Cell cycle checkpoint regulation is crucial for the prevention of carcinogenesis in mammalian cells.
Cell cycle checkpoint regulation is crucial for the prevention of carcinogenesis in mammalian cells.
To test the hypothesis that common sequence variants in the cell cycle control pathway may affect bladder cancer susceptibility, the effects of a panel of 10 potential functional single nucleotide polymorphisms (SNPs) from 7 cell cycle control genes, P53, P21, P27, CDK4, CDK6, CCND1, and STK15, were evaluated on bladder cancer risk in a case-control study of 696 bladder cancer cases and 629 healthy controls.
Overall, on individual SNP analysis only individuals with the p53 intron 3 16-bp duplication polymorphism variant allele had a significantly reduced bladder cancer risk (odds ratio [OR] = 0.74, 95% confidence interval [CI], 0.56–0.96). This effect was more evident in former smokers and younger subjects. We then applied the Classification and Regression Tree (CART) statistical approach to explore the high-order gene-gene and gene-smoking interactions. In the CART analysis, smoking status was identified as the most influential factor for bladder cancer susceptibility. The final decision tree by CART contained 6 terminal nodes. Compared with the second-lowest risk group the ORs for terminal nodes 1 and 3 to 6 ranged from 0.46 to 6.30.
These results suggest that cell cycle genetic polymorphisms may affect bladder cancer predisposition through modulation of host genome stability and confirm the importance of studying gene-gene and gene-environment interactions in bladder cancer risk assessment. Cancer 2008. © 2008 American Cancer Society.
Bladder cancer is the fourth most frequently diagnosed cancer in men; estimated totals of 67,160 new cases and 13,750 deaths are expected in the US in 2007.1 The most common type of bladder cancer is transitional cell carcinoma (TCC), which accounts for 90% of all cases. Cigarette smoking is the most prominent risk factor for bladder cancer. Smokers are more than twice as likely to develop bladder cancer as nonsmokers. Exposure to aromatic amine is another major risk factor. Other epidemiologic risk factors include work-related exposure in industries related to radioactive material, arsenic, dry cleaning fluids, dyes, rubber, or leather.2, 3
Mounting evidence indicates that, besides environmental factors, genetic components and gene-gene and gene-environment interactions also play important roles in bladder cancer development.4–6 Considering the complexity of bladder carcinogenesis, single-factor studies may not have sufficient power to detect small genetic effects on cancer risk. Studies that look at gene-gene and gene-environment interactions in the same causal pathway may provide more promising insights.7, 8
Cell cycle checkpoint functions regulate cell cycle progression and cell proliferation. Defects of cell cycle control are 1 of the hallmarks of cancer development and may have relevance in cancer diagnosis and treatment.9–11 In this study, we selected 10 single nucleotide polymorphisms (SNPs) from 7 cell cycle control pathway genes, detailed in Table 1, and studied their associations with bladder cancer risk. We applied the Classification and Regression Tree (CART) statistical model to examine the high-order gene-gene and gene-environment interactions in cell cycle control pathways.
|Gene symbol||Gene name||Chromosomal location||Amino acid change||dbSNP ID||Common/variant alleles||Control population MAF|
|P53||Tumor protein p53||17p13.1||Intron 3*||16 bp duplication||0.1537|
|CDK4||Cyclin-dependent kinase 4||12q14||Promoter||rs2072052||A/C||0.2922|
|P27||Cyclin-dependent kinase inhibitor 1B||4q43||5′UTR||rs34330||C/T||0.2386|
|P21||Cyclin-dependent kinase inhibitor 1A||6p21.2||3′UTR||rs1059234||C/T||0.0745|
|CDK6||Cyclin-dependent kinase 6||7q21–q22||3′UTR||rs2309||C/T||0.2576|
|STK15||Aurora kinase A||20q13.2–q13.3||F31I||rs2273535||T/A||0.1968|
Patients with newly diagnosed and histologically confirmed bladder cancer were recruited from the University of Texas M. D. Anderson Cancer Center and Baylor College of Medicine between 1999 and 2003. None of the case subjects had undergone chemotherapy or radiation before enrollment to the study and all were interviewed by trained M. D. Anderson Cancer Center staff. Controls were accrued from healthy individuals without a history of cancer (except nonmelanoma skin cancer) at Kelsey-Seybold Clinic, the largest private multispecialty physician group in the Houston, Texas, metropolitan area. Control subjects were frequency matched to cases by age (±5 years), sex, and ethnicity. After an initial screening comprising a short survey about willingness to participate and to provide demographic data for matching, the potential control subjects were contacted by telephone at a later date to confirm their willingness to participate and to set up an interview at a Kelsey-Seybold site convenient to the participant. The response rates for cases and controls were 92% and 75%, respectively. All the study participants provided blood samples. The bladder cancer types were as follows: 55.75% superficial, 42.53% invasive, and 1.72% unconfirmed. Among the cases, the proportions of patients with a CIS component, without CIS component, and with unknown CIS information were 24.3%, 67.2%, and 8.5%, respectively.
