Abstract

We performed a three-stage genome-wide association study (GWAS) to identify common Parkinson's disease (PD) risk variants in the European population. The initial genome-wide scan was conducted in a French sample of 1039 cases and 1984 controls, using almost 500 000 single nucleotide polymorphisms (SNPs). Two SNPs at SNCA were found to be associated with PD at the genome-wide significance level (P < 3 × 10−8). An additional set of promising and new association signals was identified and submitted for immediate replication in two independent case–control studies of subjects of European descent. We first carried out an in silico replication study using GWAS data from the WTCCC2 PD study sample (1705 cases, 5200 WTCCC controls). Nominally replicated SNPs were further genotyped in a third sample of 1527 cases and 1864 controls from France and Australia. We found converging evidence of association with PD on 12q24 (rs4964469, combined P = 2.4 × 10−7) and confirmed the association on 4p15/BST1 (rs4698412, combined P = 1.8 × 10−6), previously reported in Japanese data. The 12q24 locus includes RFX4, an isoform of which, named RFX4_v3, encodes brain-specific transcription factors that regulate many genes involved in brain morphogenesis and intracellular calcium homeostasis.

INTRODUCTION

Parkinson's disease (PD) is the second most common degenerative disease, affecting 1–2% of individuals older than 65 years. Clinical features of PD result primarily from the loss of dopaminergic neurons in the substantia nigra. Although the common form of PD is sporadic, six genes have been identified, mainly by linkage analyses of Mendelian forms of the disease. Two genes, SNCA (encoding α-synuclein) and LRRK2, have an autosomal dominant inheritance and four other genes, PARK2 (parkin), PARK6 (PINK1), PARK7 (DJ-I) and PARK13 (ATP13A2), have an autosomal recessive inheritance (1). Frequently, mutations in these genes are found in patients with early-onset PD, particularly those with autosomal recessive inheritance. However, in most populations, Mendelian forms of parkinsonism are rare when compared with the most common form of PD, a frequent and complex disorder probably explained by the interaction between genetic and environmental factors.

The first two genome-wide association studies (GWASs) in PD (2,3) provided evidence of association with several loci but most often not at the genome-wide significant level, and most initial association findings were not confirmed by subsequent replication analyses (4). Two recent GWASs (5,6) reported strong or genome-wide significant associations with one or more of the known PD genes (i.e. SNCA, MAPT and/or LRRK2). So far, only two ‘new’ loci have been identified, 1q32/PARK16 and 4p15/BST1, in the Japanese data (5). The US/UK/German GWAS (6) replicated positive association with variants at PARK16 but failed to replicate the association at BST1.

To identify additional variants that affect PD risk in the European population, we designed a three-stage GWAS of PD in three case–control samples from France, the UK and Australia (total of >13 300 subjects). A set of 50 top association signals was identified in the scan sample (1039 cases and 1984 controls from France) using the Illumina-610Quad chip. Promising and new signals were followed-up for stepwise replication in two further UK and French/Australian case–control studies (>3200 cases and 7000 controls).

RESULTS

The genome-wide association results from logistic test corrected for genomic inflation (GC) revealed two single nucleotide polymorphisms (SNPs) with PGC < 10−7, and a substantial number of SNPs with strong (PGC < 10−4) evidence of association (Fig. 1 and Table 1). For practical reasons, we focus our attention on the 50 best-associated SNPs to prioritize for immediate in silico replication (Table 1). Secondary logistic analyses, adjusted for the first two principal componenets (PCs) led to similar rank order of SNPs, albeit slightly weaker association signals (Table 1). This suggests that the significant results, revealed by our primary analyses, are not biased by residual population substructure within our French scan sample. The 50 best-associated SNPs spanned 23 distinct genomic loci, and were associated with PGC < 5.6 × 10−5. Sixteen associations were found within two well-known PD genes, i.e. SNCA (4q22, 4 SNPs) and MAPT (17q12–q21, 11 SNPs), or within BST1 (4p15, one SNP), a recently reported PD risk locus established at the genome-wide significance level in a Japanese population (5). The remaining 34 SNPs were located in 20 distinct previously unreported putative PD loci. The two genome-wide significant SNPs were located on 4q22/SNCA [rs356220, PGC = 2.82 × 10−8, OR = 1.37; 95% CI (1.22–1.53) and rs2736990, PGC = 2.88 × 10−8, OR = 1.35 (1.22–1.50)]. The next most significant SNP was on chromosome 12q21/LOC401725 [rs7954761, PGC = 2.09 × 10−7, OR = 1.34 (1.20–1.50)].

Table 1.

Top 50 SNPs in scan stage and in silico replication results

Chromosome (gene) Position (bp) SNP Stage-1: scan (France) data
 
Stage-2: replication (UK) data
 
   RAa RAFb OR PGC (two-tailed)c P2PCs (two-tailed)d RAF OR P (one-tailed)e 
Known PD genes/previously published loci 
4q22 (SNCA) 90858538 rs11931074 0.07 1.52 1.35E-05 9.04E-05 0.07 1.33 4.01E-05 
90860363 rs356220 0.35 1.37 2.82E-08 6.26E-07 0.36 1.27 2.59E-09 
90894261 rs3857059 0.07 1.54 1.00E-05 6.32E-05 0.07 1.33 3.95E-05 
90897564 rs2736990 0.44 1.35 2.88E-08 1.32E-07 0.45 1.24 3.98E-08 
17q12–21 (MAPT) 41074926 rs393152 0.75 1.32 2.68E-05 1.43E-04 0.76 1.31 2.20E-08 
41279463 rs12185268 0.76 1.32 3.44E-05 1.62E-04 0.76 1.30 3.59E-08 
41279910 rs12373139 0.76 1.33 1.81E-05 7.60E-05 0.76 1.30 2.77E-08 
41281077 rs17690703 0.72 1.34 3.94E-06 6.61E-06 0.72 1.24 1.37E-06 
41347100 rs17563986 0.75 1.34 1.30E-05 5.65E-05 0.76 1.31 2.58E-08 
41412603 rs1981997 0.76 1.33 2.20E-05 8.81E-05 0.76 1.30 4.61E-08 
41436901 rs8070723 0.75 1.33 2.19E-05 8.91E-05 0.76 1.30 2.61E-08 
41544850 rs7225002 0.59 1.27 2.72E-05 4.08E-05 0.57 1.23 1.14E-07 
41602941 rs2532274 0.75 1.33 2.21E-05 1.06E-04 0.75 1.28 2.92E-07 
41605885 rs2532269 0.75 1.33 1.90E-05 8.58E-05 0.76 1.29 1.11E-07 
41648797 rs2668692 0.76 1.33 1.97E-05 8.20E-05 0.76 1.29 1.22E-07 
4p15 (BST1) 15346446 rs4698412 0.52 1.28 6.88E-06 1.96E-06 0.55 1.08 0.0247 
Newly identified loci 
1p36.22 11880226 rs12724129 0.34 1.26 4.35E-05 4.45E-05 0.40 0.99 f 
1p36.11 26448619 rs10902724 0.05 1.55 5.13E-05 5.00E-04 0.07 0.96 f 
2q14.3 126112282 rs1365783 0.42 1.26 4.33E-05 9.10E-05 0.46 0.94 f 
2q21.3 135011650 rs621341 0.28 1.30 5.11E-06 4.37E-04 0.48 1.08 0.0277 
135196450 rs6723108 0.31 1.25 6.85E-05 3.20E-03 0.51 1.11 0.0053 
135318626 rs6729702 0.46 1.26 1.75E-05 6.02E-04 0.64 1.04 0.15 
135339278 rs6430552 0.46 1.26 1.89E-05 6.95E-04 0.64 1.05 0.14 
135367236 rs6714498 0.46 1.26 1.84E-05 6.20E-04 0.64 1.05 0.14 
2q22 136611978 rs4954564 0.52 1.27 1.21E-05 4.07E-04 0.74 1.03 0.27 
136722668 rs6430612 0.41 1.27 1.35E-05 1.89E-03 0.65 1.03 0.23 
136730076 rs10221893 0.41 1.27 1.31E-05 1.88E-03 0.65 1.03 0.22 
2q35 216463846 rs6741233 0.87 1.51 9.34E-06 4.46E-04 0.93 1.01 0.46 
4p16 11054284 rs368039 0.11 1.41 1.52E-05 2.11E-05 0.13 0.86 f 
5p15.2 10016889 rs1428954 0.53 1.27 1.65E-05 3.73E-04 0.57 1.04 0.18 
10026935 rs10072891 0.53 1.27 2.02E-05 6.68E-04 0.57 1.04 0.16 
10037418 rs38065 0.65 1.29 1.50E-05 2.19E-04 0.69 1.02 0.33 
5q12.1 60896208 rs1423326 0.60 1.27 2.10E-05 6.54E-04 0.65 1.02 0.36 
5q22.2 112814742 rs26990 0.13 1.41 6.67E-06 8.67E-06 0.19 0.94 f 
6q12 70963882 rs9360414 0.38 1.25 5.34E-05 4.00E-05 0.38 1.05 0.13 
10p14 6933911 rs10905042 0.06 1.53 3.08E-05 3.82E-04 0.07 1.02 0.42 
11p12 36589978 rs12419750 0.89 1.47 4.96E-05 1.65E-05 0.90 0.97 f 
36600652 rs1391542 0.89 1.47 5.44E-05 1.89E-05 0.90 0.98 f 
36613848 rs7128419 0.89 1.47 4.91E-05 1.71E-05 0.90 0.98 f 
36618299 rs12271660 0.90 1.50 5.64E-05 4.56E-05 0.92 0.96 f 
36684837 rs12294719 0.79 1.44 5.42E-07 6.72E-07 0.81 1.00 0.48 
36687460 rs1533588 0.79 1.41 1.79E-06 3.78E-06 0.82 0.97 f 
11q13.5 75709727 rs12295401 0.06 1.53 4.40E-05 5.16E-05 0.06 1.04 0.29 
12p13.3 760163 rs11064524 0.20 1.32 2.80E-05 1.21E-04 0.24 1.08 0.0447 
12q21.31 82691472 rs7954761 0.60 1.34 2.09E-07 2.59E-07 0.61 0.99 f 
12q24 105474117 rs4964469 0.33 1.27 2.73E-05 1.30E-04 0.37 1.11 0.0045 
105513235 rs1035767 0.11 1.42 1.50E-05 2.15E-05 0.11 0.98 f 
13q34 113253980 rs2259599 0.83 1.39 3.74E-05 5.61E-03 0.88 0.96 f 
17p13.2 4376339 rs9899558 0.73 1.34 5.04E-06 3.82E-04 0.77 0.98 f 
22q11.23 22917303 rs9608247 0.16 1.33 2.99E-05 9.67E-05 0.17 0.95 f 
Chromosome (gene) Position (bp) SNP Stage-1: scan (France) data
 
