Identification of genetic risk loci and causal insights associated with Parkinson’s disease in African and African admixed populations: a Genome-Wide Association Study

Abstract

Background

Understanding the genetic mechanisms underlying diseases in ancestrally diverse populations is an important step towards development of targeted treatments. African and African admixed populations enable mapping of complex traits, because of their genetic diversity, extensive population substructure, and distinct linkage disequilibrium patterns. We aimed to do a comprehensive genome-wide assessment in African and African admixed individuals to better understand the genetic architecture of Parkinson’s disease in these underserved populations.

Methods

We performed a genome-wide association study in people of African and African admixed ancestry with and without Parkinson’s disease. We characterised ancestry-specific risk, differential haplotype structure and admixture, coding and structural genetic variation, and enzymatic activity.

Findings

We included 197 918 individuals (1488 cases; 196 430 controls) in our genome-wide analysis. We identified a novel common risk factor for PD and age at onset at the GBA1 locus (risk, rs3115534-G; OR=1.58, 95% CI = 1.37 – 1.80, P=2.397E-14; age at onset, BETA =−2.004, SE =0.57, P = 0.0005), that was found to be rare in non-African or non-African admixed populations. Downstream short- and long-read whole genome sequencing analyses did not reveal any coding or structural variant underlying the GWAS signal. However, we identified that this signal is associated with decreased glucocerebrosidase activity levels.

Interpretation

The present study identifies a novel genetic risk factor in GBA1 in people of African ancestry, which could be a major mechanistic basis of PD in this population. This striking result contrasts to previous work in Northern European populations, both in terms of mechanism and attributable risk. This finding highlights the importance of understanding ancestry-specific genetic risk in complex diseases, a particularly crucial point as the field moves toward targeted treatments in PD clinical trials and while recognizing the need for equitable inclusion of ancestrally diverse groups in such trials. Given the distinctive genetics of these underrepresented populations, their inclusion represents a valuable step towards insights into novel genetic determinants underlying PD etiology. This opens new avenues towards RNA-based and other therapeutic strategies aimed at reducing lifetime risk.

Introduction

Genome-wide association studies (GWAS) have been instrumental for identifying common variants associated with complex diseases like Parkinson’s disease (PD), unraveling the genetics and heritability of PD in European populations. The largest published GWAS meta-analysis of PD risk to date was performed on individuals of European ancestry and identified 90 independent genome-wide significant risk signals that explain 16–22% of the heritable risk of PD^1,2. However, very little is known about the genetics of PD in non-European populations. There has been considerable ethnic variability in the distribution of monogenic causes and genetic risk variants documented across populations. For instance, the relatively common LRRK2 p.G2019S mutation remains unreported in some sub-Saharan African populations, despite being most commonly associated with familial and sporadic PD in Zambia and Northern Africa^3–6.

African and African admixed populations offer unique opportunities for studying the genetics of both monogenic and complex diseases because they contain the largest portion of the within-population genetic variability in the world, shorter linkage disequilibrium (LD) blocks, and abundant alleles that are private to these populations⁷. In addition to promoting scientific equity to address health disparities, diverse representation provides a platform for replication studies to explore the strength and relevance of findings reported from other populations. Additionally, studying diverse populations have the potential to facilitate the identification of novel or unique loci and investigate genotype-phenotype correlations that can further expand our understanding of pathological and pathogenetic disease mechanisms in PD.

This study provides the first GWAS-based insights into the genetics of PD in the African and African admixed populations. Here we performed a comprehensive genome-wide assessment of PD risk and age at onset, characterizing ancestry-specific risk, haplotype structure, and genetic admixture. Leveraging this unique population genetic structure, our analyses identified a novel association signal in GBA1, the gene encoding the lysosomal enzyme glucocerebrosidase (GCase).

