Purpose: Compelling evidence has indicated that inflammation plays an important role in cancer development. We sought to test the hypothesis that common sequence variants in the inflammation pathway modulate bladder cancer risk.

Experimental Design: We genotyped 59 potentially functional single nucleotide polymorphisms from 35 candidate inflammation genes in a case-control study including 635 Caucasian bladder cancer patients and 635 matched controls.

Results: The most significant finding was in the 3′-untranslated region of PTGS2 (exon10+837T>C, rs5275), which was associated with a significantly reduced risk (odds ratio, 0.68; 95% confidence interval, 0.54-0.87; P = 0.002) and remained significant after multiple comparison adjustment. Consistently, the most common PTGS2 haplotype containing the common allele of exon10+837T>C was associated with a significantly increased risk (odds ratio, 1.27; 95% confidence interval, 1.06-1.52; P = 0.008). In contrast, the haplotypes containing at least one variant allele of exon10+837T>C were all associated with a decreased risk. In a combined analysis to assess the cumulative effects of inflammation single nucleotide polymorphisms on bladder cancer risk, we found that in the anti-inflammation pathway, but not in the proinflammation pathway, when compared with individuals with a few adverse alleles, individuals with more adverse alleles had a significantly increased risk in a dose-dependent manner (Ptrend = 0.012). To further elucidate the functional mechanism of these associations, we redefined the adverse alleles based on literature-reported functional results and found that individuals with a higher number of inflammation-enhancing alleles in the anti-inflammation pathway exhibited a greater bladder cancer risk.

Conclusions: Our results strongly suggest that common variants in inflammation genes affect bladder cancer susceptibility individually and jointly.

Bladder cancer is a common malignancy with an estimated 67,160 new cases expected to be diagnosed in the United States in 2007 (1). Although the etiology remains enigmatic, it is generally considered a multifactorial disease involving both environmental and genetic factors (2). Although epidemiologic studies have revealed significant associations between bladder cancer and environmental factors such as tobacco smoking and occupational exposures (2), the genetic components have mostly remained elusive. Investigators have observed, however, that genetic polymorphisms in pathways controlling essential cellular activities, such as DNA damage/repair, cell cycle regulation, apoptosis, and carcinogen metabolism, may influence bladder cancer susceptibility (3, 4).

Inflammation is a crucial physiologic process initiated in humans by the immune system in response to pathogen invasion, tissue damage/repair, physical or chemical irritation, and wound healing (5). The process is triggered by the migration of various types of leukocytes to the inflammation sites, whereupon a wide spectrum of inflammatory cells is activated and recruited by a pleiotropic, interconnected network of cytokines, chemokines, and growth factors (6). Inflammation is normally self-limiting due to the production of anti-inflammatory cytokines that counterbalance the effect of proinflammatory cytokines. However, inflammation can persist and become chronic, which may cause many chronic diseases, such as cancer. It was reported that approximately one fifth of new cancers worldwide were caused by infection (7). Bladder cancer is one of the cancers that are closely associated with infection-induced inflammation. Schistosomiasis is a well-established risk factor in northern Africa (8). In the United States, a large epidemiologic study of 2,982 bladder cancer patients and 5,782 population controls showed that a history of urinary tract infection significantly increased the risk of bladder cancer, especially in individuals who had three or more infections (9).

Single nucleotide polymorphisms (SNP) in the inflammation genes have been shown to alter their expressions or functions and thus may be associated with an altered risk of multiple cancers. For instance, the variant allele of the −308G>A SNP in the promoter of tumor necrosis factor (TNF) gene has been associated with higher TNF production and an elevated risk of gastric cancer (10). However, most of the previous studies selected one or a few polymorphisms and reported modest risk associations. There have been few studies of bladder cancer (1113), including one of our own (14), which only evaluated a few SNPs in inflammation genes.

In this study, we used a pathway-based polygenic approach to assess the effects of 59 potential functional SNPs from 35 inflammation genes on bladder cancer susceptibility. In addition, we evaluated the haplotype effect and the collective effects of these polymorphisms. To our knowledge, this is one of the largest evaluations to date to investigate the associations of inflammation gene variants with bladder cancer risk.

Study population and epidemiologic data. The study population has been described elsewhere (3). Briefly, the population consisted of patients with newly diagnosed and histopathologically confirmed bladder cancer accrued at M. D. Anderson Cancer Center from year 1999 to 2006. The control subjects were selected from a large control pool recruited through the collaboration with the Kelsey-Seybold Clinic, the largest private multispecialty physician group consisting of 23 clinics and >300 physicians in the Houston metropolitan area. The potential controls were identified by reviewing short survey forms distributed to individuals coming to the clinic for annual health checkups or for addressing health concerns, but they are not inpatients. The potential control subjects were subsequently contacted by telephone to confirm their willingness to participate, and an appointment was scheduled at a Kelsey-Seybold clinic site convenient to the participant. On the day of the interview, the controls visited the clinic specifically for the purpose of participating in this study but not for any treatment purposes. At the designated meeting time and location, the control subject was greeted by the M. D. Anderson interviewer who conducted the interview and then escorted the subject to a Kelsey-Seybold phlebotomist for drawing blood samples. Controls had no cancer history (except nonmelanoma skin cancer) and were frequency matched to cases on age (±5 y), gender, and ethnicity. All subjects were interviewed based on a structured questionnaire. Each participant had a 40-mL blood sample drawn into a coded, heparinized tube, which was sent to the laboratory for immediate molecular analysis. Laboratory personnel were blinded to case-control status. All participants signed written informed consent forms, and human subject approval was obtained from both M. D. Anderson and Kelsey-Seybold institutional review boards. The response rates were 92% for cases and 75% for controls. We did an analysis on a subset of potential control subjects who refused to participate and who agreed to participate and found no difference between the two groups with regard to age, sex, ethnicity, and smoking status. Furthermore, in an analysis in which we used telephone numbers and zip codes as surrogates for residency, we found no difference as well. Hence, we are reasonably assured that our control population is not confounded by selection bias.

Polymorphism selection and genotyping. To compile a list of candidate genes in the inflammation pathway, we used “inflammation” and “inflammatory” as query terms to interrogate the Gene Ontology3

database and compiled a list of 995 genes. We then did an extensive literature review on the genes identified by the Gene Ontology database to scrutinize their cancer relevance and identified 580 genes. The resultant gene list was validated and adjusted through the application of the SNPs3D bioinformatic approach,4 a web-based literature-mining tool to select genes according to a set of user-defined query terms of human diseases or biological processes (15). Basically, most genes we identified through Gene Ontology database with cancer relevance were also included in the SNPs3D list. After this step, we identified 446 genes, including those overlapping genes between the two databases and those genes only showing up in SNPs3D results with a high priority score. We further used the HUGO name and/or the common aliases of each gene in the revised list and “polymorphism” to interrogate the PubMed database, manually scrutinized all identified results, and singled out those polymorphisms that have been reported to influence gene expression or function or to be associated with the etiology of inflammation-related diseases. We finally selected 59 potential functional SNPs in 35 important inflammation-related genes to be included in the present study. All variants in this study met at least two of the following three criteria: (a) minor allele frequency of at least 5% in Caucasians; (b) located in exons, promoters, or untranslated regions; and (c) reported to be associated with disease etiology or to influence host gene function or expression. All polymorphisms were genotyped using the SNPlex assay (16) according to the manufacturer's instructions (Applied Biosystems). Internal quality controls and negative controls were used to ensure genotyping accuracy, and 5% of all samples were randomly selected and genotyped in duplicate with 100% concordance.

