Genetic Variants of the α-Synuclein Gene SNCA Are Associated with Multiple System Atrophy

Background Multiple system atrophy (MSA) is a progressive neurodegenerative disorder characterized by parkinsonism, cerebellar ataxia and autonomic dysfunction. Pathogenic mechanisms remain obscure but the neuropathological hallmark is the presence of α-synuclein-immunoreactive glial cytoplasmic inclusions. Genetic variants of the α-synuclein gene, SNCA, are thus strong candidates for genetic association with MSA. One follow-up to a genome-wide association of Parkinson's disease has identified association of a SNP in SNCA with MSA. Methodology/Findings We evaluated 32 SNPs in the SNCA gene in a European population of 239 cases and 617 controls recruited as part of the Neuroprotection and Natural History in Parkinson Plus Syndromes (NNIPPS) study. We used 161 independently collected samples for replication. Two SNCA SNPs showed association with MSA: rs3822086 (P = 0.0044), and rs3775444 (P = 0.012), although only the first survived correction for multiple testing. In the MSA-C subgroup the association strengthened despite more than halving the number of cases: rs3822086 P = 0.0024, OR 2.153, (95% CI 1.3–3.6); rs3775444 P = 0.0017, OR 4.386 (95% CI 1.6–11.7). A 7-SNP haplotype incorporating three SNPs either side of rs3822086 strengthened the association with MSA-C further (best haplotype, P = 8.7×10−4). The association with rs3822086 was replicated in the independent samples (P = 0.035). Conclusions/Significance We report a genetic association between MSA and α-synuclein which has replicated in independent samples. The strongest association is with the cerebellar subtype of MSA. Trial Registration ClinicalTrials.gov NCT00211224. [NCT00211224]


Introduction
Multiple system atrophy (MSA) is a rare progressive neurodegenerative disease characterized by parkinsonism, cerebellar dysfunction and dysautonomia corresponding to previous diagnostic appellations of striatonigral degeneration, olivopontocerebellar atrophy and Shy-Drager syndrome. Estimates of age-adjusted prevalence vary between 1.9 and 4.4 per 100,000 persons although this may represent an underestimate as post-mortem studies have suggested that as many as 5% of cases of clinically diagnosed Parkinson's disease may in fact have MSA. The converse is also true however, and in those recruited as part of the Neuroprotection and Natural History in Parkinson's Plus Syndromes (NNIPPS) study, about 7% of patients diagnosed with MSA at the start had a different diagnosis by the end [1].
Pathologically, the hallmark of MSA is a-synuclein-immunoreactive glial cytoplasmic inclusions in oligodendrocytes throughout the brain, associated with neuronal loss, loss of myelin, astrocytosis and a marked microglial reaction most prominent in brain regions involved in motor and supraspinal autonomic control. The presence of a-synuclein-immunoreactive pathology, which places MSA amongst the broad category of synucleinopathies including Parkinson's disease and dementia with Lewy bodies, suggests that an abnormality of a-synuclein may play a role in the mechanisms of cell death in MSA.
We examined the association between SNCA variants and MSA using the DNA Bank of the NNIPPS cohort [1]. This cohort has a nearly 50% autopsy rate for MSA so that the criteria for diagnosis can be validated pathologically, and in individual cases we can be certain of the diagnosis. The NNIPPS diagnostic criteria for MSA (Table 1) show excellent convergent validity with the investigators' assessment of diagnostic probability (point-biserial correlation: MSA rpb = 0.93, P,0.0001), and excellent predictive validity against histopathology [sensitivity and specificity (95% CI); 0.96 (0.88-0.99) and 0.91 (0.86-0.93)] [1], so the sample is likely to represent a more highly homogeneous diagnostic group than is usually possible. For replication of results, we analyzed a population of 78 pathologically verified cases from the Institute of Neurology Brain Bank and 83 clinically defined samples from a Parisian study (Table 2). We found that genetic variants of SNCA were associated with MSA and more specifically, with the subset of people who had cerebellar signs.

