Identification of Genetic Risk Associated with Prostate Cancer Using

ORIGINAL ARTICLE Identiﬁcation of genetic risk associated with prostate cancer using ancestry informative markers

LJ Ricks-Santi1,2,3, V Apprey2, T Mason2, B Wilson2, M Abbas2,4, W Hernandez5, S Hooker6, M Doura7, G Bonney2, G Dunston2,4, R Kittles8 and C Ahaghotu9

BACKGROUND: Prostate cancer (PCa) is a common malignancy and a leading cause of cancer death among men in the United States with African-American (AA) men having the highest incidence and mortality rates. Given recent results from admixture mapping and genome-wide association studies for PCa in AA men, it is clear that many risk alleles are enriched in men with West African genetic ancestry. METHODS: A total of 77 ancestry informative markers (AIMs) within surrounding candidate gene regions were genotyped and haplotyped using Pyrosequencing in 358 unrelated men enrolled in a PCa genetic association study at the Howard University Hospital between 2000 and 2004. Sequence analysis of promoter region single-nucleotide polymorphisms (SNPs) to evaluate disruption of transcription factor-binding sites was conducted using in silico methods. RESULTS: Eight AIMs were significantly associated with PCa risk after adjusting for age and West African ancestry. SNP rs1993973 (intervening sequences) had the strongest association with PCa using the log-additive genetic model (P ¼ 0.002). SNPs rs1561131 (genotypic, P ¼ 0.007), rs1963562 (dominant, P ¼ 0.01) and rs615382 (recessive, P ¼ 0.009) remained highly significant after adjusting for both age and ancestry. We also tested the independent effect of each significantly associated SNP and rs1561131 (P ¼ 0.04) and rs1963562 (P ¼ 0.04) remained significantly associated with PCa development. After multiple comparisons testing using the false discovery rate, rs1993973 remained significant. Analysis of the rs156113--, rs1963562-- rs615382l and rs1993973--rs585224 haplotypes revealed that the least frequently found haplotypes in this population were significantly associated with a decreased risk of PCa (P ¼ 0.032 and 0.0017, respectively). CONCLUSIONS: The approach for SNP selection utilized herein showed that AIMs may not only leverage increased linkage disequilibrium in populations to identify risk and protective alleles, but may also be informative in dissecting the biology of PCa and other health disparities.

Prostate Cancer and Prostatic Diseases (2012) 15, 359--364; doi:10.1038/pcan.2012.19; published online 17 July 2012 Keywords: ancestry informative markers; African-American; health disparities

BACKGROUND For example, single-nucleotide polymorphisms (SNPs) in Prostate cancer (PCa) is a common malignancy in aging men and a ribonuclease L (RNASEL), vitamin D receptor (VDR) and cyto- leading cause of cancer death among men in the United States. chrome P3A5 (CYP3A5) have been the subject several meta- 2--6 There are three well-established risk factors for PCa: age, ethnicity analyses and have shown inconsistent associations with PCa 7--19 and family history,1 but the molecular mechanisms underlying risk. Previous studies in our laboratories have shown associa- 20--23 its development and progression remain poorly understood. tions with PCa in the aforementioned regions. Possible With regards to ethnicity, the incidence and mortality rate of reasons that may explain these inconsistencies include sample PCa among African-American (AA) men is twofold higher when size, ethnicity and end points. These three loci emerged before the compared with European Americans. genome-wide association studies era and indeed the literature is Recent years have seen significant progress in understanding conflicting regarding possible risk associations. Nevertheless, population-level susceptibility to common cancers. Indeed, the sophisticated efforts to tackle the ‘old’ loci may still be well worth previous candidate gene-based strategy was largely disappoint- conducting. ing, producing conflicting results and publication bias. However, The potential existence of population-specific genetic factors in genome-wide association studies utilizing large numbers of cases PCa was evaluated through the use of ancestry informative and controls, supported by appropriate validation sets, have markers (AIMs) in the aforementioned candidate genes in AA men. produced robust data in the field of PCa genetics. Therefore, we chose a panel of AIMs to explore the role of African