Stage-2: replication (UK) data
 
   RAa RAFb OR PGC (two-tailed)c P2PCs (two-tailed)d RAF OR P (one-tailed)e 
Known PD genes/previously published loci 
4q22 (SNCA) 90858538 rs11931074 0.07 1.52 1.35E-05 9.04E-05 0.07 1.33 4.01E-05 
90860363 rs356220 0.35 1.37 2.82E-08 6.26E-07 0.36 1.27 2.59E-09 
90894261 rs3857059 0.07 1.54 1.00E-05 6.32E-05 0.07 1.33 3.95E-05 
90897564 rs2736990 0.44 1.35 2.88E-08 1.32E-07 0.45 1.24 3.98E-08 
17q12–21 (MAPT) 41074926 rs393152 0.75 1.32 2.68E-05 1.43E-04 0.76 1.31 2.20E-08 
41279463 rs12185268 0.76 1.32 3.44E-05 1.62E-04 0.76 1.30 3.59E-08 
41279910 rs12373139 0.76 1.33 1.81E-05 7.60E-05 0.76 1.30 2.77E-08 
41281077 rs17690703 0.72 1.34 3.94E-06 6.61E-06 0.72 1.24 1.37E-06 
41347100 rs17563986 0.75 1.34 1.30E-05 5.65E-05 0.76 1.31 2.58E-08 
41412603 rs1981997 0.76 1.33 2.20E-05 8.81E-05 0.76 1.30 4.61E-08 
41436901 rs8070723 0.75 1.33 2.19E-05 8.91E-05 0.76 1.30 2.61E-08 
41544850 rs7225002 0.59 1.27 2.72E-05 4.08E-05 0.57 1.23 1.14E-07 
41602941 rs2532274 0.75 1.33 2.21E-05 1.06E-04 0.75 1.28 2.92E-07 
41605885 rs2532269 0.75 1.33 1.90E-05 8.58E-05 0.76 1.29 1.11E-07 
41648797 rs2668692 0.76 1.33 1.97E-05 8.20E-05 0.76 1.29 1.22E-07 
4p15 (BST1) 15346446 rs4698412 0.52 1.28 6.88E-06 1.96E-06 0.55 1.08 0.0247 
Newly identified loci 
1p36.22 11880226 rs12724129 0.34 1.26 4.35E-05 4.45E-05 0.40 0.99 f 
1p36.11 26448619 rs10902724 0.05 1.55 5.13E-05 5.00E-04 0.07 0.96 f 
2q14.3 126112282 rs1365783 0.42 1.26 4.33E-05 9.10E-05 0.46 0.94 f 
2q21.3 135011650 rs621341 0.28 1.30 5.11E-06 4.37E-04 0.48 1.08 0.0277 
135196450 rs6723108 0.31 1.25 6.85E-05 3.20E-03 0.51 1.11 0.0053 
135318626 rs6729702 0.46 1.26 1.75E-05 6.02E-04 0.64 1.04 0.15 
135339278 rs6430552 0.46 1.26 1.89E-05 6.95E-04 0.64 1.05 0.14 
135367236 rs6714498 0.46 1.26 1.84E-05 6.20E-04 0.64 1.05 0.14 
2q22 136611978 rs4954564 0.52 1.27 1.21E-05 4.07E-04 0.74 1.03 0.27 
136722668 rs6430612 0.41 1.27 1.35E-05 1.89E-03 0.65 1.03 0.23 
136730076 rs10221893 0.41 1.27 1.31E-05 1.88E-03 0.65 1.03 0.22 
2q35 216463846 rs6741233 0.87 1.51 9.34E-06 4.46E-04 0.93 1.01 0.46 
4p16 11054284 rs368039 0.11 1.41 1.52E-05 2.11E-05 0.13 0.86 f 
5p15.2 10016889 rs1428954 0.53 1.27 1.65E-05 3.73E-04 0.57 1.04 0.18 
10026935 rs10072891 0.53 1.27 2.02E-05 6.68E-04 0.57 1.04 0.16 
10037418 rs38065 0.65 1.29 1.50E-05 2.19E-04 0.69 1.02 0.33 
5q12.1 60896208 rs1423326 0.60 1.27 2.10E-05 6.54E-04 0.65 1.02 0.36 
5q22.2 112814742 rs26990 0.13 1.41 6.67E-06 8.67E-06 0.19 0.94 f 
6q12 70963882 rs9360414 0.38 1.25 5.34E-05 4.00E-05 0.38 1.05 0.13 
10p14 6933911 rs10905042 0.06 1.53 3.08E-05 3.82E-04 0.07 1.02 0.42 
11p12 36589978 rs12419750 0.89 1.47 4.96E-05 1.65E-05 0.90 0.97 f 
36600652 rs1391542 0.89 1.47 5.44E-05 1.89E-05 0.90 0.98 f 
36613848 rs7128419 0.89 1.47 4.91E-05 1.71E-05 0.90 0.98 f 
36618299 rs12271660 0.90 1.50 5.64E-05 4.56E-05 0.92 0.96 f 
36684837 rs12294719 0.79 1.44 5.42E-07 6.72E-07 0.81 1.00 0.48 
36687460 rs1533588 0.79 1.41 1.79E-06 3.78E-06 0.82 0.97 f 
11q13.5 75709727 rs12295401 0.06 1.53 4.40E-05 5.16E-05 0.06 1.04 0.29 
12p13.3 760163 rs11064524 0.20 1.32 2.80E-05 1.21E-04 0.24 1.08 0.0447 
12q21.31 82691472 rs7954761 0.60 1.34 2.09E-07 2.59E-07 0.61 0.99 f 
12q24 105474117 rs4964469 0.33 1.27 2.73E-05 1.30E-04 0.37 1.11 0.0045 
105513235 rs1035767 0.11 1.42 1.50E-05 2.15E-05 0.11 0.98 f 
13q34 113253980 rs2259599 0.83 1.39 3.74E-05 5.61E-03 0.88 0.96 f 
17p13.2 4376339 rs9899558 0.73 1.34 5.04E-06 3.82E-04 0.77 0.98 f 
22q11.23 22917303 rs9608247 0.16 1.33 2.99E-05 9.67E-05 0.17 0.95 f 

aRisk allele in stage-1 data.

