A systematic confirmation study of reported prostate cancer risk-associated SNPs in Chinese men

Abstract

More than 30 prostate cancer (PCa) risk-associated loci have been identified in populations of European descent by genome-wide association studies (GWAS). We hypothesized that a subset of these loci may be associated with PCa risk in Chinese men. To test this hypothesis, 33 single nucleotide polymorphisms (SNPs), one each from the 33 independent PCa risk-associated loci reported in populations of European descent, were investigated for their associations with PCa risk in a case-control study of Chinese men (1,108 cases and 1,525 controls). We found that 11 of the 33 SNPs were significantly associated with PCa risk in Chinese men (P < 0.05). The reported risk alleles were associated with increased risk for PCa, with allelic odds ratios ranging from 1.12 to 1.44. The most significant locus was located on 8q24 Region 2 (rs16901979, P = 5.14×10⁻⁹) with a genome-wide significance (P < 10⁻⁸), and three loci reached the Bonferroni correction significance level (P < 1.52×10⁻³), including 8q24 Region 1 (rs1447295, P = 7.04×10⁻⁶), 8q24 Region 5 (rs10086908, P = 9.24×10⁻⁴), and 8p21 (rs1512268, P = 9.39×10⁻⁴). Our results suggest that a subset of the PCa risk-associated SNPs discovered by GWAS among men of European descent is also associated with PCa risk in Chinese men. This finding provides evidence of ethnic differences and similarity in genetic susceptibility to PCa. GWAS in Chinese men are needed to identify Chinese-specific PCa risk-associated SNPs.

Introduction

Prostate cancer (PCa) is the most common noncutaneous cancer among men living in Western nations, with an estimated 192,280 new cases in 2009 in the United States.¹ Although the incidence of prostate cancer in Chinese men is much lower than that in Western men, the occurrence of this disease has rapidly increased among Chinese men in recent decades, especially in developed areas.² For instance, in Shanghai, the incidence of PCa rose from 1.6/100,000 person-years in 1972 to 7.7/100,000 person-years in 2000.³ Thus, PCa has increasingly become a public health issue in China.

Age, race, and family history are well-established risk factors for PCa.⁴ Hereditary factors are generally believed to contribute to PCa etiology, but until recently, there have been few validated genetic markers associated with prostate cancer risk. Recently, genome-wide association studies (GWAS) have identified a number of genetic markers reproducibly associated with PCa risk.^5–15 However, most of these studies have been conducted and replicated in populations of European and African descent, and thus the extent to which GWAS results are applicable to other ethnicities is largely unknown.

In an attempt to extend the GWAS findings into other ethnic populations, we and others have evaluated some of these single nucleotide polymorphisms (SNPs) in specific Asian populations. Yamada et al¹⁶ evaluated 23 SNPs in a study of Japanese men (311 cases and 1,035 controls) and found that five of the SNPs were independently associated with PCa risk. Our group conducted a small pilot study with 288 cases and 155 controls from Shanghai, China and observed significant associations for two SNPs on chromosome 8q24.¹⁷

To follow-up on and extend our findings in the Chinese population, we carried out a systematic evaluation of 33 PCa risk-associated SNPs previously reported in populations of European descent and the risk for PCa in the largest Chinese case-control study population described to date.

Subjects and Methods

Study Subjects

The subjects included in this study are part of a Chinese Consortium for Prostate Cancer Genetics (ChinaPCa), which was initiated in 2010 and was designed to understand genetic determinants of PCa risk in Chinese men. Considering that the incidence of PCa is relatively low in China, a consortium effort to recruit PCa patients is necessary to achieve large sample size. Any hospital in China that is interested in joining ChinaPCa and is approved by the Institutional Ethic Review Board is welcome. Inclusion criteria are pathologically diagnosed PCa patients (prevalent or incident cases), availability of blood samples, and Han Chinese. Because PSA screening for PCa is uncommon in China, most of the PCa patients were diagnosed due to PCa related symptoms. At the time of this study, 820 PCa patients from 8 hospitals in the Southern and Eastern parts of China were available (Set 2). In addition, 288 PCa patients from an existing PCa case-control study (Set 1), described in detail elsewhere,¹⁷ were also included. The numbers and clinicopathological characteristics of PCa patients from each of these hospitals are presented in Table 1 and Supplementary Table1. This study is the first report of the ChinaPCa study.

