Analysis of rare Parkinson’s disease variants in millions of people

doi:10.21203/rs.3.rs-2743857/v1

Download PDF

Article

Analysis of rare Parkinson’s disease variants in millions of people

https://doi.org/10.21203/rs.3.rs-2743857/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 08 Jan, 2024

Read the published version in npj Parkinson's Disease →

You are reading this latest preprint version

Objective

Although many rare variants have been reportedly associated with Parkinson’s disease (PD), many have not been replicated or have failed to replicate. Here, we conduct a large-scale replication of rare PD variants.

Methods

We assessed a total of 27,590 PD cases, 6,701 PD proxies, and 3,106,080 controls from three data sets: 23andMe, Inc., UK Biobank, and AMP-PD. Based on well-known PD genes, 834 variants of interest were selected from the ClinVar annotated 23andMe dataset. We performed a meta-analysis using summary statistics of all three studies.

Results

The meta-analysis resulted in 11 significant variants after Bonferroni correction, including variants in GBA1 and LRRK2. At least 9 previously reported pathogenic or risk variants for PD did not pass Bonferroni correction in this analysis.

Conclusions

Here, we provide the largest rare variant meta-analysis to date, providing thorough information of variants confirmed, newly identified, or rebutted for their association with PD.

Health sciences/Diseases/Neurological disorders/Parkinson's disease

Health sciences/Risk factors

Parkinson’s

rare variants

genetics

Over the past several decades, common and rare variants in multiple genes have been associated with Parkinson’s disease (PD). PD is a neurodegenerative disorder primarily affecting dopaminergic neurons in the substantia nigra, and is caused by a combination of aging, environmental factors, and genetics. Genetics account for a large proportion of PD risk (Blauwendraat, Nalls, and Singleton 2020; Bandres-Ciga et al. 2020). Several large-scale case-control genome-wide association studies (GWAS) have been performed and have identified 90 risk variants (Nalls et al. 2019), typically with moderate odds ratios (OR, mean = 1.24, range: 1.05–11.35). Many rare variants that have been associated with PD are considered either a strong risk factor (OR > 2) or a monogenic cause of disease. Pathogenic variants in GBA1 and LRRK2 are the most common high-risk genetic factors for PD, typically present in 1–10% of the PD population depending on genetic ancestry. Most other known pathogenic variants are very rare (allele frequency < 0.1%, e.g. damaging variants in SNCA and PRKN).

Typically, these rare variants are also associated with a slightly different PD phenotype, highlighting the importance of their research even more. Overall, more severe motor and non-motor symptoms are seen in GBA1 variant carriers (Lesage et al. 2020; Brockmann et al. 2015; Stoker et al. 2020), a more rapid progression to dementia in SNCA carriers (Q. Wang et al. 2016; Meeus, Theuns, and Van Broeckhoven 2012; Fuchs et al. 2007), and very early onset and an akinesia-related phenotype in PRKN carriers (Khan et al. 2003). It is important to note that the status of some reported PD genes is debated, since most associations are reported without a replication cohort or in small, often biased single-case or family studies (Blauwendraat et al. 2018; Blauwendraat, Nalls, and Singleton 2020). Previously, we investigated the SNCA p.H50Q mutation, which was identified in a pathologically-proven PD case without family segregation (Proukakis et al. 2013). However, when assessing this variant in a large-scale study including over 10,000 cases and 10,000 controls, no association was identified (Blauwendraat et al. 2018). These data do not support a pathogenic role for SNCA p.H50Q despite functional evidence supporting a potential role in disease (Boyer et al. 2019). This potentially also applies to variants in other genes including DNAJC13, EIF4G1, GIGYF2, HTRA2, LRP10, TMEM230 and UCLH1, all of which have been categorized previously as low-confidence PD genes (Blauwendraat, Nalls, and Singleton 2020). A robust assessment of high-risk and causal PD variants will be incredibly valuable for the global PD community from laboratory researchers to genetic counselors.

Here we tested PD mutations from the ClinVar database for association with PD in three large case-control cohorts (23andMe, Inc., UK Biobank, and AMP-PD) totalling over 3 million individuals (27,590 cases, 6,701 proxies, and 3,106,080 controls). The large number of participants in this study is an opportunity to assess rare variants in PD more reliably, which has not previously been possible to this extent. Our goal was to create lists of high-confidence and low-confidence PD variants. This work improves our understanding of the clinical relevance of these variants, which can then be used to improve genetic testing in patients.

Gene and variant selection

For variant selection, we annotated PD GWAS summary statistics from 23andMe using ANNOVAR (K. Wang, Li, and Hakonarson 2010) including gene-based and ClinVar (version clinvar_20220320) annotation. We investigated variants with a minor allele frequency (MAF) of less than 5% to restrict our focus to rare variants while retaining known risk variants such as GBA1 p.E365K. Only biallelic exonic or splicing variants related to “Parkinson’s disease” and/or “Lewy body dementia” were kept. We selected all coding variants from genes previously published in literature: ATP13A2, DNAJC13, DNAJC6, EIF4G1, FBXO7, GBA1, GIGYF2, HTRA2, LRP10, LRRK2, PARK7, PINK1, PLA2G6, POLG, PRKN, SNCA, SYNJ1, TMEM230, UCHL1, VPS13C, VPS35 (Blauwendraat, Nalls, and Singleton 2020) and also removed “benign” and “likely benign” variants in the GBA1 gene but also in genes identified in the keyword search “Parkinson’s disease” and “Lewy body dementia”. Synonymous, PINK1-antisense (AS), and UCHL-AS variants were also excluded from the dataset. The process of variant selection based on 23andMe data is summarized in Fig. 1.

23andMe Data

A rare variant association analysis was conducted using 3,090,507 unrelated people (3,065,473 controls, 25,034 PD cases). Data was processed as previously published in (Heilbron et al. 2021). In brief: related individuals were removed, defined as > 700cM that are identical-by-descent (~ 20% of the genome or approximately first cousins in an outbred population) (Henn et al. 2012). Ancestry composition was performed as previously reported (Durand et al., n.d.), and to minimize confounding by ancestry, only individuals with predominantly European ancestry were used.