Quantitative real-time PCR measuring PTGS2 mRNA expression. Lymphoblastoid cell lines were derived from 12 control subjects with known genotypes of the PTGS2 exon10+837T>C SNP. PTGS2 haplotypes and diplotypes were also inferred for these subjects. The cells were grown in RPMI 1640 supplemented with 10% fetal bovine serum and antibiotics. Logarithmic-phase cells were plated at 5 × 105/mL in 2 mL medium in 12-well plates. After 48 h, cells were harvested and total RNAs were extracted using Trizol reagent (Invitrogen). An aliquot of 1 μg total RNA for each sample was briefly exposed to DNase I (Invitrogen) and then reverse transcribed into cDNA using gene-specific 3′ primers of PTGS2 and β-actin and SuperScript III reverse transcriptase (Invitrogen). The PCR primers used were the following: PTGS2, 5′-GCTCAAACATGATGTTTGCATTC-3′ (forward) and 5′-GCTGGCCCTCGCTTATGA-3′ (reverse); β-actin, 5′-CGAGCGCGGCTACAGCTT-3′ (forward) and 5′-TCCTTAATGTCACGCACGATTT-3′ (reverse). All PCRs were done using ABI PRISM 7900HT Sequence Detection System (Applied Biosystems) with the SYBR Green PCR Core Reagent kit (Applied Biosystems). Each PCR was done in a final volume of 8 μL containing 1× SYBR PCR buffer, 3.5 mmol/L MgCl2, 62.5 μmol/L of each deoxynucleotide triphosphate, 0.2 μmol/L of each primer, 0.15 unit of AmpliTaq Gold DNA polymerase, and 4 μL of 1:10 dilution of each synthesized cDNA sample. The thermal cycling conditions comprised an initial denaturing step at 95°C for 10 min and 40 cycles at 95°C for 15 s and 60°C for 1 min. All reactions were run in duplicate and included no RNA, no reverse transcription, and no cDNA controls. The cycle number at which the reaction reached an arbitrarily determined threshold (CT) was recorded for both PTGS2 and β-actin, and the relative amount of PTGS2 mRNA to β-actin mRNA was described as 2CT, where CT = (CTPTGS2CTβ-actin; ref. 17). To simplify the data presentation, all relative gene expression results were multiplied by 105.

Statistical analysis. Due to the small number of minority participants, we limited all our analyses to Caucasians. Statistical analyses were done using STATA software (STATA Corp.). χ2 analysis and Fisher's exact test were used to assess the case-control differences in categorical variables. Student's t test was used to test for continuous variables. The Hardy-Weinberg equilibrium (HWE) was tested using a goodness-of-fit χ2 test. The bladder cancer risks were estimated as odds ratios (OR) and 95% confidence intervals (95% CI) using a multivariate logistic regression model adjusted for age, gender, smoking status, and pack-years, where appropriate. Definitions of smoking status were as described. Haplotypes and diplotypes were inferred using the expectation-maximization algorithm implemented in the HelixTree software (Golden Helix, Inc.). Haplotypes with a probability of <95% were excluded from the data analysis. The adjusted OR and 95% CI for each haplotype were assessed using multivariate logistic regression. We used a 1-degree-of-freedom haplotype-specific score statistic to compare each haplotype with all others combined. We also conducted the haplotype analysis using the combination of the common allele for each SNP as the reference group. For the pathway-based combined analysis, we defined the minor allele as the adverse allele except for those minor alleles associated with at least a borderline significant (P < 0.1) reduction in cancer risk. In this case, the wild-type allele was defined as the adverse allele. For those genes with multiple SNPs assayed in the pathway-based adverse allele analysis, only one SNP was included in the combined analyses because all tested SNPs in the same gene were in linkage disequilibrium (data not shown). For the genes with three or more SNPs, we chose the SNP showing the highest linkage disequilibrium with all the other polymorphisms. For the genes with two SNPs, the SNP representing the gene was randomly selected to be included in the combined analysis. The adverse alleles were collapsed together and categorized by tertiles (low, medium, and high risk) of the number of adverse alleles in controls. Because we hypothesized that inflammation-enhancing alleles increase bladder cancer risk, we also defined the adverse alleles according to the literature-reported functional results as those having a molecularly validated role of either enhancing the function or expression of proinflammation genes or repressing the function or expression of anti-inflammation genes. Those variants for which there were no published reports of their functional effects on inflammation were excluded from this analysis. All P values were two sided, and P ≤ 0.05 was considered the threshold of significance.

Multiple comparison and false-positive findings issues. Two approaches were used to address the potential false-positive findings: the false discovery rate (FDR) and the false-positive report probability (FPRP). The Benjamini-Hochberg method controls for the FDR, which is defined as the expected ratio of erroneous rejections of the null hypothesis to the total number of rejections (18). We set the global FDR value at 5% and 10% and adjusted the originally significant P values in the main analysis as well as in the stratified analyses of individual SNPs and PTGS2 haplotypes to determine if they remained significant at either level. The FPRP (19) is a Bayesian approach that assesses the posterior noteworthiness of statistically significant findings considering a priori information on specific polymorphisms/genes in the etiology of the investigated or relevant diseases. The prior probability of each SNP was determined based on the existing evidence about the association of the SNP with bladder cancer or other cancers, its biological plausibility, and the importance of the host gene in bladder cancer and relevant malignancies. As suggested by several recent cancer association studies (20, 21), to minimize the chance of false-positive findings due to a relaxed setting of prior probability, we set a stringent range of prior probability from 0.001 to 0.1. We calculated the FPRP for each SNP in the main analysis of 59 individual SNPs as well as in the stratified analyses of 5 significant SNPs by assigning a high probability score (0.1) if (a) the biological plausibility of the SNP and its gene was high and (b) strong epidemiologic evidence existed to support an association of the SNP with bladder cancer or other tumors. A score of 0.01 was assigned if only one of the two criteria was met, and a score of 0.001 was assigned if neither criterion was met. As proposed in the original publication (19) and in several reports (22, 23), we preferred to be conservative by using a FPRP rejection criterion of 0.2 as the threshold for noteworthy findings. Also following the suggestion in the original report, we used a dominant mode of inheritance and the statistical power to detect an OR of 1.5 (for alleles with an adverse effect) or 0.67 (for alleles with a protective effect).

Characteristics of the study population. The final study population consisted of 1,270 Caucasians, including 635 bladder cancer patients and 635 cancer-free controls. No significant case-control differences were identified with regard to age [cases versus controls (mean ± SD): 60.0 ± 11.1 years versus 62.9 ± 10.4 years; P = 0.058] and gender (P = 1.00). As expected, significant differences in smoking status existed between cases and controls (ever smokers: cases, 74.5%; controls, 55.1%; P < 0.001). Among the ever smokers, cases reported a significantly greater cigarette consumption than did controls, as assessed by the mean number of pack-years (cases versus controls: 45.6 ± 63.4 versus 29.8 ± 27.7; P < 0.001).

