Cancer and Prostatic Diseases (2005) 8, 95–102 & 2005 Nature Publishing Group All rights reserved 1365-7852/05 $30.00 www.nature.com/pcan Association between hormonal genetic polymorphisms and early-onset prostate cancer

MS Forrest1,2, SM Edwards3, R Houlston3, Z Kote-Jarai3, T Key4, N Allen4, MA Knowles1, F Turner2, A Ardern-Jones5, A Murkin5, S Williams3, R Oram3, CR-UK/BPG UK prostate cancer study collaborators6, DT Bishop2,7 & RA Eeles3,5,7* 1Cancer Research UK Cancer Medicine Research Division, University of Leeds, Leeds, UK; 2Cancer Research UK Genetic Epidemiology Division, University of Leeds, Leeds, UK; 3Institute of Cancer Research, Sutton, Surrey, UK; 4Cancer Research UK Epidemiology Unit, University of Oxford, Oxford, UK; and 5Royal Marsden NHS Trust, Sutton, Surrey, UK

We investigated the association between seven polymorphisms in four candidate involved in vitamin D and with early-onset prostate cancer (CaP) risk. The polymorphisms were genotyped in 288 UK males who were diagnosed with CaP at the age of 55 y or younger and up to 700 population-based controls. An additional 50 cases (not selected for age) and 76 controls were also genotyped. Short (p22 repeats) AR (CAG)n repeats were associated with a significantly reduced risk of early onset CaP (OR 0.68, 95% CI 0.50–0.91) compared with men with long (422) repeats. Men homozygous for the leucine variant of SRD5A2 p.89V4L were also found to be at a significantly increased risk of CaP compared with men who were homozygous for the valine allele (OR 1.84, 95% CI 1.15–2.98). No associations were found with the AR (GGC)n, CYP17 Msp A1 I, VDR Taq I, SRD5A2 (TA)n and p.49A4T polymorphisms and CaP risk. These findings suggest that common polymorphisms in the AR and SRD5A2 genes may be associated with early-onset CaP in British men. Prostate Cancer and Prostatic Diseases (2005) 8, 95–102. doi:10.1038/sj.pcan.4500785 Published online 15 February 2005

Keywords: case–control study; genetic polymorphisms; early onset

Introduction location (eg, Japanese-Americans have a higher risk than native Japanese)3–5 and family history.6–8 After lung cancer, prostate cancer (CaP) is the most Hormones are necessary for the growth and develop- common cause of male cancer mortality in the United ment of the prostate gland and have been implicated in Kingdom, accounting for approximately 8973 cancer the development of CaP. It has therefore been suggested deaths in 2002 alone.1 The American Cancer Society that genetic polymorphisms that encode for estimates that there will be 230 110 new cases of CaP and involved in hormone biosynthesis and metabolism or 29 900 CaP deaths in the US in 2004.2 However, little is receptor function may be related to cancer risk. Two such known about the causes of the disease, the only hormonal pathways include the vitamin D receptor established risk factors being age,1 ethnicity/geographical pathway, which maintains calcium homeostasis and exerts antiproliferative and prodifferentiative effects on many cell types including the prostate, and the androgen receptor pathway, which acts to stimulate prostate cell *Correspondence: RA Eeles, Translational Cancer Genetics Team, growth directly (see Figure 1). Institute of Cancer Research, Sutton, Surrey SM2 5PT, UK. The aim of this study is to investigate the association E-mail: [email protected] between several common polymorphisms in genes 6List available on request. 7These authors contributed equally to this work. involved in these two pathways with early-onset CaP Received 1 September 2004; revised 17 December 2004; accepted 17 risk. The roles of the genes under investigation are December 2004; published online 15 February 2005 illustrated in Figure 1 and include: CYP17 (cytochrome Hormonal genetic polymorphisms and early-onset prostate cancer MS Forrest et al

96 14–19 repeats. Two studies have also suggested a link to 20,21 earlier onset CaP for smaller (CAG)n polymorphisms. The association between the (GGC)n polymorphism and CYP11a1 CaP risk is more inconsistent, with both short and long repeats being found to predispose to CaP.15,17,22,23 HSD3B2 Progesterone A polymorphism (rs743572) has been identified in the CYP17 that contains a single base T4C transition in 0 CYP17 CYP17 the 5 UTR of exon 1, which creates a novel Sp-1 type (CCACC) motif and a Msp A1 I restriction site (commonly denoted the A2 allele).24 Although this Sp-1