Phased participant data were imputed using Minimac3. Throughout, structural variants and small indels were treated the same as SNPs. Association test results were computed by logistic regression assuming additive allelic effects. Covariates for age, sex, and the top five genetic principal components (PCs) were included to account for residual population structure, and indicators for genotype platforms to account for genotype batch effects. The association test p-value reported was computed using a likelihood ratio test. Genotyped SNPs were excluded that: 1) had a genotyping rate < 90%, 2) were only genotyped on the “v1” or “v2” 23andMe genotyping array, 3) were found on the mitochondrial chromosome or the Y-chromosome, 4) failed a test for parent-offspring transmission (p < 10^− 20), 5) had an association with genotype date (p < 10^–50 by ANOVA of SNP genotypes against a factor dividing genotyping date into 20 roughly equal-sized buckets), 6) had a large sex effect (ANOVA of SNP genotypes, r2 > 0.1), or 7) had probes matching multiple genomic positions in the reference genome. For tests using imputed data, we used the imputed dosages rather than best-guess genotypes. Imputed SNPs were excluded that: 1) had imputation r2 < 0.5 for individuals genotyped on the “v4” and “v5” arrays, or 2) had a significant batch effect between the “v4” and ”v5” genotyping arrays (p < 10^− 50 by ANOVA of SNP dosage against genotyping array). Both genotyped and imputed SNPs were removed if: 1) available sample size was less than 20% of the total GWAS sample size, or 2) the logistic regression failed to converge (absolute value of the estimated log odds ratio or standard error > 10).

Participants provided informed consent and volunteered to participate in the research online, under a protocol approved by the external AAHRPP-accredited IRB, Ethical & Independent (E&I) Review Services. As of 2022, E&I Review Services is part of Salus IRB (https://www.versiticlinicaltrials.org/salusirb). The full GWAS summary statistics for the 23andMe discovery data set will be made available through 23andMe to qualified researchers under an agreement with 23andMe that protects the privacy of the 23andMe participants. Datasets will be made available at no cost for academic use. Please visit https://research.23andme.com/collaborate/#dataset-access/ for more information and to apply to access the data.

AMP-PD and UK Biobank Data

Association results for the variants selected from the 23andMe dataset were generated from whole-genome sequencing data made available by the Accelerating Medicines Partnership - Parkinson’s disease Initiative (AMP-PD, https://amp-pd.org/) and the whole-exome sequencing data made available by the UK Biobank (https://www.ukbiobank.ac.uk/). Processing of these cohorts have been previously described elsewhere (Makarious et al. 2022). In brief, the AMP-PD datasets consist of unrelated individuals and publicly available cohorts only. After filtering and removing those recruited in a genetic study, the data set comprises a total of 4,007 people (2,197 males, 1,810 females) and 2,556 of those are controls and 1,451 are PD cases. The UK Biobank is a large-scale biomedical database containing genetic and health information from half a million participants (Bycroft et al. 2018). Similar filtering parameters were used for UK Biobank resulting in a total of 45,857 people (22,040 males, 23,817 females), of which 38,051 are controls, 1,105 cases, 6,033 individuals with a parent that is diagnosed with PD, and 668 individuals with a sibling that is diagnosed with PD (6,701 proxy cases). As previously described, Controls were filtered to exclude any individuals with an age of recruitment < 59 years, any reported nervous system disorders (Category 2406), a parent with PD or dementia (field codes: 20107 and 20110) and any reported neurological disorder (field codes: Dementia/42018, Vascular dementia/42022, FTD/42024, ALS/42028, Parkinsonism/42030, PD/42032, PSP/42034, MSA/42036) (Makarious et al. 2022).

Statistical analyses

We used three different data sets, including summary statistics from 23andMe and sequencing data from AMP-PD and UK Biobank. All data used genome build GRCh38. PLINK (v1.9, (Purcell et al. 2007)) was used to extract the variants identified in 23andMe data from AMP-PD and UK Biobank. To generate the association files for AMP-PD and UK Biobank, we then used RVTests (Zhan et al. 2016) for single variant association testing, using sex, and principal components (PC) 1 to 5 as covariates for AMP-PD, excluding age since this is an age-matched cohort. Genetic sex, age at recruitment, Townsend score, and PC 1 to 5 were used as covariates for UK Biobank. We conducted a fixed-effect inverse variance-weighted meta-analysis with the summary statistics, using METAL (version 2020-05-05 (Willer, Li, and Abecasis 2010)). Results were annotated using ANNOVAR, refGene, avsnp150, and clinvar_20220320 (K. Wang, Li, and Hakonarson 2010). Forest plots were generated using the rmeta (version 3.0) and metafor (version 3.8-1) packages in R. Power calculations were conducted using the R (v. 3.6) package genpwr (version 1.0.4), a power and sample size calculator for genetic association studies which allows for misspecification of the model of genetic susceptibility (Moore, Jacobson, and Fingerlin 2019). This package allows the assessment of allele frequencies as low as 1E-9 at OR = 1.5, and as low as 1E-04 at OR = 3. We used an additive model with an alpha value of 0.05. Since we used an additive model, it is important to note that we had less power to detect recessive associations in our analysis.

Variant selection and dataset details

We attempted to replicate 834 published associations between PD and protein-altering genetic variants (Fig. 1, see Methods). 609 (73%) of the ClinVar annotated variants had either no reported clinical significance, or had conflicting or uncertain interpretations. Of the remaining 225 variants, 133 variants (15.9%) were classified as pathogenic and/or likely pathogenic, followed by 84 variants (10.1%) classified as benign or likely benign variants. 6 variants (0.7%) were reported to be risk factors for PD, and 2 (0.2%) variants were reported to be either pathogenic/risk factors (Supplementary Table 1A). The 834 variants were located in 32 genes including GBA1 (n = 103), VPS13C (n = 101), and LRRK2 (n = 97) (Supplementary Table 2).

The list of 834 variants includes variants that did not pass 23andMe-internal QC and therefore had no summary statistics, but were still useful for variant selection in other studies. 282 (33.8%) variants were available in the AMP-PD dataset, 608 (72.9%) in UK Biobank, and 656 (78.7%) variants in 23andMe, resulting in a total of 679 unique variants. We focused on significant variants passing Bonferroni correction in each analysis, then generated a list of significant hits, calculating OR with a 95% confidence interval.

