NIH Public Access Author Manuscript Cancer Epidemiol Biomarkers Prev. Author manuscript; available in PMC 2011 October 1.

Transcription

1 NIH Public Access Author Manuscript Published in final edited form as: Cancer Epidemiol Biomarkers Prev October ; 19(10): doi: / EPI A common prostate cancer risk variant 5 of MSMB (microseminoprotein-beta) is a strong predictor of circulating MSP (microseminoprotein) in multiple populations Kevin M. Waters 1, Daniel O. Stram 1, Loic Le Marchand 2, Robert J. Klein 3, Camilla Valtonen-André 4, Mari Peltola 5, Laurence N. Kolonel 2, Brian E. Henderson 1, Hans Lilja 4,6, and Christopher A. Haiman 1 1 Department of Preventive Medicine, Keck School of Medicine, University of Southern California/ Norris Comprehensive Cancer Center, Los Angeles, California 90033, USA 2 Epidemiology Program, Cancer Research Center of Hawaii, University of Hawaii, Honolulu, Hawaii 96813, USA 3 Cancer Biology and Genetics Program, Memorial Sloan-Kettering Cancer Center, New York, NY 10065, USA 4 Department of Laboratory Medicine, Lund University, Skåne University Hospital, Malmö, Sweden 5 Department of Biotechnology, University of Turku, Turku, Finland 6 Departments of Clinical Laboratories, Surgery (Urology), Medicine (GU-Oncology), Memorial Sloan-Kettering Cancer Center, New York, NY 10065, USA Abstract Background Beta-microseminoprotein (MSP) is one of the three most abundantly secreted proteins of the prostate, and has been suggested as a biomarker for prostate cancer risk. A common variant, rs , in the 5 region of the gene which encodes MSP (MSMB), has recently been identified as a risk factor for prostate cancer. Methods We examined the association between rs genotype and MSP levels in a sample of 500 prostate cancer-free men from four racial/ethnic populations in the Multiethnic Cohort (European, African, Latinos, and Japanese ). Generalized linear models were used to estimate the association between rs genotype and MSP levels. Results We observed robust associations between rs genotype and MSP levels in each racial/ethnic population (all P<10 8 ) with carriers of the C allele having lower geometric mea levels (CC/CT/TT genotypes: European, 28.8/20.9/10.0; African, 29.0/21.9/10.9; Latinos, 29.2/17.1/8.3; and Japanese 25.8/16.4/6.7). We estimated the variant accounts for 30 50% of the variation i levels in each population. We also observed significant differences i levels between populations (P= ), with MSP levels observed to be highest in African and lowest in Japanese. Conclusions Rs genotype is strongly associated with plasma MSP levels in multiple racial/ethnic populations. Copyright 2010 American Association for Cancer Research Corresponding authors: Christopher A. Haiman, Harlyne Norris Research Tower, 1450 Biggy Street, Room 1504, Los Angeles, CA 90033, Telephone: (323) , Fax: (323) , haiman@usc.edu, Hans Lilja, Department of Clinical Laboratories, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10065, Telephone: (212) , Fax: (646) , liljah@mskcc.org. Conflict of interest statement: Dr. Hans Lilja holds patents for free PSA assays.