Methods

Study Design and Participants

An overview of the study design can be found in Figure 1. Three sources of data were included in this study: Individual level data from the International Parkinson’s Disease Genomics Consortium - Africa (IPDGCAN) and the Global Parkinson’s Disease Genetics Program (GP2), and GWAS summary statistics from 23andMe, Inc. PD cases provided from efforts in Africa are predominantly from West Africa, specifically Nigeria, and therefore unlikely to be representative of the entirety of Africa. However, the majority of controls in this study come from global efforts and have higher percentages of admixture. Some of the individuals predicted to be of African descent cannot with certainty be defined as from Nigeria, but nonetheless unmistakably African (Supplementary Figure 3). Additionally, we define African admixed as individuals ancestrally similar to the following 1000 Genomes ancestry labels: African ancestry in Southwest United States of America (abbreviated as ASW in the 1000 Genomes project) and African Caribbean in Barbados (abbreviated as ACB in the 1000 Genomes project).

Figure 1: Study design.

Figure created with BioRender.com. eQTL=expression quantitative trait locus. GWAS=genome-wide association study.

For the IPDGCAN and the GP2 cohorts, the diagnosis of PD was based on fulfillment of the United Kingdom PD Society Brain Bank criteria (excluding the requirement for not more than one affected relative)⁸. The respective ethical committees for medical research approved involvement in genetic studies, and participants gave informed written consent. All participants underwent a neurological examination conducted by a study neurologist to document clinical and neurological status. Controls were generally assessed to detect overall signs of neurological condition and samples presenting any clinical signs of neurodegenerative diseases were excluded from the control series.

For the 23andMe dataset, the diagnosis of PD was self-reported (see Supplementary Materials). Summary statistics for individuals with or without PD were provided through a collaborative agreement with 23andMe, Inc. 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; Supplementary Methods).

Statistical analyses

Genotype data generation, quality control, ancestry predictions, and imputation

The IPDGCAN and GP2 samples were genotyped using two different genotyping platforms (Table 1). The NeuroBooster array (v.1.0, Illumina, San Diego, CA) contains a backbone of 1,914,935 variants densely covering ancestry informative markers, markers for determination of identity by descent, X-chromosome SNPs for sex determination, and contains 96,517 customized variants. Samples collected as part of the GP2 initiative were genotyped on this array. Samples collected as part of the IPDGCAN initiative (Table 1) were genotyped using two different platforms; the Neurochip array, containing a backbone of 306,670 variants and customized content comprising 179,467 variants⁹, and the previously described NeuroBooster array.

Table 1:

Characteristics of cohorts in the study

	African predicted ancestry		African admixed predicted ancestry*
	Nigerian origin (IPDGC cohort)†	African, broad unspecified origin (GP2 dataset)‡	African admixed origin (GP2 dataset)§	African admixed origin (23andMe dataset)
Total participants	589	1722	1334	194 273
Recruited from Nigerian sites	589 (100%)	1330 (77%)	50 (4%)	N/A
Cases	304	711	185	288
Recruited from Nigerian sites	304 (100%)	672 (95%)	16 (9%)	N/A
Female	80 (26%)	206 (29%)	80 (43%)	N/A
Male	224 (74%)	505 (71%)	105 (57%)	N/A
Controls	285	1011	1149	193 985
Recruited from Nigerian sites	285 (100%)	658 (65%)	34 (3%)	N/A
Female	97 (34%)	448 (44%)	714 (62%)	N/A
Male	188 (66%)	563 (56%)	435 (38%)	N/A
Case age at onset, years	58·20 (9·67)	59·31 (11·37)	57·84 (14·69)	N/A
Control age at examination, years	64·4 (7·56)	65·09 (9·55)	66·34 (8·71)	N/A
Array	NeuroChip	NeuroBooster	NeuroBooster	Omni Express & GSA & 550k

Data are n, n (%) or mean (SD). N/A=not available.

^*African admixed defined as individuals ancestrally similar to the following 1000 Genomes project (https://www.internationalgenome.org/) ancestry labels: African ancestry in Southwest United States of America (abbreviated as ASW in the 1000 Genomes project) and African Caribbean in Barbados (abbreviated as ACB in the 1000 Genomes project).