Analysis

We conducted a meta-analysis using summary statistics generated by the single-variant association testing data from 23andMe, UK Biobank, and AMP-PD (Table 1). 23andMe data comprised 25,034 PD cases and 3,065,473 controls. Cases were 59.9% male with a mean age of 72.3 (± 10.9) years, controls were 43.5% male with a mean age of 50.1 (± 17.5) years. The UK Biobank dataset comprised 1,105 cases and 6,701 proxies. Cases and proxies were 45.6% male with an average age of 59.1 (± 7.1) years, 38,051 were controls (48.6% male) with a mean age of 64.1 (± 2.8) years. AMP-PD had 1,451 cases (63.7% male) with a mean age of 61.3 (± 10.2) years, and 2,556 controls (49.8% male) with a mean age of 70.7 (± 13.2) years. A summary of cohort demographics can be found in Table 1.

679 unique variants were analyzed in the meta-analysis, using 656 variants from 23andMe, 408 from UK Biobank, and 256 from AMP-PD. 249 (36.7%) variants were solely contributed by 23andMe, whereas 18 (2.7%) variants were only found in UK Biobank, and 1 (0.1%) variant was only found in AMP-PD (Supplementary Fig. 1). 68 variants passed the significance threshold of p < 0.05, and 11 variants passed Bonferroni correction of p < 7.4E-05, based on 0.05/679 (forest plots are provided in Supplementary Fig. 3). MAFs ranged from 4.0E-07 to 0.012. The results with their ClinVar designation comprised 4 pathogenic variants: PRKN p.R275 (rs34424986, OR = 1.3, 95% CI: 1.15–1.47). GBA1 p.R296Q (rs78973108, OR = 3.7, 95% CI: 2.1–6.5), GBA1 p.R502C (rs80356771, OR = 3.8, 95% CI: 2.7–5.4), and LRRK2 p.R1441H (rs34995376, OR = 43.0, 95% CI: 13.8-133.9). 2 variants were pathogenic/risk factor: GBA1 p.N409S (rs76763715, OR = 2.3, 95% CI: 2.0-2.5) and LRRK2 p.G2019S (rs34637584, OR = 11.8, 95% CI: 10.5–13.2). The 5 remaining variants were of conflicting, uncertain, or unknown clinical significance: GBA1 p.E365K (rs2230288, OR = 1.4, 95% CI: 1.3–1.5), GBA1 p.T408M (rs75548401, OR = 1.5, 95% CI: 1.4–1.6), LRRK2 p.L1795F (rs111910483, OR = 2.5, 95% CI: 1.6–3.8), GBA1 p.D179H (rs147138516, OR = 3.3, 95% CI: 2.1-5.0), and GBA1 p.S146L (rs758447515, OR = 1624, 95% CI: 57-45172).

Based on the ClinVar database, GBA1 p.D179H and GBA1 p.S146L were not previously reported to be associated with PD and it is worth noting that (as is expected for very rare variants with MAF < 1%) especially GBA1 p.S146L had wide confidence intervals. GBA1 p.S146L was identified in 0 cases, 3 proxy cases (individuals with a parent with PD), and 0 controls in UK Biobank; and did not pass 23andMe QC (Fig. 2, Table 2). Interestingly, LRRK2 p.L1795F has been identified previously in a family with PD (Nichols et al. 2007), however no segregation was shown and further reports are lacking in the literature. This variant was identified in 2 cases and 0 controls in AMP-PD, shows a significant association in the 23andMe cohort (OR: 2.3, 95% CI: 1.57–3.61, P = 0.0003), but was not present in UKB.

Power calculations of variants of interest

In addition to confirming previously identified variants and finding new associations, another aim of this work was to assess if we could successfully replicate PD-associated variants found in the literature. However, when working with such low allele frequencies, one of the main challenges is sufficient statistical power. This requires very large sample sizes to capture sufficient numbers of rare variant carriers for robust statistical estimation.

We conducted power calculations for all variants in the meta-analysis, accounting for the tool’s MAF limit of > 1E-09 (n = 669, Supplementary table 8) and had a special focus on variants with a p-value > 0.05 (n = 601) to assess whether we had > 80% statistical power to detect an association assuming OR = 1.5, alpha = 0.05, and an additive model. The parameters set for this calculation are generous, since an OR of 1.5 is insufficient to cause monogenic disease in affected individuals and we are not correcting for multiple testing. Therefore, under these parameters, any association that had > 80% power but was not statistically significant in our analysis is very likely to be a spurious association.

We reached 80% power for 82 variants at OR = 1.5 and alpha 0.05. 41 (50%) were of conflicting, uncertain or unknown clinical significance, 38 (46%) were benign, and 3 (3%) were pathogenic. The pathogenic variants were GLUD2 p.S498A (rs9697983, OR: 1.02, 95%CI: 0.97–1.07), POLG p.A467T (rs113994095, OR: 1.03, 95%CI: 0.79–1.36), and POLG p.G737R (rs121918054, OR: 1.01, 95%CI: 0.84–1.22). GLUD2 p.S498A is the only variant that has previously been associated with PD. POLG variants were associated with seizures, toe walking, progressive sclerosing poliodystrophy, and many other diagnoses. Since POLG mutations typically cause disease in autosomal recessive patterns the association here is not a surprise finding, however our results provide evidence against a pathogenic role of the GLUD2 variant in PD.

When increasing the OR to 3, we reached 80% power for 231 variants. 171 (74%) were of conflicting, uncertain or unknown significance, 51 (22%) were benign, 5 (2%) were pathogenic, and 3 (1%) were risk variants. 3 pathogenic variants were the same as the ones at OR = 1.5, however, we additionally had sufficient power for PINK1 p.T313M (rs74315359, OR: 1.30, 95% CI: 0.64–2.64), SNCB p.P123H (rs104893937, OR: 1.25, 95% CI: 0.68–2.31), and POLG p.G848S (rs113994098, OR: 1.11, 95% CI: 0.59–2.11). The SNCB variant is associated with Lewy body dementia and PINK1 p.T313M has previously been associated with PD. Similar to the POLG mutations, PINK1 mutations cause disease in autosomal recessive patterns and therefore the lack of association here is not a surprise finding.