1Division of Genetics, Department of Pediatrics and Child Health, Howard University College of Medicine, Washington, DC, USA; 2Department of Community and Family Medicine, National Human Genome Center at Howard University, Washington, DC, USA; 3College of Medicine, Howard University Cancer Center, Washington, DC, USA; 4Department of Microbiology, Howard University College of Medicine, Washington, DC, USA; 5Department of Medicine, Committee on Cancer Biology, University of Chicago, Chicago, IL, USA; 6Department of Medicine, University of Chicago, Chicago, IL, USA; 7Department of Pharmacology and Physiology, George Washington University, Washington, DC, USA; 8Division of Biological Sciences, Department of Medicine, The University of Chicago, Chicago, IL, USA and 9Department of Urology, Howard University College of Medicine, Washington, DC, USA. Correspondence: Dr LJ Ricks-Santi, Division of Genetics, Department of Pediatrics and Child Health, Howard University College of Medicine, Howard University Cancer Center, National Human Genome Center, 2041 Georgia Avenue, NW #615, Washington, DC 20060, USA. E-mail: [email protected] Received 17 January 2012; revised 13 April 2012; accepted 19 April 2012; published online 17 July 2012 Prostate cancer using AIMs LJ Ricks-Santi et al 360 ancestry in the genetic etiology of PCa in AAs. We hypothesize dominant, recessive and log-additive genetic models were performed. Odds that AIMs may help track PCa risk alleles in specific populations, as ratios (ORs) with 95% confidence intervals (CIs) were calculated and well as provide clues that may help elucidate the mechanisms of adjusting for age. Two-sided P-values of p0.05 were considered as this disease. In this study, we utilized AIMs within an average of 25 statistically significant. Using the Bonferroni test and the false discovery megabases (Mb) in either direction of RNASEL, VDR and CYP3A5 to rate (FDR) adjustment was made for multiple comparisons testing. We also help elucidate the role of ancestry in PCa susceptibility and to tested the independent effect of each SNP by including the most significant identify variants associated with PCa in AA men. SNPs with the use of a backward-selection procedure. If more than one SNP was selected for analysis, a free web-based program SNPstats25 was used as the application then assumes that haplotype analysis is appropriate. To test MATERIALS AND METHODS the combinatorial effects of the SNPs and haplotypes, frequencies were Study population estimated using the implementation of the EM algorithm coded into the 26 This study was approved by the Howard University (HU) Institutional Review haplostats package. Board. Briefly, unrelated men were enrolled at the HU Cancer Center sites for genetic association studies of risk factors for PCa. All PCa cases were between 35 and 93 years of age and were diagnosed within 1 year of RESULTS enrollment. The group of men consisted of 358 unrelated AA men recruited Demographics from the Washington, DC area through the Division of Urology at the HU The mean age of cases was 65.33 (s.d.±9.33) compared with 58.42 Hospital and PCa screening at the HU Cancer Center between 2000 and (s.d.±11.33) years among the controls (Table 1). West African 2004. All control subjects had PSA levels o4.0 ng ml--1 and normal digital ancestry was not significantly different between cases and controls rectal exams. Blood samples were collected from each subject. PCa cases (Table 1). The rate of concordance between duplicate samples was were diagnosed by trans-rectal ultrasound-guided biopsy using standard 499%. All SNPs were in Hardy--Weinberg equilibrium (P40.05). saturation technique. Genotyping results Selection of AIMs for genotyping Region 1q23--32. In the 1q23--32 region, only 1 of the 21 SNPs We selected a total of 77 AIMs as follows: 21 SNPs in 1q23--32, 36 SNPs in interrogated, rs911964, was found to be associated with PCa risk 7q21--34 and 20 SNPs in 12p12--q14. The AIMs were chosen to cover an in our population even after adjusting for age using the dominant average of 25 Mb upstream and/or downstream of RNASEL, VDR and (OR ¼ 2.13, 95% CI: 1.06--4.28; P ¼ 0.032) and log-additive models CYP3A5. Delta between parental populations for all AIMs was 40.30 (OR ¼ 2.00, 95% CI: 1.02--3.93; P ¼ 0.041). After multiple compar- (based on HapMap data) (Supplementary Table 