A Study of Prostaglandin Pathway Genes and Interactions with Current Nonsteroidal Anti-Inﬂammatory Drug Use in Colorectal Adenoma

Published OnlineFirst May 2, 2012; DOI: 10.1158/1940-6207.CAPR-11-0459

Cancer Prevention Research Article Research

A Study of Prostaglandin Pathway Genes and Interactions with Current Nonsteroidal Anti-inﬂammatory Drug Use in Colorectal Adenoma

Todd L. Edwards1,2, Martha J. Shrubsole1,3,5, Qiuyin Cai1,3, Guoliang Li1, Qi Dai1,5, Douglas K. Rex7, Thomas M. Ulbright8, Zhenming Fu1, Harvey J. Murff1,3,5, Walter Smalley4,5,6, Reid Ness4, and Wei Zheng1,3,5

Abstract Colorectal cancer (CRC) is the second leading cause of cancer-related death and usually arises from colorectal polyps. Screening and removal of polyps reduce mortality from CRC. Colorectal polyps are known to aggregate in families; however the genetic determinants for risk of polyps are unknown. In addition, it has been shown that nonsteroidal anti-inflammatory drug (NSAID) use decreases the risk of CRC and the incidence and size of polyps. In this study, we used data from the Tennessee Colorectal Polyp Study and the Tennessee–Indiana Adenoma Recurrence Study to evaluate selected genes from the prostaglandin (PG) metabolism and signaling pathways for association with risk of polyps and for interactions with NSAIDs. Our design consisted of discovery and replication phases for a total of 2,551 Caucasian polyp cases and 3,285 Caucasian controls. We carried out multivariable logistic regression to test for association in both the discovery and replication phase and further examined the results with meta-analysis. We detected association signals in the genes PGE receptor 3 (PTGER3) and 15-hydroxyprostaglandin dehydrogenase (HPGD), both strong biologic candidates for influence on polyp risk. We did not observe the previously reported effects and effect modification in PG–endoperoxide synthase 2 (PTGS2), PGE receptor 2 (PTGER2), or PGE receptor 4 (PTGER4), although we did observe a single nucleotide polymorphism in PTGER2 associated with risk of multiple adenomas. We also observed effect modification of the HPGD signal by NSAID exposure. Cancer Prev Res; 1–9. 2012 AACR.

Introduction the progressive accumulation of genetic and epigenetic Colorectal cancer (CRC) is the second most common alterations such as somatic mutations through gain-of- cause of cancer death in North America and the fourth most function, loss-of-function, and subsequent genomic diagnosed cancer (1). The vast majority of CRCs are derived instability. from neoplastic colorectal polyps (2, 3), and colorectal The COX enzymes are usually expressed in response to adenomas are commonly recognized precursors to CRC inflammation and by cancerous and precancerous tissues; (2). CRC risk has been shown to be modulated by envi- nonsteroidal anti-inflammatory drugs (NSAID) inhibit the ronmental and genetic factors, in addition to epigenetic formation of prostanoids by antagonizing the activity of phenomena that associate with tumors. In CRC pathways, COX enzymes (4). It has long been recognized that COX normal colonic epithelium is transformed as the result of upregulation increases the metastatic potential of cancer cells (5). Overexpression of COX-2 occurs in 50% of colon adenomas and 85% of colon cancers and is considered a Authors' Affiliations: 1Division of Epidemiology, Department of Medicine; key and early oncogenic event in colorectal carcinogenesis 2Center for Human Genetics Research, Department of Molecular Physiol- (6). In animal studies, knockouts of prostaglandin (PG) 3 ogy and Biophysics, Vanderbilt University; Vanderbilt-Ingram Cancer receptors have lowered rates of adenomatous polyps and Center; 4Divison of Gastroenterology, Department of Medicine, Vanderbilt University School of Medicine; 5VA-TVHS GRECC; 6Section of Gastroen- cancer (7). terology, Tennessee Valley Veterans Affairs Medical Center and Vanderbilt Inhibition of the PG pathways also have been shown to University, Nashville, Tennessee; 7Department of Medicine, Division of Gastroenterology; and 8Department of Pathology, Indiana University reduce tumor cell proliferation, increase apoptosis, and School of Medicine, Clarian Pathology Laboratory, Indianapolis, Indiana reduce angiogenesis. Previous studies have shown that Note: Supplementary data for this article are available at Cancer Prevention people and animals taking NSAIDs experience lower rates Research Online (http://cancerprevres.aacrjournals.org/). of precancerous growths, cancers, and cancer-related deaths Corresponding Author: Todd L. Edwards, Vanderbilt University Medical (8, 9). Clinical trials evaluating the use of NSAIDs and Center, 2525 West End Ave., Suite 600, Nashville, TN 37203. Phone: 615- selective COX-2 inhibitors for colorectal adenoma preven- 322-3652; Fax: 615-322-8291; E-mail: [email protected] tion yielded some promising results, in which persons doi: 10.1158/1940-6207.CAPR-11-0459 taking the drugs had as much as 45% fewer adenomas than 2012 American Association for Cancer Research. the placebo group (10–15). This reduction in risk of

www.aacrjournals.org OF1

Edwards et al.