Here, we assessed the role of rare variants and their association with PD using several large case-control datasets of European ancestry. We conducted a single rare variant association analysis of PD vs controls, including 23andMe, UK Biobank and AMP-PD; totalling over 3 million individuals comprising 27,590 cases, 6,701 PD proxy cases, and 3,106,080 controls. We provide very robust evidence of 11 high risk and causal variants for PD disease development. Additionally, we provide evidence that a large number of variants that have previously been associated with PD are likely not associated with PD and should be treated with caution.

Our findings clearly confirm the prominent role of variants in LRRK2 and GBA1 on increased PD risk, in particular: LRRK2 p.R1441H, p.L1795F and p.G2019S; and GBA1 p.S146L, p.D179H, p.R296Q, p.E365K, p.T408M, p.N409S and p.R502C. All of these variants have odds ratios close to or over the threshold of 1.5, which is considered clinically more relevant (Szumilas 2010; Norton, Dowd, and Maciejewski 2018). Additionally, here we update the risk estimates for these variants with a more reliable confidence interval than previously reported. LRRK2 p.L1795F is a lesser-known and studied variant compared to p.R1441H and p.G2019S, which are both well-known damaging variants. LRRK2 p.L1795F is located in the C-terminal of ROC B region and has been reported in a family with PD, however segregation was not shown (Ghani et al. 2015). Here, we provide evidence for LRRK2 p.L1795F as a genetic risk factor for PD with an estimated OR of 2.5. Interestingly, this variant was recently shown to have a functional effect providing more evidence for its pathogenicity (Kalogeropulou et al. 2022). GBA1, p.E365K, p.T408M and p.N409S all have been previously identified via GWAS or similar approaches (Nalls et al. 2019), with p.E365K and p.T408M being risk factors for PD, and p.N409S being a risk factor for PD with and additional association with Gaucher disease. The other GBA1 variants (p.S146L, p.D179H, p.R296Q and p.R502C) are all associated with Gaucher disease in a bi-allelic state and were robustly associated with PD in this study. In the case of GBA1 p.D179H, the association with PD was statistically evident with higher ORs and narrow 95% CIs: 23andMe reporting OR = 3.4 (95% CI: 2.2–5.2) and OR = 3.3 (95% CI:2.1-5.0) in the meta-analysis. However it is worth noting that this variant is often on the same haplotype as p.E365K (Pawliczek et al. 2018; den Heijer et al. 2020) and indeed when exploring AMP-PD data we also identified that all instances of the p.D179H allele are in cis (D'=1) with p.E365K.

Many variants, common or rare, are reported to be associated with PD, in fact the ClinVar database shows 4,320 results for the condition when searching for “Parkinson” as of January 1, 2023. To generate a list of variants that were not replicated previously in large case-control analyses, we assessed whether we had > 80% power at OR = 1.5 and 3, using a p-value threshold of p > 0.05. Most well-powered variants with meta-analysis p-value > 0.05 were of uncertain, unknown or unclear clinical significance, but quite a few were also previously associated with Parkinson’s (Supplementary Table 3). For example, 6 variants that were previously reported to be “pathogenic” or “likely pathogenic” did not show evidence of association with PD and therefore these variants should be treated with caution. Those variants were: POLG p.467T (rs113994095), p.G737R (rs121918054), and p.G848S (rs113994098); GLUD2 p.S498A (rs9697983), PINK1 p.T313M (rs74315359), and SNCB p.P123H (rs104893937). Interestingly, some of these variants always had very weak evidence for an association in previous publications but were still considered PD variants in the ClinVar database. The PINK1 p.T313M variant was only found to be associated with PD in a single Saudi Arabian family (Chishti et al. 2006; Plaitakis et al. 2010). However, due to our study design, results regarding autosomal recessive genes should be interpreted carefully.

We acknowledge that our analysis comes with some limitations. First, very large sample sizes are required to study rare variants. Investigating variants with low frequencies often results in unreliable high odds ratios with wide confidence intervals, suggesting that the population parameter is not confidently predicted. For example the GBA1 p.S146L variant (rs758447515, OR = 1624.7, 95% CI = 58-45172), whose generated statistics suggest that it plays a role in PD risk but will need to be further studied to determine more accurate risk estimates. Another example is SNCA p.A53T which is known to be causal for PD (Polymeropoulos et al. 1997), but given the extreme low frequency of this allele it was not found in AMP-PD or UK Biobank data, and had a p-value of 0.03 with a very wide risk estimate OR = 38.0, 95% CI = 2.82-510.12 in the 23andMe dataset. This clearly shows association but not enough carriers to pass multiple test correction. Additionally, some pathogenic variants passing significance at p < 0.05 but failing the multiple test correction were identified. GBA1 p.V433LF (rs80356769, meta-analysis: P = 1.6E-04, OR = 4.2, 95% CI = 1.98–8.8) was present in ClinVar as a pathogenic PD variant, but also the previously linked LRRK2 variant p.I2020T (rs35870237, P = 0.0004, OR = 21.4, 95% CI = 3.97–115.5) is considered pathogenic based on biological evidence (Ray et al. 2014). LRRK2 p.I2020T is associated with autosomal dominant PD (Paisán-Ruíz et al. 2004; Zimprich et al. 2004). These variants showcase the complexities of rare variant analysis and although we used the largest sample size for PD yet, it still shows the limited power we have to identify robust risk estimates for very rare variants.

Second, an important aspect of rare and causal variants is the allele-dosage effect, since recessive genes are only causal in a homozygous or compound heterozygous state, which is especially important for PD since several genes are known to only be causal in a recessive state. Most of our statistical models used here are based on additive effects and since only summary level data was available from 23andMe, we cannot accurately report results on autosomal recessive genes such as PRKN and PINK1. To highlight this further the PRKN p.R275W (rs34424986) variant is known to increase the risk for PD in a homozygous or compound heterozygous state (Lubbe et al. 2021). In our analysis this particular variant also made the list of robust associations with an overall OR of 1.3 in the meta-analysis. A preliminary analysis by 23andMe using a recessive model for PRKN p.R275W, showed that in a heterozygous state OR is at 1.4, whereas in a homozygous alternative state the OR is 6. This variant is under heavy debate for its association with PD risk in a heterozygous state and is likely only disease causal if another damaging PRKN is on the other allele (Zhu et al. 2022).