Demographic and smoking history data was obtained from all study participants through a 45-minute interview based on a standardized risk factor questionnaire. After completion of the interview, a 40-mL blood sample was collected into coded heparinized tubes and sent to the laboratory for immediate molecular analyses. An ‘ever smoker’ was an individual who had smoked at least 100 cigarettes in his or her lifetime. An individual who had quit smoking at least 1 year before the study was classified as a former smoker. Ever smokers included former smokers, current smokers, and recent quitters (those who had quit within the last year). All study participants signed informed consent in accordance with the Institutional Review Boards of M. D. Anderson, Baylor College of Medicine, and Kelsey-Seybold Clinic.
Genomic DNA was isolated from peripheral blood with a Qiagen (Chatsworth, Calif) kit. Genotyping was performed using the Taqman method with a 7900 HT sequence detection system (Applied Biosystems, Foster City, Calif), except for the p53 intron 3 polymorphism, which was genotyped using polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP).12 Sample DNA (5 ng), 1× Taqman buffer A, 200 μM deoxynucleotide triphosphate, MgCl2 (5 mM), AmpliTaq Gold (0.65 U), each primer (900 nM), and each probe (200 nM) constituted the amplification mixes (5 μL). The thermal cycling conditions started with 1 cycle for 10 minutes at 95°C, followed by 40 cycles of 95°C for 15 seconds, and the final 40 cycles of 60°C for 1 minute. Each plate also contained a water control, ample internal controls, and previously genotyped samples to guarantee genotyping accuracy. As a measure of quality control, 5% of the samples were randomly selected and run in duplicate, with 100% concordance.
We performed Pearson χ2 or Fisher exact tests to examine the association of patient characteristics, such as sex, smoking status, and the genotypes of each SNP with those of control subjects. For continuous variables, a Student t-test was used to assess any significant differences between case and control subjects. Odds ratios (ORs) for overall analysis and stratified analysis and 95% confidence interval (CI) were computed as estimates of relative risk. We applied unconditional multivariate logistic regression to adjust for possible confounding by age, sex, and smoking status, where appropriate. These analyses were conducted using the Stata 8.0 software package (Stata, College Station, Tex). We used the Benjamini-Hochberg method to estimate the false discovery rate (FDR).13 We set the FDR at the level of 5% and calculated the FDR-adjusted P-values at this level to assess if the resulting P-values were still significant after multiple comparisons were taken into consideration. The high-order gene-gene and gene-environmental interactions were explored by CART analysis. HelixTree Genetics Analysis Software (v. 4.1.0, Golden Helix) was utilized to perform CART analysis. CART uses recursive partitioning to build a decision tree that enables identification of subgroups of individuals at differential risks.14 We selected P-values to measure goodness of split and control tree growth (P < .05). The process starts with the root node that contains all the case and control subjects. It finds the split with the smallest P-value, ie, the most optimal split, for the root node and each subsequent node. This process continues until there is no statistically significant split or the split will result in nodes with fewer than 10 subjects. The high-order gene-gene and gene-smoking interaction identified by the CART were further analyzed using likelihood ratio test comparing the model with or without interaction terms in the multivariate logistic regression.
