|Home | About | Journals | Submit | Contact Us | Français|
We studied the independent and joint effects of the genes encoding alpha-synuclein (SNCA) and microtubule associated protein tau (MAPT) in Parkinson's disease (PD) as part of a large meta-analysis of individual data from case-control studies participating in the Genetic Epidemiology of Parkinson's Disease (GEO-PD) consortium.
Participants of Caucasian ancestry were genotyped for a total of four SNCA (rs2583988, rs181489, rs356219, rs11931074) and two MAPT (rs1052553, rs242557) SNPs. Individual and joint effects of SNCA and MAPT SNPs were investigated using fixed- and random-effects logistic regression models. Interactions were studied both on a multiplicative and an additive scale, and using a case-control and case-only approach.
Fifteen GEO-PD sites contributed a total of 5302 cases and 4161 controls. All four SNCA SNPs and the MAPT H1-haplotype defining SNP (rs1052553) displayed a highly significant marginal association with PD at the significance level adjusted for multiple comparisons. For SNCA, the strongest associations were observed for SNPs located at the 3′ end of the gene. There was no evidence of statistical interaction between any of the four SNCA SNPs and rs1052553 or rs242557, neither on the multiplicative nor on the additive scale.
This study confirms the association between PD and both SNCA SNPs and the H1 MAPT haplotype. It shows, based on a variety of approaches, that the joint action of variants in these two loci is consistent with independent effects of the genes without additional interacting effects.
The microtubule associated protein tau and α-synuclein are two abundant brain proteins that aggregate in neurodegenerative diseases such as Parkinson's disease (PD), Alzheimer disease, and progressive supranuclear palsy (PSP). There is evidence that the formation of pathological inclusions containing tau and α-synuclein is promoted by common mechanisms,1 and there are reports of concurrence of α-synuclein and tau brain pathology in autosomal dominant parkinsonism.2,3
There is increasing evidence that genetic susceptibility contributes to the etiology of PD. Using a candidate-gene approach, genetic association studies pointed towards an association between PD and both the α-synuclein (SNCA)4-6 and microtubule-associated protein tau (MAPT) genes.7-10 More recently, genome-wide association studies (GWAS) confirmed that SNCA and MAPT were two of the main common contributors to PD genetic susceptibility among Caucasians.11-14 The MAPT locus (17q21.31) contains a c.900-kb inversion polymorphism with two distinct haplotypes, H1 and H2; the major H1 haplotype is associated with PD, PSP, and other tauopathies. The SNCA gene lies in a region of relatively high linkage disequilibrium (LD) with SNPs at both 3′- and 5′-ends of the gene associated with PD. Given the multiple associations at this locus, it remains unclear whether they result from a single functional variant or whether there are different functional variants at both ends.5 Given the over-expression hypothesis for SNCA, there may well be multiple variants affecting transcription factor binding sites at the promoter or miRNA sites in the 3′UTR.15
Although the functional variant(s) have not been identified, the association between PD and both the SNCA and MAPT genes is well established. These findings raise the question of a possible gene-gene interaction between SNCA and MAPT, but few studies have investigated this question and with inconsistent findings.16-18
Sample sizes needed to detect interactions between two variables are larger than for marginal effects of similar size;19,20 therefore, larger studies are needed to investigate whether SNCA and MAPT interact and collaborative efforts are needed to reach sufficient sample sizes. We invited teams involved in the Genetic Epidemiology of Parkinson's Disease (GEO-PD) consortium to undertake a collaborative effort in order to investigate the joint effects and potential interactions between SNCA and MAPT in conferring susceptibility to PD in a large sample of cases and controls.
The aim of the GEO-PD consortium is to conduct collaborative studies of genetic risk factors in PD. Since its creation in 2004, the consortium has regularly met in order to organize scientific collaborations between participating teams. During the meeting held in Tuebingen (Germany) in 2009 and following the presentation at scientific meetings of the results of PD GWAS, attending teams were invited to participate to a collaborative effort in order to study the joint effects of SNCA and MAPT.