1). isons testing, rs911964 lost its significant association. In silico analysis of rs911964 using Alibaba 2.1 saw disappearance of an Genetic ancestry estimation hepatocyte nuclear factor-1C-binding site associated with the variant C (risk) allele. In order to control for population stratification, West African ancestry was estimated in cases and controls using AIMs. Individual ancestry was determined for each individual using 77 AIMs selected from regions within Region 7q21--34. In the 7q21--34 region, 2 of the 37 SNPs an average of 25 Mb in either direction of RNASEL, VDR and CYP3A5. Global upstream and downstream of CYP3A5 were found to be individual ancestry (% West African and % European) was calculated from associated with PCa risk (Table 2). For SNP rs219821, found in the genotype data using the Bayesian Markov Chain Monte Carlo method the intervening sequence region of 7q21, the dominant model implemented in the program STRUCTURE 2.1.24 These ancestry estimates emerged as significant (OR ¼ 2.01, 95% CI: 1.12--3.51; P ¼ 0.013), were used as covariates in the regression models. after adjusting for age. For SNP rs8177113, which is found in the first intron of the gene Ephrin type-B receptor 6 (EPHB6), the recessive model was the only one to emerge as significant before Genotyping and haplotyping and after adjusting for age (OR ¼ 2.32, 95% CI: 1.08--5.01; Genotyping was performed using pyrosequencing (Qiagen, Germantown, P ¼ 0.028). Notably, haplotype analysis revealed several areas of MD, USA) techniques. Briefly, DNA samples were PCR amplified using high linkage disequilibrium. Specifically, AIMs rs883403 and whole genome amplified DNA, forward and reverse primers, MgCl2, deoxy- rs1011024, rs7779406 and rs8177113, and AIMs rs4987682, nucleotide triphosphates and platinum Taq DNA polymerase (Invitrogen, rs4987677, rs4987657, rs4987649 and rs4987622 were all in Grand Island, NY, USA). PCR products were then pyrosequenced. Results linkage disequilibrium 40.80. Although rs8177113 was in high were analyzed with the PyroMark Q24 software (Qiagen). Duplicate test linkage disequilibrium with rs7779406, there was no association samples and negative controls were included in each 96-well plate. found with the SNP. However, after adjusting each SNP for multiple testing using the Bonferroni test, neither SNP found in In silico analysis the 7q21--34 region remained significantly associated with PCa. In silico analysis of rs8177113 revealed the emergence of a RARalpha, For SNPs found in putative promoter regions or in intervening sequences, specific protein 1 and disappearance of a CREBP transcription in silico methods using AliBaba2.1 (BIOBASE, Beverly, MA, USA) were factor binding sites associated with the variant G (risk) allele. employed to determine if any of the associated SNPs disrupted or resulted Region 12p12--q14. In the 12p12--q14 region surrounding the in the appearance putative transcription binding sites. Specifically, DNA VDR gene, there were no AIMs in linkage disequilibrium. However, sequences were gathered from NCBI (http://www.ncbi.nlm.nih.gov/). five SNPs in this region were found to be associated with PCa risk (Table 2). One of the four SNPs (rs615382) was found in a gene, Statistical analysis Rac GTPase-activating protein 1 (RacGAP1). Age-adjusted ORs for The statistical analyses for the case--control study were done with the rs615382 using the recessive and log-additive models were 0.34 SAS/STAT software, version 9.1 (SAS Institute, Cary, NC, USA) and R statistical (0.16--0.75; P ¼ 0.0062) and 0.66 (0.46--0.94; P ¼ 0.021), respec- software (version 2.9.0) (Vienna, Austria). Allele frequencies in controls were tively. tested for Hardy--Weinberg equilibrium using w2 analysis or Fisher’s exact SNPs rs1993973 and rs1561131 are both found in the intergenic test when appropriate. The association of disease status with genotype, SNP region of 12p11. The unadjusted, adjusted ORs and P-values for combination and haplotype was analyzed by logistic regression. For this rs1993973 and rs1561131 can also be found in Table 2. study, the major allele found in people of African descent was considered Specifically, SNP rs1993973 had the strongest overall association the reference allele. For each SNP genotype, tests using the genotypic, with PCa. After adjusting for age, each genetic model applied was