Main effects and stratified analyses of bladder cancer risk by individual polymorphisms. We classified the 59 SNPs into two categories: proinflammation (40 SNPs in 23 genes) and anti-inflammation (19 SNPs in 12 genes; Supplementary Table S1). The genotyping completion rate ranged from 95% to 99% for all SNPs except for iNOS 19514C>T (67%) and IL4R Glu400Ala (74%) and was similar between cases and controls. When the combined variant genotypes were compared with the homozygous wild-type genotype, the variant allele of three SNPs in proinflammation genes (MCP1 −2518A>G, PTGS2 exon10+837T>C, and PTGS2 −765G>C) was associated with a significantly altered bladder cancer risk, with ORs (95% CI) of 1.26 (1.00-1.59; P = 0.05), 0.68 (0.54-0.87; P = 0.002), and 0.76 (0.59-0.98; P = 0.032), respectively (Table 1). Of the anti-inflammation genes, IL4R Gln576Arg was associated with a significantly reduced risk (OR, 0.75; 95% CI, 0.59-0.96; P = 0.022), whereas IFNAR2 Phe10Val was associated with a significantly increased risk (OR, 1.28; 95% CI, 1.02-1.62; P = 0.035; Table 1). However, only the association of PTGS2 exon10+837T>C remained significant after the testing with both the FDR at the 5% level and the FPRP at a prior probability level of 0.1. Similar results on PTGS2 exon10+837T>C were observed when we tested the ordinal model. In stratified analyses, the reduced risk associated with PTGS2 exon10+837T>C remained significant in males, ever smokers, and heavy smokers after adjustment using both methods, with ORs of 0.71 (0.54-0.93; P = 0.013), 0.60 (0.45-0.81; P < 0.001), and 0.39 (0.22-0.72; P = 0.002; Supplementary Table S2). No other associations remained significant after FPRP adjustment.

Table 1.

Association of selected SNPs in inflammation pathway with bladder cancer risk

GenePolymorphism (rs no.; position)SeriesCommon homozygoteHeterozygoteRare homozygoteOR (95% CI)* per variant (Ptrend)Heterozygote + rare homozygoteMinor allele frequencyHWE P value in controls
MCP1 −2518A>G (rs1024611; promoter) Case 326 253 45   0.27 0.12 
  Control 352 229 51   0.26  
  OR 1.0 1.31 1.03 1.13 1.26   
  95% CI  1.03-1.67 0.66-1.60 0.95-1.36 (P = 0.17) 1.00-1.59 (P = 0.050)   
PTGS2 Exon10+837T>C (rs5275; 3′-untranslated region) Case 279 268 76   0.34 0.27 
  Control 236 312 85   0.38  
  OR 1.0 0.68 0.70 0.79 0.68   
  95% CI  0.53-0.87 0.48-1.01 0.66-0.93 (P = 0.006), 0.54-0.87 (P = 0.002),   
PTGS2 −765G>C (rs20417; promoter) Case 446 163 10   0.15 0.02 
  Control 416 200 11   0.18  
  OR 1.0 0.76 0.83 0.79 0.76   
  95% CI  0.58-0.98 0.33-2.05 0.62-0.99 (P = 0.043) 0.59-0.98 (P = 0.032)   
IL4R Gln576Arg (rs1801275; exon 11) Case 406 193 29   0.20 0.08 
  Control 374 229 22   0.22  
  OR 1.0 0.72 1.11 0.84 0.75   
  95% CI  0.56-0.93 0.61-2.01 0.68-1.03 (P = 0.087) 0.59-0.96 (P = 0.022)   
IFNAR2 Phe10Val (rs7279064;  Case 264 304 64   0.34 0.93 
  Control 300 273 60   0.31  
  OR 1.0 1.30 1.23 1.18 1.28   
 exon 2) 95% CI  1.02-1.65 0.82-1.84 0.98-1.40 (P = 0.075) 1.02-1.62 (P = 0.035)   
GenePolymorphism (rs no.; position)SeriesCommon homozygoteHeterozygoteRare homozygoteOR (95% CI)* per variant (Ptrend)Heterozygote + rare homozygoteMinor allele frequencyHWE P value in controls
MCP1 −2518A>G (rs1024611; promoter) Case 326 253 45   0.27 0.12 
  Control 352 229 51   0.26  
  OR 1.0 1.31 1.03 1.13 1.26   
  95% CI  1.03-1.67 0.66-1.60 0.95-1.36 (P = 0.17) 1.00-1.59 (P = 0.050)   
PTGS2 Exon10+837T>C (rs5275; 3′-untranslated region) Case 279 268 76   0.34 0.27 
  Control 236 312 85   0.38  
  OR 1.0 0.68 0.70 0.79 0.68   
  95% CI  0.53-0.87 0.48-1.01 0.66-0.93 (P = 0.006), 0.54-0.87 (P = 0.002),   
PTGS2 −765G>C (rs20417; promoter) Case 446 163 10   0.15 0.02 
  Control 416 200 11   0.18  
  OR 1.0 0.76 0.83 0.79 0.76   
  95% CI  0.58-0.98 0.33-2.05 0.62-0.99 (P = 0.043) 0.59-0.98 (P = 0.032)   
IL4R Gln576Arg (rs1801275; exon 11) Case 406 193 29   0.20 0.08 
  Control 374 229 22   0.22  
  OR 1.0 0.72 1.11 0.84 0.75   
  95% CI  0.56-0.93 0.61-2.01 0.68-1.03 (P = 0.087) 0.59-0.96 (P = 0.022)   
IFNAR2 Phe10Val (rs7279064;  Case 264 304 64   0.34 0.93 
  Control 300 273 60   0.31  
  OR 1.0 1.30 1.23 1.18 1.28   
 exon 2) 95% CI  1.02-1.65 0.82-1.84 0.98-1.40 (P = 0.075) 1.02-1.62 (P = 0.035)   
*

Adjusted for age, gender, and smoking status.

FPRP < 0.2, assuming a prior probability of 0.1 and estimated statistical power to detect an OR of 1.5.

Remain significant after FDR adjusted at 10% level.

Associations of PTGS2 haplotypes and diplotypes with bladder cancer risk. We further conducted haplotype analysis to identify potential interaction effect between the three PTGS2 SNPs. Table 2 summarizes the relative risks associated with common PTGS2 haplotypes. Because of the high degree of linkage disequilibrium between the three PTGS2 SNPs (D′ was 0.99 between exon10−90C>T and exon10+837T>C, 0.95 between exon10+837T>C and −765G>C, and 0.64 between exon10−90C>T and −765G>C), we only observed four haplotypes with an estimated frequency of >1%. We found that the most common H1 haplotype (WWW, in the order of −765G>C, exon10+837T>C, and exon10−90C>T), which has a wild-type allele at all three loci, was associated with a 1.27-fold (1.06-1.52; P = 0.008, significant after FDR adjustment at the 5% level) increase in risk, whereas the H3 haplotype (MMW), which has the variant allele of exon10+837T>C, was associated with a significantly reduced risk (OR, 0.78; 95% CI, 0.61-0.99; P = 0.045, significant after FDR adjustment at the 10% level). The risk associated with H1 was also more pronounced in ever smokers and heavy smokers after FDR adjustment at 10% level, with ORs of 1.36 (1.09-1.70; P = 0.007) and 1.75 (1.15-2.66; P = 0.008), respectively. We have also conducted the haplotype analysis using the combination of the common allele for each SNP as the reference group and the result was consistent with what we have reported in the original analyses (data not shown). We also conducted diplotype analyses and found that, when compared with the reference diplotype that contains wild-type allele at all three loci on both chromosomes, diplotypes containing at least one copy of the variant allele of PTGS2 exon10+837T>C were associated with a reduced bladder cancer risk (data not shown).