All studies were approved by the local ethical committees following the procedures of each country.
Participating sites were asked to contribute 250 nanograms of DNA. DNA was sent to a central laboratory (Mayo Clinic, Jacksonville, USA) where genotypes were determined blinded to case-control status.
Four SNCA (rs2583988, rs181489, rs356219, rs11931074) and two MAPT (rs1052553, rs242557) SNPs were selected for genotyping. We identified SNCA SNPs at the 5′- (rs25839885,6) or 3′-ends (rs181489,5,21 rs356219,6,13,17 rs1193107411-13,21) of the gene which had been previously associated with PD. In addition, rs11931074 (located approximately 7Kb downstream from the 3′-end) has been associated with multiple system atrophy (MSA).22 We did not select the REP1 polymorphism in the SNCA promoter because the 263 bp allele, which is more strongly associated with PD, is rare (<10%) thus leading to insufficient power for interaction analyses.
Genotyping was performed on a Sequenom MassArray iPLEX platform (San Diego, CA); primer sequences are available upon request.
We used exact tests to assess among controls of each site whether genotype distributions for each SNP violated Hardy-Weinberg equilibrium (HWE).24 Sites with a nominally significant (p<0.05) deviation from HWE were excluded. All participants were of Caucasian ancestry.
We investigated the marginal association between PD and the six SNPs by site using fixed-effects logistic regression. For SNCA SNPs, the reference allele was the major frequency allele; for MAPT SNPs we considered the minor allele as the reference in order to be in agreement with previous papers.10 Odds ratios (OR), 95% confidence intervals (CI), and p-values were computed using a dummy-coding of the genotypes (model-free analysis), as well as additive, dominant, and recessive models. The Akaike information criterion (AIC) was computed; the lowest AIC indicates the best model when both goodness of fit and parsimony are considered.
Our primary analyses of the interaction between SNPs were performed by estimating ORs for individual and joint effects and by including multiplicative terms in the models to test for statistical multiplicative interaction.25 We also tested interactions on an additive scale by estimating the relative excess risk due to interaction (RERI, also known as the interaction contrast ratio), the attributable proportion due to interaction (AP), and the synergy index (SI); RERI=AP=0 and SI=1 indicate lack of interaction on an additive scale.26,27 For interaction analyses, our primary analyses focused on MAPT rs1052553;10 we tested the interaction between each of the four SNCA SNPs and rs1052553. Additional analyses involving the other MAPT SNP (rs242557) were also performed and reported as supplementary data. Analyses unadjusted and adjusted for age and gender were performed.
The results of analyses by site are displayed as forest plots and were used to estimate between-site heterogeneity. We tested for between-site heterogeneity with the χ2-based Q statistic (significant for p<0.10) and quantified its extent with I2, which ranges from 0% to 100% and represents the proportion of between-study variability ascribed to heterogeneity rather than to chance.28,29 Values for I2 of 0–24% suggest little heterogeneity, 25–49% reflect moderate heterogeneity, 50–74% reflect large heterogeneity, and >75% reflect very large heterogeneity.
For quantitative syntheses, we used fixed- and random-effects logistic regression models. In the presence of heterogeneity, random-effects syntheses are preferable;30 because there was evidence of heterogeneity in some analyses, our primary analyses are based on random-effects models and we present the results of analyses based on fixed-effects models as supplementary data. Fixed-effects models assume that ORs are constant across sites and that observed differences are due to chance; they were implemented by including site as a categorical covariate in the models. Random-effects models allow that results might be genuinely heterogeneous across sites and take into account between-study heterogeneity by including random effects for genotypes; they were implemented using multilevel regression models.31-34 For analyses of gene-gene interactions, regression models included several random effects and we used an unstructured variance-covariance matrix for the random effects.