adenoma has been consistently observed for regular users of uary, 1996 and December, 2002 at the Vanderbilt Gastro- NSAIDs (16–19) and is more pronounced for larger ade- enterology Clinic, Veterans’ Affairs Tennessee Valley Health noma (16, 20). Interactions between NSAIDs and genetic System Nashville campus, Indiana University Hospital, the variation in COX genes have also been associated with Richard L. Roudebush Veterans Administration Medical protection from cancers and adenomas (21–23). Center, and Wishard Memorial Hospital. Excluded from In this study, we investigated whether interindividual TIARS were patients who could not speak or understand genetic variation within candidate genes in the PG pathway English, had genetic CRC syndromes (e.g., hereditary non- is a determinant of risk for adenoma formation in partici- polyposis CRC or familial adenomatous polyposis), were pants from the Tennessee Colorectal Polyp Study (TCPS) participating in an intervention trial to prevent adenoma and the Tennessee–Indiana Adenoma Recurrence Study recurrence, had a prior history of colon resection, inflam- (TIARS). Given the evidence that PG genes have a biologic matory bowel disease, adenomas, or any cancer other than role in adenoma incidence, it stands to reason that these nonmelanoma skin cancers, or were a current resident in a genes may harbor alleles that influence the fate of colorectal correctional facility. Overall, 1,643 eligible individuals were epithelial cells through a mechanism related to the activity identified. Potential participants who were not known to be of NSAIDs. We also evaluated single nucleotide polymorph- deceased were contacted first by letter and then by tele- isms (SNP) with apparent effects on adenoma risk for phone. A total of 670 participants provided written interactions with current NSAID use. informed consent. Deceased individuals (351) were also included in the study. The overall participation rate was 62.1%. A standardized telephone interview was conducted Materials and Methods by trained interviewers to obtain information on follow-up Study population and data collection examinations, medication use since baseline, demo- The TCPS was a colonoscopy-based case–control study graphics, medical history, family history, reproductive his- conducted in Nashville, TN from 2003 to 2010. Eligible tory, anthropometry, and lifestyle. Among participants, 706 participants, aged between 45 and 70 years old, were iden- (63.7%) completed the telephone interview. Beginning in tified from patients scheduled for colonoscopy at the Van- May 2004, buccal cell samples were collected from partici- derbilt Gastroenterology Clinic and the Veteran’s Affairs pants, or a saliva sample was collected using an Oragene kit. Tennessee Valley Health System Nashville Campus. Demo- A total of 532 participants (48.0%) provided a buccal and/ graphic properties of all participants are described in Table or Oragene sample. The study was approved by the Van- 1. For the purposes of the association analyses, we only derbilt University Institutional Review Board, the Veterans’ included participants of Caucasian race, although original Affairs Tennessee Valley Health System Institutional Review recruitment for TCPS was from a multiethnic population. Board, the Veterans’ Affairs Tennessee Valley Health System Excluded from the study were participants who had Research and Development Committee, and the Indiana genetic CRC syndromes, a prior history of inflammatory University Institutional Review Board Development bowel disease, prevalent adenomatous polyps, or any can- Committee. cer other than nonmelanoma skin cancer. Among eligible In both study populations, colonoscopic procedures were participants, 65% provided informed consent, and subse- carried out and reported using standard clinical protocols quently, 84% completed telephone interviews and 75% by the patient’s gastroenterologist. Any identified polyps completed a food frequency questionnaire designed for were removed using biopsy forceps or snare techniques. All the southern United States (24). Participants provided pathology diagnoses were determined by hospital pathol- DNA either before or after colonoscopy. Participants ogists and reported as part of routine care. Data were recruited before colonoscopy were asked to donate a 15 abstracted from these reports to classify study participants mL of blood sample. A total of 5,504 participants provided a into the following groups: adenomas only, hyperplastic blood sample. Buccal cell or Oragene kit samples were polyps only, presence of both adenomas and hyperplastic collected from 1,079 participants who chose not to provide polyps, and polyp-free controls. To be classified as polyp a blood sample, or if they were recruited after colonoscopy. free, the participant had to have a complete colonoscopy DNA was obtained from blood for 82.9% of participants reaching the cecum without the observation of polyps. and mouthwash buccal samples or Oragene samples for Participants with at least 2 adenomas were further classified 16.3% of participants. The study was approved by the as having multiple adenomas. An advanced adenoma was Vanderbilt University Institutional Review Board, the Veter- defined as meeting one of the following criteria: (i) size 1 ans’ Affairs Tennessee Valley Health System Institutional cm, (ii) tubulovillous or villous, or (iii) high-grade Review Board, and the Veterans’ Affairs Tennessee Valley dysplasia. Health System Research and Development Committee. Two independent samples of participants from TCPS and Participants were also included as adenoma cases from TIARS were evaluated for associations between genetic the TIARS, a retrospective cohort study conducted in Nash- variation in PG pathway genes and adenoma risk in a 2- ville, Tennessee, and Indianapolis, Indiana, United States. stage design. In the discovery phase, genotypes from a Eligible participants, aged between 40 and 75 years old, genome-wide association study (GWAS) were supplemen- were identified from patients diagnosed during colonosco- ted with additional genotyping assays to complete genomic py with an advanced or multiple adenomas between Jan- coverage of those genes and then imputed to the 1000

OF2 Cancer Prev Res; 2012 Cancer Prevention Research

NSAIDs and Adenoma

Table 1. Characteristics of study participants by phase, the TCPS (2003–2010) and TIARS (1996–2006)

Discovery (N ¼ 1,867) Replication (N ¼ 3,969)

Cases Controls Cases Controls Characteristic Total (N ¼ 958) (N ¼ 909) Pa,b Total (N ¼ 1,593) (N ¼ 2,376) Pa,b

Study population (%) TCPS 100 100 100 NA 87.5 68.8 100 <0.001 TIARS 0 0 0 12.5 31.2 0 Study site (%) <0.001 Vanderbilt University 100 100 100 NA 94.0 85 100 Indiana University 0 0 0 6.0 15 0 Age [y, mean (SD)] 58.5 (7.4) 59.0 (7.3) 58.1 (7.5) <0.001 57.1 (7.5) 58.2 (7.1) 56.6 (7.6) <0.001 Sex (female, %) 26.5 26.3 26.6 0.885 40.6 27.9 49.2 <0.001 Indications for 0.04 <0.001 colonoscopy (%)c Screening 56.7 56.6 60.3 56.7 52.8 61.0 Other 43.3 43.5 36.8 43.3 47.2 39.0 Educational 0.021 <0.001 attainment (%)c High school or less 25.9 31.7 25.6 24.3 27.0 19.8 Some college 25.1 25.7 26.6 26.5 28.4 23.7 College graduate 20.0 18.8 20.8 19.3 18.6 18.9 Graduate or 22.8 20.9 24.1 23.7 15.7 27.2 professional education Race (white, %) 100 100 100 NA 100 100 100 NA CRC family history (%)c 8.4 9.6 7.7 <0.001 8.0 8.0 7.9 0.790 Regular cigarette 58.4 64.9 54.1 <0.001 57.7 70.3 52.0 <0.001 smoking (%)c Regular alcohol 50.2 54.7 48.4 <0.001 49.3 53.7 48.5 <0.001 consumption (%)c BMI (kg/mc, mean)d 28.3 28.5 28.1 0.073 28.0 28.9 27.7 <0.001 Regularly exercised (%)c 51.4 51.6 52.3 0.83 52.9 45.8 54.2 <0.001 NSAID use (%)c 0.007 0.343 Current 39.1 40.0 38.7 45.1 34.3 50.0 Former 10.9 5.0 13.6 6.2 5.7 6.4 Never 50.0 55.0 47.7 48.7 60.0 43.6 Total energy 2,330 2,333.7 2,292.4 0.89 2,301 2,376.9 2,190.3 0.064 intake (kcal/d, mean)c

aDerived from ANOVA for continuous variables and c2 test for categorical variables. bP value for case–control comparison. cStandardized by age (40–49, 50–59, 60–64, and 65 years old) and sex distribution of all study participants. dStandardized by age distribution (40–49, 50–59, 60–64, and 65 years old) of all study participants