Table 1.

Summary of subject recruitment in ChinaPCa

Center	Area	Number of samples	Age (years)		Set
			Median	Range
Shanghai Cancer Institute	Shanghai	155 cases	73	50–100	1
		288 controls	72	36–85	1
Huashan Hospital	Shanghai	192 cases	73	56–86	2
Suzhou Municipal Hospital	Suzhou	189 cases	73	34–91	2
Changhai Hospital	Shanghai	170 cases	73	51–87	2
Xinhua Hospital	Shanghai	89 cases	74	52–88	2
Guangxi Medical University	Guangxi	87 cases	71	37–89	2
Ninth People’s Hospital	Shanghai	68 cases	74	53–97	2
Sun Yat-sen University	Guangzhou	15 cases	69	61–77	2
Pudong Gongli Hospital	Shanghai	10 cases	73	61–89	2
Fudan University	Taizhou	1,370 controls	66	53–90	2

There were two sources for the control subjects in this study. The first source was 1,370 male controls randomly selected from a community in Taizhou, Jiangsu Province in the South-East of China. As a population control, age and other factors are not matched for PCa patients. The second source was the control subjects from the existing PCa case-control study. ¹⁷ Serum level of prostate-specific antigen (PSA) is available for all control subjects.

This study was approved by the Institutional Review Board at Fudan University, the Shanghai Cancer Institute, and each participating hospital.

SNP Selection and Genotyping

PCa risk-associated SNPs were defined according to GWAS reports and following fine-mapping studies. Each of these SNPs has been independently associated with PCa risk in studies of men of European descent exceeding genome-wide significance levels in their initial reports (P < 10⁻⁷) and was replicated in independent study populations.^{5–15,18–20}

Samples were genotyped at the Center for Cancer Genomics at Wake Forest University (set 1; 288 cases and 155 controls) and in the Fudan-VARI Center for Genetic Epidemiology at Fudan University (set 2; 820 cases and 1,370 controls) (Table 2). Both sets of subjects were genotyped on the same genotyping system (MassARRAY iPLEX; Sequenom, Inc., San Diego, CA). Duplicates from 2 subjects and two water samples (negative control) were included in each 96-well plate for genotyping quality control. Genotyping was performed by technicians that were blinded to case-control status. The overall concordance rate was 99% for these 33 SNPs among the 58 duplicate samples.

Table 2.

Genotyping call rates and Hardy-Weinberg equilibrium (HWE) for 33 loci in ChinaPCa¹

Chromosome	SNP	Alleles	Call Rate		HWE in Controls2
			Set 1	Set 2	Set 1	Set 2
2	rs1465618	A>G	0.98	0.98	0.73	0.12
2	rs721048	G>A	0.98	0.97	0.68	0.25
2	rs12621278	A>G	0.98	0.99	0.75	0.12
3	rs2660753	C>T	0.98	0.98	0.44	0.14
3	rs10934853	C>A	0.97	0.98	0.64	0.31
4	rs17021918	C>T	0.97	0.98	0.29	0.07
4	rs7679673	A>C	0.98	0.99	0.03	0.82
6	rs9364554	C>T	0.98	0.97	0.82	0.75
7	rs10486567	T>C	0.98	0.99	0.72	0.45
7	rs6465657	C>T	0.98	0.98	0.17	0.14
8	rs2928679	G>A	0.98	0.98	0.58	0.32
8	rs1512268	G>A	0.98	0.99	0.11	0.03
8	rs10086908	T>C	0.97	0.99	0.97	0.17
8	rs16901979	C>A	0.97	0.97	0.99	0.74
8	rs16902094	A>G	0.97	0.98	0.29	0.94
8	rs620861	G>A	0.98	0.98	0.41	0.89
8	rs6983267	T>G	0.98	0.97	0.69	0.82
8	rs1447295	C>A	0.98	0.98	0.15	0.17
9	rs1571801	G>T	0.98	0.99	0.64	0.02
10	rs10993994	T>C	0.96	0.98	0.29	0.31
10	rs4962416	A>G	0.97	0.97	0.93	0.73
11	rs7127900	G>A	0.98	0.97	0.58	0.01
11	rs12418451	G>A	0.98	0.99	0.52	0.65
11	rs10896449	A>G	0.97	0.53	0.33	0.53
17	rs11649743	C>T	0.97	0.97	0.55	0.55
17	rs4430796	T>C	0.98	0.98	0.72	0.71
17	rs1859962	T>G	0.97	0.98	0.66	0.51
19	rs8102476	A>G	0.97	0.99	0.19	0.79
19	rs887391	T>C	0.97	0.98	0.64	0.96
19	rs2735839	G>A	0.98	0.98	0.34	0.40
22	rs9623117	T>C	0.98	0.99	0.64	0.88
22	rs5759167	G>T	0.98	0.99	0.37	0.77
X	rs5945619	A>G	0.98	0.98	--	--