Gene-gene multiplicative interactions were also investigated by looking at the association of SNCA and MAPT genotypes among cases-only using fixed- and random-effects logistic regression (secondary analysis). If genotypes are independent among controls, this approach has generally increased power compared to case-control analyses.20,35
SNCA haplotypes were defined using Thesias software, that allows to test interactions with covariates.36 We tested the interaction between SNCA haplotypes and rs1052553 or rs242557. The relative effect of SNCA SNPs was explored using a unified stepwise regression procedure.37
Among cases, we followed a similar strategy as described above to investigate, with fixed- and random-effects linear regression models, the effect of SNPs and their interaction on age at onset (AAO) of PD as a continuous outcome.
Both for case-control and case-only analyses, a Bonferroni correction was used to take into account multiple testing. In marginal-effects analyses, we considered four models for six SNPs; p≤0.0021 (0.05/24) was considered as statistically significant. For analyses of joint effects, three multiplicative interaction models (model-free, dominant, or additive coding of SNCA SNPs) were considered for four SNPs; p≤0.0042 (0.05/12) was considered as statistically significant.
The sample size needed to detect interactions on a multiplicative scale was investigated for different gene frequencies, genetic models, and effects (supplementary figure 1). For instance, to detect an interaction OR of 1.5, and assuming that the marginal OR for a SNCA SNP with a frequency of 15% is 1.20 (additive model) and that the marginal OR for a MAPT SNP with a frequency of 80% is 1.25 (recessive model), a case-control study (1:1 matching) would need to include ~1700 cases in order to reach 80% power at the two-sided 0.05 significance level. At the 0.0042 significance level, ~2900 cases would be necessary to reach 80% power.
Analyses were performed using SAS 9.1 (Cary, NC, USA) and STATA 11.0 (College Station, TX, USA).
Fifteen sites contributed a total of 5302 cases and 4161 controls. Their demographic and clinical characteristics are shown in table 1; 16% of the cases reported a positive history of PD among first-degree relatives. The distributions by site of the six SNPs in cases and controls are shown in supplementary table 1. HW equilibrium was tested among controls for each SNP and site; rs1052553 was the only SNP for which a significant departure (p=0.0008) was identified in site K. We excluded participants from this site and retained the remaining 5199 cases and 4059 controls. Genotyping call-rates were > 95% for all SNPs and sites. One site (P) contributed only cases; they were included in case-only analyses (therefore based on 5272 cases).
Table 2 shows the association between each SNP and PD using random-effects logistic regression, together with heterogeneity estimates; results from fixed-effects models are shown in supplementary table 2. Supplementary figure 2 shows forest plots under additive, dominant, and recessive genetic models. There was evidence of heterogeneity for some SNCA SNPs (rs181489, additive and dominant model; rs356219, recessive model; rs2583988, additive and dominant model; table 2), while no heterogeneity was detected for MAPT SNPs. Fixed- and random-effects models yielded similar conclusions. All SNCA SNPs displayed a highly significant association with PD at the Bonferroni-corrected significance level. For rs181489, rs356219, and rs2583988, both heterozygotes and homozygotes for the minor allele were significantly more frequent in cases than controls, but ORs increased with the number of minor alleles and the additive model displayed the lowest AIC value. For rs11931074, the association pattern was more consistent with a dominant model, but the additive model was also very close in AIC. The strongest associations were detected for rs181489 followed by rs356219 (both located at the 3′ end of the gene). Adjustment for age and sex yielded similar findings (not shown).
When all SNCA SNPs together with all possible pair-wise interactions were included in the same model and after using a backward selection procedure (supplementary table 3), main effects remained highly significant for the 3′ SNPs, indicating that they were independently associated with PD. The direction of the independent association between PD and rs356219 was reversed in this analysis compared to univariate analyses (table 2); it is likely that the strong linkage disequilibrium between SNPs in the SNCA gene leads to confounding in univariate analyses and that an independent effect of rs356219 is not apparent in univariate analyses due to its strong association with rs181489. In addition, there was a trend in favor of an interaction between rs11931074 and rs181489, suggesting that the effect of rs11931074 decreased with the number of rs181489 alleles. The main effect for rs2583988 (5′-end) was not significant once 3′-end SNPs were included in the model; however, there was a trend in favor of interaction between rs2583988 and rs356219, suggesting that rs2583988 has a small effect among carriers of the minor rs356219 allele.