2 Waters et al. Page 2 Impact This supports the hypothesis that rs may be the biologically functional allele. Keywords MSMB; beta-microseminoprotein; prostate; genetic; multiethnic Introduction Methods Study Subjects Beta-microseminoprotein (MSP) has been implicated as a biomarker of prostate cancer risk, detection and prognosis (1,2). MSP, also called prostate secretory protein of 94 amino acids (PSP94), is one of the most highly secreted proteins by the prostate gland (3). Aside from an abundant occurrence of MSMB-transcripts and MSP-protein in the prostate, abundant expression has also been found in trachea and the bronchi in the respiratory tract, and in the gastric mucosa, which may help to explain why circulating levels are quite comparable in women and in men below age 40 (4 7). MSP may have a physiologic role in fertility as it has been shown to bind sperm and inhibit the acrosome reaction (8). In the prostate, MSP expression has been observed to be higher in benign than in cancerous tissue (9). Serum MSP levels have been found to be lower in men with aggressive prostate cancer (1) as well as in men with a prostate cancer recurrence after radical prostatectomy (2,10). Studies in vivo and in vitro have also implicated MSP as having a functional role in regulating cellular growth in the prostate through apoptosis (11). Genome-wide association studies (GWAS) of prostate cancer have identified a common risk variant (rs ) in the 5 region of MSMB on chromosome 10q11.2 (the gene that encodes MSP) (12,13). We recently replicated the association of this variant with prostate cancer risk in a multiethnic population (14). The variant, rs , is located 57 bp upstream of the first exon of MSMB with the C allele (hypothesized low risk allele) found to preferentially bind the CREB transcription factor and to be associated with higher MSMB gene expression in tumor cell lines (15,16). A recent study among Chinese men, in which 60 of 251 cases and 30 of 258 controls were evaluated for the potential association between MSP levels and MSMB genotype, reported serum MSP levels to be lower in those who carried the CT or TT genotype than in those with the CC genotype among the 60 cases (17). These genetic data, together with the clinical and laboratory observations, are consistent with the hypothesis that lower MSMB expression is associated with prostate cancer development. In an attempt to clarify the biological mechanism underlying the association of genetic variation at the MSMB locus and prostate cancer risk, we examined the association between rs genotype and MSP levels in the blood plasma from prostate cancer-free men from four racial/ethnic populations (European, African, Latinos, and Japanese ). We also examined known and putative risk factors for prostate cancer as potential contributors to inter-individual variability of circulating levels of MSP in the study population. This study consisted of male participants of the Multiethnic Cohort (MEC). The MEC includes 96,810 men and 118,441 women and is comprised mainly of African, Japanese, Latinos, Native Hawaiians and European (18). Between 1993 and 1996, adults between 45 and 75 years old were enrolled by completing a 26-page, self-administered questionnaire asking detailed information about demographic factors, personal behaviors, dietary habits and history of prior medical conditions. Potential cohort

