Genome-wide association study of prostate-specific antigen levels identifies novel loci independent of prostate cancer

ARTICLE Received 29 Jul 2016 | Accepted 12 Dec 2016 | Published 31 Jan 2017 DOI: 10.1038/ncomms14248 OPEN Genome-wide association study of prostatespecific antigen levels identifies novel loci independent of prostate cancer Thomas J. Hoffmann1,2,*, Michael N. Passarelli1,*, Rebecca E. Graff1, Nima C. Emami1, Lori C. Sakoda3, Eric Jorgenson3, Laurel A. Habel3, Jun Shan3, Dilrini K. Ranatunga3, Charles P. Quesenberry3, Chun R. Chao4, Nirupa R. Ghai4, David Aaronson5, Joseph Presti5, Tobias Nordström6, Zhaoming Wang7, Sonja I. Berndt7, Stephen J. Chanock7, Jonathan D. Mosley8, Robert J. Klein9,10,11,12, Mridu Middha9,10,11,12, Hans Lilja10,11,12, Olle Melander13, Mark N. Kvale2, Pui-Yan Kwok2, Catherine Schaefer3, Neil Risch1,2,3, Stephen K. Van Den Eeden3,14 & John S. Witte1,2,14 Prostate-specific antigen (PSA) levels have been used for detection and surveillance of prostate cancer (PCa). However, factors other than PCa—such as genetics—can impact PSA. Here we present findings from a genome-wide association study (GWAS) of PSA in 28,503 Kaiser Permanente whites and 17,428 men from replication cohorts. We detect 40 genomewide significant (Po5  10  8) single-nucleotide polymorphisms (SNPs): 19 novel, 15 previously identified for PSA (14 of which were also PCa-associated), and 6 previously identified for PCa only. Further analysis incorporating PCa cases suggests that at least half of the 40 SNPs are PSA-associated independent of PCa. The 40 SNPs explain 9.5% of PSA variation in non-Hispanic whites, and the remaining GWAS SNPs explain an additional 31.7%; this percentage is higher in younger men, supporting the genetic basis of PSA levels. These findings provide important information about genetic markers for PSA that may improve PCa screening, thereby reducing over-diagnosis and over-treatment. 1 Department of Epidemiology and Biostatistics, University of California San Francisco, San Francisco, California 94158, USA. 2 Institute for Human Genetics, University of California San Francisco, San Francisco, California 94143, USA. 3 Division of Research, Kaiser Permanente, Northern California, Oakland, California 94612, USA. 4 Department of Research and Evaluation, Kaiser Permanente Southern California, Pasadena, California 91101, USA. 5 Department of Urology, Kaiser Oakland Medical Center, Northern California, Oakland, California 94612, USA. 6 Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm 17177, Sweden. 7 Laboratory of Translational Genomics, Division of Cancer Epidemiology and Genetics, Department of Health and Human Services, National Cancer Institute, National Institutes of Health, Bethesda, Maryland 20814, USA. 8 Department of Medicine, Vanderbilt University, Nashville, Tennessee 37232, USA. 9 Icahn Institute for Genomics and Multiscale Biology, Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, New York 10029 USA. 10 Departments of Laboratory Medicine, Surgery, and Medicine, Memorial Sloan-Kettering Cancer Center, New York, New York 10065, USA. 11 Nuffield Department of Surgical Sciences, University of Oxford, Oxford OX3 7LD, UK. 12 Department of Translational Medicine, Lund University, Malmö 205 02, Sweden. 13 Department of Clinical Sciences, Lund University, Malmö 205 02, Sweden. 14 Department of Urology, University of California San Francisco, San Francisco, California 94158, USA. * These authors contributed equally to this work. Correspondence and requests for materials should be addressed to S.K.V.D.E. (email: ) or to J.S.W. (email: ). NATURE COMMUNICATIONS | 8:14248 | DOI: 10.1038/ncomms14248 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14248 P rostate specific antigen (PSA) is a blood-based biomarker used for the detection and surveillance of prostate cancer (PCa)1. PCa can cause disruption of the prostate’s cellular architecture, which in turn can result in PSA leaking into circulating blood. However, PSA levels can also be affected by benign prostatic hyperplasia (BPH), local inflammation or infection, prostate volume, age2, and germline genetics. In this regard, PSA is an organ—but not cancer—specific biomarker. PSA screening for PCa has been used for over 20 years, but its use has declined recently because of concerns about overdiagnosis and over-treatment3,4. While PSA levels at mid-life may modestly predict long-term PCa risk5, and high PSA levels are correlated with more aggressive and lethal forms of disease6–8, low PSA levels do not rule out PCa, and high PSA levels have a low predictive value for PCa9. In the Prostate, Lung, Colorectal, Ovarian (PLCO) Cancer Screening Trial, which had substantial crossover, there was no appreciable reduction in mortality directly related to PSA screening10. Another randomized trial, however, showed that PSA screening may reduce PCa mortality11. Between 20 and 60% of PSA-screened PCas are estimated to be over-diagnoses, and non-aggressive PSA-detected PCas are often treated with therapy that may involve substantial side effects12,13. The value of PSA screening may be higher among individuals defined by particular characteristics, such as family history of PCa, ethnicity, age, and genetic factors. PSA is a glycoprotein enzyme encoded by kallikrein-3 (KLK3) on chromosome 19, but evidence from genetic association studies suggests that PSA levels are a complex polygenic trait, influenced by several different genes. Determining the genetic basis of PSA levels unrelated to cancer may help increase both the sensitivity and specificity of screening for PCa by adjusting PSA levels for constitutive germline genetics. Doing so could improve PSA screening performance. Clinicians could more accurately decide who should have a prostate biopsy, thereby reducing unnecessary procedures and their associated morbidities, as well as decreasing overdiagnosis14,15. Twin studies estimate that 40–45% of the variation in PSA levels can be explained by inherited factors16,17. However, the single-nucleotide polymorphisms (SNPs) that have been identified thus far14,18–24 only explain a limited percentage of the variation in PSA levels (4.2% in 4,620 subjects from Iceland, and 11.8% in 454 subjects from the UK)14. In addition, several of the loci that harbor SNPs associated with PSA levels also harbor SNPs associated with PCa, making it complicated to disentangle genetic effects on PSA levels versus PCa. PSA level associations with PCa risk variants may reflect a number of factors, including: (1) true disease-related increases in PSA levels; (2) the use of PSA levels to restrict controls in case-control studies of PCa; and/or 3) non-cancer related PSA levels that prompt additional biopsy screening (Supplementary Fig. 1). One study reported14 that correcting PSA levels using four PSA-associated variants reclassifies 3.0% of individuals as needing biopsies and 3.0% as not needing biopsies. It did not, however, assess wheth (...truncated)