Table 2.

Association of PTGS2 haplotypes and bladder cancer risk stratified by host characteristics

VariablesH1 (WWW)*
H2 (WMW)
H3 (MMW)
H4 (MMM)
Case/controlOR (95% CI)
Case/controlOR (95% CI)
Case/controlOR (95% CI)
Case/controlOR (95% CI)
PPPP
Overall         
 813/779 1.27 (1.06-1.52) 227/229 0.87 (0.71-1.07) 152/190 0.78 (0.61-0.99) 18/22 0.78 (0.40-1.53) 
  P = 0.008  P = 0.20  P = 0.045§  P = 0.47 
Gender         
    Male 618/594 1.23 (1.01-1.49) 181/199 0.90 (0.72-1.14) 123/150 0.78 (0.59-1.03) 14/15 0.93 (0.43-2.02) 
  P = 0.041  P = 0.39  P = 0.075  P = 0.85 
    Female 195/185 1.48 (0.98-2.24) 46/50 0.74 (0.46-1.20) 29/40 0.79 (0.46-1.37) 4/7 0.46 (0.12-1.75) 
  P = 0.061  P = 0.22  P = 0.40  P = 0.26 
Smoking status         
    Never 213/349 1.13 (0.84-1.51) 62/100 1.08 (0.76-1.51) 42/88 0.77 (0.51-1.16) 4/11 0.62 (0.19-1.99) 
  P = 0.428  P = 0.68  P = 0.21  P = 0.42 
    Ever 600/430 1.36 (1.09-1.70) 165/149 0.77 (0.60-1.00) 110/102 0.79 (0.58-1.07) 14/11 0.88 (0.38-2.02) 
  P = 0.007§  P = 0.051  P = 0.13  P = 0.76 
Age         
    1st quartile 209/206 1.32 (0.93-1.88) 60/61 1.03 (0.69-1.55) 31/59 0.47 (0.28-0.79) 8/4 1.82 (0.50-6.62) 
  P = 0.12  P = 0.88  P = 0.004  P = 0.37 
    2nd quartile 169/201 1.15 (0.81-1.62) 57/65 1.02 (0.66-1.55) 37/53 0.85 (0.53-1.39) 1/9 0.13 (0.02-1.15) 
  P = 0.44  P = 0.94  P = 0.52  P = 0.067 
    3rd quartile 201/199 1.22 (0.86-1.73) 47/66 0.71 (0.47-1.05) 40/35 1.31 (0.77-2.22) 4/5 0.65 (0.16-2.68) 
  P = 0.27  P = 0.086  P = 0.32  P = 0.55 
    4th quartile 234/173 1.43 (0.97-2.09) 63/57 0.81 (0.52-1.24) 44/43 0.76 (0.46-1.24) 5/4 0.89 (0.22-3.59) 
  P = 0.070  P = 0.33  P = 0.27  P = 0.87 
Smoking level         
    1st quartile 65/118 1.23 (0.71-2.13) 17/36 0.91 (0.48-1.74) 11/31 0.67 (0.31-1.45) 2/5 0.80 (0.15-4.31) 
  P = 0.45  P = 0.78  P = 0.31  P = 0.79 
    2nd quartile 106/105 1.21 (0.75-1.95) 24/35 0.72 (0.40-1.30) 19/21 1.06 (0.52-2.16) 1/0 NA 
  P = 0.44  P = 0.28  P = 0.87   
    3rd quartile 195/113 1.26 (0.83-1.91) 57/36 0.88 (0.54-1.46) 30/26 0.68 (0.38-1.22) 5/1 2.46 (0.27-22.41) 
  P = 0.29  P = 0.63  P = 0.20  P = 0.42 
    4th quartile 237/94 1.75 (1.15-2.66) 67/42 0.61 (0.39-0.96) 51/24 0.83 (0.47-1.47) 6/5 0.56 (0.16-2.04) 
  P = 0.008§  P = 0.032  P = 0.53  P = 0.38 
VariablesH1 (WWW)*
H2 (WMW)
H3 (MMW)
H4 (MMM)
Case/controlOR (95% CI)
Case/controlOR (95% CI)
Case/controlOR (95% CI)
Case/controlOR (95% CI)
PPPP
Overall         
 813/779 1.27 (1.06-1.52) 227/229 0.87 (0.71-1.07) 152/190 0.78 (0.61-0.99) 18/22 0.78 (0.40-1.53) 
  P = 0.008  P = 0.20  P = 0.045§  P = 0.47 
Gender         
    Male 618/594 1.23 (1.01-1.49) 181/199 0.90 (0.72-1.14) 123/150 0.78 (0.59-1.03) 14/15 0.93 (0.43-2.02) 
  P = 0.041  P = 0.39  P = 0.075  P = 0.85 
    Female 195/185 1.48 (0.98-2.24) 46/50 0.74 (0.46-1.20) 29/40 0.79 (0.46-1.37) 4/7 0.46 (0.12-1.75) 
  P = 0.061  P = 0.22  P = 0.40  P = 0.26 
Smoking status         
    Never 213/349 1.13 (0.84-1.51) 62/100 1.08 (0.76-1.51) 42/88 0.77 (0.51-1.16) 4/11 0.62 (0.19-1.99) 
  P = 0.428  P = 0.68  P = 0.21  P = 0.42 
    Ever 600/430 1.36 (1.09-1.70) 165/149 0.77 (0.60-1.00) 110/102 0.79 (0.58-1.07) 14/11 0.88 (0.38-2.02) 
  P = 0.007§  P = 0.051  P = 0.13  P = 0.76 
Age         
    1st quartile 209/206 1.32 (0.93-1.88) 60/61 1.03 (0.69-1.55) 31/59 0.47 (0.28-0.79) 8/4 1.82 (0.50-6.62) 
  P = 0.12  P = 0.88  P = 0.004  P = 0.37 
    2nd quartile 169/201 1.15 (0.81-1.62) 57/65 1.02 (0.66-1.55) 37/53 0.85 (0.53-1.39) 1/9 0.13 (0.02-1.15) 
  P = 0.44  P = 0.94  P = 0.52  P = 0.067 
    3rd quartile 201/199 1.22 (0.86-1.73) 47/66 0.71 (0.47-1.05) 40/35 1.31 (0.77-2.22) 4/5 0.65 (0.16-2.68) 
  P = 0.27  P = 0.086  P = 0.32  P = 0.55 
    4th quartile 234/173 1.43 (0.97-2.09) 63/57 0.81 (0.52-1.24) 44/43 0.76 (0.46-1.24) 5/4 0.89 (0.22-3.59) 
  P = 0.070  P = 0.33  P = 0.27  P = 0.87 
Smoking level         
    1st quartile 65/118 1.23 (0.71-2.13) 17/36 0.91 (0.48-1.74) 11/31 0.67 (0.31-1.45) 2/5 0.80 (0.15-4.31) 
  P = 0.45  P = 0.78  P = 0.31  P = 0.79 
    2nd quartile 106/105 1.21 (0.75-1.95) 24/35 0.72 (0.40-1.30) 19/21 1.06 (0.52-2.16) 1/0 NA 
  P = 0.44  P = 0.28  P = 0.87   
    3rd quartile 195/113 1.26 (0.83-1.91) 57/36 0.88 (0.54-1.46) 30/26 0.68 (0.38-1.22) 5/1 2.46 (0.27-22.41) 
  P = 0.29  P = 0.63  P = 0.20  P = 0.42 
    4th quartile 237/94 1.75 (1.15-2.66) 67/42 0.61 (0.39-0.96) 51/24 0.83 (0.47-1.47) 6/5 0.56 (0.16-2.04) 
  P = 0.008§  P = 0.032  P = 0.53  P = 0.38 

NOTE: Order of SNPs comprising PTGS2 haplotype: −765G>C, exon10+837T>C, and exon10−90C>T.