^†See appendix (p 37) for a complete list of Nigerian hospitals and institutions contributing to this cohort.

^‡GP2 cohorts with predicted African ancestry include Baylor College of Medicine (https://www.bcm.edu/), BioFIND (https://biofind.loni.usc.edu/), BLAAC PD (https://www.blaacpd.org/), Coriell (https://www.coriell.org/), Movement Disorders Genotypes and Phenotypes – King’s College London (MDGAP-KINGS; further details at https://gp2.org/the-components-of-gp2s-third-data-release/), PPMI (https://www.ppmi-info.org/), PAGE (https://www.pagestudy.org/), University of Maryland (https://umd.edu/), and IPDGCAF-NG (https://www.ipdgc-africa.com/).

^§GP2 cohorts with predicted African admixed ancestry include Baylor College of Medicine, BioFIND, BLAAC PD, Coriell, MDGAP-KINGS, PPMI, PAGE, University of Maryland, Systemic Synuclein Sampling Study (S4; https://pubmed.ncbi.nlm.nih.gov/28353371/), and IPDGCAF-NG.

Raw genotype data was passed through a custom ancestry prediction and pruning machine learning method as a part of the GenoTools pipeline (https://github.com/GP2code/GenoTools), as described elsewhere¹⁰. All samples underwent similar standardized quality control (QC). For additional information, please see the Supplementary Materials.

Estimation of PD risk, age at onset and admixture

To estimate risk associated with PD, imputed dosages (meaning genotype probabilities for a variant to be A/A, A/B, or B/B from 0 to 2 that account for some uncertainty) were analyzed using a logistic regression model adjusted for sex, age, and the first ten PCs as covariates. The PCs were fit on the set of overlapping SNPs between the datasets and the reference panels before being transformed by UMAP to represent the population substructure (see Supplementary Materials). Age at onset (AAO) was used for cases and age at recruitment was used for controls. In instances where AAO was not available for cases, age at recruitment was used instead (less than 6% of individuals). For individuals who had no age information provided, average age was imputed (less than 5% and 2% of cases and controls, respectively). Summary statistics were generated using PLINK 1.9 and 2.0¹¹, and filtered for inclusion after meeting a minimum imputation quality of 0.30 and MAF > 5 %. To explore the influence of genetic variation on the AAO of PD cases, a linear regression model adjusted for the same covariates was performed. Here, AAO was defined as the self-reported date of first motor symptom. Additionally, we conducted linear regression analyses to explore how potential GWAS signals would correlate with admixture levels. All the analyses were performed on Terra (https://terra.bio/). GWAS was conducted on African and African admixed ancestries independently using PLINK and a Bonferroni threshold of 5E-8 prior to meta-analysis. We utilized fixed-effects meta-analyses as implemented in METAL¹² to leverage summary statistics across all sources. Pairwise LD values were calculated using 1000 Genomes African population data through LD link (https://ldlink.nci.nih.gov/?tab=home).

Haplotype and fine-mapping analyses

Haplotype size was compared using individual level data across African, African admixed, and European PD cases. After standardizing the three datasets with the same genotyped SNPs passing identical QC steps, we determined the size of the haplotype blocks using default parameters in PLINK 1.9. This analysis estimates haplotype blocks by Haploview’s interpretation of the block definition. By default, only pairs of variants within 200 kilobases (kb) of each other were considered. Two variants are considered by this procedure to be in strong LD if the lower bound of the 90% D-prime confidence interval (CI) was >0.70, and the upper bound of the CI was at least 0.98. Fine-mapping analyses were conducted using the R package coloc (https://CRAN.R-project.org/package=coloc; Supplementary Methods).

Procedures

Short and Long read Whole Genome Sequencing

To further dissect the novel identified GWAS signal, we performed whole-genome sequencing (WGS) analyses in 206 individuals (141 cases and 65 controls) of which 39 individuals were GBA1 rs3115534-GG carriers, 69 were rs3115534-GT and 98 were rs3115534-TT carriers. Short-read WGS DNA sequencing was performed by Psomagen (detailed in Supplementary Methods).