Genomes and HapMap reference panels. In the replication PGE receptor 1 (PTGER1) were assayed for association with phase, selected SNPs were genotyped in an independent adenoma risk. This subset of all possible genes that are sample of participants, and results from both phases were involved in PG-mediated inﬂammation, NSAID metabo- combined using meta-analysis. lism, PG synthesis and catabolism, and other relevant pathways were selected to reﬁne the scope of this study to genes Genotyping most proximal to the phenomenon of adenoma chemo- Genes were selected from the PG signaling and metabo- prevention by NSAIDs. lism pathways for analysis. PGE synthase (PTGES), 15- Initial genotyping was done using the Affymetrix hydroxyprostaglandin dehydrogenase (HPGD), PG–endo- Genome-Wide Human SNP Array 5.0 (Affymetrix, Inc.) to peroxide synthase 2 (PTGS2), hydroxysteroid (11-beta) agnostically detect associations with adenoma risk through- dehydrogenase 2 (HSD11B2), PGE receptor 4 (PTGER4), out the genome. Imputation was done using IMPUTEv2.2 PGE receptor 3 (PTGER3), PGE receptor 2 (PTGER2), and (25) with reference panels of densely genotyped SNPs from

www.aacrjournals.org Cancer Prev Res; 2012 OF3

Edwards et al.

the International HapMap Project Phase 3 data (26) and evaluated for concordance among replicate QC participants data from the 1000 Genomes project (27). Genomic cov- (99%), missing data by more than 5%, HWE P < 0.001, and erage of regions of interest in the discovery sample were MAF agreement with phase I. All 8 SNPs passed QC checks. augmented with further genotyping using Sequenom iPLEX The final data for association analysis consisted of 1,593 Gold genotyping (Sequenom, Inc.) in the PTGES, HPGD, cases and 2,376 controls. and PTGER1 to ensure at least 80% coverage of known common variants in the Caucasian population. The pro- Statistical analysis portion of common SNPs from the International HapMap In data from the GWAS, we assessed the relationship r2 Project phase 2 data that were tagged with at least an of 0.8 between genetic variation in candidate genes and the risk of are given in Supplementary Table S1. Follow-up genotyping colorectal adenoma using the software package of candidate SNPs in which association signals were SNPTESTv2.2.0 with the "–method score" option, using observed was carried out using Sequenom iPLEX Gold logistic regression with frequentist tests and assuming an genotyping. additive effect of SNP alleles on risk, adjusted for age and sex (31). In genes in which there were multiple nominally Quality control significant SNPs, we conducted conditional tests of associ- Quality control (QC) procedures were conducted on CEL ation for the remaining SNPs, adjusting for the most sig- files using the Dynamic Model (DM) algorithm in the nificant SNP, age, and sex. This procedure mitigates the Affymetrix Power Tools software package. Genotypes were effect of LD-induced significance and helps identify associa- called in the remaining samples using the BRLMM-P algo- tions at SNPs that are potentially due to LD with indepen- rithm (28). The average concordance of genotypes assessed dent mutations on distinct haplotypic backgrounds. Sum- using the PLINK software package within duplicate QC maries of the LD among the index and conditional SNPs are participants was 99.9% (29). The PLINK–sex check option provided in Supplementary Tables S2A to S4B. did not discover any participants whose X-chromosome The SNPs genotyped for replications were evaluated for heterozygosity was inconsistent with their reported sex. association with adenoma risk using PLINK with logistic Sixteen participants who were first- or second-degree rela- regression, adjusting for age, body mass index (BMI: kg/ tives with other study participants were removed from m2), and sex. We also evaluated risk for multiple adenoma further analysis. A total of 165 participants who were and advanced adenoma in the SNPs that were genotyped for missing greater than 5% of their autosomal genotypes were the replication phase. In addition, models were fit to eval- removed from further analysis. Population stratification uate interactions between the 8 candidate SNPs and NSAID was assessed by comparing the study participants to refer- exposure, encoded as current versus former and never users ence panels from the HapMap Phase 3 participants using in both phases of the analysis. Use was defined as taking EIGENSTRAT (30), resulting in the removal of 22 partici- NSAIDs at least 3 times a week for at least 1 year. Former pants with apparent ancestral differences from the rest of the users had stopped NSAID use for 1 year or more but did not sample. significantly differ from never users with regard to adenoma For SNP QC, SNPs were removed if they were missing in risk in either phase, and so these 2 classes of participants greater than 5% of participants, or if the minor allele were merged in the analysis (data not shown). frequency (MAF) in the samples that passed sample QC A meta-analysis was done by combining the results from was less than 1%. After related and admixed participants both phases of the investigation for both SNPs and SNP x were removed, SNPs were removed for major deviations NSAID interactions using the software METAL for the com- P < 6 from Hardy–Weinberg equilibrium (HWE) 1 10 . bined sample size of 2,551 cases and 3,285 controls (32). In After sample and SNP QC procedures, 402,326 SNPs addition, stratified analysis of the genotypes at SNP remained in 958 adenoma cases and 909 adenoma controls. rs12647154 were conducted to estimate ORs of adenoma In the COX genes, a total of 1,145 genotyped and high- for each genotype versus the referent allele homozygotes, quality imputed SNPs remained after QC. stratified by NSAID status. All reported P values are 2-sided. In the replication phase of the study, 8 SNPs were selected from COX pathway genes based on the statistical significance of tests of association, imputation quality, and allele Results frequency. SNPs were required to be nominally significant Demographic data for tests of main effect association in primary analyses or be Analysis of demographic variables for this study identi- associated with P value less than 0.15 in conditional anal- fied several associations with risk of adenoma (Table 1). In yses adjusted for the index SNP in each gene, have an the discovery phase, cases were significantly older than imputation quality information score from SNPTEST of at controls, less educated, were more likely to have a family least 0.8, and an allele frequency of at least 10%. These SNPs history of CRC, were more likely to drink and smoke, and were genotyped using the Sequenom genotyping system in more likely to be current NSAID users. In the replication 2,028 cases and 3,087 controls. Samples were checked for phase, cases were older, less likely to be female, less edu- duplications, and 41 pairs were removed. Also participants cated, were more likely to drink and smoke, had signifi- who did not self-report as Caucasian were removed from cantly higher BMI, and were less likely to exercise. Differ- this analysis (394 cases, 618 controls). These SNPs were ences in these associations between phases are mostly