17α-Hydroxypregnenolone HSD3B2 17α-Hydroxyprogesterone motif does not appear to influence Sp-1 binding,25 the variant allele might result in increased transcriptional CYP17 CYP17 activity, and hence increased androgen production. This polymorphism has been associated with an increased risk of polycystic ovaries and male premature pattern Dehydroepiandrosterone HSD3B2 Androstenedione baldness,24 although the evidence that this polymorph- ism is related to breast cancer26,27 or CaP is inconsis- HSD17B1 HSD17B1 tent.28,29 Several SRD5A2 polymorphisms have been identified, the most studied of which include a TA dinucleotide ∆5-Androstenediol HSD3B2 repeat and the single nucleotide polymorphisms (SNPs) A49T and V89L. Davis and Russell30 originally SRD5A2 reported three different alleles of the TA repeat in the 3’UTR, 0, 9 and 18 repeats, with 96% of alleles being 30 assigned (TA)0. The observation that the rare (TA)18 30 1,25 (OH) D3 allele was limited to African-American populations, 2 who have a higher rate of CaP than Asian-American or Caucasian men, suggested that a long allele length may be associated with increased risk. However, case–control studies in other ethnic populations have not found the longer variant alleles to be associated with an increased CaP risk.31–34 A mis-sense substitution in SRD5A2 has Vitamin D Receptor Androgen Receptor been identified that substitutes alanine with threonine at codon 49 (A49T (rs9282858)). This polymorphism results in a variant protein whose in vitro enzymic activity is about five-fold higher than normal.35 Despite the low prevalence of this polymorphism (frequency of Prostate Cell Division Prostate Cell Division Decreased Increased 2%), it has been associated with an increased risk of CaP in African-American and Hispanic men35 and Figure 1 Hormonal genes involved in the regulation of prostate cell also with advanced disease in a case series of division. We investigated polymorphisms in CYP17, SRD5A2, the prostate cancers.36 However, studies in Caucasian men androgen receptor and the vitamin D receptor genes. have generally not found this polymorphism to be associated with an increased CaP risk in Caucasian men.33,37,38 A second mis-sense substitution in SRD5A2, V89L P-450 17a-hydroxylase/17,20 lyase), a gene that (rs523349), identified by Makridakis,39 substitutes ami- encodes an enzyme involved in testosterone synthesis no-acid 89 from a valine to a leucine and is much more within the testes;9 SRD5A2 (steroid 5a reductase type II), common than A49T at around 30% frequency.36,40,41 The a gene that encodes for the enzyme which converts variant L allele has been associated with reduced 5a- testosterone to dihydrotestosterone (DHT), its more reductase type II enzyme activity in vitro41 and also with active metabolite within the prostate;10,11 the AR (andro- lower circulating A-diol-g levels, a marker of 5a- gen receptor), which binds testosterone and DHT and reductase type II activity and intraprostatic androgen upregulates prostate cell growth;12 and the VDR (vitamin activity in adult men.39–41 However, this polymorphism D receptor), which binds vitamin D and downregulates has not been found to be significantly associated with prostate growth.13 Two AR gene polymorphisms have CaP risk among Caucasian men.28,40 been extensively investigated with regard to CaP risk The VDR gene encodes for the receptor that binds and include a polyglutamine (CAG) repeat and a vitamin D (1,25-dihydroxyvitamin D3), a hormone polyglycine (GGC) repeat, both of which are found in involved in prostate cell growth. The TaqI polymorphism the first exon encoding the N-terminal modulatory in codon 352 (rs731236) is silent, but has been shown to domain of the AR (the domain that binds to androgen be associated with both circulating levels of osteocalcin, response elements to regulate ). Previous a VDR-regulated protein42 and bone density in osteopo- epidemiological studies have shown that men with rotic women.43 The TaqI site is in linkage disequilibrium relatively short AR polyglutamine (CAG)n repeats are with a poly-A microsatellite polymorphism, with the at an increased risk of developing CaP or more t (presence of TaqI restriction site) and short poly-A aggressive CaP compared with men with longer (A14–A17, long poly-A ¼ A18–A22) almost always