We restricted the analysis to Caucasian subjects who comprised the majority (≈90%) of the study population. Among the 696 patients with newly diagnosed bladder cancer and 629 healthy controls, there were 56 patient subjects and 9 control subjects with missing genotypes among all the 10 SNPs. They were excluded from the analysis below and 640 patients and 620 controls were used in the final analysis. Since the study is still ongoing, we have not achieved perfect matching yet. There was a significant sex difference: cases were 78.13% male and controls were 72.58% male (P = .022). Case patients were significantly older (64.1 ± 11.2 years) than controls (62.8 ± 10.5 years; P = .03), more likely to be ever smokers (74.22%) than control subjects (54.03%; P < .001), and cigarette consumption among ever smokers was significantly higher for cases (42.8 pack-years) than for controls (28.4 pack-years; P < .001).
Table 2 summarizes the associations between individual SNPs and bladder cancer susceptibility. The age cutoff point was chosen because it was the lower two-fifths percentile of the age distribution in the control population. For the main effect of each SNP, only P53 intron 3 variant allele-containing genotypes showed a significantly protective effect (OR = 0.74, 95% CI, 0.56–0.96). In the stratified analysis, carriers of at least 1 variant allele of the P53 intron 3 polymorphism had a significantly reduced risk, with an adjusted OR of 0.49 (95% CI, 0.27–0.89) in those younger than 55 years (the number of young subjects in control/case group for no insertion, 1 insertion, and 2 insertions were as follows: 88 of 101, 37 of 27, and 4 of 1, respectively, P for trend .013; Table 2). This effect of significantly reduced risk for individuals younger than 55 was also exhibited by P53 intron 6 (OR = 0.54, 95% CI, 0.30–0.97; P for trend .03). Risk among former smokers containing variant allele genotypes of P53 intron 3 was also markedly reduced, with an adjusted OR of 0.66 (95% CI, 0.45–0.96; Table 2). Never smokers with rare homozygote of the CDK4 3′ untranslated region (UTR) showed a significantly increased risk of bladder cancer compared with never smokers with common homozygote and heterozygote of CDK4 3′ UTR (OR = 2.04, 95% CI, 1.00–4.13). Individuals who were current smokers and heterozygous for an SNP in the CDK6 3′ UTR exhibited a significantly reduced risk (OR = 0.46, 95% CI, 0.22–0.93). After adjusting for the FDR at the 5% alpha level, none of the association in overall and stratified analyses remained statistically significant (data not shown).
|Controls/cases||Overall*||Male†||Female†||Age, y < 55*||Age, y ≥ 55*||Never smoker‡||Former smoker‡||Current smoker‡|
|AG||300/293||0.87 (0.67–1.14)||0.88 (0.65–1.19)||0.83 (0.48–1.44)||0.93 (0.53–1.64)||0.85 (0.63–1.16)||1.11 (0.70–1.74)||0.80 (0.55–1.16)||0.63 (0.30–1.34)|