OF4 Cancer Prev Res; 2012 Cancer Prevention Research

NSAIDs and Adenoma

attributable to statistical significance, as the direction of HPGD after adjusting for rs6846029, 2 SNPs in the gene effects for family history, exercise, and BMI were consistent PTGER3 with adjustment for rs3765414, and 1 SNP in between phases. The direction for NSAID use is in the PTGER2 after adjusting for rs17125318. The 2 additional opposite direction, and this is mostly because of a higher SNPs in HPGD were rs3797013 with OR, 1.28 (95% CI, proportion of current NSAID users in the control group of 0.97–1.68) and rs12647154 with OR, 0.87 (95% CI, 0.73– the replication phase. 1.04). Estimates for HPGD SNPs without adjustment for rs6846029 at rs3797013 were OR, 1.34 (95% CI, 1.04– Genetic main effects 1.75; P value ¼ 0.026) and at rs12647154 OR, 0.93 (95% Referent alleles were assigned at random for analysis of CI, 0.79–1.10; P value ¼ 0.414). In PTGER3, the unadjusted SNP data, as there are not strong a priori reasons for estimates for rs41485048 were OR, 1.17 (95% CI, 0.97– specifying a particular allele at a SNP as referent in GWAS. 1.42; P value ¼ 0.101) and for rs7541963 they were OR, As a result, effect sizes may be presented as protective, but we 1.16 (95% CI ¼ 0.96–1.41; P value ¼ 0.110). At the PTGER2 only know the magnitude of the association, and not the SNP rs1254600, the unadjusted OR, 1.1 (95% CI, 0.90– true direction with regard to population prevalence without 1.35; P value ¼ 0.339). making risk estimates from prospective data. SNPs in 3 No SNPs in the remaining genes of interest in this study genes from the GWAS data were nominally significantly (PTGS2, PTGER1, HSD11B2, PTGER4,and PTGES)were associated with risk of adenoma in the discovery partici- nominally significant (data not shown). Additional pants (Table 2). No SNPs in the other 5 genes had a P value adjustment for use of NSAIDs did not substantively for association with risk of adenoma of less than 0.05. The change estimates of effects, significance, or seem to be SNP rs3765414 in PTGER3 was associated with adenoma a confounder for the association of SNPs in the candidate risk with an OR, 1.32 (95% CI, 1.04–1.68) and P value ¼ genes. 0.023. The SNP rs6846029 in the HPGD was associated with adenoma risk with OR, 0.84 (95% CI, 0.70–0.96), and P Replication of associated SNPs value ¼ 0.011. The SNP rs17125318 in the PTGER2 was The genotyped SNPs were evaluated for association associated with adenoma risk with OR, 0.85 (95% CI, 0.75– with adenoma risk in an independent set of self-reported 0.96) and P value ¼ 0.008. These SNPs were genotyped for European ancestry participants (Table 2). The same con- replication. All other SNPs in these genes with P values ditional models were fit for the replication study as were smaller than 0.05 are presented in Supplementary Tables used to identify the SNPs in the discovery phase. Two S5–S7. SNPs were nominally associated with adenoma risk; rs3797013 with OR, 0.83 (95% CI, 0.70–0.98; P value Conditional analyses ¼ 0.03), and rs12647154 was weakly associated in the Analyses adjusting for the most significant index SNP replication study with OR, 0.90 (95% CI, 0.79–1.02; P were conducted in each gene to find associations that were value ¼ 0.116). We also tested these 8 SNPs for associ- conditionally independent of the index signal. We identi- ation with risk of multiple and advanced adenomas. The fied and genotyped 3 index SNPs and 5 additional SNPs of PTGER2 SNP rs1254600 was associated with risk of interest using this approach. There were 2 SNPs in the gene multiple adenomas in the discovery phase with OR,

Table 2. Meta-analysis results with unadjusted P values for 8 selected SNPs from COX pathway genes the TCPS (2003–2010) and TIARS (1996–2006)

Discovery Replication Meta

Gene SNP Referent Allele OR (95% CI) P OR (95% CI) P OR (95% CI) P Effecta

PTGER2 rs17125318 G 0.65 (0.48–0.88) 0.005 1.01 (0.87–1.18) 0.856 0.85 (0.75–0.96) 0.008 þ PTGER2 rs1254600b G 0.69 (0.49–0.98) 0.037 0.97 (0.79–1.18) 0.762 0.84 (0.73–0.98) 0.028 PTGER3 rs3765414 T 1.32 (1.04–1.68) 0.021 1.06 (0.89–1.26) 0.492 1.12 (0.96–1.31) 0.158 þþ PTGER3 rs41485048c T 1.21 (0.99–1.46) 0.056 1.05 (0.91–1.20) 0.489 1.09 (0.97–1.24) 0.153 þþ PTGER3 rs7541936c C 1.19 (0.99–1.45) 0.062 1.05 (0.91–1.20) 0.499 1.09 (0.96–1.23) 0.176 þþ HPGD rs6846029 T 0.82 (0.70–0.96) 0.011 1.07 (0.96–1.19) 0.241 1.05 (0.97–1.15) 0.223 þ HPGD rs3797013d A 1.28 (0.97–1.68) 0.079 0.83 (0.70–0.98) 0.032 0.88 (0.77–1.00) 0.054 þ HPGD rs12647154d A 0.87 (0.73–1.04) 0.120 0.90 (0.79–1.02) 0.116 0.89 (0.80–1.01) 0.073

aIndicates whether the nonreferent allele increases risk (þ) or decreases risk () in the discovery and replication stages, respectively. bAnalyses conditioned upon PTGER2 SNP rs17125318. cAnalyses conditioned upon PTGER3 SNP rs3765414. dAnalyses conditioned upon HPGD SNP rs6846029.

www.aacrjournals.org Cancer Prev Res; 2012 OF5

Edwards et al.