Prostate Cancer and Prostatic Diseases (2012), 359 -- 364 & 2012 Macmillan Publishers Limited Prostate cancer using AIMs LJ Ricks-Santi et al 361 significant, with the most significant model being the log-additive binding protein of transcription factor-binding sites (Table 3). For model (OR ¼ 0.51, 95% CI: 0.35--0.74; P ¼ 0.0003), followed by the rs585224, in silico analysis using AliBaba2.1, revealed that the recessive (OR ¼ 0.51, 95% CI: 0.35--0.74; P ¼ 0.0016) and dominant variant A (risk) allele resulted in the appearance of a upstream (OR ¼ 0.46, 95% CI: 0.26--0.81; P ¼ 0.01) models. In silico analysis stimulatory factor transcription factor-binding site. revealed that rs1993973 variant G (risk) allele may result in the After multiple comparisons testing using the Bonferroni and emergence of binding sites for transcription factors such as FDR tests, only rs1993973 remained associated with PCa risk specific protein 1 and the disappearance of RNA polymerase (P ¼ 0.0004 and FDR P ¼ 0.03). Additionally, after testing the II-associated protein and NFkappaB (Table 3). SNP rs1561131 independent effect of each significantly associated SNP using the (recessive OR ¼ 0.50, 95% CI: 0.27--0.91; P ¼ 0.02) remained backward selection technique, SNPs rs1561131 (P ¼ 0.04) and statistically significant after adjusting for age. rs1963562 (P ¼ 0.04) were the only SNPs interrogated that SNPs rs1963562 and rs585224 can be found in the 12q13 remained independently significantly associated with PCa. region. SNPs rs1963562 (dominant OR ¼ 0.48, 95% CI: 0.26--0.90; P ¼ 0.01 and log-additive OR ¼ 0.56, 95% CI: 0.31--0.99; P ¼ 0.04) and rs585224 (dominant OR ¼ 0.50, 95% CI: 0.2--0.87; P ¼ 0.01 and Table 3. SNPs found to be associated with PCa risk and variant log-additive OR ¼ 0.61, 95% CI: 0.39--0.95; P ¼ 0.03) also remained implication statistically significant after adjusting for age. In silico analysis also revealed that rs1963562 ancestral C (risk) allele may result SNP ID Chromo- Ancestral Allele Risk Transcription factor in the emergence of an specific protein 1, upstream stimulatory somal allele most allele binding factor and activator protein-1a and disruption CCAAT-enhancer region prevalent in AAs

