www.nature.com/npjparkd ARTICLE OPEN Mitochondria function associated genes contribute to Parkinson’s Disease risk and later age at onset Kimberley J. Billingsley1,2, Ines A. Barbosa3, Sara Bandrés-Ciga2, John P. Quinn1, Vivien J. Bubb1, Charu Deshpande4, Juan A. Botia5,6, Regina H. Reynolds 6, David Zhang6, Michael A. Simpson3, Cornelis Blauwendraat2, Ziv Gan-Or7,8,9, J. Raphael Gibbs2, Mike A. Nalls2,10, Andrew Singleton2, International Parkinson’s Disease Genomics Consortium (IPDGC), Mina Ryten6 and Sulev Koks 11,12 Mitochondrial dysfunction has been implicated in the etiology of monogenic Parkinson’s disease (PD). Yet the role that mitochondrial processes play in the most common form of the disease; sporadic PD, is yet to be fully established. Here, we comprehensively assessed the role of mitochondrial function-associated genes in sporadic PD by leveraging improvements in the scale and analysis of PD GWAS data with recent advances in our understanding of the genetics of mitochondrial disease. We calculated a mitochondrial-specific polygenic risk score (PRS) and showed that cumulative small effect variants within both our primary and secondary gene lists are significantly associated with increased PD risk. We further reported that the PRS of the secondary mitochondrial gene list was significantly associated with later age at onset. Finally, to identify possible functional genomic associations we implemented Mendelian randomization, which showed that 14 of these mitochondrial function- associated genes showed functional consequence associated with PD risk. Further analysis suggested that the 14 identified genes are not only involved in mitophagy, but implicate new mitochondrial processes. Our data suggests that therapeutics targeting mitochondrial bioenergetics and proteostasis pathways distinct from mitophagy could be beneficial to treating the early stage of PD. npj Parkinson’s Disease (2019)5:8 ; https://doi.org/10.1038/s41531-019-0080-x INTRODUCTION An interest in mitochondrial function and PD began with the Parkinson’s disease (PD) is a progressive neurodegenerative observation that exposure to the drug 1-methyl-4-phenyl-1,2,3,4- movement disorder characterized pathologically by the death of tetrahydropyridine (MPTP) can cause rapid parkinsonism and dopaminergic neurons in the substantia nigra (SN) and aggrega- neuronal loss in the SN in humans, and that this is mediated tion of α-synuclein protein (encoded by SNCA), within intraneur- through inhibition of complex I of the mitochondrial electron onal inclusions called Lewy bodies. The majority of PD cases are transport chain.7,10,11 Subsequent work suggested that individuals apparently sporadic in nature (~95%). Aging is a major risk factor with sporadic PD have reduced complex I activity not only in the for disease and due to population ageing the prevalence of PD is SN, but in other brain regions and peripheral tissues.12 Genetic predicted to increase rapidly, making the identification of studies focusing on monogenic forms of PD provided further therapeutic targets a high priority.1–3 support for the involvement of mitochondrial dysfunction in the Although there have been great advances in understanding disease. Pathogenic mutations that lead to autosomal recessive both the genetic architecture and cellular processes involved in forms of PD have been reported in PINK1, PARK2, PARK7, CHCHD2, PD, the exact molecular mechanisms that underlie PD remain and VPS13C and the proteins they encode are all now known to be unknown.1 However, it is suggested that PD has a complex involved in the mitochondrial quality control system and in etiology, involving several molecular pathways, and understand- particular mitophagy.13–16 ing these specific pathways will be key to establishing mechanistic Therefore, in this paper, we aim to comprehensively assess the targets for therapeutic intervention. While several key pathways role of mitochondrial function in sporadic PD by leveraging are currently being investigated, including autophagy, endocy- improvements in the scale and analysis of PD genome-wide tosis, immune response, and lysosomal function,4–7 mitochondrial association study (GWAS) data with recent advances in our function was the first biological process to be associated with understanding of the genetics of mitochondrial disease. The PD.8,9 availability of large-scale genome-wide association data in PD 1 Department of Molecular and Clinical Pharmacology, Institute of Translational Medicine, , University of Liverpool, Liverpool, UK; 2Laboratory of Neurogenetics, National Institute on Aging, National Institutes of Health, Bethesda, MD 20892, USA; 3Department of Medical and Molecular Genetics, King’s College London School of Basic and Medical Biosciences, London SE1 9RT, UK; 4Clinical Genetics Unit, Guys and St. Thomas’ NHS Foundation Trust, London SE1 9RT, UK; 5Departamento de Ingeniería de la Información y las Comunicaciones, Universidad de Murcia, 30100 Murcia, Spain; 6Department of Neurodegenerative Disease, UCL Institute of Neurology, 10-12 Russell Square House, London, UK; 7 Montreal Neurological Institute, McGill University, Montréal, QC, Canada; 8Department of Neurology and Neurosurgery, McGill University, Montréal, QC, Canada; 9Department of Human Genetics, McGill University, Montréal, QC, Canada; 10Data Tecnica International, Glen Echo, MD 20812, USA; 11The Perron Institute for Neurological and Translational Science, 8 Verdun Street, Nedlands, WA 6009, Australia and 12Centre for Comparative Genomics, Murdoch University, Murdoch 6150, Australia Correspondence: Sulev Koks (sulev.koks@perron.uwa.edu.au) Members of the IPDGC are listed at the end of the paper. These authors contributed equally: Kimberley J. Billingsley, Ines A. Barbosa, Mina Ryten and Sulev Koks Received: 3 December 2018 Accepted: 15 April 2019 Published in partnership with the Parkinson’s Foundation K.J. Billingsley et al. 2 Fig. 1 Workflow of mitochondrial-function specific PD analysis Table 1. Cohort-level heritability analysis for the primary and secondary mitochondrial gene lists Primary Secondary Heritability estimate SE of heritability estimate Heritability estimate SE of heritability estimate WTCCC PD GWAS (PMID:21044948) 0.00321 0.00277 0.00563 0.00688 Spanish Parkinson’s (IPDGC) part2 0.00027 0.00314 0.00629 0.00932 NIA PD GWAS (PMID:19915575) 0.00945 0.00540 0.03616 0.01365 1234567890():,; Dutch GWAS (PMID:21248740) 0.00000 0.00530 0.03562 0.01681 Reporting estimates for the WTCCC PD GWAS (PMID:21044948), Spanish Parkinson’s (IPDGC) part2, NIA PD GWAS (PMID:19915575), Dutch GWAS (PMID:21248740) cohorts. Showing heritability estimates generated using GCTA and standard error of the estimates (SE) cases and the rapid identification of genetic lesions that underlie After establishing the consistency of our heritability estimates, mitochondrial disease provide an opportunity to systematically we next calculated heritability using only variants located within assess the role of genetic variability in mitochondrial linked genes genic regions specified as being of primary (n = 176) or secondary in the context of risk for PD.17 In this study we combine these new (n = 1463) importance in mitochondrial function. Initially, to assess resources with current statistical tools, such as polygenic risk the full contribution of the mitochondria pathway we ran the scoring and Mendelian randomization, to explore the role of analysis including and excluding the PD risk genes.6 However, as mitochondrial function in both PD risk and age at onset of disease shown in Supplementary Fig. 1 there was little difference overall in to obtain novel insights. the heritability estimates. Therefore, to only to catalog mitochondria-specific