Abbreviation: NA, not available.

*

W, wild-type allele; M, variant allele.

Adjusted for age, gender, smoking status, and pack-years for the overall analysis; adjusted for age, gender, and pack-years for the analysis stratified with smoking status; adjusted for age, gender, and smoking status for the analysis stratified with smoking level; adjusted for age, smoking status, and pack-years for the analysis stratified with gender; and adjusted for gender, smoking status, and pack-years for the analysis stratified with age.

Remained significant after FDR adjusted at 5% level.

§

Remained significant after FDR adjusted at 10% level.

Age was treated as a continuous variable and categorized based on the quartile distribution in controls.

Subjects were restricted to ever smokers. Smoking level was treated as a continuous variable and categorized based on the quartile distribution in controls.

Effects of PTGS2 genotypes on PTGS2 mRNA expression. To assess the functional relevance of the PTGS2 exon10+837T>C SNP, we measured the steady-state PTGS2 mRNA levels in lymphoblastoid cell lines derived from individuals with differential PTGS2 genotypes and diplotypes. The results showed that subjects with the variant genotypes of exon10+837T>C exhibited lower steady-state PTGS2 mRNA level than those with the homozygous wild-type [mean ± SE: 15.96 ± 2.82 (n = 8) versus 33.02 ± 14.66 (n = 4)]. However, due to the small number of available lymphoblastoid cell lines, the difference was not statistically significant (P = 0.140).

Combined effects of adverse alleles in proinflammation and anti-inflammation pathways. To enhance the predictive power, we collapsed the adverse alleles in the anti-inflammation or proinflammation pathway and determined their combined effects on bladder cancer risk. As shown in Table 3, in the anti-inflammation pathway, compared with individuals with less than seven adverse alleles, the risk was 1.31 (0.98-1.77; P = 0.070) for individuals with seven to eight adverse alleles and 1.61 (1.19-2.18; P = 0.002) for those with nine or more adverse alleles (Ptrend = 0.012). Each additional adverse allele was associated with a 1.07-fold increased risk. By contrast, analyses of the genes in the proinflammation pathway did not reveal such associations (Table 3). The practical application of this combined analytic approach was further validated when we restricted the analysis to the SNPs showing significant risk associations in the individual SNP analysis from Supplementary Table S1. We found a strong dose-response effect between bladder cancer risk and the number of adverse alleles, with ORs of 1.47 (1.07-2.03; P = 0.019) for the medium-risk group and 1.81 (1.29-2.54; P < 0.0001) for the high-risk group (Ptrend < 0.0001; Table 3).

Table 3.

Combined effects of adverse alleles on bladder cancer risk by inflammatory pathways and smoking status

Pathway and no. adverse alleles*Adverse alleles
Case/controlOR (95% CI)P
All SNPs in the anti-inflammatory pathway    
    <7 148/187 Reference  
    7-8 233/235 1.31 (0.98-1.77) 0.070 
    ≥9 230/190 1.61 (1.19-2.18) 0.002 
    Per allele 1.08   
    Ptrend 0.012   
All SNPs in the proinflammatory pathway    
    <14 255/252 Reference  
    14-16 211/209 1.04 (0.79-1.36) 0.79 
    ≥17 111/116 0.99 (0.72-1.38) 0.97 
    Per allele 1.00   
    Ptrend 0.89   
Selected SNPs showing significant effects in the individual SNP analysis§    
    <4 103/140 Reference  
    4-5 289/290 1.47 (1.07-2.03) 0.019 
    ≥6 220/188 1.81 (1.29-2.54) <0.0001 
    Per allele 1.16   
    Ptrend <0.0001   
Pathway and no. adverse alleles*Adverse alleles
Case/controlOR (95% CI)P
All SNPs in the anti-inflammatory pathway    
    <7 148/187 Reference  
    7-8 233/235 1.31 (0.98-1.77) 0.070 
    ≥9 230/190 1.61 (1.19-2.18) 0.002 
    Per allele 1.08   
    Ptrend 0.012   
All SNPs in the proinflammatory pathway    
    <14 255/252 Reference  
    14-16 211/209 1.04 (0.79-1.36) 0.79 
    ≥17 111/116 0.99 (0.72-1.38) 0.97 
    Per allele 1.00   
    Ptrend 0.89   
Selected SNPs showing significant effects in the individual SNP analysis§    
    <4 103/140 Reference  
    4-5 289/290 1.47 (1.07-2.03) 0.019 
    ≥6 220/188 1.81 (1.29-2.54) <0.0001 
    Per allele 1.16   
    Ptrend <0.0001   
*

Categorized based on the tertile distribution of the number of adverse alleles in controls.

If the variant allele exhibited an at least borderline significant protective effect (OR, <1; P < 0.1), the wild-type allele was considered the adverse allele.

Adjusted for age, gender, and smoking status.

§

Included IFNAR2 Phe10Val, MCP1 −2518A>G, IL4R Gln576Arg, PTGS2 exon10+837T>C, and eNOS 1251T>A. PTGS2 −765G>C and eNOS IVS1−898A>G were not included because of the high degree of linkage disequilibrium.

Combined effects of functionally validated inflammation-enhancing alleles on bladder cancer risk. To evaluate the effects of host inflammation status on bladder carcinogenesis, we used a similar approach to the “at-risk score” used by Matullo et al. (23) and redefined the adverse allele based on the results of functional assays reported previously in literature as those showing inflammation-augmenting effects either by enhancing the function or expression of proinflammation genes or by repressing the function or expression of anti-inflammation genes. When the analyses were conducted in the complete SNP panel in the anti-inflammation pathway, combined effects from eight SNPs with consistently reported functional relevance [IKB 2726C>T (24), PPARA Leu162Val (25, 26), PPARD exon4+15T>C (27), PPARG Pro12Ala (28, 29), IL4 exon1−168C>T (30, 31), IL4R Gln576Arg (32), IL10 −1082A>G (33), and IL13 Arg130Gln (34, 35)] showed a borderline significant gene-dose response (P = 0.075) for the medium-risk group (OR, 1.13; 95% CI, 0.81-1.56; P = 0.472) and the high-risk group (OR, 1.34; 95% CI, 1.02-1.77; P = 0.036). No significant association was identified by analyses of SNPs in the proinflammation pathway (data not shown). These findings were further validated when we restricted the analysis to the SNPs showing a significant effect in the individual SNP analysis. Three of these SNPs [MCP1 −2518A>G (36), PTGS2 −765G>C (37), and IL4R Gln576Arg (32)] were identified as having a functionally confirmed role in the inflammation response. Combined analysis using these three SNPs showed that, compared with the reference group with less than four inflammation-enhancing risk alleles, individuals with four risk alleles (OR, 1.09; 95% CI, 0.84-1.43; P = 0.514) and with more than four risk alleles (OR, 1.56; 95% CI, 1.15-2.12; P = 0.005) exhibited a significant gene-dosage effect (Ptrend = 0.004; Table 4).