We found a highly significant association between PD and the H1-tagging allele of rs1052553 with the same AIC values for additive and recessive models; the GA rs1052553 genotype was not associated with PD and it was only the AA genotype (i.e., carriers of the H1/H1 haplotype) that was positively associated with PD. A weaker association was found for rs242557 (dominant model). Adjustment for age and sex yielded similar findings (data not shown). After Bonferroni correction, rs1052553 remained significantly associated with PD while rs242557 did not. When both rs1052553 (recessive) and rs242557 (dominant) were included in the same model, the association between PD and rs1052553 remained virtually unchanged (fixed-effects, OR=1.25, 95% CI=1.14-1.37) while the association with rs242557 disappeared (fixed-effects, OR=0.95, 95% CI=0.84-1.08).
The genotype counts for the cross-tabulation of rs1052553 and each of the SNCA SNPs are presented in supplementary table 4 by study. Figure 1 includes forest plots of ORs corresponding to the multiplicative interaction between the H1/H1 MAPT haplotype (defined by rs1052553) and each SNCA SNP (additive coding); supplementary figure 3 shows the same analysis using a dominant coding of SNCA SNPs. Depending on the SNPs, heterogeneity measures suggested weak (rs356219, rs11931074), moderate (rs181489) or large (rs2583988) heterogeneity.
Table 3 presents the results of interaction analyses between each SNCA SNP and MAPT rs1052553; because of the large sample size, we were able to investigate the interaction between carriers of the H1/H1 MAPT haplotype and each of the genotypes of SNCA SNPs. Multiplicative interaction was tested using different codings of SNCA SNPs. All interaction tests were far from nominal significance: while both SNCA and MAPT SNPs independently increased the risk of PD, their joint effects were not different from that expected under a multiplicative model. Fixed-effects models (supplementary table 5) and adjustment for age and gender (data not shown) yielded the same conclusions. For the analyses presented in figure 2, we used a dominant coding for SNCA SNPs to show that there was no interaction using alternative codings of SNCA SNPs. In addition, RERI and AP were not different from 0 and SI was not different from 1 in all instances, thus showing that there was no interaction on an additive scale. There was no evidence of interaction using a recessive coding for SNCA SNPs17 (data not shown). No multiplicative interactions were observed between MAPT rs242557 and SNCA SNPs (supplementary table 6); there was no evidence of additive interaction either (not shown). Finally, we defined SNCA haplotypes and tested their interaction with rs1052553 or rs242557; no interactions were detected both on multiplicative or additive scales (not shown). Analyses restricted to sporadic cases yielded the same conclusions (data not shown).
Supplementary table 7 shows the association between MAPT rs1052553 and SNCA SNPs stratified by disease status. There was no association between either rs1052553 or rs242557 (not shown) and any SNCA SNP among controls or cases.
We performed additional analyses restricted to cases in order to investigate whether SNCA and MAPT SNPs influenced AAO. Table 4 shows the marginal association between each SNP and AAO. There was no significant association between any SNP and AAO at the Bonferroni-corrected significance level. Table 5 shows individual and joint effects of SNCA SNPs and rs1052553 on AAO; there was no evidence of departure from additivity in linear regression models. No interaction between SNCA SNPs and rs242557 was seen for AAO (not shown).
SNCA and MAPT have been confirmed by recent GWAS as two of the main contributors to genetic susceptibility in PD among Caucasians.11-14 α-synuclein is one of the main protein components of Lewy bodies and it has been reported that common mechanisms promote the aggregation of α-synuclein and tau,1 supporting findings of concurrent α-synuclein and tau brain pathology in autosomal dominant parkinsonism.2,3 These observations raise the possibility of an interaction between SNCA and MAPT.