|AA||113/106||0.82 (0.58–1.16)||0.86 (0.58–1.28)||0.71 (0.35–1.45)||0.75 (0.36–1.56)||0.83 (0.56–1.22)||0.92 (0.51–1.65)||0.85 (0.53–1.38)||0.60 (0.23–1.52)|
|AG/AA||0.86 (0.67–1.10)||0.87 (0.66–1.16)||0.80 (0.47–1.35)||0.87 (0.51–1.48)||0.85 (0.63–1.13)||1.05 (0.69–1.62)||0.81 (0.57–1.15)||0.62 (0.30–1.26)|
|1 insert.||154/130||0.74 (0.56–0.97)||0.74 (0.54–1.02)||0.71 (0.40–1.27)||0.53 (0.29–0.96)||0.81 (0.59–1.11)||0.69 (0.43–1.12)||0.65 (0.44–0.96)||1.42 (0.64–3.13)|
|2 insert||18/14||0.73 (0.35–1.54)||0.86 (0.40–1.87)||0.19 (0.02–1.86)||0.97 (0.43–2.20)||0.63 (0.19–2.07)||0.79 (0.27–2.29)||0.81 (0.08–8.19)|
|≥1 insert||0.74 (0.56–0.96)||0.76 (0.56–1.03)||0.67 (0.38–1.18)||0.49 (0.27–0.89)||0.82 (0.61–1.12)||0.68 (0.43–1.08)||0.66 (0.45–0.96)||1.36 (0.63–2.91)|
|GA||142/128||0.85 (0.64–1.13)||0.83 (0.60–1.16)||0.91 (0.51–1.61)||0.56 (0.31–1.02)||0.95 (0.69–1.32)||0.82 (0.51–1.32)||0.79 (0.53–1.18)||1.30 (0.56–3.01)|
|AA||18/13||0.71 (0.33–1.52)||0.83 (0.37–1.84)||0.24 (0.02–2.61)||0.91 (0.40–2.08)||0.83 (0.28–2.51)||0.58 (0.18–1.86)||0.81 (0.08–8.15)|
|GA/AA||0.84 (0.64–1.10)||0.83 (0.61–1.14)||0.85 (0.49–1.50)||0.54 (0.30–0.97)||0.95 (0.69–1.30)||0.82 (0.52–1.29)||0.77 (0.53–1.14)||1.25 (0.56–2.79)|
|GC||156/186||1.12 (0.86–1.47)||1.07 (0.79–1.45)||1.33 (0.76–2.32)||0.85 (0.48–1.52)||1.19 (0.88–1.61)||1.44 (0.91–2.26)||1.02 (0.71–1.47)||0.96 (0.47–1.94)|
|CC||52/39||0.77 (0.48–1.22)||0.66 (0.39–1.10)||1.46 (0.53–4.05)||0.59 (0.22–1.53)||0.84 (0.49–1.43)||0.59 (0.27–1.30)||0.68 (0.35–1.33)||4.04 (0.51–32.32)|
|GC/CC||1.04 (0.81–1.33)||0.97 (0.73–1.28)||1.35 (0.81–2.27)||0.78 (0.46–1.33)||1.11 (0.84–1.47)||1.16 (0.77–1.76)||0.95 (0.67–1.33)||1.15 (0.58–2.27)|
|AC||246/237||0.95 (0.74–1.23)||0.99 (0.74–1.31)||0.86 (0.52–1.45)||1.14 (0.67–1.94)||0.93 (0.70–1.25)||0.73 (0.48–1.11)||1.20 (0.84–1.72)||0.92 (0.46–1.83)|
|CC||50/63||1.20 (0.79–1.83)||1.32 (0.82–2.15)||0.91 (0.38–2.14)||1.36 (0.49–3.82)||1.17 (0.74–1.86)||1.76 (0.85–3.66)||1.06 (0.61–1.83)||0.95 (0.28–3.26)|
|AA/AC vs CC§||1.23 (0.82–1.84)||1.33 (0.84–2.12)||0.96 (0.42–2.21)||1.28 (0.47–3.48)||1.21 (0.77–1.89)||2.04 (1.00–4.13)||0.98 (0.58–1.66)||0.99 (0.30–3.26)|
|CT||216/219||0.99 (0.77–1.27)||0.89 (0.67–1.19)||1.34 (0.81–2.22)||0.71 (0.41–1.23)||1.08 (0.81–1.43)||1.00 (0.66–1.51)||1.08 (0.76–1.55)||1.02 (0.19–5.40)|
|TT||33/35||1.09 (0.65–1.83)||1.02 (0.57–1.83)||1.30 (0.43–3.94)||0.76 (0.25–2.32)||1.19 (0.66–2.15)||1.21 (0.52–2.82)||1.02 (0.50–2.08)||Ref.|
|CT/TT||1.00 (0.79–1.27)||0.91 (0.69–1.20)||1.33 (0.82–2.17)||0.72 (0.43–1.22)||1.09 (0.83–1.44)||1.02 (0.69–1.52)||1.07 (0.76–1.51)||0.73 (0.38–1.41)|
|CT||82/86||1.04 (0.74–1.46)||0.99 (0.67–1.46)||1.24 (0.62–2.50)||1.02 (0.51–2.04)||1.03 (0.70–1.53)||1.38 (0.79–2.42)||0.82 (0.51–1.32)||1.21 (0.47–3.10)|
|TT||3/3||1.33 (0.26–6.69)||1.35 (0.27–6.82)||0.99 (0.95–1.04)||0.91 (0.15–5.63)||1.61 (0.22–11.75)||1.10 (0.07–18.26)|
|CT/TT||1.05 (0.75–1.46)||1.00 (0.68–1.47)||1.06 (0.53–2.12)||1.03 (0.70–1.51)||1.39 (0.81–2.41)||0.82 (0.51–1.33)|