Table 4.

Combined effects of inflammation-enhancing alleles on bladder cancer risk

Pathway and no. adverse alleles*Adverse alleles
Case/controlOR (95% CI)P
Analysis with functionally validated SNPs in the anti-inflammatory pathway (8 SNPs)§    
    <10 188/162 Reference  
    10 148/139 1.13 (0.81-1.56) 0.472 
    ≥11 279/311 1.34 (1.02-1.77) 0.036 
    Per allele 1.07   
    Ptrend 0.075   
Analysis with functional validated SNPs showing significant effects in the individual SNP analysis (3 SNPs)    
    <4 237/213 Reference  
    4 248/234 1.09 (0.84-1.43) 0.514 
    ≥5 134/163 1.56 (1.15-2.12) 0.004 
    Per allele 1.19   
    Ptrend 0.004   
Pathway and no. adverse alleles*Adverse alleles
Case/controlOR (95% CI)P
Analysis with functionally validated SNPs in the anti-inflammatory pathway (8 SNPs)§    
    <10 188/162 Reference  
    10 148/139 1.13 (0.81-1.56) 0.472 
    ≥11 279/311 1.34 (1.02-1.77) 0.036 
    Per allele 1.07   
    Ptrend 0.075   
Analysis with functional validated SNPs showing significant effects in the individual SNP analysis (3 SNPs)    
    <4 237/213 Reference  
    4 248/234 1.09 (0.84-1.43) 0.514 
    ≥5 134/163 1.56 (1.15-2.12) 0.004 
    Per allele 1.19   
    Ptrend 0.004   
*

Categorized based on the tertile distribution of the number of adverse alleles in controls.

The alleles with a literature-reported role of enhancing the function or expression of proinflammatory genes or repressing the function or expression of anti-inflammatory genes were defined as the adverse alleles. Those variants without functional annotations in published reports were excluded from the analyses.

Adjusted for age, gender, and smoking status.

§

Included IKB 2726C>T, PPARD −420T>C, PPARA Leu162Val, IL4 338C>T, IL13 Arg130Gln, IL10 −1082A>G, PPARG Pro12Ala, and IL4R Gln576Arg. The wild-type alleles of the underlined SNPs were inflammation-enhancing alleles.

Included MCP1 −2518A>G, PTGS2 −765G>C, and IL4R Gln576Arg. The wild-type alleles of the underlined SNPs were inflammation-enhancing alleles.

Three major findings were made in this study: (a) a functional polymorphism of the PTGS2 gene, as well as haplotypes and diplotypes of the gene, was significantly associated with an altered bladder cancer risk in a smoking-dependent manner; (b) the combined effect of multiple variants on bladder cancer risk was only observed in genes in the anti-inflammation pathway, but not in the proinflammation pathway; and (c) the combined analyses based on the previously reported functional effects of the inflammation polymorphisms on the polarization of inflammatory responses revealed a positive correlation between enhanced host inflammation status and increased bladder cancer risk.

Five SNPs (TNF −857T>C, IL16 Asn446Lys, eNOS IVS1−898A>G, eNOS 1251T>A, and PTGS2 −765G>C) significantly departed from HWE. These deviations might not occur by chance because the possibility of a systematic genotyping error was ruled out by our use of strict quality controls and internal controls in the genotyping process. In addition, duplicate results were completely concordant. Moreover, only TNF −857T>C was still out of HWE when the Bonferroni-corrected significance threshold (0.05/59) was applied to test the null hypothesis of HWE according to a conservative measure of the type I error in studies with a large number of testings (data not shown; ref. 38), which was also consistent with the results obtained from the less conservative Benjamini-Hochberg step-up procedure used to control for FDR of HWE at the 5% level (data not shown).

Two PTGS2 SNPs were associated with a significantly reduced bladder cancer risk in the current study. PTGS2 is an essential enzyme in the biogenesis of inflammation-promoting prostaglandins. The overexpression of PTGS2 has been associated with advanced-stage disease, high-grade histology, and poor prognosis in patients with bladder cancer (39). The variant allele of the promoter SNP (−765G>C) displayed a protective effect against bladder cancer development. Consistently, this allele has been shown to lead to reduced PTGS2 expression through interference with the binding of transcription factors (37) and has been associated with a decreased risk of various inflammation-related diseases (40, 41). The exon10+837T>C SNP in the 3′-untranslated region of PTGS2 was shown in our study to be associated with a significantly reduced bladder cancer risk after adjustment through both FDR and FRFP, suggesting that the association is particularly robust. In accordance, the exon10+837T>C SNP was also related to a reduction in lung cancer risk in Korean and Chinese populations (42, 43). However, it has also been reported to increase the susceptibility to breast cancer in Austrians (44). Despite study design differences and population stratifications, the pivotal and complicated role of PTGS2 in tumorigenesis might partially explain the apparent discrepancies in these findings. Our functional analyses measuring PTGS2 mRNA level in lymphoblastoid cell lines suggested that the decreased risk associated with the exon10+837T>C variant-containing genotypes may be attributed to lower PTGS2 expression. Previous studies have reported that in the proximal upstream region of this SNP there is a conserved AU-rich sequence element, which mediates posttranscriptional degradation of PTGS2 mRNA (45). However, whether PTGS2 exon10+837T>C confers the altered bladder cancer risk through modulating the AU-rich sequence element–mediated mRNA decay remains to be determined. Further functional determination using luciferase reporter assay should be used to confirm our findings.

Although most of these significant associations are biologically plausible, there is the possibility that they were attained by chance, given the loss of robustness after FDR and FPRP adjustments for four of the five SNPs showing overall significance (Table 1). Moreover, considering the departure from the HWE (PTGS2 −765G<C), the lack of mechanistic clues (IFNAR2 Phe10Val), and the inconsistent trend in the risk association for heterozygotes (OR, 0.72) and rare homozygotes (OR, 1.11; IL4R Gln576Arg), the risk-conferring effects of these polymorphisms, if they exist, might be only modest or minimal. To more powerfully elucidate the influence of inflammation polymorphisms on bladder cancer risk, we used a pathway-based polygenic strategy in which we divided all SNPs into proinflammation and anti-inflammation pathways and collapsed all the adverse alleles in each pathway to assess their combined effects on bladder cancer susceptibility. This pathway-based combined analysis of a panel of polymorphisms that act in the same functional pathway may amplify the effects of single variants and therefore enhance the predictive power. Moreover, many independent studies have shown that genetic variations in different genes of modest effect on disease risk within the same biological or functional pathway may exhibit an aggregate influence, leading to different levels of disease severity or quantitative variation of target traits (46, 47). Furthermore, this approach has also been extensively used in multiple cancer association studies to identify high-prevalence but low-penetrance polymorphisms and is effective in disclosing modest cancer risk that is difficult to detect by single variant analysis (3, 48). Our results from this analysis supported the advantage of this method by showing that the combined adverse alleles in the anti-inflammation pathway, but not in the proinflammation pathway, conferred a significantly increased bladder cancer risk.