Third, our analysis started with selection of variants based on array genotype data of 23andMe and therefore we missed variants that cannot be reliably imputed or are hard to genotype (such as GBA1 p.L444P). This is a limitation that will be resolved with the availability of more sequencing data in the coming years.

Finally, the lack of diversity is a critical challenge in genetic research and limits our insights and understanding of the disease, for example current sample sizes in 23andMe for African and Asian ancestry is under 400 cases (Kim et al. 2022). However, initiatives such as the Global Parkinson’s Genetics Program (GP2) are making an active effort to increase the number of non-European underrepresented populations in genetic datasets (Global Parkinson’s Genetics Program 2021), so that in the near future, based on genetics, we can hopefully create a more representative picture of the disease.

In summary, we provide a robust list of 11 variants associated with PD and more reliable risk estimates. Additionally we provide a list of variants that were previously considered pathogenic and risk variants for PD which were not replicated in our analysis and should be treated with caution moving forward. Assessing rare variants is crucial to improve our understanding of disease development and heritability, but several complications arise when working with low frequency levels, such as lack of power. Larger data sets and more suitable tools are a critical requirement to further our understanding of rare variants, and we hope this work can be a useful tool for the wider PD community.

Acknowledgements and Funding

We would like to thank all of the subjects who donated their time and biological samples to be part of this study. This research was supported in part by the Intramural Research Program of the National Institutes of Health (National Institute on Aging and National Institute of Neurological Disorders and Stroke; project numbers: 1ZIAAG00935, 1ZIANS003154, Z01-AG000949-02). This research has been conducted using data received from 23andMe, Inc as part of a collaboration. 23andMe is a publicly held personal genomics biotechnology company based in Sunnyvale, California, US. We thank the research participants and employees of 23andMe who contributed data to make the research cited here possible.

We also used data from the UK Biobank Resource under Application Number 33601. This study used the high-performance computational capabilities of the Biowulf Linux cluster at the National Institutes of Health (http://hpc.nih.gov). Data used in the preparation of this article were obtained from the AMP-PD Knowledge Platform. For up-to-date information on the study, visit https://www.amp-pd.org. AMP-PD – a public-private partnership – is managed by the FNIH and funded by Celgene, GSK, the Michael J. Fox Foundation for Parkinson’s Research, the National Institute of Neurological Disorders and Stroke, Pfizer, and Verily. We would like to thank AMP-PD for the publicly available whole-genome sequencing data, including cohorts from the Fox Investigation for New Discovery of Biomarkers (BioFIND), the Parkinson’s Progression Markers Initiative (PPMI), and the Parkinson’s Disease Biomarkers Program (PDBP). The Parkinson's Disease Biomarker Program (PDBP) consortium is supported by the National Institute of Neurological Disorders and Stroke (NINDS) at the National Institutes of Health. A full list of PDBP investigators can be found at https://pdbp.ninds.nih.gov/policy. Harvard Biomarker Study (HBS) is a collaboration of HBS investigators (full list of HBS investigators found at https://www.bwhparkinsoncenter.org/biobank) and funded through philanthropy and NIH and Non-NIH funding sources. The HBS Investigators have not participated in reviewing the data analysis or content of the manuscript.

We would like to thank the research participants and employees of 23andMe for making this work possible. The following members of the 23andMe Research Team contributed to this study: Stella Aslibekyan, Adam Auton, Elizabeth Babalola, Robert K. Bell, Jessica Bielenberg, Katarzyna Bryc, Emily Bullis, Daniella Coker, Gabriel Cuellar Partida, Devika Dhamija, Sayantan Das, Sarah L. Elson, Nicholas Eriksson, Teresa Filshtein, Alison Fitch, Kipper Fletez-Brant, Pierre Fontanillas, Will Freyman, Julie M. Granka, Karl Heilbron, Alejandro Hernandez, Barry Hicks, David A. Hinds, Ethan M. Jewett, Yunxuan Jiang, Katelyn Kukar, Alan Kwong, Keng-Han Lin, Bianca A. Llamas, Maya Lowe, Jey C. McCreight, Matthew H. McIntyre, Steven J. Micheletti, Meghan E. Moreno, Priyanka Nandakumar, Dominique T. Nguyen, Elizabeth S. Noblin, Jared O'Connell, Aaron A. Petrakovitz, G. David Poznik, Alexandra Reynoso, Morgan Schumacher, Anjali J. Shastri, Janie F. Shelton, Jingchunzi Shi, Suyash Shringarpure, Qiaojuan Jane Su, Susana A. Tat, Christophe Toukam Tchakouté, Vinh Tran, Joyce Y. Tung, Xin Wang, Wei Wang, Catherine H. Weldon, Peter Wilton, Corinna D. Wong.

Data Availability and Ethics Statement

Accelerating Medicines Partnership in Parkinson’s Disease (AMP PD data) and quality control notebooks are access-controlled [https://amp-pd.org/], and require individual sign-up to access the data. United Kingdom Biobank (UKBiobank) data are access-controlled and require an application for access [https://www.ukbiobank.ac.uk/]. Each contributing study abided by the ethics guidelines set out by their institutional review boards, and all participants gave informed consent for inclusion in both their initial cohorts and subsequent studies. All data produced in the present work are contained in the manuscript. All authors and the public can access the statistical programming code used in this project for the analyses and results generation. VP and CB take final responsibility for the decision to submit the paper for publication. Studies that are conducted on de-identified human genetics are waived ethical approval by the NIH Intramural IRB, as they are considered non-human subjects research.

Author contributions

V.P., K.H., and C.B. contributed to the conception and design of the study.

V.P., M.B.M, M.A.N., K.H., and C.B. contributed to the acquisition and analysis of the data. V.P., M.B.M., K.H., and C.B. contributed to manuscript drafting. All authors participated in the interpretation of the results, critical revision of the manuscript, and final version approval.

