|Home | About | Journals | Submit | Contact Us | Français|
Recently, the first genome wide association study of bladder cancer identified an association of genome-wide significance between SNP rs715021 and urinary bladder cancer risk among European individuals. This SNP is in a haplotype block in linkage disequilibrium with the TP63 gene. We investigated the role of this SNP among 1,042 cases and 1,123 controls among non-Latino whites in Los Angeles County, California, USA, and among Chinese in Shanghai, China. We confirmed an association between the A allele and bladder cancer risk (log-additive per A allele OR = 1.24; 95% CI = 1.02–1.52, p = 0.032) in Los Angeles County and a similar association in Shanghai (log-additive per A allele OR = 1.21; 95% CI = 0.98–1.49, p = 0.080). These estimates did not differ by study site, smoking status, or gender. However, the effects were greater in older individuals. Analysis within non-Latino whites for whom we had histological results, revealed that this association was restricted to low risk tumors (OR = 1.49, 95% CI = 1.17–1.92, p = 0.002) and absent among high-risk tumors (OR = 1.03, 95% CI = 0.80–1.33, p = 0.790; heterogeneity p = 0.019). A positive association (OR = 1.56; 95% CI = 0.93–2.62, p = 0.089) was only observed among high-risk tumors from individuals older than 56 years old (interaction p = 0.045). Our results suggest that a TP63 gene variant may increase susceptibility for the development of urinary bladder tumors with low risk of progression.
Genome-wide association studies (GWAS) have emerged as feasible approaches for the identification of susceptibility genes. Novel information was obtained from GWAS of breast, colorectal, lung, and prostate cancer, and other common diseases (1). Recently, Kiemeney et al. (2) reported results from the first GWA study of urinary bladder cancer, conducted in Europe. They reported strong associations with single nucleotide polymorphism (SNP) rs9642880 on chromosome 8q24 and SNP rs710521 on chromosome 3q28. Whereas several markers in 8q24 have been identified to associate with various cancers, reviewed in (3), this was the first report of an association with a marker in 3q28, which maps closely to the p63 gene. We report findings from the investigation of the role of the 3q28 rs710521 SNP in bladder cancer risk among individuals in a case-control study of transitional cell carcinoma (TCC) of the urinary bladder conducted in Los Angeles County, California, USA, and Shanghai, China.
Subjects were enrolled in two population-based case-control studies at two sites: Los Angeles County, California, USA, and Shanghai, China. The characteristics of the Los Angeles County component have been previously described (4, 5). Briefly, non-Asian urinary bladder TCC cases diagnosed between 1987 and 1996, residents of Los Angeles County, 25–68 years of age at diagnosis were identified by the Los Angeles County California Cancer Surveillance Program. Controls were identified from the neighborhoods of residence of the cases using a standard procedure, and were 1:1 matched to cases by age (± 5 years), gender, and race/ethnicity (non-Latino white, Latino, or African American). In the Shanghai component, city of Shanghai Chinese residents diagnosed between 1995 and 1998 who were 25–74 years of age at diagnosis were identified by the Shanghai Cancer Registry. Patients were enrolled into the study between July 1996 and June 1999. Population-based controls were randomly selected from the city of Shanghai using an algorithm previously described (6), and were frequency matched to bladder cancer cases by gender and 5-year age groups during the recruitment period (July 1996 through June 1999). All individuals were administered an in-person questionnaire that ascertained information on demographic, lifestyle, and medical characteristics up to a reference date: two years before cancer diagnosis for cases and two years before the interview for controls (Shanghai study) or diagnosis of the index/paired case (Los Angeles County). At present, tumor pathology reports are available only for Los Angeles County site cases.
Peripheral blood samples were requested from all participants in the Shanghai study, and from Los Angeles County participants interviewed from 1992 on. In Los Angeles County, 88% of cases were non-Latino whites (89% controls), 6% of cases were Latinos (7% controls) and 6% of cases were African-American (4% controls). In Shanghai, all participants were Han Chinese. Given these numbers, we restricted analyses to non-Latino white and Chinese participants. For this report we included 1,042 cases (501 from Los Angeles County and 541 from Shanghai) and 1,123 controls (588 from Los Angeles County and 535 from Shanghai) for whom we had questionnaire data and DNA available. Studies were approved by the institutional review boards at the University of Southern California, the University of Minnesota and the Shanghai Cancer Institute, as appropriate.