¹The samples genotyped at the Center for Cancer Genomics in Wake Forest University are referred to as set 1 and the remaining samples genotyped at the Fudan-VARI Center for Genetic Epidemiology in Fudan University are considered as set 2;

²P values of tests for Hardy-Weinberg equilibrium (HWE).

Statistical Methods

Genotype frequencies for all gene variants in the control subjects were analyzed by Pearson’s Chi-square tests for Hardy-Weinberg equilibrium. The allelic odds ratios (ORs) and 95% confidence intervals (CI) were estimated using logistic regression models. All analyses were performed using Statistical Analysis System (SAS) software (version 9.2; SAS Institute, Cary, NC). All statistical tests were two-sided.

Results

The call rates for 33 SNPs in set 1 (288 cases and 155 controls) were between 96% and 98%. One SNP failed the quality check due to low call rate (53%) in set 2 (820 cases and 1,370 controls) while the call rates for the remaining 32 SNPs were between 97% and 99% (Table 2; Supplementary Table 2). None of the SNPs significantly deviated from Hardy-Weinberg equilibrium among control subjects in either sets at P < 0.01 level.

As shown in Table 3, among the 33 PCa risk-associated SNPs reported in men of European descent, 11 SNPs were significantly associated with PCa risk in Chinese men at P < 0.05. Reported risk alleles of all 11 SNPs were associated with increased risk for PCa, with ORs ranging from 1.12 to 1.44. Among these 11 SNPs, the most significant SNP was located at 8q24 Region 2 (rs16901979, P = 5.14×10⁻⁹) with a genome-wide significance (P<10⁻⁸), and three SNPs reached the reached the Bonferroni correction significance level (P < 1.52×10⁻³), including 8q24 Region 1 (rs1447295, P = 7.04×10⁻⁶), 8q24 Region 5 (rs10086908, P = 9.24×10⁻⁴), and 8p21 (rs1512268, P = 9.39×10⁻⁴). In addition, seven SNPs had P values between 1.52×10⁻³ and 0.05, and all had risk estimates in the same direction as that reported for men of European descent (Table 3).

Table 3.

PCa susceptibility loci identified in populations of European descent and their associations with PCa risk in Chinese men