Competing interests statement

M.A.N. is a consultant employed by Data Tecnica International, whose participation in this is part of a consulting agreement between the US National Institutes of Health and said company. M.A.N. also currently serves on the scientific advisory board for Clover Therapeutics and is an advisor to Neuron23 Inc. K.H. and members of the 23andMe Research Team are employees of 23andMe, Inc. and hold stock or stock options in 23andMe.

Bandres-Ciga, Sara, Monica Diez-Fairen, Jonggeol Jeff Kim, and Andrew B. Singleton. 2020. “Genetics of Parkinson’s Disease: An Introspection of Its Journey towards Precision Medicine.” Neurobiology of Disease 137 (April): 104782.
Blauwendraat, Cornelis, Demis A. Kia, Lasse Pihlstrøm, Ziv Gan-Or, Suzanne Lesage, J. Raphael Gibbs, Jinhui Ding, et al. 2018. “Insufficient Evidence for Pathogenicity of SNCA His50Gln (H50Q) in Parkinson’s Disease.” Neurobiology of Aging 64 (April): 159.e5–159.e8.
Blauwendraat, Cornelis, Mike A. Nalls, and Andrew B. Singleton. 2020. “The Genetic Architecture of Parkinson’s Disease.” Lancet Neurology 19 (2): 170–78.
Boyer, David R., Binsen Li, Chuanqi Sun, Weijia Fan, Michael R. Sawaya, Lin Jiang, and David S. Eisenberg. 2019. “Structures of Fibrils Formed by α-Synuclein Hereditary Disease Mutant H50Q Reveal New Polymorphs.” Nature Structural & Molecular Biology 26 (11): 1044–52.
Brockmann, Kathrin, Karin Srulijes, Sylvia Pflederer, Ann-Kathrin Hauser, Claudia Schulte, Walter Maetzler, Thomas Gasser, and Daniela Berg. 2015. “GBA-Associated Parkinson’s Disease: Reduced Survival and More Rapid Progression in a Prospective Longitudinal Study.” Movement Disorders: Official Journal of the Movement Disorder Society 30 (3): 407–11.
Bycroft, Clare, Colin Freeman, Desislava Petkova, Gavin Band, Lloyd T. Elliott, Kevin Sharp, Allan Motyer, et al. 2018. “The UK Biobank Resource with Deep Phenotyping and Genomic Data.” Nature 562 (7726): 203–9.
Chishti, Muhammad A., Saeed Bohlega, Maqbool Ahmed, Arslan Loualich, Pamela Carroll, Christine Sato, Peter St George-Hyslop, David Westaway, and Ekaterina Rogaeva. 2006. “T313M PINK1 Mutation in an Extended Highly Consanguineous Saudi Family with Early-Onset Parkinson Disease.” Archives of Neurology 63 (10): 1483–85.
Durand, Eric Y., Chuong B. Do, Joanna L. Mountain, and J. Michael Macpherson. n.d. “Ancestry Composition: A Novel, Efficient Pipeline for Ancestry Deconvolution.” https://doi.org/10.1101/010512.
Fuchs, J., C. Nilsson, J. Kachergus, M. Munz, E-M Larsson, B. Schüle, J. W. Langston, et al. 2007. “Phenotypic Variation in a Large Swedish Pedigree due to SNCA Duplication and Triplication.” Neurology 68 (12): 916–22.
Ghani, Mahdi, Anthony E. Lang, Lorne Zinman, Benedetta Nacmias, Sandro Sorbi, Valentina Bessi, Andrea Tedde, et al. 2015. “Mutation Analysis of Patients with Neurodegenerative Disorders Using NeuroX Array.” Neurobiology of Aging 36 (1): 545.e9–14.
Global Parkinson’s Genetics Program. 2021. “GP2: The Global Parkinson’s Genetics Program.” Movement Disorders: Official Journal of the Movement Disorder Society 36 (4): 842–51.
Heijer, Jonas M. den, Valerie C. Cullen, Marialuisa Quadri, Arnoud Schmitz, Dana C. Hilt, Peter Lansbury, Henk W. Berendse, et al. 2020. “A Large-Scale Full GBA1 Gene Screening in Parkinson’s Disease in the Netherlands.” Movement Disorders: Official Journal of the Movement Disorder Society 35 (9): 1667–74.
Heilbron, Karl, Melanie P. Jensen, Sara Bandres-Ciga, Pierre Fontanillas, Cornelis Blauwendraat, Mike A. Nalls, Andrew B. Singleton, et al. 2021. “Unhealthy Behaviours and Risk of Parkinson’s Disease: A Mendelian Randomisation Study.” Journal of Parkinson’s Disease 11 (4): 1981–93.
Henn, Brenna M., Lawrence Hon, J. Michael Macpherson, Nick Eriksson, Serge Saxonov, Itsik Pe’er, and Joanna L. Mountain. 2012. “Cryptic Distant Relatives Are Common in Both Isolated and Cosmopolitan Genetic Samples.” PloS One 7 (4): e34267.
Kalogeropulou, Alexia F., Elena Purlyte, Francesca Tonelli, Sven M. Lange, Melanie Wightman, Alan R. Prescott, Shalini Padmanabhan, Esther Sammler, and Dario R. Alessi. 2022. “Impact of 100 LRRK2 Variants Linked to Parkinson’s Disease on Kinase Activity and Microtubule Binding.” Biochemical Journal 479 (17): 1759–83.
Khan, Naheed L., Elizabeth Graham, Peter Critchley, Anette E. Schrag, Nicholas W. Wood, Andrew J. Lees, Kailash P. Bhatia, and Niall Quinn. 2003. “Parkin Disease: A Phenotypic Study of a Large Case Series.” Brain: A Journal of Neurology 126 (Pt 6): 1279–92.
Kim, Jonggeol Jeffrey, Dan Vitale, Diego Véliz Otani, Michelle Lian, Karl Heilbron, Hirotaka Iwaki, Julie Lake, et al. 2022. “Multi-Ancestry Genome-Wide Meta-Analysis in Parkinson’s Disease.” medRxiv. https://doi.org/10.1101/2022.08.04.22278432.
Lesage, Suzanne, Ariane Lunati, Marion Houot, Sawssan Ben Romdhan, Fabienne Clot, Christelle Tesson, Graziella Mangone, et al. 2020. “Characterization of Recessive Parkinson’s Disease in a Large Multicenter Study.” Annals of Neurology, May. https://doi.org/10.1002/ana.25787.
Lubbe, Steven J., Bernabe I. Bustos, Jing Hu, Dimitri Krainc, Theresita Joseph, Jason Hehir, Manuela Tan, et al. 2021. “Assessing the Relationship between Monoallelic PRKN Mutations and Parkinson’s Risk.” Human Molecular Genetics 30 (1): 78–86.