3 Waters et al. Page 3 Lab Assays Genotyping members were identified through Department of Motor Vehicles drivers license files, voter registration files and Health Care Financing Administration data files. Between 1995 and 2006, blood specimens were collected from ~67,000 MEC participants for genetic and biomarker analyses. Blood plasma MSP and PSA levels and rs genotype were measured in 500 men from four racial/ethnic groups (n=125 of each European, African, Latinos, and Japanese ). Blood samples of the men included in this study were collected between 1994 and 2004, and were cancer free, as determined through linkage of the cohort with SEER Cancer Registries in California and Hawaii, as of January 1, Men were randomly selected from the control group of the nested prostate cancer casecontrol study in the MEC (19). We limited the study to men in the age range of years old at the time of blood draw as this is the age when men are typically PSA screened for prostate cancer (20). We also limited the study to men with a self-reported BMI (kg/m2) between 19 and 35 in order to avoid subjects with extremely high or low BMI levels where the effects o are not known. Four hundred and twelve (82%) of the men had fasted 8 or more hours before blood collection. The Institutional Review Boards at the University of Southern California and University of Hawaii approved the study protocol. Archived EDTA anti-coagulated blood plasma samples (stored frozen for 3 14 years at 80 C after initial processing within 4 hours from venipuncture) were retrieved and shipped frozen on dry ice to Malmö, Sweden in fall of Analyses of free and total PSA, and MSP were performed blinded to ethnicity or genotyping data in Dr. Lilja s laboratory at the Wallenberg Research Laboratories, Department of Laboratory Medicine, Lund University, Skåne University Hospital in Malmö, Sweden during Production and purification of the polyclonal rabbit anti-msp antibody and seminal plasma MSP have been described elsewhere (21). Biotinylation and Europium labeling of the anti-msp antibody were done as previously described (22). Streptavidin coated microtiter plates were from Kaivogen Oy, Turku, Finland. Assay buffer and wash solution have been described previously (21). The MSP immunoassay of plasma MSP levels, which have previously been shown to not be significantly different from serum MSP levels (4), was done with the AutoDelfia 1235 automatic immunoassay system (Perkin-Elmer Life Sciences, Wallac, Turku, Finland). First, 150 ng of the biotinylated MSP-antibody was attached to the streptavidin coated wells in 100 µl of assay buffer during a 30 min incubation. After two washes, 100 µl of assay buffer and 10 µl of standard or sample material were added to the wells, incubated 1 h and washed twice. The Eu-labelled anti-msp antibody (50 ng) was added in 200 µl of assay buffer. After 1 h incubation, the wells were washed six times before adding 200 µl of the Delfia enhancement solution (Perkin-Elmer Life Sciences, Wallac, Turku, Finland) which was then incubated for 5 min at room temperature with shaking before the time-resolved measurement of the Eu-fluorescence. To measure free and total PSA, we used the dual-label DELFIA Prostatus total/free PSA-Assay (Perkin-Elmer, Turku, Finland) (23), which is calibrated against the WHO 96/670 (PSA-WHO) and WHO 68/668 (free PSA-WHO) standards. Samples were run in 8 batches. Mean coefficients of variation of 7.4%, 11.8%, and 7.8% were observed for MSP, total PSA, and free PSA respectively among 20 blinded duplicate samples (5 per ethnic group). Genotyping of rs and other GWAS-validated risk alleles for prostate cancer was performed using the Taqman allelic discrimination, as previously described (14,19,24). The

4 Waters et al. Page 4 Statistical analysis Results DNA for genotyping was extracted from buffy coat samples using a QIAamp 96 DNA blood kit (Qiagen, Valencia, CA, USA). Of the ~5% duplicate samples used to assess genotyping reproducibility, there was 100% concordance for rs MSP and PSA levels were log transformed to better normalize the distributions and mean levels are presented as least squares geometric means. Generalized linear models were used to examine the association between rs and MSP, adjusted for age (continuous), BMI (continuous), laboratory batch, and race/ethnicity (pooled analysis). Laboratory batch was not significantly associated with MSP levels. We also adjusted for PSA levels, as PSA was found to be modestly correlated with MSP levels. Adjustment for either total, free PSA or percent free PSA provided similar results. The least squares geometric mean MSP levels and the percent change relative to the wild-type (i.e. low risk) rs genotype (CC) was calculated in ethnic-specific analyses. The effect of global European ancestry on the association betwee levels and rs was examined in the admixed racial/ethnic populations (African and Latinos) using ancestry estimates from a previous study (19). Fasting time before blood collection was not included in the analysis because it was not significantly associated with MSP levels and had no effect on the association between rs and MSP. As the determinants of MSP levels are largely unknown, we also examined other known and suspected risk factors for prostate cancer in relation to MSP levels including demographic factors, dietary factors, and established common risk variants from GWAS. More specifically we examined age at the time of blood draw, BMI (kg/m 2 ), weight (kg), height (cm), total PSA, free PSA, percent free PSA, physical activity (hours of vigorous work and exercise), and intake of saturated fat (percentage of calories from saturated fat), red meat (calorie adjusted intake per day; g/kcal/day), lycopene (calorie adjusted intake per day; 100mcg/kcal/day), calcium (calorie adjusted intake per day; mg/ kcal/day), vitamin D (calorie adjusted intake per day; IU/kcal/day), and alcohol (grams per day). We tested the association between an increase of 1 unit of each factor and the percent change in geometric mea levels. Age, BMI, weight, and height were coded as continuous variables. A Bonferroni corrected p-value (based on ~10 independent tests) was used to determine statistical significance in the analysis of each risk factor and MSP levels. We also examined 26 other validated genetic risk variants for prostate cancer in association with MSP levels (12,13,19,24 27). Coefficients of determination (r 2 ) were estimated in multivariable models in order to assess the variance of MSP levels explained by rs genotype and other covariates. The mean ages at the time of blood draw were well matched between racial/ethnic groups; African were slightly younger on average (mean age, 60.2 years) and Japanese slightly older (mean age, 61.5 years; Table 1). There were significant differences in mean BMI among racial/ethnic groups (p= ), with Japanese having the lowest mean BMI (25.0 kg/m 2 ) and Latinos the highest (26.5 kg/m 2 ). There were also significant differences in both weight and height among the groups. We observed significant differences in geometric mea levels adjusted for age, BMI, laboratory batch, and free PSA across racial/ethnic groups (P= ). On average, European had the highest levels (20.4 ng/ml), followed by Latinos (19.7 ng/ml) and African (19.0 ng/ml) with the lowest levels observed for Japanese (14.8 ng/ml). Untransformed mea levels were highest in European (mean 24.2 ng/ml; standard deviation 15.9) and lowest in Japanese (mean 19.2; standard deviation 20.4) (Supplemental Table 1).