In this large case-control study, we confirmed the association between PD and four SNPs distributed across the SNCA gene as well as with the H1 MAPT haplotype. All SNCA SNPs and the MAPT H1 defining variant (rs1052553) affected PD susceptibility. Neither SNCA nor MAPT SNPs influenced AAO among cases. The large sample size allowed us to address the important question of whether there is a SNCA-by-MAPT interaction. We did not find any evidence in favor of interaction for susceptibility to PD or AAO.
Previously, Mamah et al.16 genotyped 557 case-control pairs for the REP1 polymorphism in the SNCA promoter and MAPT H1 haplotype, and found a marginal association between PD and both the 261/261 REP1 genotype and the H1/H1 MAPT haplotype. They investigated individual and joint effects by implementing several genetic models, without formally testing for statistical interaction, and concluded that the most likely model was one that forced the regression parameters to be equal for individual gene effects and their combination. According to this model, the combined effect of the genes is smaller than expected under a multiplicative model. A subsequent study of the relation between REP1 and PD found that it was not the 261 allele, but the less-frequent 263 allele that was associated with PD.4 Goris et al.17 genotyped 659 PD cases and 2176 controls for the H1 MAPT haplotype and one 3′ SNCA SNP (rs356219). They found a marginal association between PD and both H1/H1 MAPT and GG rs356219. In addition, there was evidence of a multiplicative interaction (p=0.03): the combined OR for H1/H1 MAPT and GG rs356219 (OR=2.14) was greater than expected under a multiplicative model given their individual ORs (H1/H1MAPT, OR=1.23; GG rs356219, OR=1.00). As a part of another study that looked at several gene-gene and gene-environment interactions (932 cases, 664 controls), McCulloch et al.18 did not find evidence of an interaction between the H1 MAPT haplotype and REP1. Finally, a GWAS did not detect an epistatic interaction between SNCA and MAPT, but without further information about the interaction models that were tested.11
Interestingly, for all SNCA SNPs except rs11931074, the best fitting marginal model was an additive model and PD risk increased with the number of minor alleles. As duplications and triplications of SNCA have been identified in families, overdosage has been postulated as a potential mechanism. In addition, additive models are typically the best-fitting for GWAS-discovered SNPs.38 For rs11931074, the best-fitting model was the dominant one, but the difference in AIC from the additive model was minimal, and it was recently associated with PD in two GWAS using an additive coding.11,12 In agreement with previous studies,5,6,11 our analysis of the relative effects of SNCA SNPs suggests that the causal variant may be located towards the 3′-end and could affect post-transcriptional RNA processing or stability, and thus gene/protein expression; however, we cannot exclude that the 5′-end may also play a role, and additional genomic capture and sequencing studies are needed to identify the causal variant(s).
Mutations in the MAPT gene cause frontotemporal dementia with parkinsonism and there is consistent evidence that PSP is associated with the MAPT H1 haplotype.23,39 There is increasing evidence from candidate gene studies10 and GWAS11-14 that the H1 MAPT haplotype is also associated with PD among Caucasians. In agreement with these studies, we found a strong association with rs1052553. The rs242557 A-allele defines the H1c PSP-associated haplotype and a recent study reported that PD may be associated with the opposite G-rs242557 allele,40 but in our study, the association between PD and rs242557 was not significant after correction for multiple testing and is likely to be accounted for by its LD with rs1052553, as demonstrated by analyses in which both SNPs were included in the same model.
One of the main strengths of our study is its large sample size that conferred sufficient power to detect even small interactions. Smaller studies suffer from insufficient numbers of subjects jointly exposed to the two variables investigated, which leads to reduced power and unstable interaction estimates with large confidence intervals. Additional strengths involve the centralized genotyping in a single laboratory, the consideration of several SNPs in the SNCA gene located both at its 5′ and 3′ ends (compared to previous studies on SNCA-by-MAPT interaction that considered single SNPs), and implementation of complementary approaches to test for interaction.