Methods
The protocol for this trial and supporting CONSORT checklist are published elsewhere [1].

Ethics Statement
Patients gave informed written consent prior to inclusion in the trial. Additional consent was obtained both for DNA sampling and post-mortem brain tissue donation. The protocol was approved by the Ethics Committees/Institutional Review Boards of each coordinating centre in the three participating countries. The trial was conducted according to International standards of Good Clinical Practice-(ICH guidelines and the Helsinki Declaration).

Study Subjects
Subjects were recruited as part of the Neuroprotection and Natural History in Parkinson's Plus Syndromes (NNIPPS) study using the NNIPPS diagnostic criteria which have been described elsewhere together with the trial design, population description and treatment results [1] ( Table 1). The 239 cases and 617 controls in this study were of European ancestry and matched for age (MSA mean age at sampling 62.84, control mean age 61.03) ( Table 2). The clinical diagnosis at the study end was used to define MSA unless a pathological diagnosis was available. For exploratory sub-group analysis, patients were stratified as MSA-P or MSA-C based on the absence or presence of cerebellar signs at entry. 11 cases and 14 controls had a low genotyping rate (,90%) and were excluded from further analysis. 101 of the remaining cases were in the MSA-C subgroup. 251 patients with progressive supranuclear palsy (PSP) recruited within the same study were also tested for MSA-associated SNPs to exclude a recruitment bias as the explanation for any findings, as PSP is not a synucleinopathy and we would not therefore expect an MSA-associated SNCA SNP to be associated with PSP.
The NNIPPS study was an academic-led, multi-centre (UK, France, Germany), phase III trial of riluzole in MSA and progressive supranuclear palsy. Controls were recruited in the UK from the neurologically normal spouses of participating patients. In France and Germany, the controls were taken from existing DNA banks of neurologically normal controls.

Replication study samples
The 78 MSA samples from the Institute of Neurology Brain Bank (Table 2) fulfilled the consensus criteria for pathologically proven MSA [2]. All were of UK ancestry. Of these, 68% had cerebellar signs. Samples were derived from blood.
The MSA samples from Paris (Table 2) were collected consecutively as part of a study on atypical Parkinsonism and cerebellar ataxia. There were 44 cases with Probable MSA and 44 with Possible MSA as defined by the consensus criteria [2]. Of 88 samples, five were not of European ancestry and were therefore excluded from further analysis. Of these, 85% had cerebellar signs as the clinic specializes in cerebellar disease. Samples were derived from blood.
Genotyping 32 SNPs in and around SNCA were selected for association testing with MSA ( Figure 1, Table 3). In addition, we genotyped the multiallelic microsatellite repeat known as NACP-Rep1 which is situated within the promoter region ,10 kbp upstream of the translational start point of SNCA at chromosome 4q21, using an ABI3130XL Genotyper (Applied Biosystems) and Genotyper v4.0 software. This complex microsatellite with sequence (TC) 10-11 TT(TC) [8][9][10][11] (TA) 7-9 (CA) [10][11] has been variably associated with PD [22,23] but not MSA [12], although the numbers studied were small. There are five common alleles, each differing in size by two nucleotides, the greatest variability being in the CA portion with subsidiary variability in the (TC) [8][9][10][11] (TA) 7-9 portion. However, ex vivo functional analysis suggests that the overall length of the microsatellite rather than its sequence, determines transcriptional regulation and hence SNCA gene expression [24]. We numbered the alleles from 1 to 5 in order of increasing size. For the NNIPPS samples, DNA was extracted from venous blood samples collected into Li-heparin tubes by standard methods. SNPs were genotyped independently of the study group and blinded to case-control status by KBiosciences (Hoddesdon Herts, EN11 0EX, UK; www.kbioscience.co.uk) using a patentprotected competitive allele specific PCR system (KASPar) and Taqman TM .
The Institute of Neurology pathological samples were genotyped for rs3822086 using Taqman TM assay (Applied Biosystems, Forster City, CA, USA).
The 83 Paris samples were genotyped for the single nucleotide polymorphism rs3822086 by sequencing in a single direction using the BigDye Terminator Cycle Sequencing kit v3.1 (Applied Biosystems, Forster City, CA, USA). Sequences were analysed using Seqscape v2.5 software.