5 Waters et al. Page 5 Discussion The risk allele at rs (T allele) was common in all populations and ranged in frequency from 0.36 in Latinos to 0.55 in African. The variant, rs , was strongly and highly statistically associated with MSP levels in each racial/ethnic group (all global P values < ; Figure 1). Compared to men with the CC genotype, men with the CT genotype had a percent change in geometric mean PSA levels between 24.3% (African ) and 41.5% (Latinos) (all P<0.05), and men with the TT genotype had changes ranging from 62.3% in African to 74.2% in Japanese (all P< 10 8 ) (Table 2). We observed little evidence of heterogeneity of the association of genotype and MSP levels between racial/ethnic groups (P=0.37; Table 2). The associations were similar when adjusting for global European ancestry in the admixed African American and Latino populations (Supplemental Table 2). Differences in the frequency of rs genotype could not explain the ethnic differences i levels (global P = ; Table 1). Following adjustment for genotype, MSP levels were highest in African (19.3 ng/ml) and lowest in Japanese (14.1 ng/ml; Table 1). We did not observe evidence that age, BMI, or PSA modified the association between MSMB genotype and geometric mea levels. We found no significant association between age and MSP levels (P=0.10), however the youngest 5-year age group (50 54 years) was noted as having lower least squares geometric mean levels than the older groups (55 69 years; Table 3). We observed inverse associations of MSP levels with BMI (P= ) and weight (P= ) and a suggestive inverse association with height (P=0.072). Least squares geometric mea levels by quintiles of these factors are shown in Table 3. Family history of prostate cancer was positively associated with MSP levels in African (P=0.023), but not in the pooled population (P=0.34). We observed modest, yet, statistically significant positive correlations betwee levels and free (P= ), and total (P= ) PSA, with Pearson correlation coefficients of 0.15 ( among racial/ethnic groups) and 0.13 ( among racial/ethnic groups), respectively. Least squares geometric mea levels by quintiles of the PSA measures are shown in Table 3. Aside from age and height in Latinos the direction of the associations with these factors was consistent across populations. We found no strong associations between suspected risk factors for prostate cancer and MSP levels (Supplemental Table 3). Of the dietary variables, we observed a 0.44% decrease in MSP levels per 1 unit (100mcg/kcal/day) increase in lycopene intake which was only nominally statistically significant (p=0.047; see Methods). In testing 26 validated risk variants (12, 13, 19, 24 27) for association with MSP levels, we observed a nominally significant inverse association with only two variants, rs on chromosome 2p15 (EHBP1 gene locus; 11.5% change i per allele P=0.013), and rs on chromosome 11q13 (7.2% change i per allele P=0.041). In unadjusted models, rs genotype explained between 30% (African ) and 50% (Japanese ) of the variation in log MSP levels. Together with age, BMI, free PSA, and laboratory batch, between 39% (African ) and 61% (Japanese ) of the variation in log MSP levels could be explained. In ethnic-pooled analysis, rs explained 38% of the variance in log MSP levels, while 48% of the variance could be explained when including age, BMI, free PSA, ethnicity, and laboratory batch. In this study, we found that rs genotype is an important and highly statistically significant predictor of circulating MSP levels. The association of genotype and MSP levels was observed in all racial/ethnic groups which supports fine-mapping studies pinpointing rs as the most strongly associated prostate cancer risk variant in the region (15,16).

