Published OnlineFirst March 2, 2010; DOI: 10.1158/0008-5472.CAN-09-3575 Prevention and Epidemiology Cancer Research Pooled Analysis of Phosphatidylinositol 3-Kinase Pathway Variants and Risk of Prostate Cancer Stella Koutros1, Fredrick R. Schumacher2, Richard B. Hayes3, Jing Ma4, Wen-Yi Huang1, Demetrius Albanes1, Federico Canzian6, Stephen J. Chanock1, E. David Crawford7, W. Ryan Diver8, Heather Spencer Feigelson8,9, Edward Giovanucci5, Christopher A. Haiman2, Brian E. Henderson2, David J. Hunter9, Rudolf Kaaks6, Laurence N. Kolonel10, Peter Kraft5, Loïc Le Marchand10, Elio Riboli11, Afshan Siddiq11, Mier J. Stampfer4,5, Daniel O. Stram2, Gilles Thomas1, Ruth C. Travis12, Michael J. Thun8, Meredith Yeager1, and Sonja I. Berndt1 Abstract The phosphatidylinositol 3-kinase (PI3K) pathway regulates various cellular processes, including cellular proliferation and intracellular trafficking, and may affect prostate carcinogenesis. Thus, we explored the as- sociation between single-nucleotide polymorphisms (SNP) in PI3K genes and prostate cancer. Pooled data from the National Cancer Institute Breast and Prostate Cancer Cohort Consortium were examined for asso- ciations between 89 SNPs in PI3K genes (PIK3C2B, PIK3AP1, PIK3C2A, PIK3CD, and PIK3R3) and prostate can- cer risk in 8,309 cases and 9,286 controls. Odds ratios (OR) and 95% confidence intervals (95% CI) were estimated using logistic regression. SNP rs7556371 in PIK3C2B was significantly associated with prostate can- – P P cer risk [ORper allele, 1.08 (95% CI, 1.03 1.14); trend = 0.0017] after adjustment for multiple testing ( adj = 0.024). Simultaneous adjustment of rs7556371 for nearby SNPs strengthened the association [ORper allele, 1.21 (95% CI, – P 1.09 1.34); trend = 0.0003]. The adjusted association was stronger for men who were diagnosed before the age of 65 years [ORper allele, 1.47 (95% CI, 1.20–1.79); Ptrend = 0.0001] or had a family history [ORper allele = 1.57 (95% CI, 1.11–2.23); Ptrend = 0.0114], and was strongest in those with both characteristics [ORper allele = 2.31 (95% CI, 1.07–5.07), P-interaction = 0.005]. Increased risks were observed among men in the top tertile of circulating insulin-like growth factor-I (IGF-I) levels [ORper allele = 1.46 (95% CI, 1.04–2.06); Ptrend = 0.075]. No differences ≥ were observed with disease aggressiveness (Gleason grade 8 or stage T3/T4 or fatal). In conclusion, we ob- served a significant association between PIK3C2B and prostate cancer risk, especially for familial, early-onset disease, which may be attributable to IGF-dependent PI3K signaling. Cancer Res; 70(6); 2389–96. ©2010 AACR. Introduction fication, and rearrangement in the PI3K pathway and its downstream targets have been observed in several cancer The phosphatidylinositol 3-kinase (PI3K) pathway regu- sites, including prostate cancer (9–11). Because of the role lates various cellular processes such as cell growth, prolif- of PI3Ks in cell proliferation, much of the research on this eration, apoptosis, motility, differentiation, survival, and pathway concerns its potential as a target for anticancer intracellular trafficking (1). PI3Ks are heterodimeric lipid therapies. kinases that are composed of regulatory and catalytic sub- Despite the well-known role of this pathway in cancer pro- units that catalyze the production of several phosphoinosi- gression, genetic variants in PI3K genes have not been well tides critical for the signal transduction in these multiple studied. One study found an association between a PIK3CA cellular processes (2). In addition, numerous growth fac- SNP (rs2865084) and endometrioid ovarian cancer but not tors signal through the PI3K pathway (3, 4), including in- overall ovarian cancer risk (12). Two studies have evaluated sulin-like growth factors (IGF), which have also been the functional variant Met326Ile (rs3730089) in PIK3R1 and linked with prostate carcinogenesis (5–8). Mutation, ampli- risk of colon (13) and prostate (14) cancers and found no Authors' Affiliations: 1Division of Cancer Epidemiology and Genetics, Department of Clinical Medicine, University of Oxford, Oxford, United National Cancer Institute, Rockville, Maryland; 2Department of Kingdom Preventive Medicine, University of Southern California, Los Angeles, Note: Supplementary data for this article are available at Cancer California; 3New York University School of Medicine, New York, New Research Online (http://cancerres.aacrjournals.org/). York; 4Channing Laboratory, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School; 5Harvard School of Corresponding Author: Stella Koutros, Occupational and Environmental Public Health, Boston, Massachusetts; 6German Cancer Research Epidemiology Branch, Division of Cancer Epidemiology and Genetics, Center (DKFZ), Heidelberg, Germany; 7Department of Urology, National Cancer Institute, 6120 Executive Boulevard, EPS 8115, MSC University of Colorado Health Sciences Center, Aurora, Colorado; 7240, Rockville, MD 20852. Phone: 301-594-6352; Fax: 301-402-1819; 8Department of Epidemiology, American Cancer Society, Atlanta, E-mail: [email protected]. Georgia; 9Kaiser Permanente, Denver, Colorado; 10Cancer Research doi: 10.1158/0008-5472.CAN-09-3575 Center, University of Hawaii, Honolulu, Hawaii; 11Imperial College, London, United Kingdom; and 12Cancer Epidemiology Unit, Nuffield ©2010 American Association for Cancer Research. www.aacrjournals.org 2389 Downloaded from cancerres.aacrjournals.org on September 30, 2021. © 2010 American Association for Cancer Research. Published OnlineFirst March 2, 2010; DOI: 10.1158/0008-5472.CAN-09-3575 Koutros et al. association with either cancer site. Although some studies Diagnostic System Laboratories. Blood levels were catego- have looked at polymorphisms in genes of downstream tar- rized into tertiles based on the distribution among the con- gets of PI3K and prostate cancer risk (15–17), there is mini- trols. Although the blood level assays were done at different mal information about the effect of variants in the PI3K gene times, batch effects did not seem to confound the results, family on prostate cancer risk. and the analyses were pooled across cohorts controlling for Given the limited evaluation of PI3K gene variants, we study. sought to further explore the association between germline Genotyping and SNP selection. Tag SNPs for the five PI3K polymorphisms, which may alter the function of these genes candidate genes potentially related to the IGF pathway were and the proteins they encode, and risk of prostate cancer. selected using the CEU HapMap data assuming a minor al- Pooled data from seven prospective studies in the National lele frequency >0.02, an Illumina design score >0.4, and an r2 Cancer Institute (NCI) Breast and Prostate Cancer Cohort >0.8 for binning.(22) Genotyping in the prostate cancer cases Consortium (BPC3; ref. 18) were examined for associations and controls was done in four laboratories (University of between 89 single-nucleotide polymorphisms (SNP) in five Southern California, Los Angeles, CA; Harvard School of Pub- PI3K pathway genes, PIK3C2B (chromosome 1), PIK3AP1 lic Health, Boston, MA; Core Genotyping Facility, National (chromosome 10), PIK3C2A (chromosome 11), PIK3CD (chro- Cancer Institute, Bethesda, MD; and Imperial College, Lon- mosome 1), and PIK3R3 (chromosome 1), and prostate can- don, United Kingdom) using Illumina GoldenGate technology cer risk in 8,751 cases and 9,742 controls. We also evaluated as part of a larger array of 1,536 SNPs in total. Each genotype the effects of these PI3K pathway SNPs and circulating IGF-I center genotyped 30 CEU HapMap trios to evaluate interla- and IGF binding protein 3 (IGFBP-3) on prostate cancer risk boratory reproducibility, and for the 1,536 SNPs that made up in a subgroup of 6,076 men using prediagnostic sera. the Illumina platform, the interlaboratory concordance was 99.5% (before excluding failed SNPs or samples). Within each Materials and Methods study, blinded duplicate samples (∼5%) were also included and concordance of these samples ranged from 97.2% to Study population. The BPC3 has been described else- 99.9% across studies. where (18). Briefly, the consortium includes large well-estab- Data filtering and imputation. Any sample where more lished cohorts assembled in the United States and Europe than 25% of the SNPs attempted on a given platform failed that have DNA for genotyping and extensive questionnaire was removed from the data set. Data were filtered by study to data from cohort members. The prostate cancer study remove poorly performing SNPs: All SNPs that failed on 25% includes seven case-control studies nested within these or more samples were excluded from the data set, as were all − cohorts: the American Cancer Society (ACS) Cancer Preven- SNPs that showed statistically significant (P <105) devia- tion Study II, the Alpha-Tocopherol, Beta-Carotene Cancer tions from Hardy-Weinberg equilibrium genotype frequen- Prevention Study (ATBC), the European Prospective Investi- cies among European-ancestry controls and all SNPs with gation into Cancer and Nutrition (EPIC), the Health Profes- minor allele frequency <1%. Any SNP that was missing or sionals Follow-up Study (HPFS), the Physicians Health Study excluded in more than three studies or exhibited large differ- (PHS), the Hawaii-Los Angeles Multiethnic Cohort (MEC), ences in European-ancestry allele frequencies across cohorts and the Prostate, Lung, Colorectal, and Ovarian Cancer (Fst > 0.02) was excluded from further analysis. The