Statistical Analysis
Quality control. Tests for Hardy-Weinberg equilibrium were performed for each SNP in controls using Fisher's exact test. A P-value,0.05 led to exclusion from further analysis. Tests for systematically missing genotypes in either cases or controls were performed to identify sample selection or genotyping problems preferentially affecting either group. Tests for missingness by genotypic status were also performed to identify errors relating to difficulty calling certain genotypes. Associated SNPs were further tested by looking for association in a tagging haplotype that excluded the associated SNP to reduce the likelihood of an artefactual association resulting from genotyping.
Power. Assuming a lifetime MSA risk of 1 in 1000, and for a P-value close to the permutation corrected threshold of 0.05 (alpha = 0.005), there is 80% power to detect an odds ratio of 1.8 for a minor allele frequency (MAF) of 0.1, 1.6 for MAF 0.2 and 1.5 for MAF.0.2.
Imputation. Genotypes for SNPs genotyped in the CEU HapMap sample but not in this study were imputed using IMPUTE [25] with a threshold of 0.9 and default values for other parameters. Subsequent analysis was performed in PLINK.
Association testing. SNP analyses were performed in PLINK (http://pngu.mgh.harvard.edu/purcell/plink/) [26]. Samples were tested for allelic association using the Mantel-Haenszel chi-squared test, stratifying by country of origin. Odds ratios were tested for homogeneity between countries using the Breslow-Day test. Association testing by Pearson chi-squared test was performed for the replication study to maximise numbers and therefore statistical power. Controls were those used in the primary analysis.   Table 3). Multiple testing correction was performed using the Max(T) permutation procedure implemented in PLINK permuted within the strata of country of origin.
A multilocus genotype test examining whole gene association was performed using step-wise permutation implemented in PLINK.
Microsatellite alleles were tested for association by x 2 test using the permutation procedure instituted in CLUMP to account for the multiple ways that microsatellite alleles can be combined to form two groups.

SNP analyses
SNP rs356220 showed missing genotypes in nearly 18% of controls compared with ,1% of cases (P = 4.73610 215 ) suggesting a systematic problem in genotyping of this SNP in control samples, and it was therefore excluded from further analyses. The remaining 31 SNPs were analysed for association with MSA.
Two SNPs showed association with MSA in the NNIPPS samples: rs3822086 (P = 0.0044), and rs3775444 (P = 0.012) ( Table 3). Only rs3822086 survived multiple testing correction for association with MSA (P = 0.047). A step-wise whole gene permutation test gave P = 0.029 with the same two SNPs retained. We repeated the analysis in the MSA-C subgroup for these two SNPs and found the association strengthened despite more than halving the number of cases: rs3822086 P = 0.0024, OR 2.153, (95% CI 1.3-3.6, Breslow-Day test for homogeneity, P = 0.81); rs3775444 P = 0.0017, OR 4.386 (95% CI 1.6-11.7, Breslow-Day test P = 0.36); step-wise whole gene permutation test P = 0.0084.
A haplotype of six surrounding SNPs was identified correlating with rs3822086 at r 2 .0.5. Without rs3822086 included, this still showed association at P = 0.048, increasing confidence that the association observed was not due to a genotyping artefact. Including rs3822086 strengthened the association with MSA-C. The most strongly associated haplotype, GGATGGT gave P = 8.7610 24 , omnibus test P = 0.027.
As expected, testing of rs3822086 and rs3775444 in PSP showed no association, with unadjusted P-values of 0.112 and 0.242 respectively.

Replication analyses
The samples from the Institute of Neurology Brain Bank and Paris confirmed association of rs3822086 with MSA (P = 0.035) with odds ratio 1.75 (95% C.I. 1.05-2.89).