6 Waters et al. Page 6 We also found MSP levels to vary significantly among racial/ethnic groups, and after conditioning on genotype, African were observed to have the highest levels and Japanese the lowest. The variant is located in a putative CREB transcription factor binding site and the C to T change at this position has been shown to influence CREB binding, with preferential binding observed with the low risk C allele. In cell lines, the C allele has also been associated with increased MSMB expression (16). A recent study in a Chinese Han population found significantly lower circulating serum MSP levels in men with prostate cancer and the CT or TT genotype (16.32 µg/l) compared to men with the CC genotype (19.33 µg/l) (P=0.022) (17). They also noted a similar but non-significant association in 30 unaffected men. In Japanese we observed geometric mean levels of 6.7, 16.4, and 25.8 in men with the TT, CT, and CC genotypes respectively. While our findings are similar in direction, it is difficult to compare the absolute MSP values between the studies since MSP values reported by our new sandwich-type immunoassay are markedly higher (close to three-fold) compared to the competitive assay-design used in prior studies (21), although MSP-values reported by the two different assay designs have excellent correlation (r 2 >0.90). Our observation among unaffected men from multiple populations provides further support for the hypothesis that variant rs is biologically functional, and that the risk for prostate cancer may be conferred by altering expression of MSP. Numerous studies have investigated the genetic basis underlying inter-individual variation in circulating levels of biomarkers with relevance to prostate cancer including PSA, SHBG, testosterone, 3α-androstanediol-glucoronide, and 17β-estradiol (28,29). The genetic variants identified to date in association with these biomarkers however explain only a small fraction of the variation in circulating levels in any population. In contrast, the variant at the MSMB locus alone may account for 30% (African ) to 50% (Japanese ) of the variation in levels in any one population. This finding supports those of Xu et al. in a population of year old Swedish men (paper co-submitted). While rs does explain a large portion of the variance i levels, a portion of the remaining variance may be explained by day-to-day inter-individual variation, environmental risk factors as well as genetic variation both local to the MSMB region and genome-wide. Extensive sequencing and fine-mapping efforts of the MSMB region (30,31) have identified additional local variation that will need to be examined for association with MSP levels in future experiments. In previous analyses of the MEC data (32), we did not replicate reports (33,34) that high lycopene intake or blood measurements may be protective for prostate cancer, and we note here that the suggestive weak inverse association betwee and lycopene intake is not consistent with such protection. However, the association with lycopene seen here (p=0.047) should be interpreted cautiously, as this is not significant after adjusting for multiple tests (Bonferroni corrected α= ). The geometric mea levels were lower in men with more risk alleles for rs which lies in an intron of the EHBP1 gene on chromosome 2, and MSP levels were higher in men with more risk alleles of rs located in a gene desert on chromosome 11. These associations also do not survive correction for multiple comparisons and will require replication. In summary, we observed a strong association between rs genotype and plasma levels of MSP in a multiethnic sample of men without prostate cancer. The association was robust and statistically significant in all four racial/ethnic groups. The variant explained a large proportion of the variability in plasma MSP levels in all groups. Examining MSP in a prospective study will be necessary to determine whether the prostate cancer risk association