Oxford Nanopore Technologies (ONT) long-read whole-genome sequencing data was generated for five GBA1 rs3115534-GG carriers, two heterozygotes and six GBA1 rs3115534-TT carriers. High molecular weight DNA was extracted from either frozen blood samples or cell-lines. Additional details are described in Supplementary Methods.

Glucocerebrosidase activity

Patient-derived lymphoblastoid cell lines (LCLs) were obtained from the Coriell repository (https://www.coriell.org/). LCLs were maintained as directed in suspension with RPMI 1640 (ThermoFisher Scientific, 11875093) containing 2mM Glutamax (ThermoFisher Scientific, 35050061), and 15% FBS (ThermoFisher Scientific, A3160501) at 37°C in 5% CO2. Protein was extracted from LCLs using a citrate-phosphate buffer (0.2 M Na2HPO4, 0.1 M citrate, protease inhibitor, pH 5.8, Millipore Sigma, 11836170001) that was activated with 0.25% Triton X-100. Cells were subjected to a 4-methylumbelliferone (4-MU, Sigma Aldrich, M1381) fluorometric glucocerebrosidase (GCase) activity assay in quadruplicate as previously reported in the literature¹³ with adjusted incubation time of 2.5 hours. A total of 5E6 cells were used per sample with protein concentrations normalized to 0.7 mg/ml via BCA Protein Assay (Thermo Fisher Scientific 23225).

Acknowledgements

We thank all the participants that contributed to this study. GP2 is funded by the Aligning Science Across Parkinson’s (ASAP) initiative and implemented by The Michael J. Fox Foundation for Parkinson’s Research (https://gp2.org). Additional funding was provided by The Michael J. Fox Foundation for Parkinson’s Research through grant MJFF-009421/17483. The funders supported clinical data collection and genotyping data generation.

Results

Here, we performed a GWAS meta-analysis (Figure 2) of the African (Supplementary Figure 4) and African admixed datasets (Supplementary Figure 5), totaling 1,488 cases and 196,430 controls. The demographic and clinical characteristics of the cohorts under study are provided in Table 1. This revealed that a total of 35 SNPs near the GBA1 gene were significantly associated with PD risk with consistent directionality of effect, the two most distant SNPs being 639,773 base pairs apart from each other (Supplementary Table 2). Conditional analyses on the top two SNPs suggested that there is only one causal signal driven by rs3115534 as the leading SNP. Of note, rs3115534-G is much more common in individuals of African or African admixed ancestry relative to other populations; allele frequency = 0.16 according to gnomAD (accessed February 2023)¹⁴ and allele frequency = 0.21 according to the African 1000 Genomes panel ¹⁵. The African and African admixed datasets used in this study yielded similar frequencies (African dataset; cohort MAF = 0.25, affected MAF = 0.33, unaffected MAF = 0.19), (African admixed datasets; cohort MAF = 0.14, affected MAF = 0.22, unaffected MAF = 0.13). Within our research cohorts, we found that rs3115534-G was more frequent in Nigerian populations (Supplementary Table 3). Linear regression analyses showed that the GBA1 rs3115534 variant was positively associated with a higher percentage of African ancestry (BETA = −0.001, SE= 0.0005, P= 0.011).

Figure 2: African and African admixed GWAS meta-analysis assessing Parkinson’s disease risk.

Manhattan plot showing the significance of the association as −log₁₀.(p value) against chromosomes at a genomic scale (Bonferroni correction highlighted in grey at 5 × 10 ³). GWAS=genome-wide association study.