Chr	SNP	Region	Gene	Alleles	European			RAF2		OR (95%CI)3	P
					R1	OR1	RAF2	Case	Control
2	rs1465618	2p21	THADA	A>G	A	1.15	0.212	0.745	0.716	1.16(1.02, 1.31)	0.020
2	rs721048	2p15	EHBP1	G>A	A	1.18	0.137	0.042	0.031	1.36(1.01, 1.83)	0.042
2	rs12621278	2q31	ITGA6	A>G	A	1.35	0.960	0.740	0.710	1.16(1.03, 1.31)	0.019
3	rs2660753	3p12		C>T	T	1.24	0.102	0.312	0.295	1.08(0.96, 1.22)	0.212
3	rs10934853	3q21		C>A	A	1.12	0.239	0.447	0.430	1.07(0.96, 1.20)	0.236
4	rs17021918	4q22	PDLIM5	C>T	C	1.14	0.646	0.649	0.652	0.99(0.88, 1.11)	0.815
4	rs7679673	4q24	TET2	A>C	C	1.14	0.624	0.221	0.192	1.20(1.05, 1.37)	9.39E-03
6	rs9364554	6q25	SLC22A3	C>T	T	1.17	0.274	0.315	0.329	0.94(0.83, 1.06)	0.282
7	rs10486567	7p15	JAZF1	T>C	C	1.16	0.752	0.159	0.141	1.15(0.98, 1.34)	0.081
7	rs6465657	7q21	LMTK2	C>T	C	1.14	0.509	0.872	0.857	1.13(0.96, 1.33)	0.134
8	rs2928679	8p21	NKX3.1	G>A	A	1.13	0.456	0.138	0.132	1.05(0.90, 1.24)	0.522
8	rs1512268	8p21	NKX3.1	G>A	A	1.17	0.420	0.314	0.272	1.23(1.09, 1.38)	9.39E-04
8	rs10086908	8q24 (Region 5)		T>C	T	1.13	0.625	0.844	0.808	1.28(1.11, 1.48)	9.24E-04
8	rs16901979	8q24 (Region 2)		C>A	A	1.82	0.031	0.322	0.248	1.44(1.28, 1.63)	5.14E-09
8	rs16902094	8q24.21		A>G	G	1.20	0.271	0.273	0.279	0.97(0.85, 1.10)	0.595
8	rs620861	8q24 (Region 4)		G>A	G	1.16	0.619	0.572	0.567	1.02(0.91, 1.14)	0.717
8	rs6983267	8q24 (Region 3)		T>G	G	1.20	0.487	0.454	0.434	1.09(0.97, 1.21)	0.150
8	rs1447295	8q24 (Region 1)		C>A	A	1.47	0.071	0.209	0.160	1.38(1.20, 1.60)	7.04E-06
9	rs1571801	9q33	DAB2IP	G>T	T	1.17	0.290	0.050	0.051	0.98(0.76, 1.27)	0.893
10	rs10993994	10q11	MSMB	T>C	T	1.25	0.341	0.519	0.490	1.12(1.01, 1.26)	0.038
10	rs4962416	10q26	CTBP2	A>G	G	1.15	0.257	0.011	0.009	1.20(0.69, 2.08)	0.521
11	rs7127900	11p15	IGF2/IGF2AS	G>A	A	1.25	0.235	0.104	0.102	1.03(0.86, 1.24)	0.752
11	rs12418451	11q13		G>A	A	1.16	0.342	0.064	0.065	0.99(0.79, 1.24)	0.926
11	rs10896449 4	11q13		A>G	G	1.16	0.532	0.111	0.086	1.31(0.82–2.10)	0.261
17	rs11649743	17q12	HNF1B	C>T	C	1.16	0.757	0.692	0.657	1.17(1.04, 1.32)	8.51E-03
17	rs4430796	17q12	HNF1B	T>C	T	1.22	0.491	0.734	0.713	1.11(0.98, 1.26)	0.088
17	rs1859962	17q24		T>G	G	1.21	0.527	0.426	0.425	1.01(0.9, 1.12)	0.934
19	rs8102476	19q13		A>G	G	1.12	0.504	0.369	0.374	0.98(0.87, 1.10)	0.721
19	rs887391	19q13		T>C	T	1.14	0.757	0.618	0.592	1.12(1.00, 1.25)	0.056
19	rs2735839	19q13	KLK2/KLK3	G>A	G	1.30	0.863	0.604	0.612	0.97(0.87, 1.09)	0.581
22	rs9623117	22q13	TNRC6B	T>C	C	1.13	0.221	0.034	0.037	0.94(0.70, 1.27)	0.681
22	rs5759167	22q13	TTLL1/BIK	G>T	G	1.18	0.549	0.730	0.694	1.19(1.06, 1.35)	4.81E-03
X	rs5945619	Xp11	NUDT10/NUDT11	A>G	G	1.27	0.385	0.084	0.068	1.25(0.94, 1.68)	0.130

¹Risk allele (RA) reported in Europeans. The ORs were derived from pooled results in populations of European descent;²⁰

²Risk allele frequency (RAF) reported in HapMap or from the results in our population;

³Allelic odds ratio (OR) and 95% confidence interval (95% CI);

⁴The results for rs10896449 were available only from set 1, and did not include set 2 due to a low call rate for set 2.

Discussion

In a large Chinese population, we systematically evaluated 33 PCa risk loci that were previously identified from studies in men of European descent. Our results show that reported risk alleles for 11 of these 33 loci were significantly associated with increased PCa risk in Chinese men with ORs of 1.12–1.44. These findings, although in need of confirmation, show that a subset of PCa genetic susceptibility markers may be shared between European and Chinese populations.