Prostate Cancer and Prostatic Diseases Hormonal genetic polymorphisms and early-onset prostate cancer MS Forrest et al coinciding.44,45 However, the epidemiological evidence genotyping. Cases and controls were recruited from 97 that these polymorphisms are associated with a reduc- Oxfordshire, West Berkshire and Leeds as described tion in CaP risk is inconsistent.44–48 The aim of this study elsewhere52 and DNA was available for 50 cases and 76 is to test the hypotheses that the CYP17 A2 allele, the controls. Age of diagnosis ranged from 55 to 75 y with a SRD5A2 V89L LL genotype and the VDR tt genotype are median age of 69.0 y. One case had a father with CaP and associated with a reduction in CaP risk, and that one case had a brother with CaP. shorter AR CAG and GGC repeats, and the SRD5A2 A49T T allele and (TA)9 allele are associated with an increased CaP risk in a unique population of early-onset (diagnosed aged 55 y or younger) CaP Genotyping of polymorphisms cases, and a smaller, unrestricted, pilot case–control population. DNA was collected and extracted by standard methods, for example, Edwards et al.23 The AR CAG and GGC repeat polymorphisms of the cases and control set A samples were genotyped as described in Edwards et al.23 Materials and methods The CYP17 Msp A1 I polymorphism was genotyped as described in Feigelson et al26 in the cases and control sets Subjects A and B. The SRD5A2 TA repeats of the cases and control sets A and B were amplified using 25 ng of DNA with Case patients were identified from the Cancer previously published primers (50-GCTGATGAAAA Research UK/British Prostate Group UK Familial CTGTCAAGCTGCTGA-30 and 50-GCCAGCTGGCAGA Prostate Cancer Study, as described elsewhere.49 DNA ACGCCAGGAGAC-30)30 at a concentration of 0.32 mM was available from 267 early-onset (aged 55 y or young- each in a total volume of 15 ml, with 0.5 units Amplitaq er) CaP cases from across the UK who were predomi- Gold (Applied Biosystems, Foster City, CA, USA) in nantly identified by clinical symptoms rather than by 1 Â PE Buffer II (Applied Biosystems), 2 mM MgCl2 and screening. Of these, 28 cases were from CaP families 1 mM total dNTP. PCR conditions used were: 951C for (families containing one or more of: (a) at least three 10 min followed by 35 cycles of 941C for 30 s, 641C for first-degree relatives with CaP; (b) three generations in 1 min, 721C for 30 s, with a final extension step of 721C either the proband’s paternal or maternal lineage; for 10 min. Products were run in 2% agarose gels and (c) two affected first-degree relatives where one of the alleles called by length of PCR product (98 bp for 0 TA relatives was age o55 y). The youngest age of diagnosis repeats, 116 bp for 9 TA repeats and 134 bp for 18 TA was 32 y, with a mean age of 51.1 y and a median age repeats). of 51.9 y. Minisequencing was used to genotype the SRD5A2 Population controls were taken from three previous p.49 A4T polymorphism in the cases and control sets A studies. The first control set (set A) was comprised of 149 and B. In all, 25 ng of genomic DNA were amplified in a male and 135 female spouses of patients with colorectal total volume of 15 ml, with 0.45 U Amplitaq Gold cancer collected from Bristol, Wessex and the Thames (Applied Biosystems) in 1 Â PE Buffer II (Applied area. These individuals were of Caucasian origin and Biosystems), 2 mM MgCl2, 0.8 mM total dNTP and had no personal or family history of cancer.50 The second 0.12 mM each of primers 50-AAGCACACGGAGATCCT set of population controls (set B) were 206 Caucasian GAATCCGGCTGCTACC-30 and 50-GGGCACCGC females from the Oxford, Leeds and South London area, GAAGGAAGGCCGCTCCTGCAGG-30. PCR conditions who were part of a breast cancer case–control study used were: 951C for 10 min followed by 40 cycles of 941C conducted by the UK National Case–Control Study for 30 s, 601C for 30 s, 721C for 20 s, with a final extension Group (UK National Case–Control Study Group, 1989). step of 721C for 10 min. PCR products were cleaned by All the women were cancer-free, randomly selected via sodium acetate precipitation and resuspended in 10 ml general practice surgeries and aged up to 45 y. This set H2O. Minisequencing was carried out using the SNaP- had previously been used as a control population for AR shot Ready Reaction Mix (Applied Biosystems) accord- (CAG)n and (GGC)n repeat length polymorphisms and ing to the manufacturer’s instructions and either primer CaP risk.23 Controls (set C) for the SRD5A2 V89L 50-CGGCGGCTACCCGCCTGCCA-30 (forward) or pri- polymorphism were derived from a previous study that mer 50-CAGGAACCAGGCGGCGCGGG-30 (reverse). Re- investigated the association between this polymorphism actions using forward and reverse minisequencing and hormone levels as the RFLP test used was not primers were loaded in alternate lanes on ABI prism successful for control sets A and B.51 This set of controls 377 DNA sequencers at 5 min intervals. As the inter- comprised over 600 Caucasian males selected from the rogated base was different depending on the DNA Oxford UK component of the European Prospective strand investigated (G/A forward, C/T reverse) it was Investigation into Cancer and Nutrition (EPIC) as possible to run 2–3 samples per lane without the risk of described elsewhere51 and was not available for use lane miscalling. Variant samples were confirmed using with the other polymorphisms investigated. The three both minisequencing primers. sets of controls provided a source of data for allele SRD5A2 V89L genotypes were determined in the cases frequencies in the UK Caucasian population for compar- and the EPIC controls as described by Lunn et al.28 VDR ison to our early-onset CaP cases. The use of female Taq 1 genotypes were determined in the cases and control controls meant that we had data on up to 341 extra AR sets A and B as described in Taylor et al.46 repeat alleles. All PCR reactions were performed in Hybaid Touch- Samples from a small case–control study of diet and down and Applied Biosystems GeneAmp PCR System CaP were also used as a pilot population for the 9700 thermocyclers.

Prostate Cancer and Prostatic Diseases Hormonal genetic polymorphisms and early-onset prostate cancer MS Forrest et al 98 Statistical methods Results Logistic regression was performed using STATA 6.0 The CYP 17 Msp A1 I, VDR Taq I and SRD5A2 V89L, (STATA Corporation) to provide odds ratios (OR), 95% A49 T and (TA)n polymorphisms were in Hardy–Wein- confidence intervals (CI) and associated P-values for the berg equilibrium for all control groups both individually risk of each polymorphism with CaP. The genotype and combined and for both sexes of control set B. Since frequencies for each polymorphism between cases and genotype frequencies were similar between the different controls were tested using standard w2 tests. The control groups for all polymorphisms, the control groups heterogeneity of polymorphisms such as the AR CAG were combined in subsequent analyses. The proportion repeat means that the large number of alleles, often at of controls that possessed the variant alleles in the genes low frequencies, greatly reduces the power of simple under investigation ranged from 8 to 48%. In all, 264/795 2 contingency table-based statistical tests such as w . The (33%) of control AR (CAG)n repeat lengths were greater 53 CLUMP program, which uses a Monte Carlo approach than 22 repeats and 41% of AR (GGC)n repeat lengths to determine the significance of allele distributions were greater than 16 repeats (see Tables 1 and 2). For the (http://www.smd.qmul.ac.uk/statgen/dcurtis/software. VDR Taq I polymorphism, 50% possessed one variant t html), was therefore used to investigate the association allele, and 17% possessed two variant alleles. For the between any one genotype or group of genotypes and SRD5A2 polymorphisms, 40% possessed one variant L CaP risk. allele, and 8% possessed two variant L alleles. For the