|CT||240/223||0.82 (0.64–1.05)||0.78 (0.58–1.03)||0.95 (0.57–1.58)||0.62 (0.36–1.09)||0.93 (0.70–1.24)||0.74 (0.48–1.12)||1.01 (0.71–1.44)||0.46 (0.23–0.93)|
|TT||32/39||1.21 (0.73–2.01)||1.24 (0.67–2.31)||1.13 (0.45–2.80)||1.36 (0.45–4.14)||1.21 (0.68–2.16)||0.61 (0.24–1.54)||1.78 (0.89–3.58)||1.95 (0.22–17.04)|
|CT/TT||0.87 (0.68–1.10)||0.82 (0.62–1.08)||0.98 (0.60–1.59)||0.70 (0.41–1.19)||0.97 (0.73–1.27)||0.72 (0.48–1.08)||1.10 (0.79–1.54)||0.52 (0.26–1.02)|
|TA||213/180||0.81 (0.63–1.04)||0.84 (0.63–1.12)||0.70 (0.41–1.19)||1.05 (0.59–1.87)||0.79 (0.59–1.06)||0.85 (0.56–1.29)||0.83 (0.58–1.19)||0.59 (0.29–1.19)|
|AA||11/20||1.59 (0.73–3.46)||1.68 (0.74–3.80)||0.99 (0.06–16.77)||3.06 (0.58–16.23)||1.22 (0.49–3.03)||1.79 (0.43–7.43)||1.56 (0.56–4.35)||1.15 (0.12–10.55)|
|TA/AA||0.85 (0.66–1.08)||0.89 (0.67–1.18)||0.71 (0.42–1.20)||1.16 (0.67–2.02)||0.81 (0.61–1.08)||0.88 (0.58–1.33)||0.85 (0.66–1.08)||0.89 (0.67–1.18)|
|GA||148/162||1.14 (0.87–1.50)||1.18 (0.86–1.62)||1.06 (0.62–1.84)||0.93 (0.51–1.70)||1.22 (0.90–1.66)||1.19 (0.76–1.87)||1.13 (0.77–1.64)||1.19 (0.54–2.62)|
|AA||13/22||1.78 (0.85–3.71)||1.59 (0.66–3.83)||2.24 (0.59–8.49)||2.71 (0.48–15.25)||1.55 (0.68–3.52)||2.34 (0.76–7.19)||1.19 (0.40–3.51)||2.60 (0.31–22.16)|
|GA/AA||1.19 (0.92–1.55)||1.21 (0.90–1.64)||1.16 (0.69–1.96)||1.03 (0.58–1.84)||1.25 (0.93–1.68)||1.28 (0.83–1.98)||1.19 (0.92–1.55)||1.21 (0.90–1.64)|
CART analysis is known for its ability to handle a large number of covariates and detect high-order gene-gene and gene-environment interactions. As shown in Figure 1, the decision tree identified smoking status as the top split, the most influential factor in bladder cancer risk. The final tree had 6 terminal nodes. The lowest-risk group was terminal node 1, comprising never smokers who were rare homozygotes for the CDK4 promoter SNP and common homozygotes for CCND1 P241P. As expected, the highest risk group was terminal node 6 with current smokers. Since the lowest-risk group, comprising never smokers who were rare homozygotes for the CDK4 promoter SNP and common homozygotes for CCND1 P241P, had only 10 subjects, we chose the second-lowest risk group consisting of never smokers who were AA/AC for CDK4 promoter as the reference. The ORs for terminal nodes 1 and 3 to 6 ranged from 0.46 to 6.30. We categorized the terminal nodes into 3 risk groups based on their ORs: low-risk group (OR < 1.50), medium-risk group (1.50 ≤ OR < 4.00), and high-risk group (OR ≥ 4.00). These 3 risk groups were associated with progressively elevated bladder cancer risk (P for trend < .001). We further tested the gene-gene interaction between CCND1 and CDK4 identified by CART using logistic regression. Interactions were significant for both overall analysis and stratified analysis for never smokers (P for interaction 0.007 and 0.009, respectively), while no significant interaction was observed among former and current smokers, Table 3.