We tested whether the effect of the risk allele was additive by calculating the frequency of homozygotes for the risk allele and heterozygotes in cases versus controls (Supplementary Table 8). As a follow-up analysis, we assessed whether this GBA1 variant is associated with AAO. Linear regression analyses in 711 African ancestry cases and 185 African admixed ancestry cases showed that GBA1 rs3115534-G is also an AAO disease modifier (African ancestry: BETA =−2.004, SE = 0.57, P = 0.0005; African-admixed: BETA = −4.15, SE =0.58, P =0.015; Meta-analysis: BETA =−3.06, SE =0.40, P = 0.008) resulting in onset of PD three years earlier per risk allele (Supplementary Figure 7).

In an attempt to further dissect the novel signal identified in the GBA1 locus, we next compared effect estimates and directionality of effect leveraging summary statistics from the largest PD GWAS meta-analysis of PD in Europeans¹, Latin American¹⁶, and East Asian populations¹⁷. The rs3115534-G allele is extremely rare in Europeans (allele frequency = 0.0015), East Asians (allele frequency = 0.0005), South Asians (allele frequency = 0.0017), and Ashkenazi Jewish populations (allele frequency = 0.0009) according to gnomAD.

The GBA1 locus in African and African admixed populations differs substantially from Europeans (Figure 3; Supplementary Figure 8), whose association with disease risk is driven by two independent signals, including rs35749011 (GBA1-E326K) and rs76763715 (GBA1-N370S). These variants are very rare in individuals of African and African admixed ancestry (Supplementary Figure 14B). Similarly, the GBA1 locus considerably differs from the East Asian population, for which the rs3115534 variant was also not imputed in the largest East Asian GWAS meta-analysis¹⁷ (Supplementary Figure 14C). These differences are less noticeable when assessing the Amerindian/Latin American and indigenous populations, which harbor higher levels of African admixture (Supplementary Figure 14D); Loesch et al. GWAS¹⁶; rs3115534-G; OR = 1.13, 95% CI =0.41–1.86, P= 0.72; Amerindian/Latin American and indigenous 23andMe GWAS; rs3115534-G; OR = 1.56, 95% CI = 1.55–1.88, P= 0.01).

Figure 3: Miami plot comparing European versus African and African admixed GWAS meta-analyses of Parkinson’s disease risk.

Randomly sampled 1200 cases and 2445 controls were included in each of the GWAS meta-analyses. The grey horizontal lines indicate the significance threshold of 5 × 10 ³. GWAS=genome-wide association study.

Interestingly, the larger sub-African population haplotypes spanning the rs3115534 variant were found in the Esan and the Yoruba in Ibadan (Nigerian) populations according to 1000 Genomes (Supplementary Figure 9), suggesting that this haplotype might have originated in these populations, given that founder effects result in decreased genetic admixture and therefore larger haplotype block sizes. Fine-mapping of this locus showed the lead SNP had a PP of 71.4% (rs3115534; Supplementary Table 4).

In an effort to identify a functional coding variant undetectable through genotyping or imputation that could explain the novel GWAS signal, we conducted short-read WGS analyses on a total of 206 individuals (141 cases and 65 controls) of which 39 individuals were GBA1 rs3115534-GG carriers, 69 were rs3115534-GT and 98 were rs3115534-TT carriers. A 96.6 % correlation was observed between short-read WGS and imputed genotyped data for rs3115534, validating the high quality of our imputed data. No differences in coding variation were observed between carriers and non-carriers of the GWAS signal (Table 2).

Table 2:

Functional coding variants identified by short-read whole genome sequencing in carriers of the novel GBA1 rs3115534 variant

Variant	Base change	Functional consequence	Genetic variant	Cases with functional variant (n)	Controls with functional variant (n)	rs3115534-GG carriers (n)	rs3115534-GT carriers (n)	rs3115534-TT carriers (n)
chr1:155236249:A:C	A→C	Non-synonymous SNV	Ile320Ser	1	0	0	1	0
rs149487315	C→T	Non-synonymous SNV	Met313Ile	1	0	0	0	1
rs143222798	C→T	Synonymous SNV	Gly277Gly	6	3	0	6	3
rs61748906	A→G	Non-synonymous SNV	Trp136Arg	1	0	1	0	0
rs368786234	G→T	Non-synonymous SNV	Ser77Arg	1	0	0	1	0
rs761621516	GTA→deleted	Non-frameshift deletion	Trp75del (222_224del)	1	0	0	1	0
rs150466109	T→C	Non-synonymous SNV	Lys13Arg	12	8	0	10	10

Analyses were done in 141 cases and 65 controls. All variants were on chromosome 1, were exonic, and were heterozygous. SNV=single nucleotide variant.