Table 3. Results from tests of interaction with unadjusted P values for selected SNPs and NSAID exposure, the TCPS (2003–2010) and TIARS (1996–2006)

Discovery Replication Meta

Gene Test OR (95% CI) P OR (95% CI) P OR (95% CI) P Effecta

PTGER2 rs17125318xNSAID 0.69 (0.38–1.25) 0.220 0.79 (0.55–1.14) 0.205 0.76 (0.59–0.95) 0.019 – PTGER2 rs1254600xNSAIDb 0.78 (0.52–1.19) 0.255 0.99 (0.74–1.33) 0.979 0.89 (0.72–1.11) 0.299 – PTGER3 rs3765414xNSAID 0.90 (0.56–1.46) 0.675 0.87 (0.58–1.28) 0.471 0.88 (0.69–1.12) 0.289 – PTGER3 rs41485048xNSAIDc 1.02 (0.69–1.49) 0.921 0.92 (0.67–1.26) 0.600 0.95 (0.76–1.20) 0.691 þ PTGER3 rs7541936xNSAIDc 0.99 (0.68–1.45) 0.963 0.93 (0.68–1.27) 0.647 0.95 (0.76–1.19) 0.655 – HPGD rs6846029xNSAID 1.02 (0.75–1.39) 0.894 0.87 (0.68–1.11) 0.261 0.91 (0.77–1.09) 0.312 þ HPGD rs3797013xNSAIDd 1.62 (0.94–2.77) 0.081 1.03 (0.71–1.49) 0.888 1.10 (0.78–1.56) 0.565 þþ HPGD rs12647154xNSAIDd 0.79 (0.56–1.10) 0.164 0.69 (0.52–0.92) 0.011 0.76 (0.62–0.85) 5 105 –

aIndicates whether the nonreferent allele increases risk (þ) or decreases risk (-) in the discovery and replication stages, respectively. bAnalyses conditioned upon PTGER2 SNP rs17125318. cAnalyses conditioned upon PTGER3 SNP rs3765414. dAnalyses conditioned upon HPGD SNP rs6846029.

0.73 (95% CI, 0.53–0.99, P value 0.046), but not in the Discussion P replication phase with OR, 0.65 (95% CI, 0.29–1.40, In this study, we have conducted a detailed 2-stage ¼ value 0.263). The meta-analysis of this SNP was also analysis of 8 candidate genes from the PG synthesis pathway P ¼ significant, with OR, 0.71 (95% CI, 0.57–0.88, value for association with risk of colorectal adenoma. We also 0.0007). No other SNP was significantly associated with evaluated interactions with NSAID exposure on adenoma risk of either multiple or advanced adenomas in either risk for the most statistically implicated SNPs and found phase (data not shown). evidence for effect modification by NSAIDs. We detected nominally significant associations in PTGER2 and an inter- Meta-analyses action in HPGD in the combined meta-analysis of both Meta-analyses of the combined discovery and replication stages. We also detected an association between PTGER2 samples for SNPs and interactions were conducted for the 8 and the risk of multiple adenomas. No statistical signal in SNPs identified for replication. For single SNPs, 5 of 8 SNPs this analysis survives a formal multiple testing correction for had effects in the same direction in both phases, and 2 SNPs all tests, although the interaction result in HPGD withstands had a P value less than 0.05 in the combined analysis a Bonferroni correction within tests of interaction. The (Table 2). For interactions with current NSAID use, 6 of 8 HPGD SNP rs12647154 had consistent direction of effect tests were in the same direction in both phases, and in both phases for both main effect and interaction analyses. HPGD SNP rs12647154 had a nominally significant The hypotheses relating the genes studied here to colorectal interaction in the meta-analysis ORmeta, 0.76 (95% CI, adenoma are well supported by biologic and epidemiologic 0.62–0.85; P value ¼ 0.00005), with a P value ¼ 0.42 for a data. Many other phenotypes have been ascribed to HPGD 1-degree-of-freedom test of heterogeneity (Table 3). For and PTGER2 in human and animal studies, showing that rs12647154, the C allele decreased risk of adenoma, and these genes are important in multiple biologic pathways, this protective phenomenon was synergistic with current from crucial steps of development and tumor suppression NSAID use. The nominally significant SNPs from the to inflammation homeostasis. meta-analysis were the PTGER2 SNPs rs17125318 and Previous studies of PG pathway genetic variants on risk of rs1254600. Two HPGD SNPs trended toward an associ- adenoma have produced generally subtle and inconsistent ation at rs3797013, ORmeta, 0.88 (95% CI, 0.77–1.00; P results. Some subtle effect modification was recently value ¼ 0.054) and at rs12647154, ORmeta, 0.89 (95% CI, observed for PTGER2 and PTGER4 SNPs and NSAID expo- 0.73–1.01; P value ¼ 0.073). ORs for adenomas were sure (33). We did not observe the previously reported effects estimated for individual genotypes at rs12647154, strat- and effect modification in PTGS2, although we did not test ified by NSAID status (Supplementary Table S8). Similar for interactions unless we observed some marginal effect of trends of effect sizes were observed in both the discovery a SNP on risk of adenoma (34–39). and replication samples, in which current NSAID users The gene product of HPGD, which oxidizes the PG carrying the C allele at rs12647154 had approximately molecule and greatly reduces inflammatory activity in vivo, 30% reduced risk of adenoma, whereas participants with is the primary enzyme for PG degradation (40). Two recent those genotypes who were not current NSAID users were genetic studies of CRC identified SNPs in HPGD as mar- not protected. ginally associated with CRC (41, 42). HPGD is highly