bRisk allele frequency in controls.

cLogistic tests corrected for genomic inflation.

dLogistic tests including 2PCs as covariates.

eP-values shown when the direction of effect in stage-1 and stage-2 data are consistent.

fOne-tailed P > 0.5.

Figure 1.

Manhattan plot of the genome-wide association results for 492 929 SNPs. Logistic analysis corrected for genomic inflation (GC results).

Figure 1.

Manhattan plot of the genome-wide association results for 492 929 SNPs. Logistic analysis corrected for genomic inflation (GC results).

The 50 top SNPs were tested for in silico replication in the WTCCC2 PD study data (Table 1). For the sake of clarity, OR values are reported as a function of the number of risk alleles as identified in the stage-1 data. Associations for all 15 SNPs in SNCA and MAPT genes were replicated at nominal P-values < 4 × 10−5. Association with the BST1 variant was also replicated but at a weaker significance level (OR = 1.08, P = 0.025). For all SNPs at SNCA, MAPT and BST1, the results in the French scan and the UK replication samples were highly congruent in terms of risk alleles and allele frequencies. As expected, ORs estimated in our scan study tend to be higher than those obtained in the replication-stage data, especially for BST1. Of the remaining 20 loci, association signals at three loci (four SNPs) were replicated with nominal P < 5% and with the same direction of effect. These SNPs were located on chromosomes 2q21.3 (rs621341, OR = 1.08, P = 0.028 and rs6723108, OR = 1.11, P = 0.005), 12p13.3 (rs11064524, OR = 1.08, P = 0.045) and 12q24 (rs4964469, OR = 1.11, P = 0.0045). Differences in allele frequencies across the data from France and UK were notable for the 2q21–q22 SNPs. Indeed, the region encompasses the LCT (lactase) gene whose SNPs are known to vary in frequency across Europe, and rs6723108 has been shown to have different allele frequencies in the French and the UK–Irish populations (7).

We further followed-up the five replicated SNPs (from three newly identified loci and from BST1) in the second replication dataset (1527 cases and 1864 controls from France and Australia) (Table 2). In stage 3, evidence of association was assessed with the Mantel–Haenszel test to control for the potential confounding owing to the different geographical origins (France versus Australia). Evidence of association was replicated for two SNPs located on 12q24 (rs4964469, OR = 1.12, P = 0.0175) and on 4p15/BST1 (rs4698412, OR = 1.10, P = 0.029). Association signals from joint analysis of the two replication datasets were improved for the same two SNPs only: at 4p15/BST1 (stage 2 + stage 3, P = 0.0033) and at 12q24 (stage 2 + stage 3, P = 0.00036) loci. Notably, joint analysis of the three datasets showed a consistently greater support for association for the newly identified locus on chromosome 12q24 (P = 2.38 × 10−7) than for 4p15/BST1 (P = 1.79 × 10−6). Additional analyses showed that our initial association signals were not confounded by age and they did not appear to be driven either by the subgroup of cases having an early age (<50) of onset of the disease or by those having a positive family history of PD (results not shown). The population-attributable risk (PAR) associated with SNCA, MAPT, BST1 and the 12q24 locus estimated in stage-1 data was 11, 20, 13 and 8%, respectively; in the combined data, PAR was 7% and 4% for BST1 and 12q24, respectively.

Table 2.

GWAS and replication: loci considered to follow-up

             Combined
 
 Stage
 
Stage-1 (France) Stage-2 (UK) Stage-3 (France/Australia) Stage 2 + 3
 
Stage 1 + 2 + 3
 
 n (case/control) (1039/1984) (1705/5200) (1527/1864) (3232/7064) (4271/9048) 
Locus Position (bp) SNP RAa RAFb %OR (95% CI) P (two-tailed) RAF OR (95% CI) P (one-tailed)c RAF OR (95% CI) P (one-tailed)c,d OR (95% CI) P (one-tailed)c,d OR (95% CI) P (two-tailed)d 
BST1 15346446 rs4698412 0.52 1.28 (1.15–1.42) 6.9E-06 0.55 1.08 (1.00–1.17) 0.0247 0.54 1.10 (1.00–1.21) 0.0290 1.09 (1.02–1.16) 0.00333 1.14 (1.08–1.20) 1.79E-06 
2q21.3 135011650 rs621341 0.28 1.30 (1.16–1.46) 5.1E-06 0.48 1.08 (1.00–1.17) 0.0277 0.35 1.02 (0.92–1.13) 0.3399 1.06 (1.00–1.13) 0.03500 1.12 (1.04–1.18) 7.50E-05 
135196450 rs6723108 0.31 1.25 (1.12–1.40) 6.9E-05 0.51 1.11 (1.02–1.20) 0.0053 0.38 1.03 (0.93–1.14) 0.3047 1.08 (1.01–1.15) 0.00916 1.12 (1.06–1.18) 2.99E-05 
12p13.3 760163 rs11064524 0.20 1.32 (1.16–1.50) 2.8E-05 0.24 1.08 (0.99–1.18) 0.0447 0.25 0.93 (0.83–1.04) e – – – – 
12q24 105474117 rs4964469 0.33 1.27 (1.13–1.41) 2.7E-05 0.37 1.11 (1.03–1.21) 0.0045 0.34 1.12 (1.01–1.24) 0.0175 1.12 (1.05–1.19) 0.00036 1.16 (1.09–1.22) 2.38E-07 
             Combined
 
 Stage
 
Stage-1 (France) Stage-2 (UK) Stage-3 (France/Australia) Stage 2 + 3
 
Stage 1 + 2 + 3
 
 n (case/control) (1039/1984) (1705/5200) (1527/1864) (3232/7064) (4271/9048) 
Locus Position (bp) SNP RAa RAFb %OR (95% CI) P (two-tailed) RAF OR (95% CI) P (one-tailed)c RAF OR (95% CI) P (one-tailed)c,d OR (95% CI) P (one-tailed)c,d OR (95% CI) P (two-tailed)d 
BST1 15346446 rs4698412 0.52 1.28 (1.15–1.42) 6.9E-06 0.55 1.08 (1.00–1.17) 0.0247 0.54 1.10 (1.00–1.21) 0.0290 1.09 (1.02–1.16) 0.00333 1.14 (1.08–1.20) 1.79E-06 
2q21.3 135011650 rs621341 0.28 1.30 (1.16–1.46) 5.1E-06 0.48 1.08 (1.00–1.17) 0.0277 0.35 1.02 (0.92–1.13) 0.3399 1.06 (1.00–1.13) 0.03500 1.12 (1.04–1.18) 7.50E-05 
135196450 rs6723108 0.31 1.25 (1.12–1.40) 6.9E-05 0.51 1.11 (1.02–1.20) 0.0053 0.38 1.03 (0.93–1.14) 0.3047 1.08 (1.01–1.15) 0.00916 1.12 (1.06–1.18) 2.99E-05 
12p13.3 760163 rs11064524 0.20 1.32 (1.16–1.50) 2.8E-05 0.24 1.08 (0.99–1.18) 0.0447 0.25 0.93 (0.83–1.04) e – – – – 
12q24 105474117 rs4964469 0.33 1.27 (1.13–1.41) 2.7E-05 0.37 1.11 (1.03–1.21) 0.0045 0.34 1.12 (1.01–1.24) 0.0175 1.12 (1.05–1.19) 0.00036 1.16 (1.09–1.22) 2.38E-07 

aRisk allele in stage-1 data.

bRisk allele frequency in controls; %odds ratio computed for the stage-1 risk allele.

cP-values shown when the direction of effect in stage-1 and each replication data are consistent.

dP-values from stratified association tests.

eOne-tailed P > 0.5.