|CCND1||CDK4||Controls||Cases||Overall||Never smoker||Former smoker||Current smoker|
|OR (95% CI)*||P||OR (95% CI)†||P||OR (95% CI)†||P||OR (95% CI)†||P|
|GG||CC||20||13||0.53 (0.25–1.12)||.095||0.41 (0.08–2.04)||.278||0.57 (0.22–1.46)||.243||0.40 (0.03–4.81)||.470|
|GA/AA||AA/AC||382||340||0.73 (0.56–0.95)||.021||0.86 (0.55–1.36)||.528||0.70 (0.48–1.02)||.060||0.55 (0.26–1.15)||.110|
|GA/AA||CC||30||50||1.31 (0.78–2.19)||.309||3.30 (1.30–8.39)||.012||0.89 (0.45–1.75)||.735||0.76 (0.18–3.21)||.707|
|Test for interaction||.007||.009||.166||.412|
The most significant finding in this study is that the use of a multifaceted analytic approach allowed identification of significant high-order gene-gene and gene-environment interactions which, compared with single-factor analyses, exhibited greater power to predict the modest risks conferred by genetic variants in the cell cycle control pathway to the overall risk of bladder cancer.
Only p53 intron 3 polymorphism was associated with an altered bladder cancer risk in the individual SNP analyses. The effect was also significant among former smokers and younger subjects (aged younger than 55 years). The rare allele (16-bp insertion) of this polymorphism has been associated with reduced steady-state p53 mRNA expression in Epstein-Barr virus-immortalized lymphoblastoid cell lines and increased risks of colorectal cancer and breast cancer.15, 16 Wang-Gohrke and Chang-Claude16 noted, however, that the increased breast cancer risk was evident only in women with a first-degree family history, indicating that the elevated breast cancer risk associated with this variant might be modulated by additional genetic elements. Consistent with this notion, Mahasneh and Abdel-Hafiz17 found that, in a Jordanian population, carriers of the 16-bp deletion of p53 intron 3 had an increase in breast cancer risk when they also carried the variant allele of the p53 intron 6 polymorphism, further suggesting the presence of other genetic components that regulate the cancer risk associated with the p53 intron 3 16-bp ins/del polymorphism. In the present study we found that the variant-containing genotypes of both p53 intron 3 and intron 6 conferred a reduced risk of bladder cancer among individuals younger than 55. Joint analysis of p53 intron 3 and intron 6 showed significantly reduced risk of bladder cancer for individuals younger than 55 who carry either variant allele from these 2 SNPs (OR = 0.51, 95% CI, 0.28–0.91) (data not shown).
A number of studies have investigated the associations between cell cycle gene polymorphisms and bladder cancer risk,18–24 with mixed results as detailed in Table 4. For instance, Wang et al.24 reported a significantly increased bladder cancer risk for carriers of the CCND1 P241P A/A genotype in a hospital-based case-control study, which could not be replicated by Cortessis et al.20 in a population-based study. Similar conflicting results have been reported for p53 R72P21–23 and STK15 F31V.25 Although these discrepant findings might result from distinctions in study design and population stratification, they might also be due to the finding that cell cycle regulation is a highly elaborate and complex process, with many relevant genes possibly playing conflicting roles even in the pathogenesis of the same malignancy, depending on the cellular or molecular milieu. Moreover, bladder cancer is a complex disease that is unlikely to be significantly influenced by a single genetic or environmental factor. Studies examining the associations between individual genetic variants and bladder cancer risk may provide limited information and sometimes conflicting outcomes resulting from a lack of adequate prediction power. These, taken together, support the employment of more comprehensive approaches to assess the interactions of multiple genes in the same carcinogenic pathway and their combined influence on bladder cancer predisposition.