rs911964 1q25 T T C ÀHNF-1C Table 1. Demographic information for the study population rs8177113 7q21 C C G +SP1, +RARalpha, ÀCRE-BP1 Controls Cases P-value rs1993973 12p12 A G G ÀRAP, ÀNFkappaB, +SP1 Total recruited 197 161 rs1963562 12q13 C C C +SP1, +USF, +AP1, Mean age 58.42±11.33 65.33±9.33 o0.01 ÀC/EBPalpha Family history of PCa 18 (9.1%) 36 (22%) o0.05 rs585224 12q13 G A A +USF ± ± Average PSA 3.11 12.18 85.01 391.58 o0.0001 Abbreviations: AA, African-American; AP1, activator protein 1; HNF-1C, ± (mean s.d.) hepatocyte nuclear factor-1C; PCa, prostate cancer; RAP, RNA polymerase % WA ancestry 80% 80% II-associated protein; SNP, single-nucleotide polymorphism; SP1, speciﬁc Abbreviations: PCa, prostate cancer; WA, West African. protein 1; USF, upstream stimulatory factor.

Table 2. SNPs associated with PCa risk

SNP ID Chromo- Contig Gene Genetic Unadjusted Adjusted some position model Genotype Controls % Cases % OR Lower Upper P- OR Lower Upper P- value value

rs911964 1q25 IVS Dominant TT 129 81.10 125 86.80 1.00 1.00 TC/CC 30 18.90 19 13.20 1.58 0.85 2.95 0.15 2.13 1.06 4.28 0.03 Log-additive 0, 1, 2 1.46 0.80 2.67 2.00 1.02 3.93 0.04 rs219821 7q21 IVS Dominant CC 55 37.70 58 48.70 1.00 1.00 CT/TT 91 62.30 61 51.30 1.50 0.92 2.43 0.10 2.01 1.12 3.51 0.01 rs8177113 7q21 3150468 EPHB6 Recessive CC/CG 160 85.10 140 92.10 1.00 1.00 C4G GG 28 14.90 12 7.90 2.07 1.02 4.23 0.04 2.32 1.08 5.01 0.03 rs1993973 12q13 19070231 IVS Dominant GG 62 38.30 32 25.20 1.00 1.00 G4A GA/AA 100 61.70 95 74.80 0.55 0.33 0.77 0.02 0.46 0.26 0.81 0.00 Recessive GG/GA 141 87.00 92 72.40 1.00 1.00 AA 21 13.00 35 27.60 0.42 0.23 0.77 0.00 0.35 0.18 0.68 0.00 Log-additive 0, 1, 2 130 44.40 163 55.60 0.59 0.42 0.82 0.00 0.51 0.35 0.74 0.00 rs1561131 12q13 22068055 IVS Recessive CC/CT 143 82.20 89 70.10 1.00 1.00 G4A TT 31 17.80 38 29.90 0.53 0.31 0.91 0.02 0.50 0.27 0.91 0.02 rs1963562 12q13 806202 IVS Dominant CC 151 85.30 109 74.20 1.00 1.00 C4T CT/TT 26 14.70 38 25.90 0.50 0.29 0.86 0.01 0.48 0.26 0.90 0.02 Log-additive 0, 1, 2 152 46.10 178 53.90 0.54 0.33 0.90 0.02 0.56 0.31 0.99 0.04 rs615382 12q13 12556217 RACGAP1 Recessive CC/CA 172 92.00126 84.00 1.00 1.00 C4A

AA 15 8.00 24 16.00 0.47 0.24 0.94 0.03 0.34 0.16 0.75 0.01 Log-additive 0, 1, 2 155 45.20 188 54.80 0.75 0.55 1.03 0.08 0.66 0.46 0.94 0.02 rs585224 12q13 IVS Dominant AA 95 67.90 69 53.10 1.00 1.00 AG/GG 45 32.10 61 46.90 0.55 0.34 0.90 0.02 0.50 0.28 0.87 0.01 Log-additive 0, 1, 2 0.68 0.46 1.01 0.05 0.61 0.39 0.95 0.03 Abbreviations: EPHB6, Ephrin type-B receptor 6; IVS, intervening sequence; OR, odds ratio; PCa, prostate cancer; RACGAP1, Rac GTPase-activating protein 1; SNP, single-nucleotide polymorphism.