Finally, we also examined 18 SNPs from five loci, previously reported to be associated with PD at a genome-wide significance level from two published GWASs of PD (5,6) (Table 3). We added three SNPs (at SNCA and BST1) that were found strongly associated in our stage-1 data. The table also shows the results for a suggestive PD risk locus (GAK) reported by the published GWAS of PD from familial cases (8). As for the previously reported PD loci, two loci only (SNCA and MAPT) have been identified with genome-wide significance at the screen stage: SNCA in both the Japanese and European populations and MAPT in the European population only. The two new PD risk loci (BST1 and PARK16) were identified in the Japanese population: association signals were strong (P < 10−6) in the discovery sample, and exceeded P < 10−8 in the combined data (5). As already reported here, our GWAS provided genome-wide significance for two SNCA variants and replicated positive associations for variants at MAPT and BST1 loci. It is worth noting that allele frequencies may differ markedly between the Japanese and the European datasets, especially for SNPs at the SNCA and BST1 loci. Saliently, the directions of effects (i.e. risk allele) and effect sizes at SNCA variants are rather congruent across the European and Japanese datasets. For the remaining three PD loci, evidence of association was nominal (PARK16, PGC = 0.03; LRKK2, PGC = 0.04; GAK PGC = 0.008) in the France-GWAS data.

Table 3.

Association results of previously reported PD loci

  Japanese samples
 
European samples
 
   Scan phase
 
Combined (scan + replication)  Scan phase
 
Combined (scan + replication) France-GWAS 
SNP bp GWAS Risk all RAF OR Pa OR P GWAS Risk all RAF OR P OR P Risk all RAF OR PGC 
(A) Genome-wide significant loci 
 4q22 (SNCA)                    
 rs11931074 90858538 S1 0.58 1.50 6.2E-13 1.37 7.4E-17 S2 0.07 1.49 4.8E-08 1.46 1.6E-14 0.07 1.52 1.3E-05 
 rs356220 90860363 NA       NA       0.35 1.37 2.82E-08 
 rs3857059 90894261 S1 0.59 1.49 1.2E-12 1.36 5.7E-16 S2 0.07 1.49 3.6E-08 1.48 3.7E-15 0.07 1.54 1.0E-05 
 rs2736990 90897564 NA       S2 0.46 1.27 5.7E-09 1.23 2.2E-16 0.44 1.35 2.9E-08 
vrs6532194 90999925 S1 0.6 1.44 7.0E-11 1.32 4.2E-13 NA       0.09 1.22 0.028 
17q12-q22 (MAPT) 
 rs393152 41074926 NA       S2 0.78 1.32 1.4E-07 1.30 2.0E-16 0.75 1.32 2.7E-05 
 rs17563986 41347100 NA       S2 0.78 1.30 3.4E-07 1.28 1.7E-14 0.75 1.34 1.3E-05 
 rs199533 42184098 NA       S2 0.80 1.33 5.1E-08 1.28 1.1E-14 0.78 1.28 3.0E-04 
4p15 (BST1) 
 rs3213710 15326419 NA       NA       0.50 1.24 7.7E-05 
 rs4698412 15346446 S1 0.33 1.25 5.3E-05 1.24 1.8E-08 S2 0.56 1.07 0.09 1.06 0.03 0.52 1.28 6.9E-06 
 rs4538475 15347035 S1 0.36 1.25 4.1E-05 1.24 3.9E-09 NA       0.83 1.15 0.07 
 rs12646913 15348374 NA       S2 0.92 1.18 0.04 1.09 0.03 0.91 1.19 0.08 
 rs4698120 15352430 NA       NA       0.53 1.24 7.4E-05 
1q32 (PARK16) 
 rs16856139 203905087 S1 0.86 1.50 2.6E-06 1.46 1.0E-07 NA       0.06 1.10 
 rs947211 204019288 S1 0.52 1.23 1.2E-04 1.3 1.5E-12 NA       0.78 1.10 0.14 
 rs823156 204031263 S1 0.83 1.40 1.2E-05 1.37 3.6E-09 S2 0.82 1.12 4.3E-03 1.12 7.6E-04 0.81 1.17 0.03 
 rs708730 204044403 S1 0.82 1.37 2.6E-05 1.33 2.4E-08 NA       0.82 1.12 0.13 
12q12 (LRRK2) 
 rs11564162 38729159 NA       S2 0.81 1.28 4.0E-05 1.15 9.5E-05 0.17 1.11 
 rs2708453 38764919 S1 0.08 1.41 7.5E-05 1.38 9.7E-08 NA       0.16 1.16 0.04 
 rs2896905 38779683 NA       S2 0.60 1.22 5.0E-06 1.07 7.8E-03 0.64 1.02 0.75 
 rs1491923 38877384 NA       S2 0.31 1.20 2.2E-04 1.14 1.6E-05 0.31 1.10 0.09 
(B) Suggestive loci 
 4p16 (GAK)                    
 rs1564282 842 313 NA       S3 0.09 1.7 6.0E-06   0.09 1.27 0.008 
 rs11248060 954 359 NA       S3 0.10 1.69 3.4E-06   0.11 1.18 0.05 
  Japanese samples
 
European samples
 
   Scan phase
 
Combined (scan + replication)  Scan phase
 
Combined (scan + replication) France-GWAS 
SNP bp GWAS Risk all RAF OR Pa OR P GWAS Risk all RAF OR P OR P Risk all RAF OR PGC 
(A) Genome-wide significant loci 
 4q22 (SNCA)                    
 rs11931074 90858538 S1 0.58 1.50 6.2E-13 1.37 7.4E-17 S2 0.07 1.49 4.8E-08 1.46 1.6E-14 0.07 1.52 1.3E-05 
 rs356220 90860363 NA       NA       0.35 1.37 2.82E-08 
 rs3857059 90894261 S1 0.59 1.49 1.2E-12 1.36 5.7E-16 S2 0.07 1.49 3.6E-08 1.48 3.7E-15 0.07 1.54 1.0E-05 
 rs2736990 90897564 NA       S2 0.46 1.27 5.7E-09 1.23 2.2E-16 0.44 1.35 2.9E-08 
vrs6532194 90999925 S1 0.6 1.44 7.0E-11 1.32 4.2E-13 NA       0.09 1.22 0.028 
17q12-q22 (MAPT) 
 rs393152 41074926 NA       S2 0.78 1.32 1.4E-07 1.30 2.0E-16 0.75 1.32 2.7E-05 
 rs17563986 41347100 NA       S2 0.78 1.30 3.4E-07 1.28 1.7E-14 0.75 1.34 1.3E-05 
 rs199533 42184098 NA       S2 0.80 1.33 5.1E-08 1.28 1.1E-14 0.78 1.28 3.0E-04 
4p15 (BST1) 
 rs3213710 15326419 NA       NA       0.50 1.24 7.7E-05 
 rs4698412 15346446 S1 0.33 1.25 5.3E-05 1.24 1.8E-08 S2 0.56 1.07 0.09 1.06 0.03 0.52 1.28 6.9E-06 
 rs4538475 15347035 S1 0.36 1.25 4.1E-05 1.24 3.9E-09 NA       0.83 1.15 0.07 
 rs12646913 15348374 NA       S2 0.92 1.18 0.04 1.09 0.03 0.91 1.19 0.08 
 rs4698120 15352430 NA       NA       0.53 1.24 7.4E-05 
1q32 (PARK16) 
 rs16856139 203905087 S1 0.86 1.50 2.6E-06 1.46 1.0E-07 NA       0.06 1.10 
 rs947211 204019288 S1 0.52 1.23 1.2E-04 1.3 1.5E-12 NA       0.78 1.10 0.14 
 rs823156 204031263 S1 0.83 1.40 1.2E-05 1.37 3.6E-09 S2 0.82 1.12 4.3E-03 1.12 7.6E-04 0.81 1.17 0.03 
 rs708730 204044403 S1 0.82 1.37 2.6E-05 1.33 2.4E-08 NA       0.82 1.12 0.13 
12q12 (LRRK2) 
 rs11564162 38729159 NA       S2 0.81 1.28 4.0E-05 1.15 9.5E-05 0.17 1.11 
 rs2708453 38764919 S1 0.08 1.41 7.5E-05 1.38 9.7E-08 NA       0.16 1.16 0.04 
 rs2896905 38779683 NA       S2 0.60 1.22 5.0E-06 1.07 7.8E-03 0.64 1.02 0.75 
 rs1491923 38877384 NA       S2 0.31 1.20 2.2E-04 1.14 1.6E-05 0.31 1.10 0.09 
(B) Suggestive loci 
 4p16 (GAK)                    
 rs1564282 842 313 NA       S3 0.09 1.7 6.0E-06   0.09 1.27 0.008 
 rs11248060 954 359 NA       S3 0.10 1.69 3.4E-06   0.11 1.18 0.05 