We leveraged existing whole blood expression quantitative trait locus (eQTL) summary statistics from Mak et al., 2021 based on RNA sequencing from 2,733 samples of predominantly African American and Indigenous American ancestries¹⁸. Of note, we identified a strong eQTL signal at rs3115534, located 8,821 bp from the canonical transcription start site (Figure 4; MAF = 0.15; P= 9.99E-25, BETA = 0.238, SE = 0.022). The rs3115534-G risk allele was found to be associated with increased GBA1 expression levels. We questioned whether this observation could be explained by the existence of multi-mapping reads between GBA1 and its pseudogene, GBAP1, which are often discarded in standard processing and do not contribute to gene-level quantification of expression in many publicly available datasets like GTEx (https://gtexportal.org/). Indeed, transcript diversity is a common and known biological phenomena¹⁹ that could explain the fact that rs3115534-G may increase the expression of a non-functional transcript that in turn would decrease the levels of the transcript responsible for optimal production of the protein isoform with GCase activity.

Figure 4. LocusZoom plot displaying African and African admixed Parkinson’s disease GWAS meta-analysis summary statistics versus African American eQTL summary statistics from blood.

eQTL=expression quantitative trait locus. GWAS=genome-wide association study.

Our data suggests a decreasing trend in GCase activity estimates when comparing rs3115534-GG homozygous risk allele (762.50 ± 273.50 U) versus rs3115534-GT heterozygous carriers (2743.76 ± 1960.83 U); (Welch Two Sample t-test - GG versus GT; t = −4.3138, df = 21.583, p-value = 0.00029) and rs3115534-TT homozygous non-risk allele carriers (1879.94 ± 1010.84 U) versus rs3115534-GG homozygous risk allele carrier; (Welch Two Sample t-test - GG versus TT; t = −4.7564, df = 18.363, p-value = 0.00014)(Supplementary Figures 11 and 12).

The largest PD-GWAS and multi-ancestry GWAS meta-analyses to date identified a total of 104 independent significant PD risk variants ^1,17,20. Out of the 104 variants, 91 variants passed QC, imputation filters, and were present in the African and African admixed GWAS meta-analysis (Supplementary Figure 15; Table 2). Out of the 91 variants, 16 variants were nominally significant (p < 0.05; Supplementary Table 5) in the African and African admixed meta-GWAS reported here.

Discussion

Although there have been a number of published studies exploring PD genetics in the African and African admixed populations^{21,22,23,24,25,26,27,28,6,29,30,31,32,33,34}, in the present study, we have gathered the largest collection of PD patients and controls from African and African admixed ancestry populations to comprehensively assess the genetic architecture of PD on a genome-wide scale. Here, we identified a novel African-specific GWAS signal on the GBA1 locus, significantly associated with PD risk and AAO, to be the most important risk factor for PD in these African and African admixed populations. Remarkably, almost a four times larger sample size in cases was required to nominate GBA1 as one of the major PD risk factors in the European ancestry population through GWAS³⁵, showing the power and benefit of using diverse ancestry data.

GBA1 is a classic pleomorphic locus, showing coding, structural, and non-coding variants that exert different degrees of risk³⁶. Despite the large effect size driven by this signal, our study did not identify an association with any previously reported or new GBA1 coding or structural aberration that could explain this signal^37,38,39,40.