Table 1 Involvement of polymorphisms in the androgen and vitamin D receptors in early onset CaP risk Control set A Control set B Controls Cases (%) OR (95% CI) Pilot controls Pilot cases OR (95% CI) (%) (%) (%) (%) (%)

AR (CAG)n p22 272 (67) 259 (66) 531 (67) 151 (58) 0.68 (0.50–0.91) 422 133 (33) 131 (34) 264 (33) 111 (42) 1.0

AR (CAG)n 21/22 151 (37) 119 (31) 270 (34) 66 (25) 1.0 o21/422 254 (63) 271 (69) 525 (66) 196 (75) 1.53 (1.10–2.13)

AR (GGC)n p16 244 (60) 168 (59) 412 (60) 158 (61) 1.06 (0.78–1.44) 47 (62) 32 (64) 1.10 (0.52–2.30) 416 165 (40) 116 (41) 282 (40) 102 (39) 1.0 29 (38) 18 (36) 1.0

VDR Taq I TT 93 (35) 51 (29) 144 (32) 90 (34) 0.91 (0.65–1.28) 20 (30) 15 (33) 0.68 (0.23–2.01) Tt 130 (49) 94 (53) 224 (50) 128 (49) 0.93 (0.59–1.46) 38 (57) 23 (50) 0.84 (0.26–2.70) tt 45 (17) 31 (18) 76 (17) 44 (17) 1.0 9 (13) 8 (17) 1.0

The AR (CAG)n polymorphism was shown to be associated with early onset CaP risk (data shown in bold).

Table 2 Involvement of polymorphisms in the steroid 5-alpha-reductase type II and cytochrome P450 17A1 in early-onset CaP risk Control set A Control set B Combined Cases OR Pilot ctrls Pilot cases OR (%) (%) controls (%) (%) (95% CI) (%) (%) (95% CI)

CYP17 Msp A1 I A1/A1 112 (41) 73 (39) 185 (40) 98 (37) 1.0 23 (31) 20 (42) 1.0 A1/A2 130 (47) 92 (49) 222 (48) 132 (50) 1.12 (0.81–1.55) 43 (57) 20 (42) 0.53 (0.24–1.19) A2/A2 32 (12) 23 (12) 55 (12) 32 (12) 1.10 (0.67–1.81) 9 (12) 8 (17) 1.02 (0.33–3.15)

SRD5A2 A49T AA 252 (93) 183 (95) 435 (94) 250 (94) 1.0 65 (90) 47 (100) AT 20 (7) 10 (5) 30 (6) 15 (6) AT+TT:a 7 (10) 0 TT 0 0 0 1 (0.4) 0.93 (0.50–1.74) 0 0

SRD5A2 (TA)n 0/0 222 (80) 154 (75) 376 (78) 195 (76) 1.0 59 (78) 39 (78) 1.0 0/9 51 (18) 45 (22) 96 (20) 57 (22) 0/9+9/9:a 15 (20) 10 (20) 0/9+9/9:a 9/9 3 (1) 5 (2) 8 (2) 3 (1) 1.11 (0.77–1.60) 2 (3) 1 (2) 0.98 (0.41–2.31)

SRD5A2 V89L EPIC Controls (%) VV 317 (52) 126 (48) 1.0 41 (55) 18 (37) 1.0 VL 245 (40) 98 (38) 1.01 (0.74–1.38) 31 (41) 25 (51) 1.84 (0.86–3.95) LL 49 (8) 36 (14) 1.84 (1.15–2.98) 3 (4) 6 (12) 4.56 (1.02–20.27)

There was a significant association between individuals homozygous for the SRD5A2 89L allele and early-onset CaP risk (data shown in bold). aHeterozygotes and homozygotes for rare alleles were combined for analyses.