S1: Table 1 from Satake et al., Nat. Genet., 2009; Japanese data – scan phase (1078 PD/2521 controls). S2: Table 2 from Simon-Sanchez et al., Nat. Genet., 2009; US/GE/UK data – scan phase (1745 PD/4047 controls). S3: Table 2 from Pankratz et al., Hum. Genet., 2009; US (PROGENI + GenePD) data – (857 familial PD cases/867 controls).

aGenome-wide significant values (<10−7) are italicized.

DISCUSSION

Our genome-wide association analyses in the French scan data revealed two SNPs with genome-wide significance (PGC < 10−7), and a number of additional SNPs with suggestive evidence (PGC < 10−4). Here, we focused on the 50 top associated SNPs for immediate replication in two independent case–control samples. We used a stepwise replication design. To refine the set of most promising results, we first conducted in silico replication for the 50 SNPs in the WTCCC2 PD data (1705 cases and 5200 WTCCC controls). Replicated SNPs were genotyped and tested in a third dataset of 1527 cases and 1864 controls from France and Australia. Our scan stage showed genome-wide significance of association with PD for two SNPs at the 4q22/SNCA locus (PGC < 2.88 × 10−8). Indeed, out of the 50 top associated SNPs, 15 are located in genomic regions of two known PD genes (SCNCA, MAPT) and one is located on 4p15/BST1, a risk locus recently reported with genome-wide significance in Japanese samples. SNPs at SNCA and MAPT were all significantly associated with PD in the UK-GWAS data (SNCA, P < 8 × 10−5; MAPT, P < 2.75 × 10−6). Evidence of association with 4p15/BST1 was also replicated in the UK sample but at a lower (rs4698412, P = 0.025) significance level. Out of the remaining 34 SNPs, four SNPs (three loci) showed significant (P < 0.05) and consistent evidence of association in the UK data. A total of five SNPs (four loci: 2q21.3, 12p13.3, 12q24 and BST1) were followed-up for replication in the third case–control sample. Two of the four tested regions were replicated: 4p15/BST1 (P = 0.03) and 12q24 (P = 0.018). Of the four regions, only one (12p13.3) showed no evidence of association from the combined analysis of the two replication datasets. Overall, evidence of association was consistently stronger with the region of the newly identified PD risk locus than with BST1, in each replication sample as well as in the combined (genome-wide and two replication samples) data (12q24, P = 2.38 × 10−7; BST1, P = 1.79 × 10−6).

The evidence of association (PGC < 1.35 × 10−5, Tables 1 and 3) that we detected with several SNPs in the 3′ block of linkage disequilibrium (LD) of the SNCA locus (Fig. 2A), including the two SNPs reaching genome-wide significance in our scan sample, is highly consistent with previous PD GWAS studies (5,6).

Figure 2.

Regional association plots and LD structure for the four PD risk loci (A) 4q22/SNCA, (B) 17q12–q22/MAPT, (C) 4p15/BST1 and (D) 12q24/RFX4. The –log10 P-values (logistic regression tests corrected for genomic inflation) in the GWAS stage. In each panel, the blue horizontal line indicates a P-value of 5 × 10−5. Pairwise linkage disequilibrium (D′) values are displayed and the SNPs with the strongest association signals are circled. SNPs are color-coded for LD relationships (r2) to the best (colored in black) SNP: red, 0.8 ≤ r2 < 1; green, 0.6 ≤ r2 < 0.8; gray, 0.4 ≤ r2 < 0.6; blue, 0 ≤ r2 < 0.4. Positions are NCBI build 36 coordinates. Intron and exon structures of genes are taken from the UCSC Genome Browser.

Figure 2.

Regional association plots and LD structure for the four PD risk loci (A) 4q22/SNCA, (B) 17q12–q22/MAPT, (C) 4p15/BST1 and (D) 12q24/RFX4. The –log10 P-values (logistic regression tests corrected for genomic inflation) in the GWAS stage. In each panel, the blue horizontal line indicates a P-value of 5 × 10−5. Pairwise linkage disequilibrium (D′) values are displayed and the SNPs with the strongest association signals are circled. SNPs are color-coded for LD relationships (r2) to the best (colored in black) SNP: red, 0.8 ≤ r2 < 1; green, 0.6 ≤ r2 < 0.8; gray, 0.4 ≤ r2 < 0.6; blue, 0 ≤ r2 < 0.4. Positions are NCBI build 36 coordinates. Intron and exon structures of genes are taken from the UCSC Genome Browser.

MAPT is located in a large block of LD on chromosome 17q12–q22, which contains several additional genes (Fig. 2B). Previous studies (9,10) have identified a large haplotypic block associated with PD, with H1 and H2 being the at-risk and the protective haplotype, respectively. Our two most associated SNPs in the 17q12–q22 region are located within this haplotypic block: rs17690703 (PGC = 3.9 × 10−6) and rs17563986 (PGC = 1.3 × 10−5), the latter being at MAPT. In addition, H2 is tagged by the minor alleles of four of our genotyped SNPs: rs12185268/G, rs12373139/A, rs1981997/A and rs8070723/G. In our scan data, we found the same H1/H2 association signals, with all minor alleles of these four SNPs being significantly associated (PGC < 3.44 × 10−5) with a decreased risk of PD (Table 1).

The BST1 gene has previously been associated with PD in a Japanese GWAS at a genome-wide significance level (5). Strong evidence of association for rs4698412 was found in the Japanese scan (P = 5.3 × 10−5, OR = 1.25) and in the combined (scan + replication) data (P = 1.8 × 10−8, OR = 1.24) (5). A much weaker signal was obtained in the US/UK/German data, in both the scan (P = 0.09, OR = 1.07) and the combined (P = 0.03, OR = 1.06) data (6). Here, we report strong evidence of association of PD with BST1 (combined P = 1.79 × 10−6, OR = 1.14). The most associated SNP (rs4698412) maps to a 15 kb LD-block (Fig. 2C) and is in high LD (r2 = 0.74/0.79) with the next top two BST1 variants (Table 3). Despite the variation in the allele frequency of the risk allele between the Japanese (RAF = 0.33) and the European (RAF = 0.52–0.56) samples (Tables 2 and 3), we found marked homogeneity in the direction of effects across the groups, but effect sizes seemed to be lower in European than in Japanese samples. BST1 has been proposed to play a role in generating cyclic ADP-ribose that serves as a second messenger for Ca2+ mobilization in endoplasmic reticulum and thus Ca homeostasis-related BST1 could be a cause of selective vulnerability of dopaminergic neurons in PD (11).