OF6 Cancer Prev Res; 2012 Cancer Prevention Research

NSAIDs and Adenoma

expressed in normal colon mucosa, but expression is lost in on risk. Only controls with complete colonoscopies were human colon cancer cells (43, 44). A recent study showed included in the analysis, protecting against misclassification that the adenoma-preventive activity of the NSAID cele- of cases and controls. Additional SNPs were also assayed in coxib is abrogated in HPGD knockout animals, and that addition to the GWAS SNPs in the discovery phase of the participants who develop adenoma in clinical trials inves- study to provide adequate coverage of the candidate genes tigating chemoprevention by celecoxib have lower average before imputation was carried out. The final analysis con- colonic HPGD expression levels than participants who do sisted of 2,551 cases and 3,285 controls, a sample size not develop adenoma (45). Whether the interaction sufficient to detect a modest effect of OR between 1.1 and observed here is a result of linkage disequilibrium with a 1.5 with a MAF of 0.1 or larger at multiple comparisons mutation affecting gene expression or secondary to a change corrected levels of significance. For interactions, the final in HPGD catalytic activity is unknown and requires further analysis had 80% power to detect an interaction OR of 1.5 to study, in which the primary hypothesis of interest is that the 1.8 with MAF of 0.1 or more. high-risk A allele at rs12647154 is associated with lower In summary, modest direct effects on risk of adenoma levels of HPGD expression. We do not know of any study in were detected in the combined sample for HPGD and the literature that specifically evaluates this mechanism. PTGER2 variants. Effect modification by NSAID exposure Similarly, PTGER2 has been implicated in several was observed, suggesting that the role of these genes in cancer traits and cancer-related phenomena. PTGER2 adenoma formation is modified by PG inhibition. Further overexpression in CRC has been associated with micro- studies may be required to completely identify the roles of satellite instability, independent of the CpG island these genes in adenoma formation. methylator phenotype (46). Polymorphisms in PTGER2 have also been associated with CRC risk in humans (42). Disclosure of Potential Conflicts of Interest We found in this study PTGER2 variants associated with No potential conflicts of interest were disclosed. the risk of multiple adenomas only. The biology of PTGER3 is also complex and related to Authors' Contributions PTGER3 Conception and design: M.J. Shrubsole, R. Ness, W. Zheng cancer risk and progression. The gene encodes one Development of methodology: T.L. Edwards, M.J. Shrubsole, R. Ness, W. of several PG receptors. PTGER3 has been shown to regulate Zheng the ability of breast cancer tumors to undergo angiogenesis Analysis and interpretation of data (e.g., statistical analysis, biosta- tistics, computational analysis): T.L. Edwards, M.J. Shrubsole, Z. Fu, and tumor growth (47–49). Mitogenic activity in CRC cells H.J. Murff, W. Smalley, W. Zheng has also been linked to PTGER3 (50). Studies of tumor cell Acquisition of data (provided animals, acquired and managed patients, migrations have suggested that PTGER3 contributes to provided facilities, etc.): M.J. Shrubsole, Q. Cai, G. Li, D. Rex, T.M. Ulbright, R. Ness, W. Zheng metastasis by upregulation of VEGF receptor signaling (51). Writing, review, and/or revision of the manuscript: T.L. Edwards, M.J. Although there is a definite role for PG biology in the Shrubsole, Q. Dai, D. Rex, T.M. Ulbright, Z. Fu, H.J. Murff, R. Ness, W. Zheng Administrative, technical, or material support (i.e., reporting or orga- pathophysiology of colorectal adenoma, the effects of nizing data, constructing databases): M.J. Shrubsole, Q. Dai, W. Zheng genetic polymorphisms in PG pathway genes on risk of Study supervision: M.J. Shrubsole, W. Zheng adenoma are modest. The effects we describe here are subtle and do not explain a large proportion of risk. A recent well- Grant Support powered meta-analysis of CRC GWAS did not identify PG This study was supported through the National Cancer Institute grants pathway genes with genome-wide levels of significance, P50CA95103, R01CA121060, and R01AT004660. T.L. Edwards is supported by a Vanderbilt Clinical and Translational Research Scholar Award which may have occurred because of the subtle influence 5KL2RR024975 (to T.L. Edwards, in part). The TCPS was conducted by of these genes, or effect modification that was not modeled the Survey and Biospecimen Shared Resource supported in part by the in those studies (52). However, even modest effects of Vanderbilt-Ingram Cancer Center (P30 CA 68485). The costs of publication of this article were defrayed in part by the genetic variants can provide some insight into biologic payment of page charges. This article must therefore be hereby marked mechanisms. advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate The TCPS and TIARS together are one of the largest this fact. colonoscopy-based case–control studies of colorectal Received October 5, 2011; revised March 26, 2012; accepted March 28, polyps, providing adequate power to detect subtle effects 2012; published OnlineFirst May 2, 2012.

References 1. Siegel R, Ward E, Brawley O, Jemal A. Cancer statistics, 2011: The 5. Tsujii M, Kawano S, DuBois RN. Cyclooxygenase-2 expression in impact of eliminating socioeconomic and racial disparities on prema- human colon cancer cells increases metastatic potential. Proc Natl ture cancer deaths. CA Cancer J Clin 2011;61:212–36. Acad Sci U S A 1997;94:3336–40. 2. Fearon ER, Vogelstein B. A genetic model for colorectal tumorigenesis. 6. Gupta RA, DuBois RN. Colorectal cancer prevention and treatment Cell 1990;61:759–67. by inhibition of cyclooxygenase-2. Nat Rev Cancer 2001;1: 3. Vakiani E, Yantiss RK. Pathologic features and biologic importance of 11–21. colorectal serrated polyps. Adv Anat Pathol 2009;16:79–91. 7. Sonoshita M, Takaku K, Sasaki N, Sugimoto Y, Ushikubi F, Narumiya 4. Patrono C, Patrignani P, Garcia Rodriguez LA. Cyclooxygenase-selec- S, et al. Acceleration of intestinal polyposis through prostaglandin tive inhibition of prostanoid formation: transducing biochemical selec- receptor EP2 in Apc(Delta 716) knockout mice. Nat Med 2001;7: tivity into clinical read-outs. J Clin Invest 2001;108:7–13. 1048–51.

www.aacrjournals.org Cancer Prev Res; 2012 OF7

Edwards et al.