The relevant scale to test for statistical interaction has been a subject of intense debate in the epidemiological community and there is no consensus about the most appropriate method.41,42 The multiplicative scale has been most often used for dichotomous outcomes mainly due to an easier implementation using logistic regression. It has been argued, however, that biologic interactions are more likely to lead to departure from additive effects of two variables.25,26,43 As we were concerned that we may have failed to detect an interaction on the additive scale, we investigated additive interactions and found that the pattern of association was not suggestive of an interaction on this scale either. In addition, it has been pointed out that the same word, epistasis, i.e., interaction between genes, has been used in the literature to describe different concepts,44 and Phillips45 recently described three main categories: functional, compositional, and statistical epistasis. The extent to which statistical interaction implies functional or compositional epistasis, and vice-versa, is unclear,44,45 but empirical tests have been proposed as a way to detect compositional epistasis (also termed epistasis in the sense of masking);46,47 our RERI estimates were clearly not in agreement with this type of mechanism either.
Weaknesses of our study include lack of standardized inclusion/exclusion criteria for cases or controls, different diagnostic criteria across studies, and lack of standard definition of AAO; however, random-effects models, which take into account heterogeneity between studies, yielded the same results as fixed-effects models and analyses stratified by diagnostic criteria showed no differences across strata.
In conclusion, this study confirms the association between PD and both SNCA SNPs and the H1 MAPT haplotype, with similar size effects as previous studies, and it shows, based on a variety of approaches, that the association between PD and SNCA is not modified by MAPT, and vice-versa. Thus, the joint action of variants in these two loci is consistent with independent effects. While these findings do not support a strong gene-gene interaction between MAPT and SNCA at the level of epidemiological risk, they do not rule out functional interactions at the protein level and further in vitro and in vivo studies will be necessary to address this question.
The authors acknowledge the following GEO-PD collaborators and funding sources: France: Philippe Amouyel (Inserm, U744, Lille, France); Christophe Tzourio (Inserm, U708, Paris, France); Claire Mulot (Inserm, U775; CRB Epigenetec, Paris, France). Funding: Agence Nationale de la Recherche (Santé-environnement 2005, Maladies neurologiques, 2009), Agence française de sécurité sanitaire de l'environnement et du travail (Afsset), France Parkinson. USA: Justin A. Bacon, BS, and Stephanie A. Cobb, BA (Mayo Clinic, Jacksonville). Funding: The Michael J. Fox Foundation (O.A. Ross & M.J. Farrer); American Parkinson's Disease Association Research Grant (O.A. Ross); Mayo Clinic Jacksonville is a Morris K. Udall Parkinson's Disease Research Center of Excellence (NINDS P50 #NS40256); NIH grant 2R01ES10751 (D.M. Maraganore). Greece/USA: Funding: Scientific support for this project to J.P.A. Ioannidis was provided through the Tufts Clinical and Translational Science Institute (Tufts CTSI) under funding from the National Institute of Health/National Center for Research Resources (UL1 RR025752, PI: Harry Selker). Points of view or opinions in this paper are those of the authors and do not necessarily represent the official position or policies of the Tufts CTSI. Australia (site A): Greg T. Sutherland, Gerhard A. Siebert, Nadeeka Dissanayaka, John D. O'Sullivan, Richard S. Boyle. Funding: National Health and Medical Research Council (Australia) Project Grant #401537; Geriatric Medical Foundation of Queensland; Princess Alexandra