Our most associated SNP, on 12q24 (combined P = 2.38 × 10−7, OR = 1.16), is 26 kb centromeric of RFX4 (Regulatory factor X4) (Fig. 2D). Two other close genes, POLR3B (Polymerase RNA III polypeptide B) and RIC8B (Resistance to inhibitors of cholinesterase 8 homolog B), are 200 kb centromeric and telomeric of the 12q24 SNP, respectively. The RFX proteins belong to the winged-helix subfamily of helix–turn–helix transcription factors. The RFX4_v3 transcript variant is the only RFX4 isoform that is significantly expressed in the fetal and adult brain, and its expression is restricted to the brain. In addition, it has a role in the transcription of many genes involved in brain morphogenesis, such as the signaling components in the wnt, bone morphogenetic protein (BMP) and retinoic acid (RA) pathways. In particular, cx3cl1, a CX3C-type chemokine gene, which is highly expressed in brain in response to injury or infection and regulates intracellular calcium concentration, was downregulated in RFX4_v3-null mice (12). This allows speculation that RFX4 and BST1 are functionally linked and indirectly involved in the regulation of intracellular Ca2+ concentrations, which plays an important role in various cellular functions and cell death. Finally, polymorphisms in RFX4 have been shown to be risk factors for the bipolar disorder, manic-depressive illness (13). A recent study showed that a substantial proportion (10–15%) of top GWAS hits, so far identified, are e-quantitative trait loci (eQTLs), i.e. associated with gene expression levels (14). We have initiated eQTL analysis using an existing brain expression database (15), but so far failed to identify any association of the PD-associated rs4964469 SNP with the expression of known genes contained within the 12q24 region.

In conclusion, we have conducted a large GWAS of PD in three case–control samples from France, the UK and Australia. The GWAS stage has 75% and 33% power to detect the loci of the effect sizes observed in stage-1 data for the 12q24 variant (OR = 1.27) at a significance of P < 5 × 10−5 and P < 10−7, respectively. In the scan-step, we detected genome-wide significance of association with PD for two SNPs on 4q22, and strong evidence of association with 17q12–q22 SNPs. The two regions encompass previously reported loci: SNCA and MAPT, respectively. In addition, we confirmed, for the first time in subjects of European ancestry, the association of PD with 4p15/BST1, recently identified in Japanese samples. Finally, we identified a new locus on 12q24, potentially associated with PD. Further replication studies conducted in large case–control samples are warranted to evaluate the contribution of this locus to PD risk.

MATERIALS AND METHODS

Sample ascertainment and diagnostic criteria

The main characteristics of the three case–control samples are shown in Table 4.

Table 4.

Samples used (post-QC) in this study

 Stage-1 Stage-2 Stage-3 Total 
 Scan Replication Replication  
Center French UK French–Australian  
Genotyping platform Illumina 610-Quad Illumina 650Y Illumina GoldenGate  
Cases 1039 1705 1527 4271 
Sex ratio: M/F 1.42 1.37 1.42  
Age: mean ± SD (n)a 57.5 ± 16.6 (1003) NA 69.0 ± 12.7 (1365)  
AOO: mean ± SD (n)a 48.9 ± 12.8 (970) 65.2 ± 11.3 (1109) 61.3 ± 12.4 (1351)  
FH+ (%) 47 17  
Controls 1984 5200 1864 9048 
Sex ratio (M/F) 1.33 1.02 1.05  
Age: mean ± SD (n)a 73.7 ± 5.4 (1984) 51 68.1 ± 10.0  
Total 3023 6905 3391 13 319 
 Stage-1 Stage-2 Stage-3 Total 
 Scan Replication Replication  
Center French UK French–Australian  
Genotyping platform Illumina 610-Quad Illumina 650Y Illumina GoldenGate  
Cases 1039 1705 1527 4271 
Sex ratio: M/F 1.42 1.37 1.42  
Age: mean ± SD (n)a 57.5 ± 16.6 (1003) NA 69.0 ± 12.7 (1365)  
AOO: mean ± SD (n)a 48.9 ± 12.8 (970) 65.2 ± 11.3 (1109) 61.3 ± 12.4 (1351)  
FH+ (%) 47 17  
Controls 1984 5200 1864 9048 
Sex ratio (M/F) 1.33 1.02 1.05  
Age: mean ± SD (n)a 73.7 ± 5.4 (1984) 51 68.1 ± 10.0  
Total 3023 6905 3391 13 319 

aNumber of subjects for which age/age of onset of disease is known.

Stage-1 subjects

The total number of cases and controls from France included in stage 1 was 1070 and 2023 controls, respectively.

  • PD subjects: Patients were recruited through the French network for the study of Parkinson's disease Genetics (PDG) that comprises 15 university hospitals across France. Definite and probable PD was defined according to standard criteria. Definite PD required at least two of three cardinal signs (akinesia and/or rigidity and/or tremor) and absence of exclusion criteria (ophthalmoplegia, pyramidal or cerebellar signs, early dementia, urinary incontinence or postural instability and prior exposure to neuroleptic drugs), and a positive and sustained response to levodopa therapy. Probable PD required at least two of the five following criteria: the parkinsonian triad, a good response to levodopa therapy and asymmetrical onset. Most (>80%) of the PD cases fulfilled the criteria for definite PD. Patients were selected in an effort to enrich for individuals who may have greater genetic predisposition to PD, through selection of cases with a positive family history of PD (Table 4). Cases were of European origin, mostly French (n = 930). Subjects diagnosed genetically with known PARK mutations (SNCA, LRRK2, parkin and PINK1) were excluded.

  • 3C neurologically normal controls: The French Three-City (3C) cohort is a population-based, prospective (4-year follow-up) study of the relationship between vascular factors and dementia, carried out in three French cities: Bordeaux (Southwest France), Dijon (central eastern France) and Montpellier (Southeast France) (16). Participants (>9000) are non-institutionalized subjects, over 65 years of age, randomly selected from the electoral rolls of each city. Patients with Alzheimer's disease or other types of dementia, and individuals for whom information on their dementia status during the 4-year follow-up was missing were further excluded. Here, we used a sample of 2023 neurologically normal subjects matched on gender with PD cases, randomly selected from all the participants.

Stage-2 subjects

In silico replication sample: we exchanged genome-wide association data with the WTCCC2 PD study group (Spencer et al., submitted). This case–control study consisted of 1705 PD cases and 5200 controls from the 1958 Birth Cohort and from the OK Blood Services Controls (17).

Stage-3 subjects

De novo genotyping was conducted in two independent case–control datasets from France (872 PD, 1440 controls) and Australia (655 PD, 424 controls). The subjects from France were combined from three French studies: TERRE (207 cases, 468 controls), PARTAGE (313 cases, 593 controls) and an extension of PDG (352 cases, 378 controls). The extension PDG study includes patients who were not available at the time of the stage-1 genotyping execution and neurologically normal spouses of PDG patients. In cases, the mean age at examination and the mean age of onset of PD is 59 (30–86) and 50 (20–84) years, respectively. The mean age of controls is 60 (31–85) years. In PARTAGE, patients and controls were identified among affiliates of the Mutualité Sociale Agricole (MSA) from five French districts. Parkinsonism was defined as the presence of at least two cardinal signs (rest tremor, bradykinesia, rigidity, impaired postural reflexes); PD was defined as the presence of parkinsonism after exclusion of other causes of parkinsonism. Controls were randomly selected from all MSA affiliates in the same districts and matched for sex and age (±2 years). DNA was collected from saliva (Oragene kit). Cases and controls have a mean age of 67 (37–79) years, and the mean age of onset of disease is 63 (35–75) years. TERRE is based on a similar protocol (18), but DNA was collected from blood; the mean age in cases and controls is 73 (46–82) years, and the mean age of onset is 66 (39–80) years in cases.

Australian study: Subjects with PD were recruited from one private and two public movement disorder clinics in Brisbane. Controls were electoral roll volunteers and patient spouses, excluding the subjects demonstrating signs of parkinsonism (19). The mean age is 72 (34–105) and 74 (33–107) years in controls and cases, respectively; the mean age of onset is 59 (23–96) years in cases. Only Caucasian subjects were included in stage 3; in the Australian study, analyses were restricted to participants who reported having four European grandparents (>85% British). There was no overlap between the subjects used in the replication datasets and those included in the stage-1 data. Written informed consent was obtained for all participating subjects and research protocols were approved by local ethics committees.