8. Brown JR, DuBois RN. COX-2: a molecular target for colorectal cancer 28. BRLMM-P: a Genotype Calling Method for the SNP 5.0 Array. Affy- prevention. J Clin Oncol 2005;23:2840–55. metrix, Inc. [cited May 7 2012]. Available from: http://media.affymetrix. 9. Narayanan BA, Narayanan NK, Pittman B, Reddy BS. Regression of com/support/technical/whitepapers/brlmmp_whitepaper.pdf. 2012. mouse prostatic intraepithelial neoplasia by nonsteroidal anti-inflam- 29. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, matory drugs in the transgenic adenocarcinoma mouse prostate et al. PLINK: a tool set for whole-genome association and population- model. Clin Cancer Res 2004;10:7727–37. based linkage analyses. Am J Hum Genet 2007;81:559–75. 10. Arber N, Eagle CJ, Spicak J, Racz I, Dite P, Hajer J, et al. Celecoxib for 30. Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, Reich the prevention of colorectal adenomatous polyps. N Engl J Med D. Principal components analysis corrects for stratification in genome- 2006;355:885–95. wide association studies. Nat Genet 2006;38:904–9. 11. Baron JA, Cole BF, Sandler RS, Haile RW, Ahnen D, Bresalier R, et al. A 31. Marchini J, Howie B, Myers S, McVean G, Donnelly P. A new multipoint randomized trial of aspirin to prevent colorectal adenomas. N Engl J method for genome-wide association studies by imputation of geno- Med 2003;348:891–9. types. Nat Genet 2007;39:906–13. 12. Baron JA, Sandler RS, Bresalier RS, Quan H, Riddell R, Lanas A, et al. A 32. WillerCJ,LiY,AbecasisGR.METAL:fastandefficient meta-analy- randomized trial of rofecoxib for the chemoprevention of colorectal sis of genomewide association scans. Bioinformatics 2010;26: adenomas. Gastroenterology 2006;131:1674–82. 2190–1. 13. Benamouzig R, Deyra J, Martin A, Girard B, Jullian E, Piednoir B, et al. 33. Poole EM, Hsu L, Xiao L, Kulmacz RJ, Carlson CS, Rabinovitch PS, Daily soluble aspirin and prevention of colorectal adenoma recurrence: et al. Genetic variation in prostaglandin E2 synthesis and signaling, one-year results of the APACC trial. Gastroenterology 2003;125: prostaglandin dehydrogenase, and the risk of colorectal adenoma. 328–36. Cancer Epidemiol Biomarkers Prev 2010;19:547–57. 14. Bertagnolli MM, Eagle CJ, Zauber AG, Redston M, Solomon SD, Kim K, 34. Barry EL, Sansbury LB, Grau MV, Ali IU, Tsang S, Munroe DJ, et al. et al. Celecoxib for the prevention of sporadic colorectal adenomas. N Cyclooxygenase-2 polymorphisms, aspirin treatment, and risk for Engl J Med 2006;355:873–84. colorectal adenoma recurrence–data from a randomized clinical trial. 15. Sandler RS, Halabi S, Baron JA, Budinger S, Paskett E, Keresztes R, Cancer Epidemiol Biomarkers Prev 2009;18:2726–33. et al. A randomized trial of aspirin to prevent colorectal adenomas in 35. Gong Z, Bostick RM, Xie D, Hurley TG, Deng Z, Dixon DA, et al. Genetic patients with previous colorectal cancer. N Engl J Med 2003;348: polymorphisms in the cyclooxygenase-1 and cyclooxygenase-2 883–90. genes and risk of colorectal adenoma. Int J Colorectal Dis 2009;24: 16. Cole BF, Logan RF, Halabi S, Benamouzig R, Sandler RS, Grainge 647–54. MJ, et al. Aspirin for the chemoprevention of colorectal adenomas: 36. Peters WH, te Morsche RH, Roelofs HM, Mathus-Vliegen EM, Berkh- meta-analysis of the randomized trials. J Natl Cancer Inst out M, Nagengast FM. COX-2 polymorphisms in patients with familial 2009;101:256–66. adenomatous polyposis. Oncol Res 2009;17:347–51. 17. Dube C, Rostom A, Lewin G, Tsertsvadze A, Barrowman N, Code C, 37. Siezen CL, van Leeuwen AI, Kram NR, Luken ME, van Kranen HJ, et al. The use of aspirin for primary prevention of colorectal cancer: a Kampman E. Colorectal adenoma risk is modified by the interplay systematic review prepared for the U.S. Preventive Services Task between polymorphisms in arachidonic acid pathway genes and fish Force. Ann Intern Med 2007;146:365–75. consumption. Carcinogenesis 2005;26:449–57. 18. Gao F, Liao C, Liu L, Tan A, Cao Y, Mo Z. The effect of aspirin in the 38. Ueda N, Maehara Y, Tajima O, Tabata S, Wakabayashi K, Kono S. recurrence of colorectal adenomas: a meta-analysis of randomized Genetic polymorphisms of cyclooxygenase-2 and colorectal adenoma controlled trials. Colorectal Dis 2009;11:893–901. risk: the Self Defense Forces Health Study. Cancer Sci 2008;99: 19. Rostom A, Dube C, Lewin G, Tsertsvadze A, Barrowman N, Code C, 576–81. et al. Nonsteroidal anti-inflammatory drugs and cyclooxygenase-2 39. Ulrich CM, Whitton J, Yu JH, Sibert J, Sparks R, Potter JD, et al. PTGS2 inhibitors for primary prevention of colorectal cancer: a systematic (COX-2) -765G > C promoter variant reduces risk of colorectal ade- review prepared for the U.S. Preventive Services Task Force. Ann noma among nonusers of nonsteroidal anti-inflammatory drugs. Can- Intern Med 2007;146:376–89. cer Epidemiol Biomarkers Prev 2005;14:616–9. 20. Tangrea JA, Albert PS, Lanza E, Woodson K, Corle D, Hasson M, et al. 40. Pichaud F, Delage-Mourroux R, Pidoux E, Jullienne A, Rousseau- Non-steroidal anti-inflammatory drug use is associated with reduction Merck MF. Chromosomal localization of the type-I 15-PGDH gene to in recurrence of advanced and non-advanced colorectal adenomas 4q34-q35. Hum Genet 1997;99:279–81. (United States). Cancer Causes Control 2003;14:403–11. 41. Frank B, Hoeft B, Hoffmeister M, Linseisen J, Breitling LP, Chang- 21. Lin HJ, Lakkides KM, Keku TO, Reddy ST, Louie AD, Kau IH, et al. Claude J, et al. Association of hydroxyprostaglandin dehydrogenase Prostaglandin H synthase 2 variant (Val511Ala) in African Americans 15-(NAD) (HPGD) variants and colorectal cancer risk. Carcinogenesis may reduce the risk for colorectal neoplasia. Cancer Epidemiol Bio- 2011;32:190–6. markers Prev 2002;11:1305–15. 42. Hoeft B, Linseisen J, Beckmann L, Muller-Decker K, Canzian F, Husing 22. Macarthur M, Sharp L, Hold GL, Little J, El-Omar EM. The role of A, et al. Polymorphisms in fatty-acid-metabolism-related genes are cytokine gene polymorphisms in colorectal cancer and their interaction associated with colorectal cancer risk. Carcinogenesis 2010;31: with aspirin use in the northeast of Scotland. Cancer Epidemiol 466–72. Biomarkers Prev 2005;14:1613–8. 43. Backlund MG, Mann JR, Holla VR, Buchanan FG, Tai HH, Musiek ES, 23. Sansbury LB, Bergen AW, Wanke KL, Yu B, Caporaso NE, Chatterjee et al. 15-Hydroxyprostaglandin dehydrogenase is down-regulated in N, et al. Inflammatory cytokine gene polymorphisms, nonsteroidal anti- colorectal cancer. J Biol Chem 2005;280:3217–23. inflammatory drug use, and risk of adenoma polyp recurrence in the 44. Yan M, Rerko RM, Platzer P, Dawson D, Willis J, Tong M, et al. 15- polyp prevention trial. Cancer Epidemiol Biomarkers Prev 2006;15: Hydroxyprostaglandin dehydrogenase, a COX-2 oncogene antago- 494–501. nist, is a TGF-beta-induced suppressor of human gastrointestinal 24. Buchowski MS, Schlundt DG, Hargreaves MK, Hankin JH, Signorello cancers. Proc Natl Acad Sci U S A 2004;101:17468–73. LB, Blot WJ. Development of a culturally sensitive food frequency 45. Yan M, Myung SJ, Fink SP, Lawrence E, Lutterbaugh J, Yang P, et al. questionnaire for use in the Southern Community Cohort Study. Cell 15-Hydroxyprostaglandin dehydrogenase inactivation as a mecha- Mol Biol (Noisy-le-grand) 2003;49:1295–304. nism of resistance to celecoxib chemoprevention of colon tumors. 25. Howie BN, Donnelly P, Marchini J. A flexible and accurate genotype Proc Natl Acad Sci U S A 2009;106:9409–13. imputation method for the next generation of genome-wide associa- 46. Baba Y, Nosho K, Shima K, Goessling W, Chan AT, Ng K, et al. PTGER2 tion studies. PLoS Genet 2009;5:e1000529. overexpression in colorectal cancer is associated with microsatellite 26. The International HapMap Consortium. HapMap Homepage. [cited instability, independent of CpG island methylator phenotype. Cancer May 7 2012]. Available from: http://hapmap.ncbi.nlm.nih.gov/. 2012. Epidemiol Biomarkers Prev 2010;19:822–31. 27. The 1000 Genomes Project Consortium. 1000 Genomes Project. [cited 47. Katoh H, Hosono K, Ito Y, Suzuki T, Ogawa Y, Kubo H, et al. COX-2 and May 7 2012]. Available from: http://www.1000genomes.org/. 2012. prostaglandin EP3/EP4 signaling regulate the tumor stromal