7 Waters et al. Page 7 with rs is mediated through influence o levels, as well as the population risk associated with MSP levels. Supplementary Material Acknowledgments References Refer to Web version on PubMed Central for supplementary material. We thank Gun-Britt Eriksson and Mona Hassan Al-Battat for expert assistance with immunoassays. The MEC was supported by National Cancer Institute grants CA63464 and CA This research was supported in part by the National Cancer Institute [P30 CA to MSKCC]; National Cancer Institute Specialized Programs of Research Excellence [P50-CA92629]; Swedish Cancer Society [ ]; Swedish Research Council [Medicine: K X ]; Sidney Kimmel Center for Prostate and Urologic Cancers, and David H. Koch through the Prostate Cancer Foundation, Academy of Finland (Project ), and Fundación Federico SA. 1. Nam RK, Reeves JR, Toi A, et al. A novel serum marker, total prostate secretory protein of 94 amino acids, improves prostate cancer detection and helps identify high grade cancers at diagnosis. J Urol 2006;175: [PubMed: ] 2. Reeves JR, Dulude H, Panchal C, Daigneault L, Ramnani DM. Prognostic value of prostate secretory protein of 94 amino acids and its binding protein after radical prostatectomy. Clin Cancer Res 2006;12: [PubMed: ] 3. Lilja H, Abrahamsson PA. Three predominant proteins secreted by the human prostate gland. Prostate 1988;12: [PubMed: ] 4. Abrahamsson PA, Andersson C, Bjork T, et al. Radioimmunoassay of beta-microseminoprotein, a prostatic-secreted protein present in sera of both men and women. Clin Chem 1989;35: [PubMed: ] 5. Ulvsback M, Lindstrom C, Weiber H, Abrahamsson PA, Lilja H, Lundwall A. Molecular cloning of a small prostate protein, known as beta-microsemenoprotein, PSP94 or beta-inhibin, and demonstration of transcripts in non-genital tissues. Biochem Biophys Res Commun 1989;164: [PubMed: ] 6. Weiber H, Andersson C, Murne A, et al. Beta microseminoprotein is not a prostate-specific protein. Its identification in mucous glands and secretions. Am J Pathol 1990;137: [PubMed: ] 7. Weiber H, Lindstrom C, Lilja H, Bjartell A, Fernlund P. Immunohistochemical and in situ hybridization studies of beta-microseminoprotein in the human gastric mucosa. Histochem J 1997;29: [PubMed: ] 8. Anahi Franchi N, Avendano C, Molina RI, et al. beta-microseminoprotein in human spermatozoa and its potential role in male fertility. Reproduction 2008;136: [PubMed: ] 9. Imasato Y, Xuan JW, Sakai H, et al. PSP94 expression after androgen deprivation therapy: a comparative study with prostate specific antigen in benign prostate and prostate cancer. J Urol 2000;164: [PubMed: ] 10. Bjartell AS, Al-Ahmadie H, Serio AM, et al. Association of cysteine-rich secretory protein 3 and beta-microseminoprotein with outcome after radical prostatectomy. Clin Cancer Res 2007;13: [PubMed: ] 11. Garde SV, Basrur VS, Li L, et al. Prostate secretory protein (PSP94) suppresses the growth of androgen-independent prostate cancer cell line (PC3) and xenografts by inducing apoptosis. Prostate 1999;38: [PubMed: ] 12. Eeles RA, Kote-Jarai Z, Giles GG, et al. Multiple newly identified loci associated with prostate cancer susceptibility. Nat Genet 2008;40: [PubMed: ] 13. Thomas G, Jacobs KB, Yeager M, et al. Multiple loci identified in a genome-wide association study of prostate cancer. Nat Genet 2008;40: [PubMed: ]