& 2012 Macmillan Publishers Limited Prostate Cancer and Prostatic Diseases (2012), 359 -- 364 Prostate cancer using AIMs LJ Ricks-Santi et al 362 Table 4. Haplotypes associated with PCa risk adjusted for age

Haplotype rs1561131 rs1963562 rs615382 Frequency Odds ratio (95% CI) P-value

1 (Ancestral) C C C 0.37 1.00 2 T C C 0.29 1.00 (0.58--1.73) 1.00 3 T C A 0.13 0.67 (0.35--1.28) 0.23 4 C C A 0.11 0.84 (0.38--1.86) 0.67 5 T T A 0.05 0.31 (0.11--0.90) 0.03 6 T T C 0.03 1.67 (0.36--7.72) 0.51 7 C T A 0.02 0.67 (0.12--3.70) 0.65 8 C T C 0.01 0.03 (0--820--31) 0.50 Global haplotype association P-value ¼ 0.14

rs1993973 rs585224

1 G A 0.49 1.00 2 A A 0.28 0.53 (0.32--0.90) 0.02 3 (Ancestral) A G 0.15 0.40 (0.23--0.71) 0.00 4 G G 0.08 0.70 (0.28--1.78) 0.46 Global haplotype association P-value ¼ 0.0016

rs911964 rs219821 rs8177113

1 (Ancestral) T C C 0.38 1.00 2 T T C 0.24 1.83 (1.03--3.24) 0.04 3 T C G 0.22 1.91 (1.07--3.39) 0.03 4 T T G 0.08 1.55 (0.70--3.45) 0.28 5 C T C 0.04 1.69 (0.58--4.99) 0.34 6 C C C 0.03 3.87 (1.00--14.94) 0.05 7 C T G 0.01 18 Â 106 o0.0001 Global haplotype association P-value ¼ 0.014

Abbreviations: CI, conﬁdence interval; PCa, prostate cancer.

Haplotype analysis. Haplotype analysis was undertaken for sig- (ORs ¼ 1.83 (1.03--3.24) and 1.91 (1.07--3.39), respectively). The nificantly associated AIMs in the 12p12--q14 region. All haplotypes SNP combination association with PCa is 0.014. were compared with the most frequently found haplotype in AAs. Specifically, SNPs rs1561131, rs1963562 and rs615382 were examined together because for these SNPs the most frequently DISCUSSION found alleles in this population were the ancestral alleles. In this study, we identified one SNP, rs1993973, associated with Notably, these alleles were also found to be the risk alleles in this PCa risk after multiple comparisons testing. However, when not population (C, C and C, respectively) (Table 4). The haplotype adjusting for multiple comparisons testing, SNPs rs1993973 (log- encompassing the combination of protective alleles was one of the additive genetic model P ¼ 0.002), rs1561131 (genotypic, least frequently found haplotype in this population and was P ¼ 0.007), rs1963562 (dominant, P ¼ 0.01) and rs615382 (reces- significantly associated with a decreased risk of PCa (OR ¼ 0.31 95% sive, P ¼ 0.009) were significantly associated with PCa risk after CI: 0.11--0.90; P ¼ 0.032). For AIMs rs1993973 and rs585224, also adjusting for only age and ancestry. Additionally, after testing the found in the aforementioned region, the ancestral alleles (A and G, independent effect of each significantly associated SNP, respectively) were one of the least frequently found in this rs1561131 (P ¼ 0.04) and rs1963562 (P ¼ 0.04) remained signifi- population. When compared with the most frequently found cantly associated with PCa. Additionally, haplotype analysis haplotype, the AG haplotype was also significantly associated with revealed that in region 12p12--q14, the least frequently found decreased risk of PCa (OR ¼ 0.40 95% CI: 0.23--0.71; P ¼ 0.0017) alleles were associated with decreased risk of PCa. (Table 4). AIMs are genetic markers with significant allele frequency To test the combinatorial effect of the SNPs on PCa, the other differences between African and Caucasian populations. Herein, three SNPs on regions 1q and 7q were examined jointly. For AIMs AIMs were used to explore the role of African ancestry in the rs911964 (region 1q23--32), rs219821 (region 7q21--34) and etiology of PCa in AAs and to identify genetic variants associated rs8177113 (region 7q21--34), the ancestral alleles (T, C and C, with PCa in AAs. Of the SNPs evaluated, eight were found to be respectively) were also the most frequently found allele in this significantly associated with risk. Six of the SNPs were found in population. However, the variant alleles were found to be the risk intergenic or non-genic regions of the chromosomes and the alleles in our SNP analysis and remarkably the least frequently other two were found in specific genes. Although tests for found in this population. In addition, analysis of these AIMs multiple comparisons testing, Bonferroni and FDR diminished the revealed that the most frequently found SNP combination com- statistical significance for most SNPs, for exploratory reasons, we prised the most frequently found alleles (TCC frequency ¼ 0.38) are still reporting the clinical implications of such genetic while the least frequently found SNP combination comprised the associations, which warrant additional investigations. variant alleles (CTG frequency ¼ 0.0103). This SNP combination SNPs in RacGAP1 and EPHB6 may be likely candidates for conferred the highest risk (Table 4). SNP combinations TTC and predisposition to PCa given their function and association with TCG were also significantly associated with risk in our population other cancers.27--34 RacGAP1 is part of a family of Rho GTPase