Makarious, Mary B., Julie Lake, Vanessa Pitz, Allen Ye Fu, Joseph L. Guidubaldi, Caroline Warly Solsberg, Sara Bandres-Ciga, et al. 2022. “Large-Scale Rare Variant Burden Testing in Parkinson’s Disease Identifies Novel Associations with Genes Involved in Neuro-Inflammation.” medRxiv. https://doi.org/10.1101/2022.11.08.22280168.
Meeus, Bram, Jessie Theuns, and Christine Van Broeckhoven. 2012. “The Genetics of Dementia with Lewy Bodies: What Are We Missing?” Archives of Neurology 69 (9): 1113–18.
Moore, Camille M., Sean A. Jacobson, and Tasha E. Fingerlin. 2019. “Power and Sample Size Calculations for Genetic Association Studies in the Presence of Genetic Model Misspecification.” Human Heredity 84 (6): 256–71.
Nalls, Mike A., Cornelis Blauwendraat, Costanza L. Vallerga, Karl Heilbron, Sara Bandres-Ciga, Diana Chang, Manuela Tan, et al. 2019. “Identification of Novel Risk Loci, Causal Insights, and Heritable Risk for Parkinson’s Disease: A Meta-Analysis of Genome-Wide Association Studies.” Lancet Neurology 18 (12): 1091–1102.
Nichols, W. C., V. E. Elsaesser, N. Pankratz, M. W. Pauciulo, D. K. Marek, C. A. Halter, A. Rudolph, C. W. Shults, T. Foroud, and Parkinson Study Group-PROGENI Investigators. 2007. “LRRK2 Mutation Analysis in Parkinson Disease Families with Evidence of Linkage to PARK8.” Neurology 69 (18): 1737–44.
Norton, Edward C., Bryan E. Dowd, and Matthew L. Maciejewski. 2018. “Odds Ratios-Current Best Practice and Use.” JAMA: The Journal of the American Medical Association 320 (1): 84–85.
Paisán-Ruíz, Coro, Shushant Jain, E. Whitney Evans, William P. Gilks, Javier Simón, Marcel van der Brug, Adolfo López de Munain, et al. 2004. “Cloning of the Gene Containing Mutations That Cause PARK8-Linked Parkinson’s Disease.” Neuron 44 (4): 595–600.
Pawliczek, Piotr, Ronak Y. Patel, Lillian R. Ashmore, Andrew R. Jackson, Chris Bizon, Tristan Nelson, Bradford Powell, et al. 2018. “ClinGen Allele Registry Links Information about Genetic Variants.” Human Mutation 39 (11): 1690–1701.
Plaitakis, Andreas, Helen Latsoudis, Konstantinos Kanavouras, Beate Ritz, Jeff M. Bronstein, Irene Skoula, Vasileios Mastorodemos, et al. 2010. “Gain-of-Function Variant in GLUD2 Glutamate Dehydrogenase Modifies Parkinson’s Disease Onset.” European Journal of Human Genetics: EJHG 18 (3): 336–41.
Polymeropoulos, M. H., C. Lavedan, E. Leroy, S. E. Ide, A. Dehejia, A. Dutra, B. Pike, et al. 1997. “Mutation in the Alpha-Synuclein Gene Identified in Families with Parkinson’s Disease.” Science 276 (5321): 2045–47.
Proukakis, Christos, Christopher G. Dudzik, Timothy Brier, Donna S. MacKay, J. Mark Cooper, Glenn L. Millhauser, Henry Houlden, and Anthony H. Schapira. 2013. “A Novel α-Synuclein Missense Mutation in Parkinson Disease.” Neurology 80 (11): 1062–64.
Purcell, Shaun, Benjamin Neale, Kathe Todd-Brown, Lori Thomas, Manuel A. R. Ferreira, David Bender, Julian Maller, et al. 2007. “PLINK: A Tool Set for Whole-Genome Association and Population-Based Linkage Analyses.” American Journal of Human Genetics 81 (3): 559–75.
Ray, Soumya, Samantha Bender, Stephanie Kang, Regina Lin, Marcie A. Glicksman, and Min Liu. 2014. “The Parkinson Disease-Linked LRRK2 Protein Mutation I2020T Stabilizes an Active State Conformation Leading to Increased Kinase Activity.” The Journal of Biological Chemistry 289 (19): 13042–53.
Stoker, Thomas B., Marta Camacho, Sophie Winder-Rhodes, Ganqiang Liu, Clemens R. Scherzer, Thomas Foltynie, Jonathan Evans, David P. Breen, Roger A. Barker, and Caroline H. Williams-Gray. 2020. “Impact of Variants on Long-Term Clinical Progression and Mortality in Incident Parkinson’s Disease.” Journal of Neurology, Neurosurgery, and Psychiatry 91 (7): 695–702.
Szumilas, Magdalena. 2010. “Explaining Odds Ratios.” Journal of the Canadian Academy of Child and Adolescent Psychiatry = Journal de l’Academie Canadienne de Psychiatrie de L'enfant et de L'adolescent 19 (3): 227–29.
Wang, Kai, Mingyao Li, and Hakon Hakonarson. 2010. “ANNOVAR: Functional Annotation of Genetic Variants from High-Throughput Sequencing Data.” Nucleic Acids Research 38 (16): e164.
Wang, Quanbao, Qian Tian, Xiaojie Song, Yunyong Liu, and Wei Li. 2016. “SNCA Gene Polymorphism May Contribute to an Increased Risk of Alzheimer’s Disease.” Journal of Clinical Laboratory Analysis 30 (6): 1092–99.
Willer, Cristen J., Yun Li, and Gonçalo R. Abecasis. 2010. “METAL: Fast and Efficient Meta-Analysis of Genomewide Association Scans.” Bioinformatics 26 (17): 2190–91.
Zhan, Xiaowei, Youna Hu, Bingshan Li, Goncalo R. Abecasis, and Dajiang J. Liu. 2016. “RVTESTS: An Efficient and Comprehensive Tool for Rare Variant Association Analysis Using Sequence Data.” Bioinformatics 32 (9): 1423–26.
Zhu, William, Xiaoping Huang, Esther Yoon, Sara Bandres-Ciga, Cornelis Blauwendraat, Kimberly J. Billingsley, Joshua H. Cade, et al. 2022. “Heterozygous PRKN Mutations Are Common but Do Not Increase the Risk of Parkinson’s Disease.” Brain: A Journal of Neurology 145 (6): 2077–91.
Zimprich, Alexander, Saskia Biskup, Petra Leitner, Peter Lichtner, Matthew Farrer, Sarah Lincoln, Jennifer Kachergus, et al. 2004. “Mutations in LRRK2 Cause Autosomal-Dominant Parkinsonism with Pleomorphic Pathology.” Neuron 44 (4): 601–7.