DNA was extracted from peripheral blood lymphocytes by standard methods. Los Angeles County study samples were whole genome amplified using REPLI-g mini kit from Qiagen (Valencia, CA). We genotyped for rs710521 using a Taqman assay from Applied Biosystems (Foster City, CA). For quality control 5% samples were duplicated. We used an ABI 7900HT Sequence Detection and Scoring System for allele scoring. Duplicate samples showed 100% concordance, the overall call rate was >98%.
We checked among non-Latino white and Chinese controls separately for deviations from expected genotypic frequencies under Hardy-Weinberg Equilibrium using chi-square tests. We observed no differences among non-Latino white controls. Among Chinese controls, we observed a statistically significant reduced number of heterozygous genotypes (p = 0.02). The genotyping data showed clear genotype separation, call rate >98%, and 100% concordance. Therefore, it is unlikely that we have genotyping error. Given the homogeneity of the population ascertained in the Shanghai study population stratification is an unlikely cause of the observed deviations. We cannot offer other explanations for this discrepancy.
We created a variable that classified subjects according to their reference age (<45, 45–49, 50–54, 55–59, >=60 for non-Latino whites, and <45, 45–49, 50–54, 55–59, 60–64, ≥65 for Chinese), gender, and study site (Los Angeles or Shanghai), and used it to group subjects in conditional logistic regression models to estimate relative risks with odds ratios (ORs) and 95% confidence intervals (95%CI). We estimated per allele ORs assuming a log-additive mode of risk for the A allele to agree with those reported by Kiemeney et al. (2). We tested for gene×exposure interactions on a multiplicative scale and assuming a log-additive mode of risk for the A allele using likelihood ratio tests. We considered the following categorical smoking variables: smoking status (never, former, current), smoking intensity (never, ≤ 20, >20 cigarettes per day), smoking duration (never, ≤ 30, > 30 years of smoking), and pack-years of smoking (never, ≤ 22.5, >22.5 pack-years). For interaction analyses with age, we considered a categorical variable that used as cutpoint the median among Los Angeles County (56 years) or Shanghai (64 years) controls. We tested for heterogeneity of the gene effects across both study sites using likelihood ratio tests. Among Los Angeles County participants we estimated genotype effects within different tumor subtypes using conditional logistic models, as described above. We subtyped tumors by risk of progression (low or high), as defined by by Kiemeney (low risk = Ta tumors with WHO 1973 differentiation grade 1 or 2; high risk = T1–T4 stage tumors, carcinoma in situ (CIS), and all tumors with WHO 1973 differentiation grades 3–4)(2). We tested for heterogeneity of the genotype effect across different histological subtypes using case-only analyses. All statistical tests were two-sided and all analyses were conducted using STATA (STATA Corporation, College Station, TX).
Key characteristics of the two study populations are summarized in Table 1. Among non-Latino white controls the allelic frequency for the rs710521 A allele was 72%, which is comparable to the frequencies reported by Kiemeney et al. (2) among Europeans. Among Shanghai Chinese controls the corresponding allelic frequency was 79%, which is comparable to the prevalence reported for Han Chinese in HapMap (76%, HapMap release 27) and identical to the one recently reported in another case-control study among Chinese (7).
Among non-Latino whites we observed a statistically significant positive association between the A allele and bladder cancer risk (per A allele OR = 1.24; 95% CI = 1.02–1.52, p = 0.032) (Table 2), which is comparable to the estimates reported by Kiemeney et al (2). We observed a similar OR for the corresponding association among Shanghai Chinese, albeit not statistically significant (per A allele OR = 1.21; 95% CI = 0.98–1.49, p = 0.080). We saw no evidence of heterogeneity of risk estimates between non-Latino whites and Chinese.
The association between the rs710521 and bladder cancer risk did not differ by smoking status or gender. However, it was greater in older individuals in both the Los Angeles County and Shanghai data, albeit this difference did not reach statistical significance (combined interaction p = 0.06), data not shown.