genetic risk outside of known loci we focused on the analysis excluding these genes. The heritability RESULTS estimate using a random-effects meta-analysis for the primary Common variation within mitochondria function genes gene list excluding the PD genes was estimated to be a modest contributes to the heritable component of PD 0.26% (95% CI −0.11–0.66, p = 0.166). However, the heritability The general workflow for the genetic analysis used in this study is estimate using a random-effects meta-analysis for the secondary shown in Fig. 1. First, to study the importance of mitochondrial list, namely genes implicated in mitochondrial function or function in sporadic PD, we investigated the heritability of PD morphology, as well as disease, was estimated to be 1.67% specifically within genomic regions that contained genes anno- (95% CI −0.07–0.32, p = 0.041). tated as important in mitochondrial function. The construction of this annotation was driven by the principle that genomic regions, Mitochondria function-specific polygenic risk score is significantly which are known to be the sites of mutations in individuals with associated with disease status rare mitochondrial diseases or are candidate regions for such Next, we calculated PRS to capture the addictive effect of all mutations provide the best evidence for involvement in mito- common variants within genes implicated in mitochondria chondrial function. function on PD risk. PRS is a particularly powerful approach in GCTA is a statistical method that estimates phenotypic variance this context because it is able to efficiently incorporate informa- of complex traits explained by genome-wide SNPs, including tion from all hits, including sub-significant hits, which may those not associated with the trait in a GWAS. Using GCTA, nonetheless be etiologically relevant. Again, initially we ran the heritability estimates were first calculated for the four largest analysis including and excluding the PD risk genes and the IPDGC GWAS datasets, including all variants (WTCCC PD GWAS comparison can be seen in Supplementary Fig. 2. However, to only (PMID:21044948), Spanish Parkinson’s (IPDGC) part2, NIA PD report novel associations we focused on the lists excluding the PD GWAS (PMID:19915575), Dutch GWAS (PMID:21248740). Owing risk genes. to the low number of included cases, the heritability estimates in Using this approach, the primary and secondary mitochondrial the other IPDGC datasets were deemed less reliable. Consistent genomic annotations were found to be significantly associated with previous heritability estimates from both Keller et al.(24%) with PD disease status. Remarkably, the primary gene list and Chang et al. (21%), our random-effects meta-analysis for the consisting of only 176 genes implicated in Mendelian mitochon- four datasets identified 23% (95% CI 12–34, p = 2.72E-05) pheno- drial disorders was associated with PD with an odds ratio of 1.12 typic variance associated with all PD samples (Tables 1 and 2). per standard deviation increase in the PRS from the population There was a high degree of consistency across the cohorts. mean (random-effects p-value = 6.00E-04, beta = 0.11, SE = 0.03). npj Parkinson’s Disease (2019) 8 Published in partnership with the Parkinson’s Foundation K.J. Billingsley et al. 3 Table 2. Summary of random-effects meta-analysis for the primary and secondary mitochondrial gene lists Heritability estimate Lower 95% Upper 95% p-value from Heterogeneity of variance Heterogeneity from random-effects confidence interval confidence interval random effects from random effects (%) p-value All SNPs 0.2313 0.1233 0.3393 2.72E-05 0.0100 3.00E-03 Primary 0.0026 −0.0011 0.0062 1.66E-01 0.0000 4.85E-01 Secondary 0.0167 0.0007 0.0328 4.10E-02 0.0001 9.63E-02 Here, we show the random-effects meta-analysis of heritability estimates for all SNPs in the genome (All SNPs), estimate calculated with for the SNPs within the primary mitochondria list genes (Primary), and the SNPs within the secondary mitochondria list genes (Secondary) Fig. 2 Forest plots of PRS for Parkinson’s Disease across cohorts. Random effect meta-analysis results are shown as red diamonds and fixed effects are shown as blue, with the centerline of each diamond representing the summary PRS for that dataset. IPDGC NeuroX = (Nalls et al. 2015, PMID:25444595), OSLO = Oslo Parkinson’s Disease Study, PDBP = Parkinson’s Disease Biomarker’s Program, PPMI = Parkinson’s Progression Markers Initiative, Baylor = Baylor College of Medicine/University of Maryland, German GWAS = (PMID:19915575), VANCE = Vance (dbGap phs000394) The secondary gene list, which also included genes implicated in AAO6,26–28 with higher genetic risk scores being significantly mitochondria function or morphology, was associated with PD associated with an overall trend for earlier AAO of disease. While with a higher odds ratio of 1.28 per standard deviation increase in the primary mitochondrial gene list was not significantly the PRS from the population mean (random-effects p-value = associated with AAO of disease, the secondary gene list was 1.9E-22, beta = 0.25, SE = 0.03) (Fig. 2). Altogether, these analyses correlated with AAO. Contrary to expectation, the cumulative not only provide further support for importance of mitochondrial burden of common variants within the 1326 genes comprising the processes in PD, but potentially provide a tool for identifying PD PRS for PD risk were positively correlated with AAO of PD. After patients most likely to benefit from treatments specifically meta-analyzing, we found that each 1SD increase in PRS, led to a targeting mitochondrial function. 0.55 year increase in the AAO of disease (summary effect = 0.211, 95% CI (0.141–0.970),|2 = 68.49%, p-value = 9.00E-03, Fig. 3). As Mitochondria function-specific polygenic risk score is significantly the forest plots demonstrate, although there was a relatively high associated with later age at onset heterogeneity across studies, the directionality and magnitude of Although multiple lines of evidence point to the importance of effect on AAO were in concordance with the meta-analysis with mitochondrial dysfunction as a primary cause of PD, impaired the exception of the Oslo cohort. This finding could suggest that mitochondrial dynamics appears to be common to a wide range firstly, disease causation and progression are genetically separable of neurodegenerative diseases, including Huntington’s dis- processes in PD and that secondly the role of mitochondrial ease,18,19 amyotrophic lateral sclerosis,20,21 and Alzheimer’s dysfunction in PD is likely to be highly complex with multiple disease.22–25 The latter suggests that even when impaired distinct mitochondrial processes likely to be involved at different mitochondrial function is not the primary event in disease disease stages. pathogenesis, it is a common outcome and could contribute to disease progression. We sought to test this hypothesis by Mendelian randomization suggests potential causal association of investigating the importance of common variation within our 14 novel mitochondria function genes with PD risk mitochondrial gene lists in determining the age at onset of PD Given the robust evidence for the involvement of mitochondrial (AAO). Given the significant lag period between PD pathophysiol- function in sporadic PD, next we used two-sample MR analysis to ogy and symptoms, AAO was used as an indirect measure of identify specific genes likely to be important in PD risk. MR uses disease progression. This analysis was performed using PRS since genetic variants to identify if an observed association between a it has been consistently found to be the main genetic predictor of risk factor and an outcome is