Prostate Cancer and Prostatic Diseases Hormonal genetic polymorphisms and early-onset prostate cancer MS Forrest et al 99 p.A49A4T polymorphism, no controls and only one 0.25 case was homozygous for the variant allele, and so these were grouped together with the heterozygous carriers, which comprised 6% of the population. Similarly, 0.2 the prevalence of the (TA)n repeat polymorphism was low, with eight controls and three cases possessing the 9/9 genotype, and these individuals were also grouped with the heterozygotes for analysis, which com- 0.15 prised 22% of the population. For the CYP17 Msp A1 I polymorphism, 48% of the controls possessed one allele, and 12% possessed two copies of the variant Frequency 0.1 allele. Since the average control AR CAG repeat number was 21.6, repeat lengths p22 were classified as ‘short’, and repeat lengths longer than 22 CAG repeats were 0.05 classified as ‘long’. Compared with men with long (422) CAG repeats, men with short (p22) AR CAG repeats had a significantly reduced risk of early onset 0 CaP (OR 0.68, 95% CI 0.50–0.91, P ¼ 0.007), as shown in 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 Table 1. Each additional CAG repeat was associated with CAG repeat number an increased risk of 4% (95% CI, À1 to 10%), but this was Figure 2 Frequencies of AR (CAG)n repeat polymorphisms in UK early- not statistically significant (P ¼ 0.09). The strongest onset CaP cases (white bars) and population controls (black bars). Logistic association between (CAG) repeat length and CaP risk regression showed a significant decrease in early-onset CaP risk associated was found for men with (CAG)21 repeat lengths, as with p22 CAG repeats (OR 0.68, 95% CI 0.50–0.91). Linear regression derived from CLUMP analysis, based on 1000 simula- analysis estimated a 4% increase in CaP risk for each extra CAG repeat tions (w2 of 6.6, which was achieved 87 times in 1000 (OR 1.04, s.e. 0.027, t ¼ 1.69, P4|t| 0.09, 95% CI 0.99–1.10). CLUMP analysis suggested that the strongest association with CaP risk was with simulations (P ¼ 0.088 after correction for multiple 2 2 (CAG)21 which gave a w of 6.6 (P ¼ 0.088, uncorrected P ¼ 0.01). A w testing) and corresponded to an uncorrected P-value of test of (CAG)21/22 vs (CAG)o21/422 gave an OR of 1.53 (95% CI 1.10– 0.01). The most common control (CAG)n repeat lengths 2.13, P ¼ 0.008), suggesting that both long and short AR CAG repeats were 21 and 22 (see Figure 2), so it was decided to group predispose to early-onset CaP. these together and see if, as suggested by the CLUMP analysis, having a middling CAG repeat length had an 2 effect on CaP risk. A w test of (CAG)o21/422 vs (CAG)21/ possessed an SRD5A2 49T allele, and a risk estimate was 22 gave an OR of 1.53 (95% CI 1.10–2.13, P ¼ 0.008), therefore not calculated for this population. suggesting that both long and short AR CAG repeats predispose to early onset CaP. No association with early onset CaP risk was found for the AR (GGC)n repeat length polymorphism or the VDR polymorphism. Discussion Associations between the SRD5A2 p.89V4L, SRD5A2 A49T and CYP17 Msp A1 I polymorphisms and early- This is the largest study to investigate the association onset CaP risk are shown in Table 2. Men homozygous between polymorphisms in a range of candidate genes for the SRD5A2 V89L leucine variant were found to be at and early-onset CaP. This study has found that short almost twice the risk of developing early-onset CaP (p22 repeats) AR (CAG)n was associated with a compared with men who were homozygous for the statistically significant 32% reduced risk of early onset valine variant (OR 1.84, 95% CI 1.15–2.98, P ¼ 0.012). CaP and that the L/L genotype of the SRD5A2 V89L There was no significant difference between valine polymorphism was associated with a statistically sig- homozygotes and valine/leucine heterozygotes (OR nificant 85% increased risk of early-onset CaP. There was 1.01, 95% CI 0.74–1.38, P ¼ 0.97). An increase in risk no evidence that other polymorphisms including the AR associated with the L allele was also observed in the pilot (GGC)n, CYP17 Msp A1 I, VDR taq I, SRD5A2 A49T or the study population. Men who were homozygous for the SRD5A2 (TA)n were associated with early-onset CaP in leucine variant had a four-fold increased risk compared this study population. with men who were homozygous for the valine variant Our finding that short (p22 repeats) AR (CAG)n was of SRD5A2 V89L (OR 4.56, 95% CI 1.02–20.27, P ¼ 0.046). associated with a significant reduction in the risk of Men who had a V/L genotype were at an intermediate early-onset CaP are unexpected since previous case– risk (OR 1.84, 95% CI 0.86–3.95, P ¼ 0.12), but this control studies have found shorter CAG repeats to difference was not statistically significant from that of predispose towards CaP and in particular, aggressive the valine homozygotes. CaP14–19 and earlier age at diagnosis.20,21 No significant associations with CaP risk were found The precise role of the AR and other genes in prostate for, SRD5A2 (TA)n, SRD5A2 A49T and CYP17 Msp A1 I growth is not known. One explanation for our finding (Table 2). One control individual was homozygous for could be that a gene predisposing to CaP may interact the T variant allele and was therefore included in the AT with or be influenced by the AR gene. Men with genotype group for analysis. Excluding the 28 cases mutations in this gene and longer CAG repeat lengths known to be from CaP families did not alter the findings may be predisposed to early-onset CaP; in men without for any of the polymorphisms investigated (see Supple- the mutation, shorter CAG repeats may be associated mental Tables S1 and S2). In the pilot study, no cases with an increased risk. This may explain why previous

Prostate Cancer and Prostatic Diseases Hormonal genetic polymorphisms and early-onset prostate cancer MS Forrest et al