Although numerous studies have implicated inflammation in cancer development, none has compared the specific effects of proinflammation and anti-inflammation genes. Our study is therefore the first to show that genetic variants in anti-inflammation genes might play a more prominent role in bladder cancer development than those in proinflammation genes. This finding fits in with the postulated concept of chronic inflammation-mediated tumorigenesis (6, 49). In particular, the proinflammation pathway is mainly involved in acute inflammation, which self-resolves through the activation of a large variety of anti-inflammation genes. However, under some rare circumstances, for example, when the foreign invader cannot be eliminated or the wound does not heal over a prolonged period, the acute reaction may transform into chronic inflammation through the inhibition of anti-inflammation effects (6, 49). Consistent with this notion, our study showed that the aggregation of multiple adverse alleles in anti-inflammation genes, but not proinflammation genes, was associated with a significantly enhanced bladder cancer risk. Additional support for this observation comes from the study of Lan et al. (50), who observed in a case-control study examining 20 inflammation genes involved in both Th1 (proinflammation) and Th2 (anti-inflammation) pathways that only Th2 gene SNPs seemed to be associated with a risk of non–Hodgkin's lymphoma. Another finding in our study of particular note was that the risk of bladder cancer rose with an increase in the number of inflammation-enhancing alleles. We used a similar approach to the at-risk score used by Matullo et al. (23) and redefined the adverse alleles based on literature-reported functional data showing consistent effects of these SNPs on the functions of their host genes. When this analysis was confined to anti-inflammation variants, strong dose-response trends were noted, with individuals with the most inflammation-enhancing alleles showing the highest risk. The absence of such effects in the proinflammation pathway (data not shown) further supported our belief that anti-inflammation genes are more important in bladder tumorigenesis.

Our study has several strengths. It is one of the largest case-control studies of the effects of inflammation polymorphisms on bladder cancer predisposition yet done. Although we could not rule out the possibility of chance finding given the large number of SNPs and genes evaluated, we used stringent criteria to identify potential false-positive findings resulting from multiple comparisons. Through the use of a pathway-based polygenic strategy and differential definition of adverse alleles, we were able to show that multiple variants in the anti-inflammation pathway had a combined effect on bladder tumorigenesis. Moreover, we are the first, from the standpoint of molecular epidemiology, to show a positive correlation between bladder cancer risk and an augmented inflammation status through our use of the pathway-based approach with a novel definition of the risk allele based on the results of previous functional studies. However, it should be noted that this is a hypothesis-driven candidate gene study. The sample size and multiple comparison issue would not allow us to do a comprehensive genotyping of all inflammation-related genes. If sample size permits, an unbiased selection of all inflammation-related genes and haplotype-tagging SNP-based genotyping would be desired to provide more complete understanding of the associations between inflammation and bladder carcinogenesis. In addition, this is a hospital-based case control study and inflammation-related processes among controls may bias results. To address this concern, we have collected the information on a panel of inflammation-related urinary tract diseases in both cases and controls. Except for kidney infection, which was inversely associated with bladder cancer risk, there were no significant differences in the distribution of other inflammation-related conditions (including cystitis, prostate infection, bladder stone, and renal stone) between cases and controls in our population. Adjustment of kidney infection in the multivariate logistic regression model did not result in significant changes to our original reports (data not shown).

In conclusion, we found significant individual and joint effects between inflammation polymorphisms and bladder cancer risk. Our study singled out PTGS2 as a promising bladder cancer susceptibility gene and supported a role for common anti-inflammation genetic polymorphisms in modulating bladder cancer predisposition.

Grant support: NIH grants CA74880, CA91846, and CA130063.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).