Among Los Angeles County individuals, we investigated potential heterogeneity of the association between rs710521 and bladder cancer risk by risk of progression, as defined by Kiemeney et al (2). The association between rs710521 and cancer risk was restricted to low-risk tumors (per A allele OR = 1.49; 95% CI = 1.17–1.92) and absent in high-risk tumors (per A allele OR = 1.03; 95% CI = 0.80–1.33)(heterogeneity test p = 0.019) (Table 3). We found no evidence of effect modification by gender or smoking among low- or high-risk tumors. As noted above, the risk associated with allele A was greater in individuals older than 56 years, and this was seen in low-risk tumors (≤ 56 years, per A allele OR = 1.29, 95% CI = 0.92–1.82; >56 years per A allele OR = 1.77, 95% CI = 1.22–2.56) and high-risk (≤56 years, per A allele OR = 0.81, 95% CI = 0.56–1.15; >56 years, per A allele OR = 1.36, 95% CI = 0.94–1.98). These differences only achieved statistical significance in the high-risk group (p = 0.045), but not the low-risk group (p = 0.221).
Our results suggest that the rs710521 A allele may increase susceptibility for the development of urinary low-risk bladder tumors. In contrast, Kiemeney et al (2) did not find heterogeneity for the association between rs710521 and cancer risk when subtyping tumors by risk of progression and comparing the allelic frequencies. However, they did report a slightly higher frequency of the A allele among low-risk than high-risk individuals in their GWAS first stage. Whereas this difference did not reach statistical significance (p = 0.129), it indicates a similar trend in their data to ours. We note that the average age of study participants in Los Angeles was 54 (range 25–68) years old, whereas the average age at diagnosis for study participants in the “discovery” phase of their GWAS was 67 (range 22–94) years old. The older average age of the European study might have diminished the differences in the association between SNP rs710521 and bladder cancer risk when considering tumor sub-types, given that these differences seem to be most relevant among younger individuals. In our study, we did observe a non-statistically significant association between rs710521 and high-risk bladder cancer among individuals older than 56 years. We recognize that our analyses by risk of progression are based on limited numbers in some categories; therefore, replication in larger studies will be needed. Recently, a case-control study conducted in a Chinese population reported a positive association between the rs710521 A allele and bladder cancer risk (per A allele OR = 1.23; 95% CI = 0.98–1.55) of similar magnitude to the one reported by Kiemeney et al among whites, and the one reported here among Shanghai Chinese (7).
The rs710521 SNP maps between the TP63 and LEPREL1 genes in 3q28. There is little evidence of linkage disequilibrium (LD) between rs710521 and LEPREL1. Among both Caucasians and Han Chinese, rs710521 is in a haplotype block in LD with TP63 (e.g. for CEU population in HapMap, D’ = 1, r2 = 1 for the pairwise comparison between rs710521 and rs4687100, which maps within the last intron of TP63). The p63 protein is a transcription factor that transactivates p53-responsive genes, induces cell cycle arrest, apoptosis, and regulates epithelial differentiation (8, 9). Alternative splicing of TP63 generates isoforms that either retain the ability to transactivate p53-responsive genes (TAp63 isoforms) or lack the transactivation domain (ΔNp63 isoforms). TP63 is rarely mutated in bladder tumors (10); however, reduced p63 expression correlates with progression to invasive tumors (11). Tumors that retain p63 high levels are more likely to express the ΔNp63 isoform (12), which is expressed strongly in normal bladder urothelium basal and intermediate cell layers and in non-invasive tumors, with reduced levels in invasive tumors of worse prognosis (11–13). Overexpression of ΔNp63 is proposed to keep cells in a stem-cell like state, promoting growth and proliferation, inhibiting apoptosis, but reducing invasion potential (11). We speculate that the rs710521 A allele might tag a TP63 variant that increases expression of ΔNp63 relative to the TAp63 isoform, thus contributing to development of low-risk tumors, with low invasion potential (12). Kiemeney et al (2) did not find a correlation between the rs710521 genotype and TP63 mRNA expression in whole blood and adipose tissue. However, that study did not consider the different p63 isoforms and expression levels in the urothelium.
In conclusion, our study confirms a role for a variant in 3q28 as a putative susceptibility locus for low-risk Ta bladder cancer. Our finding of similar effects of the rs710521 A allele among cigarette smokers and non-smokers would suggest that risk factors other than tobacco smoke may contribute to bladder cancer risk, possibly through a p63-dependent pathway.
The authors thank Ms. Susan Roberts for oversight of study enrollment, Drs. Amit D. Joshi, Dr. Xuejuan Jiang, and Dr. Juan Pablo Lewinger for manuscript preparation, and Ms. Peggy Wan for data analysis and management.
NCI grants 1P01CA86871 and 1R01CA114665.