consistent with causal effect.29 This Published in partnership with the Parkinson’s Foundation npj Parkinson’s Disease (2019) 8 K.J. Billingsley et al. 4 to the role of multiple mitochondrial processes in PD distinct from mitophagy. Exploring the expression of the novel mitochondria risk genes provides additional support for their role in PD We leveraged publicly available cell-specific and tissue-specific gene expression data to investigate the 14 mitochondria genes implicated in PD through MR. First, we used enrichment-weighted cell-type enrichment (EWCE) to determine whether the expression of mitochondrial PD-associated genes (as identified through MR and described above, n = 14) was enriched within a specific cell- type class or their subtypes. No significant enrichment of these genes was found in any of the tested neuronal and glial cell-type classes (Supplementary Table 6 and Supplementary Fig. 3). Next, we used co-expression network analysis to identify possible functional interactions between the 14 novel mitochondrial genes identified through MR and genes implicated in monogenic forms of PD. We found that five of the 14 genes assessed, CLN8, FASN, Fig. 3 Forest plots of PRS for the age at onset of Parkinson’s Disease MPI, MRPL43, and MRPS3, were highly co-expressed with at least across cohorts. Random effect meta-analysis results are shown as one gene already implicated in monogenic PD in multiple brain red diamonds and fixed effects are shown as blue, with the centerline of each diamond representing the summary PRS for that regions ( > 3 brain regions, Supplementary Table 7). Interestingly, dataset. OSLO = Oslo Parkinson’s Disease Study, PDBP = Parkinson’s in the case of CLN8, MRPL43, and MRPS4, our novel genes were co- Disease Biomarker’s Program, PPMI = Parkinson’s Progression Mar- expressed with a monogenic PD gene already implicated in kers Initiative, German GWAS = (PMID:19915575), VANCE = Vance mitochondrial function such as PARK7. Furthermore, with the exception of CLN8 (FASN, MPI, MRPL43 and MRPS3), the novel method has been implemented in several recent genetic studies mitochondrial gene was assigned to a co-expression module to identify association between expression quantitative trait loci enriched for neuronal markers (Supplementary Table 8). (QTL), to more accurately nominate candidate genes within risk loci. Therefore, for this study, in the aim of identifying whether DISCUSSION changes in expression of mitochondria function genes are potentially causally related to PD risk, two-sample MR was We first demonstrate that a proportion of the “missing heritability” implemented. of sporadic PD can be explained by additive common genetic Since we wanted to identify novel associations, we excluded variation within genes implicated in mitochondrial function, even after exclusion of genes previously linked to PD through linkage genes already linked to PD through the most recent GWAS meta- disequilibrium with the top risk variants.4–6,30–34 We identify that analysis.6 This resulted in the exclusion of 31 genes linked to the overall heritability of PD is approximately 23% (Table 2) and mitochondrial function and in linkage disequilibrium with the top that when heritability is calculated for the secondary gene list PD risk variants. Analysis of the remaining 1432 genes (generated regions alone we can attribute around 7% of the overall by combining the primary and secondary gene lists) resulted in heritability (23%) to common variation within the mitochondria the identification of 14 novel genes linked to mitochondrial function associated genes. Although this initially looks like a low function and causally associated with PD risk (Table 3). Of the 14 estimate, to put this into context, when heritability was calculated genes, the expression of five genes (CLN8, MPI, LGALS3, CAPRIN2, in the most recent meta-analysis using the top PD risk hits, this and MUC1) was positively associated with PD risk in blood. estimate explained only 26–36% of the overall heritability.6 In Similarly, in brain PD risk was associated with increased expression addition, using PRS, which efficiently incorporates information of ATG14, E2F1, and EP300 in brain. However, negative associations from sub-significant hits, we showed that cumulative small effect in brain and blood expression were observed for MRPS34 and variants within only 196 genes linked to monogenic mitochondrial PTPN1 and LMBRD1, respectively. Finally, elevated methylation of disease significantly associated with increased PD risk (with odds FASN in the brain was found to be positively associated with PD ratios of 1.12 per standard deviation increase in PRS from the risk but elevated methylation of CRY2 was found to be inversely population mean). These findings are important for two main correlated. reasons. Firstly, given that risk profiling performed in the recent Six of the fourteen novel PD risk genes we identified (CLN8, PD meta-analysis did not identify a significant association with EP300, LMBRD1, MPI, MRPS34, and MUC1) are already associated mitochondrial function.4–6,17,30–34 Secondly, since the primary with a monogenic disorder (Table 3). We noted that neurological gene list consisted solely of the 196 genes mutated in monogenic abnormalities were a feature of the condition in five of the six mitochondrial disorders, this analysis highlights the increasingly cases with Combined Oxidative Phosphorylation Deficiency 32 close relationship between Mendelian and complex disease.7 due to biallelic mutations in MRPS34 being perhaps of particular In order to maximize the utility of this study, we used MR, which interest. In common with PD, this condition is associated with identified 14 specific mitochondrial genes of interest with putative abnormalities of movement, including dystonia and choreoathe- functional consequences in PD risk. We found that although a toid movements. Mutations causing this condition result in number of the genes we identified are clearly linked to known PD- decreased levels of MRPS34 protein causing destablization of the related pathways, such as lysosomal dysfunction in the case of small mitochondrial ribosome subunit and suggesting the CLN8 and LMBRD1 or autophagy in the case of ATG14, others involvement of mitochondrial processes distinct from mitophagy appeared to point towards new processes. In particular, this and mitochondrial homeostasis in PD. In this context, it is analysis highlighted the mitochondrial ribosome through the noteworthy that MRPL43, another nuclear gene encoding for a identification of the genes, MRPL43 and MRPS34, encoding component of the large mitochondrial ribosome subunit is also components of the large and small mitochondrial ribosome highlighted by the MR analysis. Thus, this analysis not only subunits, respectively. Interestingly, biallelic mutations in MRPS34 enabled us to identify specific genes of interest, but also pointed are known to cause a form of Leigh syndrome, characterized by npj Parkinson’s Disease (2019) 8 Published in partnership with the Parkinson’s Foundation Table 3. Significant functional associations of mitochondrial function genes via two-sample Mendelian randomization Genes Beta SE P, FDR, Probe Data source Analyte Top QTL SNP CHR, top BP, top Associated phenotype Neurological phenotypic Treatment adjusted QTL SNP QTL SNP in OMIM features response LMBRD1 −0.173 0.050 4.71E-04 ENSG00000117245 Expression57 Brain rs7763213 6 70416294 Methylmalonic Hypotonia, lethargy, Responsive to aciduria and developmental delay, vitamin B12 homocystinuria, CBLF Impaired coordination therapy type; MAHCF ATG14 0.113 0.028 4.21E-05 ENSG00000171612 Expression57 Brain rs11621265 14 55822095 NA NA NA E2F1 0.121 0.032 1.52E-04 