H. Yang and J. Gu contributed equally to this work.

1
Jemal A, Siegel R, Ward E, et al. Cancer statistics, 2007.
CA Cancer J Clin
2007
;
57
:
43
–66.
2
Burin GJ, Gibb HJ, Hill RN. Human bladder cancer: evidence for a potential irritation-induced mechanism.
Food Chem Toxicol
1995
;
33
:
785
–95.
3
Wu X, Gu J, Grossman HB, et al. Bladder cancer predisposition: a multigenic approach to DNA-repair and cell-cycle-control genes.
Am J Hum Genet
2006
;
78
:
464
–79.
4
Garcia-Closas M, Malats N, Silverman D, et al. NAT2 slow acetylation, GSTM1 null genotype, and risk of bladder cancer: results from the Spanish Bladder Cancer Study and meta-analyses.
Lancet
2005
;
366
:
649
–59.
5
Whiteside TL. The role of immune cells in the tumor microenvironment.
Cancer Treat Res
2006
;
130
:
103
–24.
6
Lu H, Ouyang W, Huang C. Inflammation, a key event in cancer development.
Mol Cancer Res
2006
;
4
:
221
–33.
7
American Cancer Society. Cancer facts and figures 2005. Atlanta (GA): American Cancer Society; 2005.
8
Bedwani R, Renganathan E, El Kwhsky F, et al. Schistosomiasis and the risk of bladder cancer in Alexandria, Egypt.
Br J Cancer
1998
;
77
:
1186
–9.
9
Kantor AF, Hartge P, Hoover RN, et al. Urinary tract infection and risk of bladder cancer.
Am J Epidemiol
1984
;
119
:
510
–5.
10
Machado JC, Figueiredo C, Canedo P, et al. A proinflammatory genetic profile increases the risk for chronic atrophic gastritis and gastric carcinoma.
Gastroenterology
2003
;
125
:
364
–71.
11
Tsai FJ, Chang CH, Chen CC, et al. Interleukin-4 gene intron-3 polymorphism is associated with transitional cell carcinoma of the urinary bladder.
BJU Int
2005
;
95
:
432
–5.
12
Bid HK, Manchanda PK, Mittal RD. Association of interleukin-1Ra gene polymorphism in patients with bladder cancer: case control study from North India.
Urology
2006
;
67
:
1099
–104.
13
Kang S, Kim YB, Kim MH, et al. Polymorphism in the nuclear factor κ-B binding promoter region of cyclooxygenase-2 is associated with an increased risk of bladder cancer.
Cancer Lett
2005
;
217
:
11
–6.
14
Leibovici D, Grossman HB, Dinney CP, et al. Polymorphisms in inflammation genes and bladder cancer: from initiation to recurrence, progression, and survival.
J Clin Oncol
2005
;
23
:
5746
–56.
15
Yue P, Melamud E, Moult J. SNPs3D: candidate gene and SNP selection for association studies.
BMC Bioinformatics
2006
;
7
:
166
.
16
Tobler AR, Short S, Andersen MR, et al. The SNPlex genotyping system: a flexible and scalable platform for SNP genotyping.
J Biomol Tech
2005
;
16
:
398
–406.
17
Schmittgen TD, Jiang J, Liu Q, et al. A high-throughput method to monitor the expression of microRNA precursors.
Nucleic Acids Res
2004
;
32
:
e43
.
18
Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing.
J R Stat Soc B
1995
;
57
:
289
–300.
19
Wacholder S, Chanock S, Garcia-Closas M, et al. Assessing the probability that a positive report is false: an approach for molecular epidemiology studies.
J Natl Cancer Inst
2004
;
96
:
434
–42.
20
Hill DA, Wang SS, Cerhan JR, et al. Risk of non-Hodgkin lymphoma (NHL) in relation to germline variation in DNA repair and related genes.
Blood
2006
;
108
:
3161
–7.
21
Rothman N, Skibola CF, Wang SS, et al. Genetic variation in TNF and IL10 and risk of non-Hodgkin lymphoma: a report from the InterLymph Consortium.
Lancet Oncol
2006
;
7
:
27
–38.
22
Hung RJ, Brennan P, Canzian F, et al. Large-scale investigation of base excision repair genetic polymorphisms and lung cancer risk in a multicenter study.
J Natl Cancer Inst
2005
;
97
:
567
–76.
23
Matullo G, Guarrera S, Sacerdote C, et al. Polymorphisms/haplotypes in DNA repair genes and smoking: a bladder cancer case-control study.
Cancer Epidemiol Biomarkers Prev
2005
;
14
:
2569
–78.
24
Romzova M, Hohenadel D, Kolostova K, et al. NFκB and its inhibitor IκB in relation to type 2 diabetes and its microvascular and atherosclerotic complications.
Hum Immunol
2006
;
67
:
706
–13.
25
Vohl MC, Lepage P, Gaudet D, et al. Molecular scanning of the human PPARα gene: association of the L162v mutation with hyperapobetalipoproteinemia.
J Lipid Res
2000
;
41
:
945
–52.
26
Flavell DM, Pineda Torra I, Jamshidi Y, et al. Variation in the PPARα gene is associated with altered function in vitro and plasma lipid concentrations in type II diabetic subjects.
Diabetologia
2000
;
43
:
673
–80.
27
Skogsberg J, Kannisto K, Cassel TN, et al. Evidence that peroxisome proliferator-activated receptor δ influences cholesterol metabolism in men.
Arterioscler Thromb Vasc Biol
2003
;
23
:
637
–43.
28
Deeb SS, Fajas L, Nemoto M, et al. A Pro12Ala substitution in PPARγ2 associated with decreased receptor activity, lower body mass index and improved insulin sensitivity.
Nat Genet
1998
;
20
:
284
–7.
29
Yen CJ, Beamer BA, Negri C, et al. Molecular scanning of the human peroxisome proliferator activated receptor γ (hPPARγ) gene in diabetic Caucasians: identification of a Pro12Ala PPARγ2 missense mutation.
Biochem Biophys Res Commun
1997
;
241
:
270
–4.
30
Kawashima T, Noguchi E, Arinami T, et al. Linkage and association of an interleukin 4 gene polymorphism with atopic dermatitis in Japanese families.
J Med Genet
1998
;
35
:
502
–4.
31
Suppiah V, Goris A, Alloza I, et al. Polymorphisms in the interleukin-4 and IL-4 receptor genes and multiple sclerosis: a study in Spanish-Basque, Northern Irish and Belgian populations.
Int J Immunogenet
2005
;
32
:
383
–8.
32
Hershey GK, Friedrich MF, Esswein LA, et al. The association of atopy with a gain-of-function mutation in the α subunit of the interleukin-4 receptor.
N Engl J Med
1997
;
337
:
1720
–5.
33
Suarez A, Castro P, Alonso R, et al. Interindividual variations in constitutive interleukin-10 messenger RNA and protein levels and their association with genetic polymorphisms.
Transplantation
2003
;
75
:
711
–7.
34
Hoffjan S, Nicolae D, Ober C. Association studies for asthma and atopic diseases: a comprehensive review of the literature.
Respir Res
2003
;
4
:
14
.
35
Vladich FD, Brazille SM, Stern D, et al. IL-13 R130Q, a common variant associated with allergy and asthma, enhances effector mechanisms essential for human allergic inflammation.
J Clin Invest
2005
;
115
:
747
–54.
36
Rovin BH, Lu L, Saxena R. A novel polymorphism in the MCP-1 gene regulatory region that influences MCP-1 expression.
Biochem Biophys Res Commun
1999
;
259
:
344
–8.
37
Papafili A, Hill MR, Brull DJ, et al. Common promoter variant in cyclooxygenase-2 represses gene expression: evidence of role in acute-phase inflammatory response.
Arterioscler Thromb Vasc Biol
2002
;
22
:
1631
–6.
38
Chen YC, Giovannucci E, Lazarus R, et al. Sequence variants of Toll-like receptor 4 and susceptibility to prostate cancer.
Cancer Res
2005
;
65
:
11771
–8.
39
Wild PJ, Kunz-Schughart LA, Stoehr R, et al. High-throughput tissue microarray analysis of COX2 expression in urinary bladder cancer.
Int J Oncol
2005
;
27
:
385
–91.
40
Abdullah L, Ait-Ghezala G, Crawford F, et al. The cyclooxygenase 2 −765 C promoter allele is a protective factor for Alzheimer's disease.
Neurosci Lett
2006
;
395
:
240
–3.
41
Ulrich CM, Whitton J, Yu JH, et al. PTGS2 (COX-2) −765G > C promoter variant reduces risk of colorectal adenoma among nonusers of nonsteroidal anti-inflammatory drugs.
Cancer Epidemiol Biomarkers Prev
2005
;
14
:
616
–9.
42
Hu Z, Miao X, Ma H, et al. A common polymorphism in the 3′UTR of cyclooxygenase 2/prostaglandin synthase 2 gene and risk of lung cancer in a Chinese population.
Lung Cancer
2005
;
48
:
11
–7.
43
Park JM, Choi JE, Chae MH, et al. Relationship between cyclooxygenase 8473T>C polymorphism and the risk of lung cancer: a case-control study.
BMC Cancer
2006
;
6
:
70
.
44
Langsenlehner U, Yazdani-Biuki B, Eder T, et al. The cyclooxygenase-2 (PTGS2) 8473T>C polymorphism is associated with breast cancer risk.
Clin Cancer Res
2006
;
12
:
1392
–4.
45
Dixon DA, Kaplan CD, McIntyre TM, et al. Post-transcriptional control of cyclooxygenase-2 gene expression. The role of the 3′-untranslated region.
J Biol Chem
2000
;
275
:
11750
–7.
46
Mohrenweiser HW, Wilson DM III, Jones IM. Challenges and complexities in estimating both the functional impact and the disease risk associated with the extensive genetic variation in human DNA repair genes.
Mutat Res
2003
;
526
:
93
–125.
47
Nelson MR, Kardia SL, Ferrell RE, et al. A combinatorial partitioning method to identify multilocus genotypic partitions that predict quantitative trait variation.
Genome Res
2001
;
11
:
458
–70.
48
Millikan RC, Hummer A, Begg C, et al. Polymorphisms in nucleotide excision repair genes and risk of multiple primary melanoma: the Genes Environment and Melanoma Study.
Carcinogenesis
2006
;
27
:
610
–8.
49
Coussens LM, Werb Z. Inflammation and cancer.
Nature
2002
;
420
:
860
–7.
50
Lan Q, Zheng T, Rothman N, et al. Cytokine polymorphisms in the Th1/Th2 pathway and susceptibility to non-Hodgkin lymphoma.
Blood
2006
;
107
:
4101
–8.

Supplementary data