100 studies have found a consistent increase in CaP for short did not account for age at all. Secondly, rates of CaP vary repeat lengths in men with a late age of onset. Indeed, the between different countries5 and differences in genetic, CLUMP analysis found both shorter CAG repeats and dietary and/or environmental factors between popula- longer CAG repeats to be associated with an increased tions might account for our findings. However, few risk CaP, compared with average repeat lengths. Longer factors for CaP have been established, making it unlikely (X28) CAG repeats have also been associated with a that adjustment for other factors would have largely lower age at diagnosis of breast cancer in women with altered the results. Thirdly, the lack of widespread PSA germline BRCA1 mutations,54 although this was not testing in the UK means that the majority of early-onset substantiated in a study of early-onset breast cancer.55 cases in this study population presented with clinical Alternatively, the finding that shorter CAG repeats is symptoms. In general, PSA testing, which is widespread associated with a reduced risk might be related to an as in the US, has led to an earlier age at diagnosis, with yet unknown negative feedback mechanism, whereby some estimating that this test lowers the age at diagnosis reduced growth stimuli caused by longer AR CAG by X5 years on average.63 The present study population repeats disrupt this loop causing cells to overcompen- may therefore be clinically different to that of other CaP sate, thus leading to earlier onset CaP. populations that have been studied thus far. The finding that the L/L genotype of the SRD5A2 The finding that the AR CAG repeat polymorphism V89L polymorphism was associated with a statistically and the SRD5A2 V89L polymorphism were associated significant 85% increased risk of early-onset CaP was with a reduced risk of early-onset CaP raises a number of unexpected, although in line with our AR (CAG)n interesting questions about the nature of early-onset CaP. findings (as shown in Figure 1, these two genes It is interesting that a recent clinical trial using effectively work in series to control prostate cell growth). , the SRD5A2 inhibitor, as a chemopreventa- The variant allele has been associated with lower levels tive agent for CaP, found that while CaP was less likely of circulating A-diol-g, a serum marker of 5a-reductase in those treated, there was a significantly higher type II activity and intraprostatic DHT production.39,51 prevalence of high grade (Gleason 7–10) disease One might therefore expect this polymorphism to be (P ¼ 0.005) than in untreated men.58 If inhibiting SRD5A2 associated with a reduction in CaP risk, although has this effect, then perhaps low levels of activity of previous studies have failed to find a significant SRD5A2 and AR over a longer period of time will association.28,36,40 Nevertheless, there is evidence that produce similar effects, with fewer, but more aggressive men with low testosterone levels have higher Gleason cases of CaP resulting which present at an earlier age. grades and worse outcomes than the prostate cancers Our results could thus be due to ascertainment bias. that develop in men with normal testosterone.56–59 More work is required to determine if early-onset CaP is Unfortunately, we do not have clinical data for our cases, a different disease to CaP detected at a later age and but it seems reasonable to assume that as they were hence different polymorphisms are involved in the predominantly detected by clinical symptoms, they had disease, or if the same polymorphisms play different more advanced forms of CaP. Longer AR CAG repeats roles. reduce the transactivation ability of the AR, and the leucine variant of SRD5A2 V89L decreases the amount of DHT available to bind to the AR (see Figure 1), suggesting that there could be a common mechanism for early onset CaP predisposition involving reduced References transcription of genes that are regulated by the AR. The 1 Office for National Statistics. Mortality statistics: Cause (Series fact that not just one, but two polymorphisms examined DH2) No. 29. http://www.statistics.gov.uk; 2002. gave the opposite, yet complementary, results to those 2 American Cancer Society. Cancer facts & figures 2004. http:// expected suggests that our findings are not anomalous— www.cancer.org; 2004. one contrary result could be due to chance, but two is 3 Breslow N et al. Latent carcinoma of prostate at autopsy in seven unlikely. areas. The International Agency for Research on Cancer, Lyons, We found no evidence to suggest that the AR (GGC)n, France. Int J Cancer 1977; 20: 680–688. VDR taq I, CYP17 Msp A1 I A2 or the SRD5A2 A49T 4 Shimizu H et al. Cancers of the prostate and breast among polymorphisms were associated with early-onset CaP. Japanese and white immigrants in Los Angeles County. Br J Previous studies on the association between CaP risk and Cancer 1991; 63: 963–966. the AR GGC repeat polymorphism,15,17,23 the VDR taq I 5 Boring CC, Squires TS, Tong T, Montgomery S. Cancer statistics, 1994. CA Cancer J Clin 1994; 44: 7–26. polymorphism,44–48 the CYP17 Msp A1 I polymorph- 28,29,60,61 6 Steinberg GD et al. Family history and the risk of prostate cancer. ism and the SRD5A2 A49T polymorph- Prostate 1990; 17: 337–347. 33,35,37,38 ism have been highly inconsistent. However, 7 Spitz MR et al. Familial patterns of prostate cancer: a case– the low prevalence of some of these variant alleles control analysis. JUrol1991; 146: 1305–1307. (particularly the SRD5A2 A49 T T allele) limits the power 8 Carter BS et al. Hereditary prostate cancer: epidemiologic and to detect a significant association, especially in studies clinical features. JUrol1993; 150: 797–802. with relatively few cases. 9 Picado-Leonard J, Miller WL. Cloning and sequence of the There may be several explanations for the inconsis- human gene for P450c17 (steroid 17 alpha-hydroxylase/17,20 tencies between the results from this study compared lyase): similarity with the gene for P450c21. DNA 1987; 6: 439–448. with previous studies. Firstly, this study population 10 Coffey DS. In: Lepor H, Lawson RK (eds). Prostate Diseases.W.B. consisted of men diagnosed with CaP at the age of 55 or Saunders, Philadelphia, 1993, pp 28–56. younger, an age group which accounts for only one 11 di Salle E et al. PNU 157706, a novel dual type I and II 62 percent of UK CaP cases. No other published popula- 5alpha-reductase inhibitor. J Steroid Biochem Mol Biol 1998; 64: tion has selected as stringently for age of onset and most 179–186.