ENSG00000159423 Expression57 Brain rs57668191 20 32289763 NA NA NA EP300 0.151 0.042 2.80E-04 ENSG00000090432 Expression57 Brain rs139486 22 41627654 Rubinstein-taybi Behavioral difficulties, mental NA syndrome 2; RTS2 retardation (mild to moderate), low-normal intelligence, Autism spectrum disorder (in some patients), delayed psychomotor development, delayed gross motor development, speech delay Published in partnership with the Parkinson’s Foundation MRPS34 −0.353 0.100 4.04E-04 ILMN_2210482 Expression58 Blood rs2575369 16 1817431 Combined oxidative Delayed psychomotor NA phosphorylation development, lack of speech, deficiency 32; inability to walk, spasticity, COXPD32 hyperreflexia, dystonia, choreoathetoid movements, abnormal T2-weighted signals in the basal ganglia and brainstem PTPN1 −0.090 0.022 5.27E-05 ILMN_1681591 Expression58 Blood rs17788127 20 49166548 NA NA NA MRPL43 0.047 0.014 8.78E-04 ILMN_1652147 Expression58 Blood rs2863095 10 102746503 NA NA NA CLN8 0.148 0.047 1.54E-03 ILMN_1701094 Expression58 Blood rs3812477 8 1734452 Ceroid lipofuscinosis, Developmental regression, NA neuronal, 8; CLN8 seizures, ataxia, speech and language difficulties, K.J. Billingsley et al. myoclonus, EEG abnormalities, cerebral atrophy, cerebellar atrophy, autofluorescent l ipopigment in neurons MPI 0.214 0.068 1.52E-03 ILMN_1761262 Expression58 Blood rs4886636 15 75196176 Congenital disorder of Hypotonia Responsive to glycosylation, type oral mannose Ib; CDG1B therapy LGALS3 0.307 0.085 3.14E-04 ILMN_1803788 Expression58 Blood rs7157678 14 55548739 NA NA NA CAPRIN2 0.337 0.101 8.49E-04 ILMN_2345739 Expression58 Blood rs11051061 12 30914668 NA NA NA MUC1 0.487 0.118 3.39E-05 ILMN_1756992 Expression58 Blood rs6427184 1 155122783 Medullary cystic NA NA kidney disease 1; MCKD11 CRY2 −0.054 0.015 1.82E-04 ch.11.939596F Methylation57 Brain rs72902436 11 45881792 NA NA NA FASN 0.068 0.019 4.47E-04 cg03407524 Methylation57 Brain rs9905991 17 80052073 NA NA NA Multi-SNP eQTL Mendelian randomization results focusing on the mitochondria-associated genes (combining the primary and secondary gene lists). Showing the 14 mitochondria function-associated genes npj Parkinson’s Disease (2019) 8 that are significantly associated with PD risk after FDR adjustment. 5 K.J. Billingsley et al. 6 neurodegeneration in infancy with dystonia and choreoathetoid functional studies will be crucial to validate how these genes movements due to basal ganglia dysfunction. Furthermore, we contribute to disease risk. Finally, it is possible that our focus on note that a recent study that utilized whole exome sequencing one specific pathway could infer a selection bias to our analyses. (WES) data from two PD cohorts to investigate35,36 rare variation in However, the fact that our results are consistent across cohorts nuclear genes associated with distinct mitochondrial processes, and the fact that this significant association is observed in multiple not only provided support for the involvement of mitochondrial tests, adds validity to our data. In light of this, however, a large- function in sporadic PD, but also identified the gene, MRPL43, scale unbiased approach should be the focus of future pathway which encodes a component of the large mitochondrial ribosomal related studies to avoid this potential bias, although this will be subunit.37 Interestingly, MRPL43 and MRPS34 were amongst five difficult given the scope of work and sample size needed. genes, which were also highly co-expressed in human brain with In summary, in this study we provide robust evidence for the genes already known to cause monogenic forms of PD. Whereas involvement of mitochondrial processes in sporadic PD, as MRPL43 and MRPS34 were highly co-expressed with PARK7 in opposed to its defined and well-established role in the monogenic modules enriched for neuronal markers, FASN and MPI were co- forms of the disease. In relation to the 14 novel mitochondrial expressed with ATP13A2, and CLN8 was located in modules function genes that we have identified, our data suggests that it is containing FBXO7 and enriched for oligodendrocyte markers. not only mitochondrial quality control and homeostasis, which While this form of analysis does not provide information at the contributes to PD risk but other key mitochondrial processes, such single-cell level, it points to the possibility of pathway interactions as the function of mitochondrial ribosomes, mirroring the between these gene sets. However, most importantly it implicates biological complexity of mitochondrial disorders. Thus, this study entirely distinct mitochondrial processes in PD risk. opens the way for the identification of novel drug targets in PD Finally, and perhaps most remarkably using our mitochondrial causation and progression. gene lists we observe clear differences between disease causation and AAO in PD. Although PRS of the primary mitochondrial gene list was not significantly associated with AAO, the PRS of the METHODS secondary mitochondrial gene list was positively correlated (p- Samples and quality control of IPDGC datasets value = 3.6E-05), indicating association with later age at onset. All genotyping data was obtained from previously generated IPDGC However, given these findings it seems plausible that some datasets, consisting of 41,321 individuals (18,869 cases and 22,452 mitochondrial processes may contribute to PD risk. Thus, this controls) of European ancestry. For the IPDGC datasets all participants analysis is consistent with the findings of the most recent and donated DNA samples and provided informed consent for participation in largest AAO PD GWAS, which reported that not all the well- genetics studies, which was approved by the local ethic committee for established risk loci are associated with AAO and suggested a each of the datasets used (National Institutes of Health, Department of different mechanisms for PD causation and AAO.38 This has also Health and Human Services; project ZO1 AG000949). Detailed demo- been shown in previous studies that have shown no association of graphic and clinical characteristics are given in Supplementary Table 1 and risk loci with AAO in sporadic PD.35,39 are explained in further detail in refs. 6,38 along with detailed quality control (QC) methods. For sample QC, in short, individuals with low call Although in this study we have comprehensively analyzed the rate ( < 95%), discordance between genetic and reported sex, hetero- largest PD datasets currently available with very specific and zygosity outliers (F-statistic cutoff of > −0.15 and < 0.15) and ancestry inclusive mitochondrial function gene lists, there are a number of outliers ( + /− 6 standard deviations from means of eigenvectors 1 and 2 limitations to our analyses. Firstly, there was a relative amount of of the 1000 Genomes phase 3 CEU and TSI populations from principal heterogeneity in AAO within the AAO GWAS studies used. This components40) were excluded. Further, for genotype QC, variants with a was due to certain cohorts AAO being self-reported and other missingness rate of > 5%, minor allele frequency < 0.01, exhibiting cohorts specifically recruiting younger onset cases. Nonetheless, Hardy–Weinberg Equilibrium (HWE) < 1E-5 and palindromic SNPs were the highly significant p-value we obtain for the association excluded. Remaining samples were imputed using the Haplotype mitochondrial genes and AAO of PD (p-value = 3.56E-05) and the Reference Consortium (HRC) on the University of Michigan imputation recognized importance of mitochondrial function in aging would server under default settings with Eagle v2.3 phasing based on reference panel HRC r1.1 2016.41,42 suggest that this finding is likely to be robust. Furthermore, it is