Genotyping

Stage-1 genotyping

DNA samples of PDG cases and 3C controls were transferred to the French Centre National de Génotypage. First-stage samples that passed DNA quality control (QC) (1064 PD cases and 2023 controls) were genotyped with Illumina Human610-Quad BeadChip and subjected to standard QC procedures.

Stage-2 genotyping

This WTCCC2 PD study sample was genotyped by the Welcome Trust Case–Control Consortium using the Illumina 650Y genotyping array (Spencer et al., submitted).

Stage-3 genotyping

Genotyping in the extended PDG sample was carried out in the UMR/S 975 laboratory, using predesigned TaqMan probes (C_537709_10/ rs621341; C_29330880_10/ rs6723108; C_12096605_10/ rs11064524; C_2775670_10/ rs4964469; C_1216796_10/ rs4698412) on an ABI 7500 Real-Time PCR system Applied Biosystems, Foster City, CA, USA), according to the manufacturer's instructions. Data were then analyzed using the 7500 software v.2.0.1. The TERRE/PARTAGE and Australian samples were genotyped using the Sequenom MassARRAY platform, with the iPLEX protocol (Genoscreen, France). The basic protocol involves a multiplex primer extension followed by matrix-assisted laser desorption ionization-time of flight mass spectroscopy detection. In order to avoid any genotyping bias, cases and controls were randomly mixed when genotyping and, laboratory personnel were blinded to case–control status.

Quality control of France GWAS scan data

Various stringent QC filters were applied to remove poorly performing SNPs and samples using tools implemented in PLINK version 1.7 (20).

SNP QC: Markers were removed if they had a genotype-missing rate >0.03 or a minor allele frequency (MAF) < 0.05 or a Hardy–Weinberg P ≤ 10−5. This SNP QC step led to the removal of 74 660 autosomal SNPs. Thus, subsequent analyses were based on 492 929 SNPs.

Individual QC: Samples were removed based on standard exclusion criteria: call rate of <96% (22 subjects), inconsistencies between reported gender and genotype-determined gender (11 subjects) and genetic relatedness (identity-by-descent estimate >0.14; 6 subjects). Applying these QC filters led to the removal of 39 subjects (14 cases, 25 controls).

Population stratification and principal component analysis: To detect individuals of non-European ancestry, we thinned the SNPs to reduce LD to a set of 55 193 SNPs. To this end, we removed SNPs from the extensive regions of LD (CHR2, CHR5, CHR6, CHR8, CHR11) (21), and excluded SNPs if any pair within a 1000-SNP window had r2> 0.2. Our stage-1 genotype data were then merged with genotypes at the same SNPs from 381 unrelated European (CEU), Yoruban (YRI) and Asian (CHB and JPT) samples from the HapMap project. Principal component analysis was applied using EIGENSTRAT (22). The two PCs clearly separated the HapMap data into three distinct clusters according to ancestry, and most of our stage-1 samples were clustered with the HapMap European samples (Fig. 3). Thirty-two samples appeared to be ethnic outliers (including one subject clearly sharing African ancestry) from the European cluster and were excluded from further analysis. The final post-QC scan sample comprised 1039 PD cases and 1984 controls.

Figure 3.

Principal components for our genome-wide stage 1. Plot of the first two principal components from the analysis of our stage-1 (post-QC) data combined with HapMap data. Ethnicity of HapMap samples indicated by color: Africa (YRI) in green, Japan (JPT) in brown, Chinese (6) in yellow and Europe (CEU) in red. Study samples identified to be non-European or not clustering with European samples (outliers) are colored in blue and the remaining study samples assumed to be of European origin are colored in black.

Figure 3.

Principal components for our genome-wide stage 1. Plot of the first two principal components from the analysis of our stage-1 (post-QC) data combined with HapMap data. Ethnicity of HapMap samples indicated by color: Africa (YRI) in green, Japan (JPT) in brown, Chinese (6) in yellow and Europe (CEU) in red. Study samples identified to be non-European or not clustering with European samples (outliers) are colored in blue and the remaining study samples assumed to be of European origin are colored in black.

Statistical analysis

Association analysis of the genotype data was conducted with PLINK (20).

Stage-1 association analyses

Logistic regression was used to study the allelic association between each SNP and PD assuming an additive genetic model. Our analysis was based on 492 929 SNPs, and on a conservative genome-wide significance threshold of 0.05/492 929 = 10−7. The distribution of the association results was found to be marginally inflated (median χ2 = 0.521); genomic inflation factor λ = 1.14 (λ1000 = 1.10). Logistic regression analysis adjusted for the two first PCs of the EIGENSTRAT analysis revealed a genomic inflation of 1.03 (median χ2 = 0.472). As for our primary analyses, we applied the genomic inflation correction method (23); the median of the GC-corrected χ2 value was 0.447.

Sensitivity analyses: Two further analyses were conducted to assist in the interpretation of results of the identified GWAS SNPs. We performed age-adjusted regression analysis and conducted subgroup analyses of two subtypes of cases against all controls. Cases with a disease onset before 50 years (n = 428) were classified as ‘early AOO', and cases having at least one first-degree relative with PD (n = 452) were classified as ‘FH+'.

Stage-2 in silico association analyses

Statistical data (ORs, effective sample sizes and nominal P-values for each of the 50 top SNPs) in the UK sample were obtained from the WTCC2 PD study group that used similar analytical methods (Spencer et al., submitted).

Stage-3 association analyses

For the de novo replication stage, we computed association statistics with the Mantel–Haenszel test to control for the potential confounding owing to the geographical center (France versus Australia) for the five SNPs replicated at stage-2. Using raw genotypes from all the study samples, we computed similar stratified (France versus Australia versus UK) association statistics in the combined (stage-2 + stage-3 and stage-1 + stage-2 + stage-3) data.

The PAR associated with the detected variants was estimated with the following formula: PAR = p (OR-1)/[p(OR-1) +1], where p is the frequency of the risk allele in controls, and OR is the odds ratio associated with the risk allele.

AUTHOR CONTRIBUTIONS

S.L. supervised DNA sampling; J.C.C., M.V., E.B., F.D., P.P., P.D., F.T., A.D. and A.B. recruited patients; J.C.C., A.D. and AB supervised clinical work; D.Z. and M.L. supervised PD and 3C GWAS genotyping and DNA QC work; S.L. and J.C.L. supervised genotyping of stage-3 samples. A.E., J.C.L., M.A.L., C.T., G.D.M. and P.A.S. contributed to stage-3 replication; M.S., A.S.P. and M.M. executed QC analyses and performed statistical association analyses; A.E., A.B. and M.M. were involved in obtaining funding; M.M. drafted the manuscript and S.L., A.B. and A.E. contributed to the writing of the final version; A.B. and M.M. conceived and oversaw the design and execution of the GWAS.

FUNDING

This work was supported by the French National Agency of Research (ANR-08-MNP-012).

ACKNOWLEDGEMENTS

The authors are grateful to the patients and their families. They thank the DNA and Cell Bank of UMR_S975 for sample preparation. We thank the members of the 3C consortium: Drs Annick Alpérovitch, Claudine Berr and Jean-Francois Dartigues for giving us the possibility to use part of the 3C cohort. This study makes use of data generated by the Wellcome Trust Case–Control Consortium. A full list of the investigators who contributed to the generation of the data is available from www.wtccc.org.uk.

Conflict of Interest statement. The authors declare no competing financial interests.