8 Waters et al. Page Waters KM, Le Marchand L, Kolonel LN, et al. Generalizability of associations from prostate cancer genome-wide association studies in multiple populations. Cancer Epidemiol Biomarkers Prev 2009;18: [PubMed: ] 15. Chang BL, Cramer SD, Wiklund F, et al. Fine mapping association study and functional analysis implicate a SNP in MSMB at 10q11 as a causal variant for prostate cancer risk. Hum Mol Genet 2009;18: [PubMed: ] 16. Lou H, Yeager M, Li H, et al. Fine mapping and functional analysis of a common variant in MSMB on chromosome 10q11.2 associated with prostate cancer susceptibility. Proc Natl Acad Sci U S A 2009;106: [PubMed: ] 17. Xu B, Wang J, Tong N, et al. A functional polymorphism in MSMB gene promoter is associated with prostate cancer risk and serum MSMB expression. Prostate 2010;70: [PubMed: ] 18. Kolonel LN, Henderson BE, Hankin JH, et al. A multiethnic cohort in Hawaii and Los Angeles: baseline characteristics. Am J Epidemiol 2000;151: [PubMed: ] 19. Haiman CA, Patterson N, Freedman ML, et al. Multiple regions within 8q24 independently affect risk for prostate cancer. Nat Genet 2007;39: [PubMed: ] 20. Schroder FH, Hugosson J, Roobol MJ, et al. Screening and prostate-cancer mortality in a randomized European study. N Engl J Med 2009;360: [PubMed: ] 21. Valtonen-Andre C, Savblom C, Fernlund P, Lilja H, Giwercman A, Lundwall A. Betamicroseminoprotein in serum correlates with the levels in seminal plasma of young, healthy males. J Androl 2008;29: [PubMed: ] 22. Vaisanen V, Peltola MT, Lilja H, Nurmi M, Pettersson K. Intact free prostate-specific antigen and free and total human glandular kallikrein 2. Elimination of assay interference by enzymatic digestion of antibodies to F(ab')2 fragments. Anal Chem 2006;78: [PubMed: ] 23. Mitrunen K, Pettersson K, Piironen T, Bjork T, Lilja H, Lovgren T. Dual-label one-step immunoassay for simultaneous measurement of free and total prostate-specific antigen concentrations and ratios in serum. Clin Chem 1995;41: [PubMed: ] 24. Eeles RA, Kote-Jarai Z, Al Olama AA, et al. Identification of seven new prostate cancer susceptibility loci through a genome-wide association study. Nat Genet 2009;41: [PubMed: ] 25. Gudmundsson J, Sulem P, Rafnar T, et al. Common sequence variants on 2p15 and Xp11.22 confer susceptibility to prostate cancer. Nat Genet 2008;40: [PubMed: ] 26. Gudmundsson J, Sulem P, Steinthorsdottir V, et al. Two variants on chromosome 17 confer prostate cancer risk, and the one in TCF2 protects against type 2 diabetes. Nat Genet 2007;39: [PubMed: ] 27. Sun J, Zheng SL, Wiklund F, et al. Evidence for two independent prostate cancer risk-associated loci in the HNF1B gene at 17q12. Nat Genet 2008;40: [PubMed: ] 28. Ahn J, Berndt SI, Wacholder S, et al. Variation in KLK genes, prostate-specific antigen and risk of prostate cancer. Nat Genet 2008;40: author reply [PubMed: ] 29. Ahn J, Schumacher FR, Berndt SI, et al. Quantitative trait loci predicting circulating sex steroid hormones in men from the NCI-Breast and Prostate Cancer Cohort Consortium (BPC3). Hum Mol Genet 2009;18: [PubMed: ] 30. Yeager M, Deng Z, Boland J, et al. Comprehensive resequence analysis of a 97 kb region of chromosome 10q11.2 containing the MSMB gene associated with prostate cancer. Hum Genet Kote-Jarai Z, Leongamornlert D, Tymrakiewicz M, et al. Mutation analysis of the MSMB gene in familial prostate cancer. Br J Cancer 2010;102: [PubMed: ] 32. Gill JK, Franke AA, Steven Morris J, et al. Association of selenium, tocopherols, carotenoids, retinol, and 15-isoprostane F(2t) in serum or urine with prostate cancer risk: the multiethnic cohort. Cancer Causes Control 2009;20: [PubMed: ] 33. Giovannucci E, Ascherio A, Rimm EB, Stampfer MJ, Colditz GA, Willett WC. Intake of carotenoids and retinol in relation to risk of prostate cancer. J Natl Cancer Inst 1995;87: [PubMed: ]