important to recognize that our understanding of mitochondrial biology is far from complete and this is made evident by the fact Curation of genes implicated in mitochondrial disorders and that many individuals with probable genetic forms of mitochon- associated with mitochondrial function drial disease remain undiagnosed. This in conjunction with the Gene lists were built to encompass different levels of evidence for fact we also removed the mitochondria function-associated genes involvement of the respective protein products in disease phenotypes that that are known risk hits (such as SNCA), suggests that our analysis relate to mitochondrial function. The list of genes implicated in genetic likely represents an underestimate of the overall contribution of mitochondrial disorders (“primary” gene list, n = 196) has the most the mitochondria pathway to sporadic PD. This underestimate is stringent criteria of evidence that the respective genes is related to mitochondrial dysfunction. It consists of 102 nuclear genes listed in illustrated in Supplementary Figs. 1 and 2, which shows that MITOMAP (downloaded 2015) and 94 sourced from literature review as heritability estimates and PRS scores are higher when the PD risk containing mutations that cause with mitochondrial disease. genes are included. However, we note that our ultimate goal for The list of genes implicated in mitochondrial function (“secondary” gene this study was to catalog mitochondria-specific genetic risk list, n = 1487) was constructed using the OMIM API to identify all genes for outside of known risk loci, which we have subsequently reported. which the word “mitochondria” (or derivatives) appeared in the free-text Another possible limitation of this present study is that the description, and by combining this information with MitoCarta v2.0 genes statistical tools we have used in these analyses are currently with no OMIM phenotype. This therefore gathered a list of plausible limited. For example, MR ultimately relies on the availability of biological candidate genes, i.e., genes that are functionally implicated in sufficient quantities of high-quality eQTL data. However, as there mitochondrial function and morphology for which we may lack genetic is a future focus to; increase dataset sample size, report and evidence for mitochondrial disease association. Next, to identify novel PD-associated genes, the 349 genes identified to characterize phenotypes such as AAO more accurately and be in LD with the PD risk variants of interest in the most recent PD meta- continue to increase the number of identified mitochondrial analysis6 were removed from both lists (removed genes listed in disease and function genes, we will be able to further explore the Supplementary Table 2). The final “primary” and “secondary” gene lists role of specific mitochondrial processes in more detail and identify are given in Supplementary Table 3 and Supplementary Table 4 and their distinct contribution to disease causation and progression. In following the removal the PD-associated genes were n = 178 and n = regard to the MR nominated genes in particular, further follow-up 1328, respectively. npj Parkinson’s Disease (2019) 8 Published in partnership with the Parkinson’s Foundation K.J. Billingsley et al. 7 Cohort-level heritability estimates and meta-analysis with a nearby gene. Using the default SMR protocols, linkage pruning and Genome-wide complex trait analysis (GCTA) was used to calculate clumping were implemented. Finally, for each dataset p-values were heritability estimates for the four largest IPDGC cohorts (WTCCC PD GWAS adjusted by false discovery rate to account for multiple testing. (PMID:21044948), Spanish Parkinson’s (IPDGC) part2, NIA PD GWAS (PMID:19915575), Dutch GWAS (PMID:21248740) using non-imputed genotyping data for all variants within both mitochondria gene lists using Co-expression network analysis the same workflow as in ref. 43 Genetic relationship matrices were Co-expression network analysis was used to determine whether mitochon- calculated for each dataset to identify the genetic relationship between drial genes associated with PD using the SMR analysis are co-expressed pairs of individuals. Genetic relationship matrices were then input into with genes associated with monogenic forms of PD in human brain. This restricted maximum likelihood analyses to produce estimates of the analysis was performed by using GTEx V6 gene expression data51 to proportion of phenotypic variance explained by the SNPs within each generate co-expression networks for each of the 13 brain tissues included subset of data. Principal components (PCs) were generated for each within the GTEx study. The raw FPKM (Fragments Per Kilobase of transcript dataset using PLINK (version 1.9). In order to adjust for factors accounting per Million mapped reads) values were corrected for known batch effects, to possible population substructure, the top 20 generated eigenvectors age at death, sex and post-mortem interval, as well as unknown effects. from the PC analysis, age, sex, and prevalence were used as basic The unknown effects were detected with the Surrogate Variable Analysis covariates. Disease prevalence standardized for age and gender based on epidemiological reports was specified at 0.002.43–47 Summary statistics (SVA) R Package52 and correction was performed using ComBat53. The from these estimates were produced for all four datasets and were resulting residuals were used to create a signed network using the included in the meta-analyses. Random-effects meta-analysis using the blockwiseConsensusModules R function from the WGCNA R package54 for residual maximum likelihood method was performed using R (version each of the 13 tissues. Next, the modules obtained in each of the 13 3.5.1) package metafor to calculate p-values and generate forest plots.48 networks were assigned to cell types using the user List Enrichment R function implemented in the WGCNA R package, which measures enrichment between module-assigned genes and defined brain-related Risk profiles versus disease status and age at onset lists using a hypergeometric test. The same approach was used to annotate Previous risk profiling methods have calculated polygenic risk scores (PRS) modules with Gene Ontology, REACTOME55 and KEGG56 terms. using only variants that exhibit genome-wide significant associated with disease risk. However, in the most recent PD meta-analysis, it is shown that using variants at thresholds below genome-wide significance improves Expression-weighted cell-type enrichment (EWCE): evaluating genetic predictions of disease risk.6,43 Mirroring this workflow, but instead enrichment of mitochondrial genes associated with PD risk using only variants within gene regions outlined in both the primary and secondary gene lists, the R package PRsice2 was used to carry out PRS Expression-weighted cell-type enrichment (EWCE) (https://github.com/ profiling in the standard weighted allele dose manner. In addition, PRsice2 NathanSkene/EWCE) (EWCE) was used to determine whether mitochon- performs permutation testing and p-value aware LD pruning to facilitate drial genes associated with PD using the MR analysis have higher identifying the best p-value threshold for variant inclusion to construct the expression within a particular cell type than expected by chance. The input PRS. External summary statistics utilized in this phase of analysis included for the analysis was (1) neuronal and glial clusters of the central nervous data from leave-one-out meta-analyses (LOOMAs) that exclude the study system (CNS) identified in the Linnarsson single-cell RNA sequencing in which the PRS was being tested, avoiding overfitting/circularity to some dataset (amounting to a subset of 114 of the original 265 clusters degree. LD clumping was implemented under default settings (window identified) (http://mousebrain.org/) and (2) our list of mitochondrial genes size = 250 kb, r2 > 0.1) and for each dataset 10,000 permutations of highlighted through the MR analysis (see Supplementary Table 6 for the phenotype-swapping were used to generate empirical p-value estimates full list of Linnarsson CNS neuronal clusters used). for each GWAS derived p-value threshold ranging from 5E-08 to 0.5, at a For each gene in the Linnarsson dataset, cell-type specificity was minimum increment of 5E-08. Each permutation test in each dataset estimated (the proportion of a gene’s total expression in one cell-type provided a Nagelkerke’s pseudo r2 after adjustment for an estimated compared to all cell types) using the ‘generate.celltype.data’’ function of prevalence of 0.005 and study-specific PCs 1–5, age and sex as covariates. the EWCE package. EWCE with the target list was run with 100,000 GWAS derived p-value threshold with the highest pseudo r2 was selected bootstrap replicates, which were sampled from a background list of genes for further analysis. Summary statistics were meta-analyzed using random that excluded all genes without a 1:1 mouse:human ortholog. In addition, effects (REML) per study-specific dataset using PRSice2.49 For the age at transcript length and GC-content biases were controlled for by selecting onset risk profiling, the same workflow was followed, however instead, age bootstrap lists with comparable properties to the target list. The analysis at onset was used as a continuous variable, as previously reported.38 To remove possible confounders that could possibly drive a false association was performed using major cell-type classes (e.g., “telencephalon with age we removed APOE and FOXO3, which are general markers for inhibitory interneurons”, “telencephalon projecting excitatory neurons”, aging.50 etc.) and subtypes of these classes (e.g., TEINH6 [“Interneuron-selective interneurons, cortex/hippocampus”], TEINH7 [“Interneuron-selective inter- neurons, hippocampus”] etc.). Data are displayed as standard deviations Mendelian randomization to explore possible causal effect of from the mean, and any values < 0, which reflect a depletion of expression, mitochondria function genes are displayed as 0. p-values were corrected for multiple testing using the Both mitochondria gene lists were combined, and all genes already Benjamini–Hochberg method over all cell types and gene lists displayed. associated with PD (i.e., that have been identified to be in LD with PD risk loci in the last meta-analysis) were removed, leaving 1432 unique mitochondria function gene regions. OMIM data We utilized four large-scale methylation and expression datasets Phenotype relationships and clinical synopses of all OMIM genes were through the summary-data-based Mendelian randomization (SMR) downloaded from http://api.omim.org on the 29th of May 2018. OMIM (http://cnsgenomics.com/software/smr) resource. Summary statistics were genes were filtered to exclude provisional, non-disease and susceptibility compared to PD outcome summary statistics for the mitochondria variants phenotypes retaining 2898 unique genes that were confidently associated of interest (extracted from refs. 4–6,30–34) to identify possible association using the R package TwoSampleMR. to 4034 Mendelian diseases. The phenotypic information relating to all Tissues were selected based on their relevance to the pathobiology of genes associated with mitochondrial disorders was collated. PD, which ultimately consisted of tissues from ten brain regions, whole blood, skeletal muscle, and nerve (a full list of the tissues utilized can be found in Supplementary Table 5). For the methylation QTLs “middle age” data was used, which was the oldest available time point. For each DATA AVAILABILITY mitochondria function variant of interest (considered here the instru- For all datasets included in this study GWAS summary statistics are available at: mental variable), wald ratios were generated for each variable that tagged https://drive.google.com/file/d/1FZ9UL99LAqyWnyNBxxlx6qOUlfAnublN/view? a cis-QTL (probes within each gene and meeting a QTL p-value of at least usp=sharing Underlying participant level IPDGC data is available to potential 5E-08 in the original QTL study) and for a methylation or expression probe collaborators, please contact ipdgc.contact@gmail.com. Published in partnership with the Parkinson’s Foundation npj Parkinson’s Disease (2019) 8 K.J. Billingsley et al. 8 ACKNOWLEDGEMENTS 20. Atsumi, T. The ultrastructure of intramuscular nerves in amyotrophic lateral This work was supported in part by the Intramural Research Program of the National sclerosis. Acta Neuropathol. 55, 193–198 (1981). Institute on Aging, National Institutes of Health, Department of Health and Human 21. Sasaki, S. & Iwata, M. Mitochondrial alterations in the spinal cord of patients with Services; project ZO1 AG000949. sporadic amyotrophic lateral sclerosis. J. Neuropathol. Exp. Neurol. 66, 10–16 (2007). 22. Moreira, P. I., Cardoso, S. M., Santos, M. S. & Oliveira, C. R. The key role of mito- AUTHOR CONTRIBUTIONS chondria in Alzheimer’s disease. J. Alzheimers Dis. 9, 101–110 (2006). 23. Nunomura, A. et al. Oxidative damage is the earliest event in Alzheimer disease. J. Data generation and analysis: K.J.B., I.A.B., S.B.-C., C.D., J.A.B., R.H.R., D.Z., M.A.S., C.B., Z. Neuropathol. Exp. Neurol. 60, 759–767 (2001). G.-O., J.R.B., M.A.N., A.S., M.R., S.K. Study concept and design: K.J.B., I.A.B., S.B.-C., J.Q., V. 24. Moreira, P. I., Duarte, A. I., Santos, M. S., Cristina Rego, A. & Oliveira, C. R. An J.B., C.D., J.A.B., R.H.R., M.A.S., J.R.B., M.A.N., A.S., M.R., S.K. Critical revision of the integrative view of the role of oxidative stress, mitochondria and insulin in Alz- manuscript for important intellectual content: all authors. heimer’s disease. J. Alzheimers Dis. 16, 741–761 (2009). 25. Fang, E. F. et al. Mitophagy inhibits amyloid-β and tau pathology and reverses cognitive deficits in models of Alzheimer's disease. Nat. Neurosci. 22, 401–412 ADDITIONAL INFORMATION (2019). Supplementary Information accompanies the paper on the npj Parkinson’s Disease 26. Nalls, M. A. et al. Genetic risk and age in Parkinson’s disease: Continuum not website (https://doi.org/10.1038/s41531-019-0080-x). stratum. Mov. Disord. 30, 850–854 (2015). 27. Escott-Price, V. et al. Polygenic risk of Parkinson disease is correlated with disease Competing interests: Mike A. Nalls’ participation is supported by a consulting age at onset. Ann. Neurol. 77, 582–591 (2015). contract between Data Tecnica International and the National Institute on Aging, NIH, 28. Lill, C. M. et al. Impact of Parkinson’s disease risk loci on age at onset. Mov. Disord. Bethesda, MD, USA, as a possible conflict of interest Dr. Nalls also consults for 30, 847–850 (2015). Neuro23 Inc, Lysosomal Therapeutics Inc, the Michael J. Fox Foundation, Illumina Inc. 29. Lawlor, D. A., Harbord, R. M., Sterne, J. A. C., Timpson, N. & Davey Smith, G. and Vivid Genomics among others. Dr. Gan-or also consults for for Lysosomal Mendelian randomization: using genes as instruments for making causal infer- Therapeutics Inc., Denaly, Prevail Therapeutics, Idorsia, and Allergan. The other ences in epidemiology. Stat. Med. 27, 1133–1163 (2008). authors declare no competing interests. 30. International Parkinson Disease Genomics Consortium et al. Imputation of sequence variants for identification of genetic risks for Parkinson’s disease: a Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims meta-analysis of genome-wide association studies. Lancet 377, 641–649 (2011). in published maps and institutional affiliations. 