Prostate Cancer and Prostatic Diseases (2012), 359 -- 364 & 2012 Macmillan Publishers Limited Prostate cancer using AIMs LJ Ricks-Santi et al 363 proteins that have a role in management of actin and microtubule identifying clinically insignificant tumors (that is, over detection). dynamics, myosin activity and cell adhesion. Notably, RacGAP1 Nevertheless, the authors recognize the limitations of the negative has been shown to regulate cell migration and motility in cells by (and positive) predictive value of PSA as a screening tool for promoting lamellipodial protrusions in migrating cells.35 One PCa. And actually, PSA velocity is probably a better predictor of could speculate that increased expression of this gene in cancer clinically relevant tumors.39 Despite increasing data to support may lead to increased cell migration and ultimately malignancy lower PSA cutoffs, there are no update ‘standards.’ This preli- and metastases. Remarkably, in hepatocellular cancer, high minary work is an important step toward a better understanding expression of RacGAP1 was found and silencing was associated of AIMs in conferring PCa risk in AA and CEU populations. with inhibited cell migration and invasion.30 Hu et al.27 also Ultimately, the goal of this study was to address the potential demonstrated that expression of RacGAP1 was associated with genetic contributions to the disproportionate PCa incidence rates neuroendocrine development. It is noteworthy that increased in AAs. We showed that the identification of gene variants neuroendocrine cell differentiation is associated with aggressive, disproportionately distributed among populations may be a useful androgen-independent PCa.36 The increased frequency of the method that could help prioritize the selection of SNPs to be rs615382 variant, located in the 50UTR of RacGAP1, in the AA interrogated in studies of health disparities. The latter underscores population (0.67 vs 0.06, AA vs Utah residents with northern and the possible relevance of population-based differences in SNPs in western European acestry (CEU), respectively) may potentially refining the identification of individuals at risk for disease and explain in part why AAs may be susceptible to a more aggressive potentially, disease diagnosis, treatment and prevention in the type of PCa at a younger age compared with Caucasians. However, emergent era of personalized medicine. further functional studies of the consequences of this variant are warranted to test those hypotheses. CONFLICT OF INTEREST For EPHB6, the biology of their association may not be that The authors declare no conflict of interest. clear. The Eph family of receptors is involved in a variety of functions, including mediating numerous developmental pro- cesses, particularly in the nervous system and alterations in the ACKNOWLEDGEMENTS expression of EPH receptors have been observed in several This project was supported in whole or in part with Federal funds from the National 31--34 cancers. Two other SNPs in the intergenic region, rs1993973 Center for Research Resources (NCRR) (UL1RR031975), National Institutes of health and rs1963562, could potentially be associated with the biology of (NIH), through the Clinical and Translational Science Awards Program (CTSA), from PCa given their effects transcription-binding sites in putative the RCMI Program at Howard University (G12 RR003048), Division of Research promoters of downstream genes. However, the effects of these Infrastructure, NCRR, NIH and the Howard University Cancer Center/Johns Hopkins SNPs on actual transcription factor binding will also require Cancer Center Partnership (U54 CA091431), NCI, NIH. functional studies. Regarding haplotypes, the haplotypes investigated herein have Author contributions never been explored elsewhere. Our studies showed that the most LJR-S, VA, TM and GB participated in the design of the study, common haplotypes were associated with risk while the least performed the statistical analysis and helped draft the manuscript. common haplotypes were associated with protection. Conversely, TM, BW, MA, WH, SH and MD carried out the molecular genetic when examining the combinatorial effects of SNPs on different studies, participated in SNP annotation, and helped draft the chromosomal loci, the most frequently found alleles were found manuscript. CA, GD and RK conceived the study, participated in its to be protective. Yet, these results indicate that use of AIMs may design and coordination and helped to draft the manuscript. help identify potential genetic contributions to the disproportionate morbidity and mortality rates in AAs because of PCa, which could provide clues that may help elucidate the mechanisms of REFERENCES this disease. 1 Rodriguez C, Calle EE, Miracle-McMahill HL, Tatham LM, Wingo PA, Thun MJ et al. In assessing these study results, it is important to consider Family history and risk of fatal prostate cancer. Epidemiology 1997; 8: 653--657. the limitations and strengths of the study. Studying PCa in 2 Berndt SI, Dodson JL, Huang WY, Nicodemus KK. A systematic review of vitamin D the AA male population is a very important subject. However, receptor gene polymorphisms and prostate cancer risk. J Urol 2006; 175: there were only a limited number of specimens available for 1613--1623. 3 Ntais C, Polycarpou A, Ioannidis JP. Vitamin D receptor gene polymorphisms and research because of the low volume of men who agreed to risk of prostate cancer: a meta-analysis. Cancer Epidemiol Biomarkers Prev 2003; 12: participate or provide biospecimen in the study herein. Therefore, 1395--1402. for some of the AIMs chosen, the population of 161 cases and 197 4 Kostner K, Denzer N, Mu¨ller CS, Klein R, Tilgen W, Reichrath J. The relevance of controls in this study may have limited our capacity to detect vitamin D receptor (VDR) gene polymorphisms for cancer: a review of the differences. Nonetheless, we are devising ways to increase literature. Anticancer Res 2009; 29: 3511--3536. participation of AA men, such as collection of saliva as an 5 Raimondi S, Johansson H, Maisonneuve P, Gandini S. Review and meta-analysis on alternative to collection of blood, while continuing to recruit for vitamin D receptor polymorphisms and cancer risk. Carcinogenesis 2009; 30: the PCa study. This work will require further investigation of 1170--1180. candidate genes and expansion to include post-genome-wide 6 McCullough ML, Bostick RM, Mayo TL. Vitamin D gene pathway polymorphisms and risk of colorectal, breast, and prostate cancer. Annu Rev Nutr 2009; 29: association studies loci, such as 8q24. 111--132. Additionally, ascertainment of controls status may have resulted 7 Rebbeck TR, Rennert H, Walker AH, Panossian S, Tran T, Walker K et al. Joint effects in the misclassification of men according to disease status. Using a of inflammation and androgen metabolism on prostate cancer severity. --1 PSA o4ngml and normal digital rectal exam as the criteria for Int J Cancer 2008; 123: 1385--1389. defining men as controls from a screening population at the HU 8 Sun J, Turner A, Xu J, Gro¨ nberg H, Isaacs W. Genetic variability in inflammation Cancer Center may have resulted in men with disease being pathways and prostate cancer risk. Urol Oncol 2007; 25: 250--259. included in the control group, introducing misclassification bias 9 Li H, Tai BC. RNASEL gene polymorphisms and the risk of prostate cancer: a meta- into this case--control study. However, during the recruitment analysis. Clin Cancer Res 2006; 12: 5713--5719. period, 4.0 ng ml--1 was still generally accepted as a general cutoff 10 Vaarala MH, Mattila H, Ohtonen P, Tammela TL, Paavonen TK, Schleutker J. The interaction of CYP3A5 polymorphisms along the androgen metabolism pathway point for ‘abnormal’ results. Thus, these patients would not have in prostate cancer. Int J Cancer 2008; 122: 2511--2516. undergone prostate biopsy to rule out cancer. Although Morgan 11 Zhenhua L, Tsuchiya N, Narita S, Inoue T, Horikawa Y, Kakinuma H et al. CYP3A5 37 38 et al. and Moul et al. suggest that lowering PSA cutoffs to 2.5 gene polymorphism and risk of prostate cancer in a Japanese population. Cancer may increase sensitivity of detection, they raise concerns about Lett 2005; 225: 237--243.