Imputation analyses
After imputation of genotypes at ungenotyped SNPs there were a total of 318 SNPs tested, with genotypes predicted at 97.2% accuracy. SNP rs3822086 remained the most strongly associated with MSA. An MSA-associated SNP identified in another study [21], rs11931074, was also associated with MSA in this study (P = 0.025). For the MSA-C subgroup, rs3775444 and rs3822086 again remained the most strongly associated SNPs. As for the SNPs in this study, rs11931074 was more strongly associated with MSA-C than with MSA alone (P = 0.0125).

Discussion
We have found two positive associations in the SNCA gene with MSA, one with rs3822086 and the other with rs3775444, although only one survives correction for multiple testing. The association is significant because of the role of a-synuclein in the underlying pathology of MSA. Because the phenotypic spectrum of MSA is broad and the relative frequency of cerebellar and parkinsonian presentations appears to be influenced by ethnicity [27] we stratified our cases as MSA-C and MSA-P to explore possible differences in association between these phenotypes in our sample. We observed a stronger association of SNCA variants with MSA-C than with MSA as a whole. This is intriguing and requires exploration in other studies, but may indicate that interactions between SNCA gene variants and other risk factors influence phenotype in MSA.
Until recently, studies of polymorphisms of SNCA had failed to demonstrate any association with MSA either by sequencing the entire coding region in a small number of Japanese cases [11] or by genotyping two microsatellite markers, one in the promoter and the other an intron of SNCA in 47 cases from the UK [12,19]. A study examining SNPs prioritized in a genome-wide association study of Parkinson's disease for follow-up testing in MSA has now identified a strong association of homozygosity for the T allele of SNP rs11931074 and MSA [21]. This SNP has a similar minor allele frequency (0.056) to rs3822086 (0.058) and the two SNPs are in strong linkage disequilibrium (D9 0.977, r 2 0.922). It is therefore not surprising that rs11931074 shows association with both MSA and MSA-C in our study, although the association is not as strong as for rs3822086. It would be of interest to impute rs3822086 in the genome-wide association study samples to see if the stronger association of this SNP persists in that sample set. Nevertheless, this suggests that both rs3822086 and rs11931074 are tagging the same functional genetic variation. Given that both have a relatively low minor allele frequency, this may be a rare haplotype or variant.
The associated variants we have identified are both intronic, and neither is close to a splice site, nor are they predicted to disrupt or alter splicing. They are not in intronic splice enhancer sites, nor alternative cassette exons. They are not especially conserved and are not obviously regulatory. It is therefore likely that a different SNP tagged by these variants is the functional variant, and this is further supported by the haplotypic association we identified for SNPs surrounding rs3822086, which persists even when this SNP is excluded.
There are several factors that could confound this study. For example, population stratification can lead to false positive results. We used three strategies to minimize this. Firstly, the samples were all of European ancestry; secondly we stratified the primary analysis by country of origin; and thirdly we replicated the findings in an independent set of cases. The associated SNP survived multiple testing correction, increasing confidence in the association. Furthermore, the global gene-wise evidence for association was significant and requires no further multiple testing correction making a false positive result less likely. Phenotypic heterogeneity was minimized by using validated clinical criteria with excellent inter-observer agreements, and pathological verification where possible. The possibility of bias inherent in the selection process for enrolment to the NNIPPS study was excluded by testing the MSAassociated SNP for association with the different disease PSP, also using the NNIPPS study samples. PSP is not expected to have an abnormality in the SNCA gene since it is not a synucleinopathy: no association was detected. Systematic genotyping errors can lead to false positive results, but we observed strict, comprehensive quality control measures. Furthermore, a haplotype using a set of SNPs surrounding the associated SNP and which correlated with the genotype, also detected the association suggesting that this is not an artefact at the tested SNP.
Our findings show that genetic variation in the a-synuclein gene is associated with sporadic MSA, and suggest that susceptibility to MSA-C may be the greater risk of SNCA variation. These results suggest that a-synuclein can play a primary role in the pathogenesis of MSA.