31. Satake, W. et al. Genome-wide association study identifies common variants at four loci as genetic risk factors for Parkinson’s disease. Nat. Genet. 41, 1303–1307 (2009). REFERENCES 32. Simón-Sánchez, J. et al. Genome-wide association study reveals genetic risk underlying Parkinson’s disease. Nat. Genet. 41, 1308–1312 (2009). 1. Singleton, A. & Hardy, J. The evolution of genetics: Alzheimer’s and Parkinson’s 33. Pihlstrøm, L. et al. Supportive evidence for 11 loci from genome-wide association Diseases. Neuron 90, 1154–1163 (2016). studies in Parkinson’s disease. Neurobiol. Aging 34, 1708.e7–1708.e13 (2013). 2. Gasser, T. Genetics of parkinson’s disease. Ann. Neurol. 44, S53–S57 (1998). 34. Lill, C. M. et al. Comprehensive research synopsis and systematic meta-analyses in 3. Reeve, A., Simcox, E. & Turnbull, D. Ageing and Parkinson’s disease: why is Parkinson’s disease genetics: The PDGene database. PLoS Genet. 8, e1002548 advancing age the biggest risk factor? Ageing Res. Rev. 14, 19–30 (2014). (2012). 4. Chang, D. et al. A meta-analysis of genome-wide association studies identifies 17 35. Hill-Burns, E. M. et al. Identification of genetic modifiers of age-at-onset for new Parkinson’s disease risk loci. Nat. Genet. 49, 1511–1516 (2017). familial Parkinson’s disease. Hum. Mol. Genet. 25, 3849–3862 (2016). 5. Nalls, M. A. et al. Large-scale meta-analysis of genome-wide association data 36. Brockmann, K. et. al. SNCA: major genetic modifier of age at onset of Parkinson’s identifies six new risk loci for Parkinson’s disease. Nat. Genet. 46, 989–993 (2014). disease. -PubMed- NCBI. Available at: https://www.ncbi.nlm.nih.gov/pubmed/ 6. Nalls, M. A. et al. Parkinson’s disease genetics: identifying novel risk loci, pro- 23674386. (Accessed 1 Mar 2019) viding causal insights and improving estimates of heritable risk (2018). https:// 37. Gaare, J. J. et al. Rare genetic variation in mitochondrial pathways influences the doi.org/10.1101/388165 risk for Parkinson’s disease. Mov. Disord. 33, 1591–1600 (2018). 7. Robak, L. A. et al. Excessive burden of lysosomal storage disorder gene variants in 38. Blauwendraat, C. et al. Parkinson disease age of onset GWAS: defining heritability, Parkinson’s disease. Brain 140, 3191–3203 (2017). genetic loci and a-synuclein mechanisms. (2018). https://doi.org/10.1101/424010 8. Hardy, J. Genetic analysis of pathways to Parkinson disease. Neuron 68, 201–206 39. Huang, Y. et al. SNCAGene, but NotMAPT, influences onset age of Parkinson’s (2010). disease in Chinese and Australians. BioMed. Res. Int. 2015, 1–6 (2015). 9. Billingsley, K. J., Bandres-Ciga, S., Saez-Atienzar, S. & Singleton, A. B. Genetic risk 40. 1000 Genomes Project Consortium. et al. A global reference for human genetic factors in Parkinson’s disease. Cell Tissue Res. 373, 9–20 (2018). variation. Nature 526, 68–74 (2015). 10. Langston, J. W., Ballard, P., Tetrud, J. W. & Irwin, I. Chronic Parkinsonism in 41. McCarthy, S. et al. A reference panel of 64,976 haplotypes for genotype impu- humans due to a product of meperidine-analog synthesis. Science 219, 979–980 tation. Nat. Genet. 48, 1279–1283 (2016). (1983). 42. Das, S. et al. Next-generation genotype imputation service and methods. Nat. 11. Langston, J. W. & Ballard, P. A. Parkinson's disease in a chemist working with 1- Genet. 48, 1284–1287 (2016). methyl-4-phenyl-l,2,5,6-tetrahydropyridine. N. Engl. J. Med. 309, 310–310 (1983). 43. Keller, M. F. et al. Using genome-wide complex trait analysis to quantify ‘missing 12. Parker, W. D., Parks, J. K. & Swerdlow, R. H. Complex I deficiency in Parkinson's heritability’ in Parkinson’s disease. Hum. Mol. Genet. 21, 4996–5009 (2012). disease frontal cortex. Brain Res. 1189, 215–218 (2008). 44. Wirdefeldt, K., Gatz, M., Reynolds, C. A., Prescott, C. A. & Pedersen, N. L. Heritability 13. Canet-Avilés, R. M. et al. The Parkinson’s disease protein DJ-1 is neuroprotective of Parkinson disease in Swedish twins: a longitudinal study. Neurobiol. Aging 32, due to cysteine-sulfinic acid-driven mitochondrial localization. Proc. Natl Acad. Sci. 1923.e1–8 (2011). USA 101, 9103–9108 (2004). 45. Gasser, T. Genetics of Parkinsonʼs disease. Curr. Opin. Neurol. 18, 363–369 (2005). 14. Funayama, M. et al. CHCHD2 mutations in autosomal dominant late-onset Par- 46. Wickremaratchi, M. M. et al. Prevalence and age of onset of Parkinson’s disease in kinson’s disease: a genome-wide linkage and sequencing study. Lancet Neurol. Cardiff: a community based cross sectional study and meta-analysis. J. Neurol. 14, 274–282 (2015). Neurosurg. Psychiatry 80, 805–807 (2009). 15. Burchell, V. S. et al. The Parkinson’s disease-linked proteins Fbxo7 and Parkin 47. Porter, B., Macfarlane, R., Unwin, N. & Walker, R. The prevalence of Parkinson’s interact to mediate mitophagy. Nat. Neurosci. 16, 1257–1265 (2013). disease in an area of North Tyneside in the north-east of England. Neuroepide- 16. Lesage, S. et al. Loss of VPS13C function in autosomal-recessive Parkinsonism miology 26, 156–161 (2006). causes mitochondrial dysfunction and increases PINK1/Parkin-dependent mito- 48. Viechtbauer, W. Conducting meta-analyses in R with the meta for Package. J. Stat. phagy. Am. J. Hum. Genet. 98, 500–513 (2016). Softw 36, 1–48 (2010). 17. Gorman, G. S. et al. Mitochondrial diseases. Nat. Rev. Dis. Prim. 2, 16080 (2016). 49. Euesden, J., Lewis, C. M. & O’Reilly, P. F. PRSice: polygenic risk score software. 18. Goebel, H. H., Heipertz, R., Scholz, W., Iqbal, K. & Tellez-Nagel, I. Juvenile Hun- Bioinformatics 31, 1466–1468 (2014). tington chorea: Clinical, ultrastructural, and biochemical studies. Neurology 28, 50. Broer, L. et al. GWAS of longevity in CHARGE consortium confirms APOE and 23–23 (1978). FOXO3 candidacy. J. Gerontol. A Biol. Sci. Med. Sci. 70, 110–118 (2015). 19. Carmo, C., Naia, L., Lopes, C. & Rego, A. C. Mitochondrial Dysfunction in Hun- 51. Carithers, L. J. et al. A Novel Approach to High-Quality Postmortem Tissue Pro- tington’s Disease. In Advances in Experimental Medicine and Biology 59–83 curement: The GTEx Project. Biopreserv. Biobank 13, 311–319 (2015). (2018). npj Parkinson’s Disease (2019) 8 Published in partnership with the Parkinson’s Foundation K.J. Billingsley et al. 9 52. Leek, J. T. & Storey, J. D. Capturing heterogeneity in gene expression studies by Open Access This article is licensed under a Creative Commons surrogate variable analysis. PLoS Genet. 3, 1724–1735 (2007). Attribution 4.0 International License, which permits use, sharing, 53. Johnson, W. E., Li, C. & Rabinovic, A. Adjusting batch effects in microarray adaptation, distribution and reproduction in any medium or format, as long as you give expression data using empirical Bayes methods. Biostatistics 8, 118–127 (2007). appropriate credit to the original author(s) and the source, provide a link to the Creative 54. Langfelder, P. & Horvath, S. WGCNA: an R package for weighted correlation Commons license, and indicate if changes were made. The images or other third party network analysis. BMC Bioinform 9, 559 (2008). material in this article are included in the article’s Creative Commons license, unless 55. Fabregat, A. et al. The Reactome Pathway Knowledgebase. Nucleic Acids Res 46, indicated otherwise in a credit line to the material. If material is not included in the D649–D655 (2018). article’s Creative Commons license and your intended use is not permitted by statutory 56. Kanehisa, M., Sato, Y., Kawashima, M., Furumichi, M. & Tanabe, M. KEGG as a regulation or exceeds the permitted use, you will need to obtain permission directly reference resource for gene and protein annotation. Nucleic Acids Res 44, from the copyright holder. To view a copy of this license, visit http://creativecommons. D457–D462 (2016). org/licenses/by/4.0/. 