REFERENCES

1
Lesage
S.
Brice
A.
Parkinson's disease: from monogenic forms to genetic susceptibility factors
Hum. Mol. Genet.
 , 
2009
, vol. 
18
 (pg. 
R48
-
R59
)
2
Maraganore
D.M.
De Andrade
M.
Lesnick
T.G.
Strain
K.J.
Farrer
M.J.
Rocca
W.A.
Pant
P.V.
Frazer
K.A.
Cox
D.R.
Ballinger
D.G.
High-resolution whole-genome association study of Parkinson disease
Am. J. Hum. Genet.
 , 
2005
, vol. 
77
 (pg. 
685
-
693
)
3
Fung
H.C.
Scholz
S.
Matarin
M.
Simon-Sanchez
J.
Hernandez
D.
Britton
A.
Gibbs
J.R.
Langefeld
C.
Stiegert
M.L.
Schymick
J.
, et al.  . 
Genome-wide genotyping in Parkinson's disease and neurologically normal controls: first stage analysis and public release of data
Lancet Neurol.
 , 
2006
, vol. 
5
 (pg. 
911
-
916
)
4
Myers
R.H.
Considerations for genomewide association studies in Parkinson disease
Am. J. Hum. Genet.
 , 
2006
, vol. 
78
 (pg. 
1081
-
1082
)
5
Satake
W.
Nakabayashi
Y.
Mizuta
I.
Hirota
Y.
Ito
C.
Kubo
M.
Kawaguchi
T.
Tsunoda
T.
Watanabe
M.
Takeda
A.
, et al.  . 
Genome-wide association study identifies common variants at four loci as genetic risk factors for Parkinson's disease
Nat. Genet.
 , 
2009
, vol. 
41
 (pg. 
1303
-
1307
)
6
Simon-Sanchez
J.
Schulte
C.
Bras
J.M.
Sharma
M.
Gibbs
J.R.
Berg
D.
Paisan-Ruiz
C.
Lichtner
P.
Scholz
S.W.
Hernandez
D.G.
, et al.  . 
Genome-wide association study reveals genetic risk underlying Parkinson's disease
Nat. Genet.
 , 
2009
, vol. 
41
 (pg. 
1308
-
1312
)
7
Tian
C.
Kosoy
R.
Nassir
R.
Lee
A.
Villoslada
P.
Klareskog
L.
Hammarstrom
L.
Garchon
H.J.
Pulver
A.E.
Ransom
M.
, et al.  . 
European population genetic substructure: further definition of ancestry informative markers for distinguishing among diverse European ethnic groups
Mol. Med.
 , 
2009
, vol. 
15
 (pg. 
371
-
383
)
8
Pankratz
N.
Wilk
J.B.
Latourelle
J.C.
Destefano
A.L.
Halter
C.
Pugh
E.W.
Doheny
K.F.
Gusella
J.F.
Nichols
W.C.
Foroud
T.
, et al.  . 
Genomewide association study for susceptibility genes contributing to familial Parkinson disease
Hum. Genet.
 , 
2009
, vol. 
124
 (pg. 
593
-
605
)
9
Healy
D.G.
Abou-Sleiman
P.M.
Lees
A.J.
Casas
J.P.
Quinn
N.
Bhatia
K.
Hingorani
A.D.
Wood
N.W.
Tau gene and Parkinson's disease: a case-control study and meta-analysis
J. Neurol. Neurosurg. Psychiatry
 , 
2004
, vol. 
75
 (pg. 
962
-
965
)
10
Zhang
J.
Song
Y.
Chen
H.
Fan
D.
The tau gene haplotype h1 confers a susceptibility to Parkinson's disease
Eur. Neurol.
 , 
2005
, vol. 
53
 (pg. 
15
-
21
)
11
Zhang
D.
Stumpo
D.J.
Graves
J.P.
Degraff
L.M.
Grissom
S.F.
Collins
J.B.
Li
L.
Zeldin
D.C.
Blackshear
P.J.
Identification of potential target genes for RFX4_v3, a transcription factor critical for brain development
J. Neurochem.
 , 
2006
, vol. 
98
 (pg. 
860
-
875
)
12
Chan
C.S.
Gertler
T.S.
Surmeier
D.J.
Calcium homeostasis, selective vulnerability and Parkinson's disease
Trends Neurosci.
 , 
2009
, vol. 
32
 (pg. 
249
-
256
)
13
Glaser
B.
Kirov
G.
Bray
N.J.
Green
E.
O'donovan
M.C.
Craddock
N.
Owen
M.J.
Identification of a potential bipolar risk haplotype in the gene encoding the winged-helix transcription factor RFX4
Mol. Psychiatry.
 , 
2005
, vol. 
10
 (pg. 
920
-
927
)
14
Cookson
W.
Liang
L.
Abecasis
G.
Moffatt
M.
Lathrop
M.
Mapping complex disease traits with global gene expression
Nat. Rev. Genet.
 , 
2009
, vol. 
10
 (pg. 
184
-
194
)
15
Myers
A.J.
Gibbs
J.R.
Webster
J.A.
Rohrer
K.
Zhao
A.
Marlowe
L.
Kaleem
M.
Leung
D.
Bryden
L.
Nath
P.
, et al.  . 
A survey of genetic human cortical gene expression
Nat. Genet.
 , 
2007
, vol. 
39
 (pg. 
1494
-
1499
)
16
Alperovitch
A.
Vascular factors and risk of dementia: design of the Three-City Study and baseline characteristics of the study population
Neuroepidemiology
 , 
2003
, vol. 
22
 (pg. 
316
-
325
)
17
Wellcome Trust Case Control Consortium
Genome-wide association study of 14 000 cases of seven common diseases and 3000 shared controls
Nature
 , 
2007
, vol. 
447
 (pg. 
661
-
678
)
18
Elbaz
A.
Clavel
J.
Rathouz
P.J.
Moisan
F.
Galanaud
J.P.
Delemotte
B.
Alperovitch
A.
Tzourio
C.
Professional exposure to pesticides and Parkinson disease
Ann. Neurol.
 , 
2009
, vol. 
66
 (pg. 
494
-
504
)
19
Sutherland
G.T.
Halliday
G.M.
Silburn
P.A.
Mastaglia
F.L.
Rowe
D.B.
Boyle
R.S.
O'sullivan
J.D.
Ly
T.
Wilton
S.D.
Mellick
G.D.
Do polymorphisms in the familial Parkinsonism genes contribute to risk for sporadic Parkinson's disease? Mov
Disord.
 , 
2009
, vol. 
24
 (pg. 
833
-
838
)
20
Purcell
S.
Neale
B.
Todd-Brown
K.
Thomas
L.
Ferreira
M.A.
Bender
D.
Maller
J.
Sklar
P.
De Bakker
P.I.
Daly
M.J.
, et al.  . 
PLINK: a tool set for whole-genome association and population-based linkage analyses
Am. J. Hum. Genet.
 , 
2007
, vol. 
81
 (pg. 
559
-
575
)
21
Price
A.L.
Weale
M.E.
Patterson
N.
Myers
S.R.
Need
A.C.
Shianna
K.V.
Ge
D.
Rotter
J.I.
Torres
E.
Taylor
K.D.
, et al.  . 
Long-range LD can confound genome scans in admixed populations
Am. J. Hum. Genet.
 , 
2008
, vol. 
83
 (pg. 
132
-
135
author reply 135–139
22
Price
A.L.
Patterson
N.J.
Plenge
R.M.
Weinblatt
M.E.
Shadick
N.A.
Reich
D.
Principal components analysis corrects for stratification in genome-wide association studies
Nat. Genet.
 , 
2006
, vol. 
38
 (pg. 
904
-
909
)
23
Devlin
B.
Roeder
K.
Genomic control for association studies
Biometrics
 , 
1999
, vol. 
55
 (pg. 
997
-
1004
)

Author notes

The authors wish it to be known that, in their opinion, the first two authors should be regarded as joint First Authors.
The French Parkinson's Disease Genetics Study Group includes Y. Agid, M. Anheim, A.-M. Bonnet, M. Borg, A. Brice, E. Broussolle, J.-C. Corvol, Ph. Damier, A. Destée, A. Dürr, F. Durif, S. Klebe, E. Lohmann, M. Martinez, C. Penet, P. Pollak, O. Rascol, F. Tison, C. Tranchant, M. Vérin, F. Viallet and M. Vidailhet.