9 Waters et al. Page Gann PH, Ma J, Giovannucci E, et al. Lower prostate cancer risk in men with elevated plasma lycopene levels: results of a prospective analysis. Cancer Res 1999;59: [PubMed: ]

10 Waters et al. Page 10 Figure 1. The least squares geometric mean of log MSP levels for each racial/ethnic group (European, African, Latinos, and Japanese ) by the genotype of rs (n=500). The error bars represent the upper and lower 95% confidence limits of the mean. The values are adjusted for age (continuous), BMI (continuous), free PSA level, and laboratory batch. P values are from a global ANCOVA test (Type III sum of squares F-test).

11 Waters et al. Page 11 Table 1 Descriptive characteristics by ethnicity (n=500). European African Latinos Japanese p value N Age (yr) 61.1 ± ± ± ± * BMI (kg/m 2 ) 26.1 ± ± ± ± * Weight (kg) 84 ± ± ± ± * Height (cm) 177 ± ± ± ± * MSP * MSP MSP PSA total * PSA total ** PSA free * PSA free ** rs Frequency of T allele * Global test, ANOVA. Global test, ANCOVA Unadjusted crude least squares geometric means. Least squares geometric means adjusted for laboratory batch, free PSA, age, and BMI. Least squares geometric means adjusted for laboratory batch, free PSA, age, BMI, and rs genotype. ** Least squares geometric means adjusted for laboratory batch, MSP, age, and BMI.

12 Waters et al. Page 12 Table 2 Least squares geometric mea levels and percent change by rs genotype. rs Genotype European African Latinos Japanese n Mean * % change P value n Mean * % change P value n Mean * % change P value n Mean * % change P value CC Ref ref ref ref ref ref ref ref CT % % % % TT % % % % r r * Values adjusted for age, BMI, free PSA, and laboratory batch. Univariate model with rs (categorized as genotype class) as independent variable. Multivariable model with rs (categorized as genotype class), age, BMI, and laboratory batch as independent variables.

13 Waters et al. Page 13 Table 3 The association of geometric mea levels with age, anthropometic variables and PSA levels. European African Latinos Japanese Pooled (n=500) Group * Age (yrs) P value BMI (kg/m 2 ) Quintile Quintile Quintile Quintile Quintile P value Weight (kg) Quintile Quintile Quintile Quintile Quintile P value Height (cm) Quintile Quintile Quintile Quintile Quintile P value

14 Waters et al. Page 14 European African Latinos Japanese Pooled (n=500) Group * Total PSA Quintile Quintile Quintile Quintile Quintile P value Free PSA Quintile Quintile Quintile Quintile Quintile P value * Aside for age, the groups showed are quintiles based in the pooled sample. For some covariates, the number of subjects in each quintile is not the same because many subjects share the same value. The cutpoints used are: BMI (<23.13 kg/m 2, , , , ; weight (<156 kg, , , , 205); height (<67 cm, 67 68, 69, 70 71, 72), total PSA (<0.54 ng/ml, , , , 1.85), and; free PSA (<0.18 ng/ml, , , , 0.52). Least squares geometric mean of MSP adjusted for BMI, free PSA, laboratory batch, rs , and race/ethnicity (pooled). Global test, ANCOVA. Least squares geometric mean of MSP adjusted for age, free PSA, laboratory batch, rs , and race/ethnicity (pooled). Least squares geometric mean of MSP adjusted for age, BMI, laboratory batch, rs , and race/ethnicity (pooled).