The incidence of PCa is markedly lower in Asian countries, especially in China, than that in western countries.² One reason for this large difference in PCa incidence may be due to different genetic backgrounds, which has been well supported by epidemiologic studies of immigrant populations²¹ and by a recent somatic genomic alteration analysis.²² Therefore, it is important to evaluate whether the PCa risk loci indentified in European populations are also relevant in Chinese populations. In a previous small study of Chinese men from Shanghai, China (288 cases and 155 controls), we assessed 17 PCa risk loci identified in men of European descent and found that 2 of the loci (rs1447295 and rs16901979 at 8q24 regions 1 and 2, respectively) were significantly associated with PCa risk.¹⁷ In a Southern Chinese population (251 cases and 258 controls), Xu et al²³ found a positive association between rs10993994 at 10q11 (MSMB locus) with risk of PCa. Consistent with these 2 previous studies in Chinese men, the current study confirmed the associations at the aforementioned 3 PCa risk loci and found associations in an additional 8 loci for Chinese men. The large sample size allowed for better statistical power to detect associations with modest effects. These PCa susceptibility loci could be critical to future efforts for building a genetic risk model to identify Chinese men at high-risk for PCa.

It is also important to note that not all PCa risk SNPs identified in men of European descent were replicated in Chinese men. Recently, in a GWAS of Japanese population, Takata et al ²⁴ also evaluated 31 SNPs that were reported in GWAS of European descent in 1583 PCa cases and 3386 controls and found significant association for 19 SNPs but not for remaining 12 SNPs. There may be several reasons for a failure to replicate association findings. First, there may be some genetic heterogeneity that exists with respect to PCa susceptibility between various ethnic populations. For instance, differences in LD patterns between the causal loci and tagging markers may exist between various ethnic populations, and this may lead to varying levels of tagging in different populations. Second, different environmental exposures may affect the degree to which genes are activated and thus may modulate the impact that genetic susceptibility loci may have on disease risk. Third, a sample size that is larger than the current study may be necessary to provide sufficient power to detect the effect of loci with more modest effects or have lower frequencies. Thus future larger studies are warranted to confirm our findings.

In previous studies of men of European descent, none of the previously reported PCa risk loci have been consistently associated with aggressive PCa, though some studies have reported stronger effects in more aggressive as compared to less aggressive tumors.^25–26 In the Japanese study, none of the SNPs found to be associated with PCa risk were significantly associated with aggressiveness of PCa.¹⁶ In the present study, we did not perform an analysis of PCa aggressiveness because three different staging systems for PCa were used by the recruiting hospitals. However, most of the PCa cases in our study had aggressive disease because PSA screening is not common practice in China and thus prostate cancer diagnoses among Chinese patients are almost always driven by symptoms, which results in very limited numbers of nonaggressive patients. Therefore, the significant loci identified in the current study may be more clinically relevant than the results derived from populations comprised primarily of less aggressive patients. At the same time, these findings provide evidence of ethnic differences in genetic susceptibility to PCa. GWAS in Chinese men are needed to identify Chinese-specific PCa risk-associated SNPs.

There are several important limitations in this study. First, the clinical characterization of PCa was not consistent across hospitals, as discussed previously, which limited our ability to conduct further analysis based on different phenotypes. Second, the cases were recruited from hospitals and communities whereas the controls were recruited from communities; these recruitment methods may have resulted in selection bias. And finally, although our study is the largest to date, we had 80% or higher statistical power to evaluate the association of 22 SNPs with PCa risk (Supplementary Table 3) and larger sample sizes are needed to replicate and extend these findings; in comparison, studies in men of European descent used to identify the 33 PCa risk loci are larger than the current study.

In conclusion, this study represents a systematic evaluation of 33 loci related to European PCa risk in Chinese men. This approach allowed us to identify genetic markers for PCa susceptibility among Chinese men, and to observe differences in genetic determinants of PCa between the Chinese population and other populations.

Supplementary Material

Supp Table S1-S3

Supplementary Table 1: The clinicopathological characteristics in this study

Supplementary Table 2: The genotype frequency for 33 SNPs between cases and controls

Supplementary Table 3: The statistical power of the sample size in this study

Abbreviations

ChinaPCa

Chinese Consortium for Prostate Cancer Genetics

CI

confidence interval

GWAS

genome-wide association study

HWE

Hardy-Weinberg equilibrium

LD

linkage disequilibrium

OR

odds ratio

PCa

prostate cancer

RA

risk allele

RAF

risk allele frequency

SNP

single nucleotide polymorphism