OF8 Cancer Prev Res; 2012 Cancer Prevention Research

NSAIDs and Adenoma

proangiogenic microenvironment via CXCL12-CXCR4 chemokine 50. Lofﬂer I, Grun M, Bohmer FD, Rubio I. Role of cAMP in the promotion of systems. Am J Pathol 2010;176:1469–83. colorectal cancer cell growth by prostaglandin E2. BMC Cancer 48. Ogawa Y, Suzuki T, Oikawa A, Hosono K, Kubo H, Amano H, et al. Bone 2008;8:380. marrow-derived EP3-expressing stromal cells enhance tumor-asso- 51. Fujino H, Toyomura K, Chen XB, Regan JW, Murayama T. Prosta- ciated angiogenesis and tumor growth. Biochem Biophys Res Com- glandin E regulates cellular migration via induction of vascular endo- mun 2009;382:720–5. thelial growth factor receptor-1 in HCA-7 human colon cancer cells. 49. Robertson FM, Simeone AM, Lucci A, McMurray JS, Ghosh S, Cris- Biochem Pharmacol 2011;81:379–87. tofanilli M. Differential regulation of the aggressive phenotype of 52. Peters U, Hutter CM, Hsu L, Schumacher FR, Conti DV, Carlson CS, inﬂammatory breast cancer cells by prostanoid receptors EP3 and et al. Meta-analysis of new genome-wide association studies of EP4. Cancer 2010;116:2806–14. colorectal cancer risk. Hum Genet 2012;131:217–34.

www.aacrjournals.org Cancer Prev Res; 2012 OF9

A Study of Prostaglandin Pathway Genes and Interactions with Current Nonsteroidal Anti-inflammatory Drug Use in Colorectal Adenoma

Todd L. Edwards, Martha J. Shrubsole, Qiuyin Cai, et al.

Cancer Prev Res Published OnlineFirst May 2, 2012.

Updated version Access the most recent version of this article at: doi:10.1158/1940-6207.CAPR-11-0459

Supplementary Access the most recent supplemental material at: Material http://cancerpreventionresearch.aacrjournals.org/content/suppl/2012/05/02/1940-6207.CAPR -11-0459.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerpreventionresearch.aacrjournals.org/content/early/2012/05/10/1940-6207.CAPR- 11-0459. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.