the longest quartile of telomere length.