Discovery and functional prioritization of Parkinson's disease candidate genes from large-scale whole exome sequencing

Abstract

Background: Whole-exome sequencing (WES) has been successful in identifying genes that cause familial Parkinson's disease (PD). However, until now this approach has not been deployed to study large cohorts of unrelated participants. To discover rare PD susceptibility variants, we performed WES in 1148 unrelated cases and 503 control participants. Candidate genes were subsequently validated for functions relevant to PD based on parallel RNA-interference (RNAi) screens in human cell culture and Drosophila and C. elegans models.

Results: Assuming autosomal recessive inheritance, we identify 27 genes that have homozygous or compound heterozygous loss-of-function variants in PD cases. Definitive replication and confirmation of these findings were hindered by potential heterogeneity and by the rarity of the implicated alleles. We therefore looked for potential genetic interactions with established PD mechanisms. Following RNAi-mediated knockdown, 15 of the genes modulated mitochondrial dynamics in human neuronal cultures and four candidates enhanced α-synuclein-induced neurodegeneration in Drosophila. Based on complementary analyses in independent human datasets, five functionally validated genes-GPATCH2L, UHRF1BP1L, PTPRH, ARSB, and VPS13C-also showed evidence consistent with genetic replication.

Conclusions: By integrating human genetic and functional evidence, we identify several PD susceptibility gene candidates for further investigation. Our approach highlights a powerful experimental strategy with broad applicability for future studies of disorders with complex genetic etiologies.

Keywords: Animal model; Functional screening; Genomics; Loss-of-function; Mitochondria; Parkin; Parkinson’s disease; Rare variants; Whole-exome sequencing; α-synuclein.

Fig. 1

Flowchart explaining multiple filtering steps to select LoF variants with assumed recessive inheritance pattern. Functional annotation was performed with transcripts of RefSeq and UCSC databases. MAF annotations were based on 1000 Genomes project, Exome variant Server, and the ExAC database. Seventeen genes harbored homozygous variants causing stopgain or loss and one gene contained a homozygous splicing variant. For the putative compound heterozygous genes, six genes were selected based on the presence of two LoF variants, and three genes were based on the presence of one LoF variant and one missense variant (predicted to belong to the 1% most harmful variants of the genome)

Fig. 2

High-content assay for mitochondrial morphology. Effect of DNM1L shRNA (a, b) and UHRF1BP1L shRNA (c, d). BE(2)M17 cells stained with Hoechst (blue; nuclei), MitoTracker CMXros, and MitoTracker Deepred (yellow; mitochondria). a Cells infected with shRNA encoding a scrambled sequence (SCR, left panel) and decrease in mitochondrial axial length ratio and roundness for DNM1L (positive control, right panel). b The graph displays normalized mitochondrial roundness. c Cells infected with shRNA encoding a SCR sequence (left panel) and decrease in number of mitochondria per cell, mitochondrial axial length ratio, and roundness for UHRF1BP1L (right panel). d The graph displays normalized mitochondrial roundness. Data are median values ± median absolute deviation (MAD) of N = 6 measurements. *p < 0.05 and **p < 0.01, Mann–Whitney U test (see “Methods”). All values were normalized to the negative control (infected with SCR shRNA) and all shRNA clones that meet the cutoff criteria are shown (b, d)

Fig. 3

High content assay for Parkin translocation. Effect of PINK1 shRNA (a, b) and GPATCH2L shRNA (c, d). a, c Cells are labeled for nuclei (blue; Hoechst), Parkin-GFP (green), mitochondria (red, Mitotracker Deepred). Untreated cells infected with shRNA encoding a scrambled sequence show absence of puncta (left panel). Cells infected with a scrambled sequence but treated with CCCP show a significant increase in puncta formation (middle panel). Infection of cells with shRNA targeting PINK1 or GPATCH2L prevents the accumulation of Parkin on mitochondrial (right panel). b, d The graph displays the normalized ratio of cells positive for translocation and cells negative for parkin translocation. All values were normalized to the negative control (CCCP treated infected with shRNA encoding a scrambled sequence). Data are median values ± median absolute deviation (MAD) of N = 6 measurements. *p < 0.05, **p < 0.01, and ***p < 0.001, Mann–Whitney U test (see “Methods”). All shRNA clones that meet the cutoff criteria (see “Methods”) are shown