Prostate Cancer and Prostatic Diseases Hormonal genetic polymorphisms and early-onset prostate cancer MS Forrest et al 101 12 Lu S, Tsai SY, Tsai MJ. Regulation of androgen-dependent 35 Makridakis NM et al. Association of mis-sense substitution in prostatic cancer cell growth: androgen regulation of CDK2, SRD5A2 gene with prostate cancer in African-American and CDK4, and CKI p16 genes. Cancer Res 1997; 57: 4511–4516. Hispanic men in Los Angeles, USA. Lancet 1999; 354: 975–978. 13 Peehl DM et al. Antiproliferative effects of 1,25-dihydroxy- 36 Jaffe JM et al. Association of SRD5A2 genotype and pathological vitamin D3 on primary cultures of human prostatic cells. Cancer characteristics of prostate tumors. Cancer Res 2000; 60: 1626–1630. Res 1994; 54: 805–810. 37 Mononen N et al. A missense substitution A49T in the steroid 5- 14 Giovannucci E et al. The CAG repeat within the androgen alpha-reductase gene (SRD5A2) is not associated with prostate receptor gene and its relationship to prostate cancer. Proc Natl cancer in Finland. Br J Cancer 2001; 84: 1344–1347. Acad Sci USA 1997; 94: 3320–3323. 38 Soderstrom T et al. 5alpha-reductase 2 polymorphisms as risk 15 Hakimi JM et al. Androgen receptor variants with short factors in prostate cancer. Pharmacogenetics 2002; 12: 307–312. glutamine or glycine repeats may identify unique subpopula- 39 Makridakis N et al. A prevalent missense substitution that tions of men with prostate cancer. Clin Cancer Res 1997; 3: modulates activity of prostatic steroid 5alpha-reductase. Cancer 1599–1608. Res 1997; 57: 1020–1022. 16 Ingles SA et al. Association of prostate cancer risk with genetic 40 Febbo PG et al. The V89L polymorphism in the 5alpha-reductase polymorphisms in vitamin D receptor and androgen receptor. J type 2 gene and risk of prostate cancer. Cancer Res 1999; 59: Natl Cancer Inst 1997; 89: 166–170. 5878–5881. 17 Stanford JL et al. Polymorphic repeats in the androgen receptor 41 Makridakis NM, di Salle E, Reichardt JK. Biochemical and gene: molecular markers of prostate cancer risk. Cancer Res 1997; pharmacogenetic dissection of human steroid 5alpha-reductase 57: 1194–1198. type II. Pharmacogenetics 2000; 10: 407–413. 18 Ekman P et al. Links between genetic and environmental factors 42 Morrison NA et al. Contribution of trans-acting factor alleles to and prostate cancer risk. Prostate 1999; 39: 262–268. normal physiological variability: vitamin D receptor gene 19 Hsing AW et al. Polymorphic CAG and GGN repeat lengths in polymorphism and circulating osteocalcin. Proc Natl Acad Sci the androgen receptor gene and prostate cancer risk: a USA 1992; 89: 6665–6669. population-based case–control study in China. Cancer Res 2000; 43 Morrison NA et al. Prediction of bone density from vitamin D 60: 5111–5116. receptor alleles. Nature 1994; 367: 284–287. 20 Hardy DO et al. Androgen receptor CAG repeat lengths in 44 Correa-Cerro L et al. Vitamin D receptor polymorphisms as prostate cancer: correlation with age of onset. J Clin Endocrinol markers in prostate cancer. Hum Genet 1999; 105: 281–287. Metab 1996; 81: 4400–4405. 45 Blazer III DG et al. Vitamin D receptor polymorphisms and 21 Bratt O et al. CAG repeat length in the androgen receptor gene is prostate cancer. Mol Carcinog 2000; 27: 18–23. related to age at diagnosis of prostate cancer and response to 46 Taylor JA et al. Association of prostate cancer with vitamin D endocrine therapy, but not to prostate cancer risk. Br J Cancer receptor gene polymorphism. Cancer Res 1996; 56: 4108–4110. 1999; 81: 672–676. 47 Ingles SA et al. Strength of linkage disequilibrium between two 22 Irvine RA, Yu MC, Ross RK, Coetzee GA. The CAG and GGC vitamin D receptor markers in five ethnic groups: implications microsatellites of the androgen receptor gene are in linkage for association studies. Cancer Epidemiol Biomarkers Prev 1997; 6: disequilibrium in men with prostate cancer. Cancer Res 1995; 55: 93–98. 1937–1940. 48 Ingles SA et al. Association of prostate cancer with vitamin D 23 Edwards SM et al. Androgen receptor polymorphisms: associa- receptor haplotypes in African-Americans. Cancer Res 1998; 58: tion with prostate cancer risk, relapse and overall survival. Int J 1620–1623. Cancer 1999; 84: 458–465. 49 Eeles RA, Group, t. U. F. P. S. C.