57. Qi, T. et al. Identifying gene targets for brain-related traits using transcriptomic and methylomic data from blood. Nat. Commun. 9, 2282 (2018). 58. Westra, H.-J. et al. Systematic identification of trans eQTLs as putative drivers of © The Author(s) 2019 known disease associations. Nat. Genet. 45, 1238–1243 (2013). INTERNATIONAL PARKINSON’S DISEASE GENOMICS CONSORTIUM (IPDGC) A. Noyce13, A. Tucci14, B. Middlehurst1, D. Kia15, M. Tan16, H. Houlden14, H. R. Morris16, H. Plun-Favreau14, P. Holmans17, J. Hardy14, D. Trabzuni14,18, J. Bras19, K. Mok14, K. Kinghorn20, N. Wood15, P. Lewis21, R. Guerreiro14,19, R. Lovering22, L. R’Bibo14, M. Rizig14, V. Escott-Price23,24, V. Chelban14, T. Foltynie6, N. Williams24, A. Brice25, F. Danjou25, S. Lesage25, M. Martinez26, A. Giri27,28, C. Schulte27,28, K. Brockmann27,28, J. Simón-Sánchez27,28, P. Heutink27,28, P. Rizzu28, M. Sharma29, T. Gasser27,28, A. Nicolas2, M. Cookson2, F. Faghri2,30, D. Hernandez2, J. Shulman31,32, L. Robak33, S. Lubbe34, S. Finkbeiner35,36,37, N. Mencacci38, C. Lungu39, S. Scholz40, X. Reed2, H. Leonard2, G. Rouleau7, L. Krohan41, J. van Hilten42, J. Marinus42, A. Adarmes-Gómez43, M. Aguilar44, I. Alvarez44, V. Alvarez45, F. Javier Barrero46, J. Bergareche Yarza47, I. Bernal-Bernal43, M. Blazquez45, M. Bonilla-Toribio Bernal43, M. Boungiorno44, Dolores Buiza-Rueda43, A. Cámara48, M. Carcel44, F. Carrillo43, M. Carrión-Claro43, D. Cerdan49, J. Clarimón50,51, Y. Compta48, M. Diez-Fairen44, O. Dols-Icardo50,51, J. Duarte49, R. l. Duran52, F. Escamilla-Sevilla53, M. Ezquerra48, M. Fernández48, R. Fernández-Santiago48, C. Garcia45, P. García-Ruiz54, P. Gómez-Garre43, M. Gomez Heredia55, I. Gonzalez-Aramburu56, A. Gorostidi Pagola57, J. Hoenicka58, J. Infante51,56, S. Jesús43, A. Jimenez-Escrig59, J. Kulisevsky51,60, M. Labrador-Espinosa43, J. Lopez-Sendon59, A. López de Munain Arregui59, D. Macias43, I. Martínez Torres61, J. Marín51,60, M. Jose Marti48, J. Martínez-Castrillo59, C. Méndez-del-Barrio43, M. Menéndez González43, A. Mínguez53, P. Mir43, E. Mondragon Rezola57, E. Muñoz48, J. Pagonabarraga51,60, P. Pastor44, F. Perez Errazquin55, T. Periñán-Tocino43, J. Ruiz-Martínez57, C. Ruz52, A. Sanchez Rodriguez56, M. Sierra56, E. Suarez-Sanmartin4, C. Tabernero59, J. Pablo Tartari44, C. Tejera-Parrado43, E. Tolosa48, F. Valldeoriola48, L. Vargas-González43, L. Vela62, F. Vives52, A. Zimprich63, L. Pihlstrom64, P. Taba65, K. Majamaa66,67, A. Siitonen66, N. Okubadejo68 and O. Ojo68 13 Preventive Neurology Unit, Wolfson Institute of Preventive Medicine, QMUL, London, UK; 14Department of Molecular Neuroscience, UCL Institute of Neurology, London, UK; 15 UCL Genetics Institute; and Department of Molecular Neuroscience, UCL Institute of Neurology, London, UK; 16Department of Clinical Neuroscience, University College London, London, UK; 17Biostatistics & Bioinformatics Unit, Institute of Psychological Medicine and Clinical Neuroscience, MRC Centre for Neuropsychiatric Genetics & Genomics, Cardiff, UK; 18Department of Genetics, King Faisal Specialist Hospital and Research Centre, Riyadh 11211, Saudi Arabia; 19UK Dementia Research Institute at UCL and Department of Molecular Neuroscience, UCL Institute of Neurology, London, UK; 20Institute of Healthy Ageing, University College London, London, UK; 21University of Reading, Reading, UK; 22 University College London, London, UK; 23MRC Centre for Neuropsychiatric Genetics and Genomics, Cardiff, UK; 24Cardiff University School of Medicine, Cardiff, UK; 25Institut du Cerveau et de la Moelle épinière, ICM, Inserm U 1127, CNRS, UMR 7225, Sorbonne Universités, UPMC University Paris 06, UMR S 1127, AP-HP, Pitié-Salpêtrière Hospital, Paris, France; 26INSERM UMR 1220; and Paul Sabatier University, Toulouse, France; 27Department for Neurodegenerative Diseases, Hertie Institute for Clinical Brain Research, University of Tübingen, Tübingen, Germany; 28DZNE, German Center for Neurodegenerative Diseases, Tübingen, Germany; 29Centre for Genetic Epidemiology, Institute for Clinical Epidemiology and Applied Biometry, University of Tubingen, Tubingen, Germany; 30Department of Computer Science, University of Illinois at Urbana-Champaign, Urbana, IL, USA; 31 Departments of Neurology, Neuroscience, and Molecular & Human Genetics, Baylor College of Medicine, Houston, TX, USA; 32Jan and Dan Duncan Neurological Research Institute, Texas Children’s Hospital, Houston, TX, USA; 33Baylor College of Medicine, Houston, TX, USA; 34Ken and Ruth Davee Department of Neurology, Northwestern University Feinberg School of Medicine, Chicago, IL, USA; 35Departments of Neurology and Physiology, University of California, San Francisco, CA, USA; 36Gladstone Institute of Neurological Disease, San Francisco, CA, USA; 37Taube/Koret Center for Neurodegenerative Disease Research, San Francisco, CA, USA; 38Northwestern University Feinberg School of Medicine, Chicago, IL, USA; 39National Institutes of Health Division of Clinical Research, NINDS, National Institutes of Health, Bethesda, MD, USA; 40Neurodegenerative Diseases Research Unit, National Institute of Neurological Disorders and Stroke, Bethesda, MD, USA; 41Department of Human Genetics, McGill University, Montréal, QC H3A 0G4, Canada; 42 Department of Neurology, Leiden University Medical Center, Leiden, Netherlands; 43Instituto de Biomedicina de Sevilla (IBiS), Hospital Universitario Virgen del Rocío/CSIC/ Universidad de Sevilla, Seville, Spain; 44Fundació Docència i Recerca Mútua de Terrassa and Movement Disorders Unit, Department of Neurology, University Hospital Mutua de Terrassa, Terrassa, Barcelona, Spain; 45Hospital Universitario Central de Asturias, Oviedo, Spain; 46Hospital Universitario Parque Tecnologico de la Salud, Granada, Spain; 47Instituto de Investigación Sanitaria Biodonostia, San Sebastián, Spain; 48Hospital Clinic de Barcelona, Barcelona, Spain; 49Hospital General de Segovia, Segovia, Spain; 50Memory Unit, Department of Neurology, IIB Sant Pau, Hospital de la Santa Creu i Sant Pau, Universitat Autònoma de Barcelona, Barcelona, Spain; 51Centro de Investigación Biomédica en Red en Enfermedades Neurodegenerativas (CIBERNED), Madrid, Spain; 52Centro de Investigacion Biomedica, Universidad de Granada, Granada, Spain; 53Hospital Universitario Virgen de las Nieves, Instituto de Investigación Biosanitaria de Granada, Granada, Spain; 54Instituto de Investigación Sanitaria Fundación Jiménez Díaz, Madrid, Spain; 55Hospital Universitario Virgen de la Victoria, Malaga, Spain; 56Hospital Universitario Marqués de Valdecilla-IDIVAL, Santander, Spain; 57Instituto de Investigación Sanitaria Biodonostia, San Sebastián, Spain; 58Institut de Recerca Sant Joan de Déu, Barcelona, Spain; 59Hospital Universitario Ramón y Cajal Madrid, Madrid, Spain; 60Movement Disorders Unit, Department of Neurology, IIB Sant Pau, Hospital de la Santa Creu i Sant Pau, Universitat Autònoma de Barcelona, Barcelona, Spain; 61Department of Neurology, Instituto de Investigación Sanitaria La Fe, Hospital Universitario y Politécnico La Fe, Valencia, Spain; 62Department of Neurology, Hospital Universitario Fundación Alcorcón, Madrid, Spain; 63Department of Neurology, Medical University of Vienna, Vienna, Austria; 64Department of Neurology, Oslo University Hospital, Oslo, Norway; 65Department of Neurology and Neurosurgery, University of Tartu, Tartu, Estonia; 66Institute of Clinical Medicine, Department of Neurology, University of Oulu, Oulu, Finland; 67Department of Neurology and Medical Research Center, Oulu University Hospital, Oulu, Finland and 68University of Lagos, Yaba, Lagos State, Nigeria Published in partnership with the Parkinson’s Foundation npj Parkinson’s Disease (2019) 8