& 2012 Macmillan Publishers Limited Prostate Cancer and Prostatic Diseases (2012), 359 -- 364 Prostate cancer using AIMs LJ Ricks-Santi et al 364 12 Zeigler-Johnson C, Friebel T, Walker AH, Wang Y, Spangler E, Panossian S et al. 24 Falush D, Stephens M, Pritchard JK. Inference of population structure using CYP3A4, CYP3A5, and CYP3A43 genotypes and haplotypes in the etiology and multilocus genotype data: linked loci and correlated allele frequencies. Genetics severity of prostate cancer. Cancer Res 2004; 64: 8461--8467. 2003; 164: 1567--1587. 13 Plummer SJ, Conti DV, Paris PL, Curran AP, Casey G, Witte JS. CYP3A4 and CYP3A5 25 Sole X, Guino´ E, Valls J, Iniesta R, Moreno V. SNPStats: a web tool for the analysis of genotypes, haplotypes, and risk of prostate cancer. Cancer Epidemiol Biomarkers association studies. Bioinformatics 2006; 22: 1928--1929. Prev 2003; 12: 928--932. 26 Schaid DJ, Sinnwell JP, Thibodeau SN. Robust multipoint identical-by-descent 14 Risio M, Venesio T, Kolomoets E, Armaroli P, Gallo F, Balsamo A, et al., BECaP mapping for affected relative pairs. Am J Hum Genet 2005; 76: 128--138. Working Group. Genetic polymorphisms of CYP17A1, vitamin D receptor and 27 Hu Y, Ippolito JE, Garabedian EM, Humphrey PA, Gordon JI. Molecular androgen receptor in Italian heredo-familial and sporadic prostate cancers. Cancer characterization of a metastatic neuroendocrine cell cancer arising in the Epidemiol 2011; 35: e18--e24. prostates of transgenic mice. J Biol Chem 2002; 277: 44462--44474. 15 Meyer MS, Penney KL, Stark JR, Schumacher FR, Sesso HD, Loda M et al. Genetic 28 Lu KH, Patterson AP, Wang L, Marquez RT, Atkinson EN, Baggerly KA et al. Selection variation in RNASEL associated with prostate cancer risk and progression. of potential markers for epithelial ovarian cancer with gene expression arrays and Carcinogenesis 2010; 31: 1597--1603. recursive descent partition analysis. Clin Cancer Res 2004; 10: 3291--3300. 16 Martinez-Fierro ML, Leach RJ, Gomez-Guerra LS, Garza-Guajardo R, Johnson-Pais T, 29 Stone II R, Sabichi AL, Gill J, Lee IL, Adegboyega P, Dai MS et al. Identification of Beuten J et al. Identification of viral infections in the prostate and evaluation of genes correlated with early-stage bladder cancer progression. Cancer Prev Res their association with cancer. BMC Cancer 2010; 10: 326. (Phila) 2010; 3: 776--786. 17 Beuten J, Gelfond JA, Franke JL, Shook S, Johnson-Pais TL, Thompson IM et al. 30 Wang SM, Ooi LL, Hui KM. Upregulation of Rac GTPase-activating protein 1 is Single and multivariate associations of MSR1, ELAC2, and RNASEL with prostate significantly associated with the early recurrence of human hepatocellular cancer in an ethnic diverse cohort of men. Cancer Epidemiol Biomarkers Prev 2010; carcinoma. Clin Cancer Res 2011; 17: 6040--6051. 19: 588--599. 31 Surawska H, Ma PC, Salgia R. The role of ephrins and Eph receptors in cancer. 18 Shea PR, Ishwad CS, Bunker CH, Patrick AL, Kuller LH, Ferrell RE. RNASEL and Cytokine Growth Factor Rev 2004; 15: 419--433. RNASEL-inhibitor variation and prostate cancer risk in Afro-Caribbeans. Prostate 32 Tang XX, Zhao H, Robinson ME, Cnaan A, London W, Cohn SL et al. Prognostic 2008; 68: 354--359. significance of EPHB6, EFNB2, and EFNB3 expressions in neuroblastoma. Med 19 Chen L, Davey Smith G, Evans DM, Cox A, Lawlor DA, Donovan J et al. Genetic Pediatr Oncol 2000; 35: 656--658. variants in the vitamin d receptor are associated with advanced prostate 33 Tang XX, Zhao H, Robinson ME, Cohen B, Cnaan A, London W et al. Implications of cancer at diagnosis: findings from the prostate testing for cancer and treatment EPHB6, EFNB2, and EFNB3 expressions in human neuroblastoma. Proc Natl Acad study and a systematic review. Cancer Epidemiol Biomarkers Prev 2009; 18: Sci USA 2000; 97: 10936--10941. 2874--2881. 34 Hafner C, Bataille F, Meyer S, Becker B, Roesch A, Landthaler M et al. Loss of EphB6 20 Bonilla C, Hooker S, Mason T, Bock CH, Kittles RA. Prostate cancer susceptibility expression in metastatic melanoma. Int J Oncol 2003; 23: 1553--1559. loci identified on chromosome 12 in African Americans. PLoS One 2011; 6: 35 Raftopoulou M, Hall A. Cell migration: Rho GTPases lead the way. Dev Biol 2004; e16044. 265: 23--32. 21 Robbins CM, Hernandez W, Ahaghotu C, Bennett J, Hoke G, Mason T et al. 36 Dema A, Raica M, Tudose N. Prognostic significance of neuroendocrine Association of HPC2/ELAC2 and RNASEL non-synonymous variants with prostate differentiation in carcinoma of the prostate. Rom J Morphol Embryol 1996; 42: cancer risk in African American familial and sporadic cases. Prostate 2008; 68: 83--88. 1790--1797. 37 Morgan TO, Jacobsen SJ, McCarthy WF, Jacobson DJ, McLeod DG, Moul JW. Age- 22 Kittles RA, Baffoe-Bonnie AB, Moses TY, Robbins CM, Ahaghotu C, Huusko P et al. specific reference ranges for prostate-specific antigen in black men. N Engl J Med A common nonsense mutation in EphB2 is associated with prostate cancer 1996; 335: 304--310. risk in African American men with a positive family history. J Med Genet 2006; 43: 38 Moul JW, Sesterhenn IA, Connelly RR, Douglas T, Srivastava S, Mostofi FK et al. 507--511. Prostate-specific antigen values at the time of prostate cancer diagnosis in 23 Kittles RA, Baffoe-Bonnie AB, Moses TY, Robbins CM, Ahaghotu C, Huusko P African-American men. JAMA 1995; 274: 1277--1281. et al. CYP3A4-V and prostate cancer in African Americans: causal or con- 39 D’Amico AV, Cote K, Loffredo M, Renshaw AA, Chen MH. Pretreatment predictors founding association because of population stratification? Hum Genet 2002; of time to cancer specific death after prostate specific antigen failure. J Urol 2003; 110: 553--560. 169: 1320--1324.

Supplementary Information accompanies the paper on the Prostate Cancer and Prostatic Diseases website (http://www.nature.com/pcan)

Prostate Cancer and Prostatic Diseases (2012), 359 -- 364 & 2012 Macmillan Publishers Limited