Exome sequencing in Asian populations identifies low-frequency and rare coding variation influencing Parkinson's disease risk

Abstract

Parkinson's disease (PD) is an incurable, progressive and common movement disorder that is increasing in incidence globally because of population aging. We hypothesized that the landscape of rare, protein-altering variants could provide further insights into disease pathogenesis. Here we performed whole-exome sequencing followed by gene-based tests on 4,298 PD cases and 5,512 controls of Asian ancestry. We showed that GBA1 and SMPD1 were significantly associated with PD risk, with replication in a further 5,585 PD cases and 5,642 controls. We further refined variant classification using in vitro assays and showed that SMPD1 variants with reduced enzymatic activity display the strongest association (<44% activity, odds ratio (OR) = 2.24, P = 1.25 × 10^-15) with PD risk. Moreover, 80.5% of SMPD1 carriers harbored the Asian-specific p.Pro332Arg variant (OR = 2.16; P = 4.47 × 10^-8). Our findings highlight the utility of performing exome sequencing in diverse ancestry groups to identify rare protein-altering variants in genes previously unassociated with disease.

Fig. 1. Exome-wide significant association of GBA1, HLA-DRB1 and SMPD1 in the Discovery exome sequencing study.

a, Quantile–Quantile plot of observed and expected CMH P for autosomal genes with OR > 1. b, Forest plot of SMPD1 association in assessed populations. CMH without continuity correction was used across strata. Data are presented as OR and 95% CI. All P values presented are unadjusted. Exome-wide significance was set at P < 2.5 × 10⁻⁶ (two-tailed), taking into account multiple hypothesis testing correction for an estimated 20,000 protein-coding genes in the human genome.

Fig. 2. ASM activity levels of SMPD1 variants in Discovery and Replication datasets.

a, Activity levels are presented as percentage of wild-type ASM activity; error bars represent standard deviation. Each variant was assayed in technical replicates of 3–6. b, Predicted deleterious variants (75 variants) have significantly lower ASM activity than predicted benign variants (47 variants). Median, 25th–75th quartiles; whiskers of 1.5× IQR shown in boxplot. *** Two-sided t-test P < 0.001. Red dotted line: 100% activity of wild-type variants. Blue dotted line: activity level (43.58%) significantly associated with variant carrier frequency in case and control. Details are provided in Supplementary Table 17. c, SMPD1 wild-type protein localizes to the lysosome. Representative image from three captured fields of view is shown (Supplementary Fig. 11). Yellow signal: SMPD1. Magenta signal: lysosome. Cyan signal: nuclei. Scale bar, 10 µm. IQR, interquartile range; WT wild-type.

Fig. 3. Replication of SMPD1 association in Chinese and European datasets for rare variants with ≤43.58% ASM activity level.

a, Forest plot of SMPD1 in assessed populations. CMH without continuity correction was used across strata. Data are presented as OR and 95% CI. b, Distribution of rare variants with ≤43.58% ASM activity level in case and control from the Discovery cohort.

Extended Data Fig. 1. Quantile-Quantile plot of all autosomal genes from the Discovery dataset.

A stratified CMH test without continuity correction was used to evaluate gene-based burden across the exomes of the study participants from the 5 countries studied (“P_CMH”) and indicated by blue circles. SKAT-O gene-based burden after adjustment for principal components (PC1-PC3), per sample variant counts and average per sample coverage (“P_adj”) indicated by red circles. Details in Supplementary Table 22.

Extended Data Fig. 2. Replication of GBA1 association in Chinese and European datasets based on rare predicted pathogenic variants.

CMH without continuity correction used across strata. Data presented as OR and 95% confidence interval (CI).

Extended Data Fig. 3. Distribution of ASM enzymatic activity levels of rare SMPD1 variants in relation to protein domain localisation and tertiary structure integrity.

(a) Each rare SMPD1 variant was classified by its localization to protein domains across the ASM protein, namely the signalling peptide (12 variants), saponin (12 variants), proline linker (7 variants), metallophosphate (70 variants), and C terminus (19 variants). (b) Each rare SMPD1 variant was classified by its effect on ASM protein tertiary structure integrity as benign (15 variants), deleterious (27 variants) and uncertain (7 variants). Analysis for both panels are based on protein crystallography of ASM tertiary structure (PDB: 5i81). Median, 25th-75th quartiles, whiskers of 1.5× IQR shown in boxplot. Two-sided t-test: P < 0.05 indicated by *, P < 0.01 indicated by **; P < 0.001 indicated by ***.

Extended Data Fig. 4. Forest plot of PD associations with rare SMPD1 variants from Discovery and Replication datasets that (a) affect ASM tertiary structure, (b) localized to the signaling peptide, (c) localized to the saponin domain, (d) localized to the proline-rich linker, (e) localized to the catalytic metallophosphate domain, and (f) localized to the C terminus.

Effect of variants on ASM tertiary structure and localisation based on protein crystallography of ASM tertiary structure (PDB: 5I85). CMH without continuity correction used across strata. Data presented as OR and 95% confidence interval (CI).

Extended Data Fig. 5. Onset ages of Parkinson’s disease cases that carry rare deleterious variants in GBA1 and SMPD1.

(a) Carriers of GBA1 qualifying variants were diagnosed with Parkinson’s disease almost 7 years younger than non-carriers (carriers: median = 55.0 years of age, range = 29-78 years of age; non-carriers: median = 62.0 years of age, range = 22-92 years of age; t-test P = 3.32 × 10^-11). (b) Carriers of SMPD1 qualifying variants had similar onset age as non-carriers (carriers: median = 60.0 years of age, range = 32-86 years of age; non-carriers: median = 62.0 years of age, range = 22-92 years of age; t-test P = 0.426). (c) Onset ages in carriers of both GBA1 and SMPD1 qualifying variants. Median, 25th-75th quartiles, whiskers of 1.5× IQR shown in boxplot. *** indicates two-sided t-test P = 3.32 ×10^-11.

Extended Data Fig. 6. Polygenic risk scores of individuals based on SMPD1 functionally deficient variant carrier status.

Polygenic risk scores of 60 cases and 27 controls which carry SMPD1 functionally deficient variant shown. Median, 25th-75th quartiles, whiskers of 1.5× IQR shown in boxplot.

Extended Data Fig. 7. Forest plots of PD associations with (a) HLA-DRB1 qualifying variants and (b) ATP5B qualifying variants.

CMH without continuity correction used across strata. Data presented as OR and 95% confidence interval (CI).

Extended Data Fig. 8. Statistical power to identify Parkinson’s disease-associated genes at exome-wide significance.

Statistical power to identify PD-associated genes at exome-wide significance (alpha = 2.5 × 10^-6) is dependent on sample size, the combined carrier frequency in the population and effect size (odds ratio) of the variants in each gene. Based on our current Discovery cohort size of 4,298 PD cases and 5,512 controls, we had >90% power to detect genes like GBA1 with large effect sizes (OR > 5) and at combined carrier frequencies ~ 0.5% in the population, and also genes like SMPD1 with smaller effect size (OR ~ 2) but at higher combined carrier frequencies of 1-1.5% in the population (contributed in part by the SMPD1 p.Pro332Arg variant). Genes with similar variant effect sizes as SMPD1 may be missed if they are present at lower combined carrier frequencies of 0.9% or less. Similarly, genes with variant effect sizes comparable with GBA1 will be missed if they are present at combined carrier frequencies of 0.1% or less.