|Author||Year||Study population||Cases/controls||SNP||Minor allele||MAF controls||Result OR (95% CI)|
|Biro||2000||Caucasians, Slovak||50/145||P53 intron 6||A1*||0.152||1.20 (0.61–2.33) for A2** allele|
|P53 R72P||Pro||0.324||1.44 (0.82–2.27) for Arg allele|
|Toruner||2001||Turkish||121/114||P53 R72P||Pro||0.39||No significant association|
|Soulitzis||2002||White Caucasian Greek||55/99||P53 R72P||Pro||0.435||2.67 (1.38–5.20) for Arg allele; 4.69 (2.13–10.41) for Arg/Arg|
|Wang||2002||Japanese||222/317||CCND1||A||0.43||1.76 (1.09–2.84) for A/A; 2.53 (1.28–4.51) for A/A+ nonsmoking case|
|Cortessis||2003||Non-Hispanic White||515/612||CCND1||A||0.47||1.00 (0.72–1.38) for A/G; 0.90 (0.60–1.33) for A/A|
|Mabrouk||2003||Tunisian||47/34||P53 R72P||Pro||0.34||No significant association|
|Kuroda||2003||Japanese||112/175||P53 R72P||Pro||0.42||2.28 (1.12–4.66) for Pro/Pro+smokers; 6.83 (0.63–73.49) for Pro/Pro+light smokers|
|Garcia-Closas||2007||Spanish||1086/1033||CCND1||NA||NA||No significant association|
|STK15 F31I||NA||NA||No significant association|
|STK15 I57V||NA||NA||No significant association|
For high-order gene-gene and gene-environment interactions in a case-control study, traditional logistic regression may not be applicable because of the sparseness of high-dimensional data. Contingency table cells with small numbers of observations will result in unreliable parameter estimates, ie, large standard errors.26 Alternative exploratory strategies such as CART were developed to address these situations. In the present study, CART analysis showed that current smokers had the highest percentage of bladder cancer cases (78%), while never smokers with the CDK4 homozygote variant genotype and CCND1 homozygote wildtype genotype exhibited the lowest risk. Never smokers with the CDK4 homozygote variant genotype and a CCND1 variant-containing genotype had the second-highest percentage of bladder cancer cases (67%). The variant allele of CCDN1 P241P encodes an alternatively spliced transcript which, compared with the transcript encoded by the wildtype allele, exhibits greater nuclear accumulation during G1-S transition, and thus higher oncogenic potential.27 Moreover, this SNP has been associated with increased risk of various malignancies such as colorectal cancer and esophageal cancer. The CART analysis showed consistently that, among never smokers with the homozygous variant genotype of CDK4 promoter SNP, carriers of the variant-containing genotypes of CCND1 P241P had a 7-fold excess in bladder cancer risk compared with CCND1 homozygous wildtype carriers, suggesting the presence of strong gene-gene and gene-environmental interactions. Furthermore, the CCND1-CDK4 interactions are biologically plausible since CCND1 is the most important cyclin in mediating the G1-S cell cycle transition, mainly through interaction with CDK4 and CDK6.26, 28–33 No investigation has reported the association of the CDK4 promoter SNP with disease etiology and, therefore, its functional relevance remains to be assessed. Further functional studies are warranted to determine whether the variant allele influences the level of CDK4 mRNA. Moreover, the CDK4-CCND1 interaction was identified only in never smokers, not in ever smokers, indicating the potential existence of higher-order gene-gene and gene-environmental interactions involved in the regulation of bladder cancer development. Alternatively, this difference may result from the strong smoking exposure, which may have overwhelmed the modest genetic effects caused by cell cycle sequence variations.34
The strength of this study rests on the employment of a nonparametric statistical approach to analyze the higher-order interactions between tobacco exposure and a panel of potentially important cell cycle gene variants. The interactions identified through the CART approach reflect the complex nature of bladder carcinogenesis. However, the CART method is a machine-learning approach that builds the model on the basis of the available dataset. These results should be interpreted with caution and validated by an independent dataset.
One limitation of this study is that this is a hospital-based case-control study in which the controls may be slightly different from the general population. In addition, the cases are from a large referral center and have a disproportionate number of patients with invasive bladder cancers, which could have a slightly different etiology than superficial bladder cancers.
In conclusion, our study results suggest that multiple common genetic variants in the cell cycle control pathway may modulate the risk of bladder cancer in conjunction with tobacco smoking. Consistent with the thinking that bladder cancer is a multifactoral and multistep disease, our results highlight the necessity of performing pathway-based polygenetic studies that can better reveal the genetic influences on cancer susceptibilities through consideration of higher-order interactions among multiple risk elements.