Fig. 4

α-synuclein-induced retinal degeneration and screening assays in Drosophila transgenic animals. Tangential sections through the fly retina stained with hematoxylin and eosin reveal the ordered ommatidial array in control animals (a Rh1-GAL4 / +). Each ommatidia consists of a cluster of eight photoreceptive neurons (seven visible at the level examined). The photoreceptors each contain a single rhabdomere, the specialized organelle subserving phototransduction, giving the ommatidia cluster its characteristic appearance (arrowhead). Expression of α-synuclein in adult photoreceptors (b, c Rh1-GAL4 / +; UAS-α-synuclein / +) causes age-dependent, progressive retinal degeneration. Compared to one-day-old Rh1 > α-synuclein flies (b), histologic sections in 30-day-old animals (c) demonstrate rhabdomere/cell loss and substantial vacuolar changes (asterisk). The pseudopupil preparation allows visualization of rhabdomeres (arrowhead) in intact, unfixed intact fly heads, permitting medium-throughput screening for progression of α-synuclein-induced retinal pathology. Compared to controls (d Rh1-GAL4 / +), in 30-day-old α-synuclein transgenic animals (e Rh1-GAL4 / +; UAS-α-synuclein / +) rhabodomeres frequently appear indistinct (arrowhead) and vacuolar changes disrupt light refraction (asterisk). Representative control histology (a) and pseudopupil images (d) are shown for 15-day-old animals, the timepoint used for screening, in order to facilitate comparison with Fig. 5. Scale bar: 20 μm

Fig. 5

PD gene candidates harboring LoF variants enhance α-synuclein toxicity in Drosophila. Conserved fly orthologs of human genes discovered from WES analysis were targeted with RNAi (IR) and screened for enhancement of α-synuclein pathology using the pseudopupil assay (a top row). For each line evaluated, the severity of retinal degeneration was scored based on penetrance of the α-synuclein pseudopupil phenotype and enhancers required consistent results for at least two independent RNAi lines (see Additional file 1: Table S8). Representative results from the primary screen are shown for controls (Rh1-GAL4 / +; UAS-α-synuclein / +) and one IR line each for the implicated enhancers [Human Gene-Fly Ortholog (experimental genotype shown)]: ARSB-CG32191 (Rh1-GAL4 / +; UAS-α-synuclein / UAS-CG32191.IR.v14294), TMEM134-CG12025 (Rh1-GAL4 / UAS-CG12025.IR.v104336; UAS-α-synuclein / +), PTPRH-Ptp10D (Rh1-GAL4 / UAS-Ptp10D.IR.v1102; UAS-α-synuclein / +), and VPS13-Vps13 (Rh1-GAL4 / UAS-Vps13.IR.HMS02460; UAS-α-synuclein / +). At the 15-day-old time point, Rh1 > α-synuclein causes a weakly-penetrant pseuodopupil phenotype and mild histopathologic changes which are amenable to modifier screening (compare with Fig. 4, panels c and e). Enhancers identified in the primary screen were confirmed based on retinal histology (a middle row) and demonstrated increased tissue destruction and disorganization. Activation of RNAi was not associated with any significant retinal degeneration in the absence of α-synuclein co-expression (a bottom row, Rh1-GAL4 / IR transgene). Scale bars: 20 μm. b Enhancement of α-synuclein-induced retinal degeneration was quantified based on the extent of vacuolar changes (area occupied by vacuoles / total retinal area). For quantification, three animals were examined per genotype. For PTPRH, additional confirmation was obtained by evaluating flies doubly heterozygous for strong alleles of the paralogs Ptp10D and Ptp4e (see also Additional file 2: Figure S5). Statistical comparisons were made using unpaired t-tests. Error bars are based on Standard Error of the Mean. *p < 0.05; **p < 0.01