Table 1

Cohort demographics
	23andMe	AMP-PD	UKB	Total
Cases, n	25,034	1,451	1,105 (+ 6,701 proxies)#	27,590 (+ 6,701 proxies)
Age, mean ± SD	72.3 (± 10.9)~	61.3 (± 10.2)*	59.1 (± 7.1)^
Male %	59.9	63.7	45.6
Controls	3,065,473	2,556	38,051	3,106,080
Age, mean ± SD	50.1 (± 17.5)	70.7 (± 13.2)*	64.1 (± 2.8)^
Male %	43.5	49.8	48.6
Nominated variants	692	301	644
Type of data	Imputed genotype data	Whole genome sequencing	Whole exome sequencing

UKB: UK Biobank, AMP-PD: Accelerating Medicines Partnership Parkinson’s disease

# proxies in UK Biobank data include: 6,033 parents and 668 siblings

~ Current age at time of analysis

* Age at analysis for AMP-PD

^ Age at recruitment for UK Biobank data

Table 2

List of 11 variants passing Bonferroni correction in meta-analysis.
rsID	Variant name	Ref	Alt	MAF 23andMe	MAF gnomAD	CLNSIG	Power OR 1.5	Power OR 3	Effect	SE	OR	L95% CI	U95% CI
rs80356771	GBA1 p.R502C	G	A	0.00019929	0	P	23%	96%	1.33	0.18	3.80	2.66	5.43
rs76763715	GBA1 p.N409S	T	C	0.00311143	0.0018	P	100%	100%	0.82	0.05	2.26	2.05	2.50
rs2230288	GBA1 p.E365K	C	T	0.01182293	0.0141	C/U/U	100%	100%	0.36	0.03	1.44	1.34	1.54
rs34637584	LRRK2 p.G2019S	G	A	0.00062618	3.00E-04	P/R	59%	100%	2.46	0.06	11.76	10.48	13.20
rs75548401	GBA1 p.T408M	G	A	0.00685015	0.0115	C/U/U	100%	100%	0.39	0.04	1.48	1.36	1.60
rs147138516	GBA1 p.D179H	C	G	0.00012406	1.00E-04	C/U/U	16%	82%	1.19	0.22	3.27	2.14	5.00
rs34995376	LRRK2 p.R1441H	G	A	3.47E-06	0	P	5%	100%	3.76	0.58	43.03	13.83	133.94
rs34424986	PRKN p.R275W	G	A	0.0024927	0.0033	P	99%	100%	0.26	0.06	1.30	1.15	1.47
rs111910483	LRRK2 p.L1795F	G	T	8.48E-05	NA	C/U/U	13%	100%	0.91	0.21	2.49	1.65	3.75
rs78973108	GBA1 p.R296Q	C	T	1.81E-05	1.00E-04	P	7%	*	1.31	0.29	3.71	2.12	6.49
rs758447515	GBA1 p.S146L	G	A	4.03E-07	NA	C/U/U	5%	*	7.39	1.70	1,624.73	58.44	45,172.33

Ref: reference allele, Alt: alternate allele, MAF: minor allele frequency, CLNSIG: clinical significance based on ClinVar, SE: standard error, L95% CI: Lower boundary 95% confidence interval, OR: odds ratio, U95% CI: upper boundary 95% confidence interval, P: pathogenic, P/R: pathogenic/risk factor, C/U/U: conflicting, uncertain, unknown.

* Minor allele frequencies too low for the tool to calculate statistical power at OR = 3 and alpha 0.05 in an additive model

# gnomAD frequencies based on NFE (non-finnish European)

Variant name indicates variants not previously associated with PD according to ClinVar

There is a conflict of interest M.A.N. is a consultant employed by Data Tecnica International, whose participation in this is part of a consulting agreement between the US National Institutes of Health and said company. M.A.N. also currently serves on the scientific advisory board for Clover Therapeutics and is an advisor to Neuron23 Inc. K.H. and members of the 23andMe Research Team are employees of 23andMe, Inc. and hold stock or stock options in 23andMe.

Download PDF

Journal Publication

published 08 Jan, 2024

Read the published version in npj Parkinson's Disease →

Editorial decision: revise
26 May, 2023
Review #2 received at journal
22 May, 2023
Reviewer #2 agreed at journal
18 May, 2023
Review #1 received at journal
07 May, 2023
Reviewer #1 agreed at journal
21 Apr, 2023
Reviewers invited by journal
07 Apr, 2023
Submission checks completed at journal
30 Mar, 2023
Editor assigned by journal
27 Mar, 2023
First submitted to journal
27 Mar, 2023

You are reading this latest preprint version

Analysis of rare Parkinson’s disease variants in millions of people

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Methods

Results

Analysis

Discussion

Declarations

Acknowledgements and Funding

Data Availability and Ethics Statement

Author contributions

Competing interests statement

References

Tables

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1