-o. & Collaborators, t. C. B. U. F. 24 Carey AH et al. Polycystic ovaries and premature male pattern P. C. S. Genetic predisposition to prostate cancer. Prostate Cancer baldness are associated with one allele of the steroid metabolism Prostatic Diseases 1999; 2: 9–15. gene CYP17. Hum Mol Genet 1994; 3: 1873–1876. 50 Popat S et al. Prevalence of the APC E1317Q variant in colorectal 25 Nedelcheva Kristensen V et al. CYP17 and breast cancer risk: the cancer patients. Cancer Lett 2000; 149: 203–206. polymorphism in the 50 flanking area of the gene does not 51 Allen NE, Forrest MS, Key TJ. The association between influence binding to Sp-1. Cancer Res 1999; 59: 2825–2828. polymorphisms in the CYP17 and 5alpha-reductase(SRD5A2) 26 Feigelson HS et al. A polymorphism in the CYP17 gene increases genes and serum androgen concentrations in men. Cancer the risk of breast cancer. Cancer Res 1997; 57: 1063–1065. Epidemiol Biomarkers Prev 2001; 10: 185–189. 27 Huang CS et al. Breast cancer risk associated with genotype 52 Key TJ et al. A case–control study of diet and prostate cancer. Br J polymorphism of the -metabolizing genes CYP17, Cancer 1997; 76: 678–687. CYP1A1, and COMT: a multigenic study on cancer suscep- 53 Sham PC, Curtis D. Monte Carlo tests for associations between tibility. Cancer Res 1999; 59: 4870–4875. disease and alleles at highly polymorphic loci. Ann Hum Genet 28 Lunn RM et al. Prostate cancer risk and polymorphism in 17 1995; 59: 97–105. hydroxylase (CYP17) and (SRD5A2). Carcino- 54 Rebbeck TR et al. Modification of BRCA1-associated breast genesis 1999; 20: 1727–1731. cancer risk by the polymorphic androgen-receptor CAG repeat. 29 Gsur A et al. A polymorphism in the CYP17 gene is associated Am J Hum Genet 1999; 64: 1371–1377. with prostate cancer risk. Int J Cancer 2000; 87: 434–437. 55 Given HF et al. The androgen receptor exon 1 trinucleotide 30 Davis DL, Russell DW. Unusual length polymorphism in human repeat does not act as a modifier of the age of presentation in steroid 5 alpha-reductase type 2 gene (SRD5A2). Hum Mol Genet breast cancer. Eur J Cancer 2000; 36: 533–534. 1993; 2: 820. 56 Ishikawa S, Soloway MS, Van der Zwaag R, Todd B. Prognostic 31 Kantoff PW et al. A polymorphism of the 5 alpha-reductase gene factors in survival free of progression after androgen deprivation and its association with prostate cancer: a case-control analysis. therapy for treatment of prostate cancer. JUrol1989; 141: Cancer Epidemiol Biomarkers Prev 1997; 6: 189–192. 1139–1142. 32 Hsing AW et al. Polymorphic markers in the SRD5A2 gene and 57 Prehn RT. On the prevention and therapy of prostate cancer by prostate cancer risk: a population-based case-control study. androgen administration. Cancer Res 1999; 59: 4161–4164. Cancer Epidemiol Biomarkers Prev 2001; 10: 1077–1082. 58 Thompson IM et al. The influence of finasteride on the 33 Latil AG et al. Prostate carcinoma risk and allelic variants of development of prostate cancer. N Engl J Med 2003; 349: 215–224. genes involved in androgen biosynthesis and metabolism 59 Schatzl G et al. Serum androgen levels in men: impact of health pathways. Cancer 2001; 92: 1130–1137. status and age. Urology 2003; 61: 629–633. 34 Margiotti K et al. Evidence for an association between the 60 Habuchi T et al. Increased risk of prostate cancer and benign SRD5A2 (type II steroid 5 alpha-reductase) locus and prostate prostatic hyperplasia associated with a CYP17 gene polymorph- cancer in Italian patients. Dis Markers 2000; 16: 147–150. ism with a gene dosage effect. Cancer Res 2000; 60: 5710–5713.

Prostate Cancer and Prostatic Diseases Hormonal genetic polymorphisms and early-onset prostate cancer MS Forrest et al 102 61 Wadelius M et al. Prostate cancer associated with CYP17 Prostate Group United Kingdom Familial Prostate Cancer Study genotype. Pharmacogenetics 1999; 9: 635–639. Collaborators. Cancer Res 2000; 60: 4513–4518. 62 Gayther SA et al. The frequency of germ-line mutations in the 63 Carter HB et al. Longitudinal evaluation of prostate-specific breast cancer predisposition genes BRCA1 and BRCA2 in antigen levels in men with and without prostate disease. JAMA familial prostate cancer. The Cancer Research Campaign/British 1992; 267: 2215–2220.

Supplementary Information accompanies the paper on Prostate Cancer website (http://www.nature.com/PCAN).

Prostate Cancer and Prostatic Diseases