Hospital; Royal Brisbane and Women's Hospital Foundations. France (site B): Florence Pasquier and Régis Bordet (Steering Commitee of the Paradigme study); Jean-Philippe Legendre (CHRU-Lille). Funding: Fondation de la Recherche Médicale (Convergence 2006/413), PHRC (Parkfanord 005/1913, Paradigme 2002/1918), Univ Lille 2 (Synucleothèque), Inserm-French Research Ministry for support for the biological resources centers (CHRU Lille and Institut Pasteur de Lille). Germany (site C): Georg Auburger; Rüdiger Hilker; Nadine Abahuni (Goethe University Frankfurt); Christof Geisen (Blood Bank Frankfurt). Funding: BMBF NGFNplus 01GS08138. Germany (site D): Susen Winkler, BS. Funding: Volkswagen Foundation and Hermann and Lilly Schilling Foundation. Germany (site E): Thomas Gasser, Olaf Riess, Daniela Berg, Claudia Schulte. Funding: Germany: Work of was supported by grants from the German Research Council [DFG, KR2119/3-2] to R. Krüger; the Federal Ministry for Education and Research [BMBF, NGFNplus; 01GS08134] to R. Krüger, T. Gasser, O. Riess; and the Michael J Fox Foundation to M. Sharma, R. Krüger, T. Gasser. Greece (site F): Demitris Vassilatis, Eleftherios Stamboulis. Funding: Hellenic Secretariat of Research and Technology (ΠENEΔ 2003). Greece (site G): Efthimios Dardiotis, Department of Neurology, Faculty of Medicine, University of Thessaly, Larissa, Greece); Ioanna Patramani, MD (Department of Neurology, Faculty of Medicine, University of Thessaly, Larissa, Greece); Persa-Maria Kountra, MD (Department of Neurology, Laboratory of Neurogenetics, Faculty of Medicine, University of Thessaly, Larissa, Greece); Christina Vogiatzi (Department of Neurology, Faculty of Medicine, University of Thessaly, Larissa, Greece); Katerina Markou, MD (Department of Neurology, Faculty of Medicine, University of Thessaly, Larissa, Greece). Funding: University of Thessaly, Research Committee (Code: 2845); Intsitute of Biomedical Research & Technology, CERETETH (Code: 01-04-207). Italy (site I): Patrizia Tarantino, Ferdinanda Annesi (Institute of Neurological Sciences, National Research Council, Cosenza, Italy). Italy (site J): Anna Rita Bentivoglio, MD, PhD; Arianna Guidubaldi, MD; Matilde Caccialupi, MD; Francesca De Nigris, MD (Institute of Neurology, Catholic University, Rome, Italy). Funding: Italian Ministry of Health (Ricerca Corrente 2010; Ricerca Finalizzata 2006; Progetto Giovani Ricercatori 2008). Italy (site K): Chiara Riva, PhD (Department of Neuroscience and Biomedical Technologies, University of Milano-Bicocca, Monza, Italy). Funding: FIRB 2003 GENOPOLIS Project; grant number: RBLA038RMA_003. Norway (site L): Funding: The Research Council of Norway, Project 10314000. Sweden (site N): Nancy L Pedersen. Funding: National Institutes of Health (ES10758), the Swedish Research Council, the Swedish Society for Medical Research, the Swedish Society of Medicine, funds from the Karolinska Institutet, and the Parkinson Foundation in Sweden. Sweden (site P): Other contributors: Karin Nilsson, RN (Department of Clinical Science, Section of Geriatric Psychiatry, Lund University, Sweden); Jan Reimer, RN (Department of Neurology, Skåne University Hospital, Lund, Sweden; Region Skåne Competence Centre, Skåne University Hospital, Malmö, Sweden). Funding: Swedish Parkinson Academy, AFA Insurance Research Grant (Sweden), Elsa Schmitz Foundation (Sweden), Apotekare Hedberg's Foundation for Medical Research (Sweden), The Swedish Parkinson Fund, The Royal Physiographic Society in Lund (Sweden), Lund University Hospital Research Fund (Sweden). USA (site O): Jay Van Gerpen, MD; Jennifer Lash; Jill Searcy; Audrey Strongosky. Funding: Mayo Clinic Jacksonville Morris K. Udall Parkinson's Disease Research Center of Excellence NIH/NINDS P50 #NS40256, NIH/NINDS R01 NS057567, NIH/NINDS RC2 NS070276-01, Mayo Clinic Florida Research Committee CR programs, and gift from Carl Edward Bolch, Jr. and Susan Bass Bolch.