Strikingly, by leveraging existing eQTL data predominantly of African American ancestry, we found the rs3115534-G risk allele to be associated with increased GBA1 expression levels in whole blood, but paradoxically linked with a trend towards decreased GCase activity, which may be due to challenges with RNA-seq in this locus. Future large scale single cell expression studies should investigate in which brain cell types these expression differences are most prominent. This potential novel mechanism opens new avenues towards efficient RNA-based therapeutic strategies, such as antisense oligonucleotides or short interfering RNAs aimed at reducing lifetime risk.

Interestingly, given the high population frequency of the identified signal and the phenotypic characteristics of the homozygous Africans and African admixed carriers, our study does not support the notion that this variant causes Gaucher disease. Furthermore, the rs3115534 variant has been found to be extremely rare in non-African/African admixed populations. These findings suggest an African founder effect, and reinforce that the genetic architecture of this locus and its influence in risk and onset is different across populations. Interestingly, rs3115534 was also found to be associated with PD AAO in our study.

Here, we produce crucial insights into targeted construction of African ancestral haplotypes and potential novel pathogenic mechanisms underlying PD etiology. The utility of genetically characterizing populations of African and African admixed ancestry is unquestionable. This study demonstrates the importance of haplotype substructure discoveries for future fine-mapping efforts, showing how leveraging unique populations can benefit our understanding of complex diseases.

Overall, addressing the genetic complexity underlying these underrepresented populations, our study represents a valuable resource for identifying and tracking GBA1 carriers that may prove relevant for enrollment in target-specific PD clinical trials. We envisage that these data generated under the Global Parkinson’s Genetics Program initiative will be key to shed light on the molecular mechanisms involved in the disease process and might pave the way for future clinical trials and therapeutic interventions.

Limitations

Although we have made progress in assessing genetic risk factors for PD in an African-specific manner, there are a number of limitations to our study. Unraveling additional susceptibility genetic risk and phenotypic relationships would have been possible if a larger cohort had been analyzed. Considering our limited sample size, we lacked statistical power to detect common genetic variants of smaller effect sizes (Supplemental Figure 13). Additionally, an important proportion of the genetic risk contributing to the missing heritability of PD in the African and African admixed populations might result from rare alleles and structural variants that have not been assessed in the present study. Due to the lack of well-powered and African or African admixed RNA sequencing datasets, the added complexity of multi-mapping reads between GBA and GBAP1 and the limited number of LCLs to explore GCase activity in a large scale manner, we assume the limitation that this potential novel functional mechanism merits further study. We are aware that although this represents the first PD GWAS in the African and African admixed populations, two-thirds of the cases are of Nigerian descent, therefore likely unrepresentative of the substantial genetic diversity across the continent.

Data Sharing

All GP2 data is hosted in collaboration with the Accelerating Medicines Partnership in Parkinson’s disease, and is available via application on the website (https://amp-pd.org/register-for-amp-pd). The GWAS summary statistics from this study, excluding 23andMe, are available as of GP2’s release 5. 23andMe summary statistics are available via application on the website (https://research.23andme.com/dataset-access/). Genotyping imputation, quality control, ancestry prediction, and processing was performed using GenoTools v1.0, publicly available on GitHub (https://github.com/GP2code/GenoTools). All scripts for analyses are publicly available on GitHub [https://github.com/GP2code/GP2-AFR-AAC-metaGWAS; 10.5281/zenodo.7888141].

Ethics Statement

All cohorts recruited to the GP2 initiative undergo a thorough review of the consent forms in the Operations and Compliance working group, ensuring that 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 within local law constraints. All GP2 data is hosted in collaboration with the Accelerating Medicines Partnership in Parkinson’s disease, and is available via application on the website (https://amp-pd.org/register-for-amp-pd).

Summary statistics for individuals with or without PD were provided through a collaborative agreement with 23andMe, Inc. 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). 23andMe summary statistics are available via application on the website (https://research.23andme.com/dataset-access/).

Research in Context.

Evidence Before this Study

Our current understanding of Parkinson’s disease (PD) is disproportionately based on studying populations of European ancestry, leading to a significant gap in our knowledge about the genetics, clinical characteristics, and pathophysiology in underrepresented populations. This is particularly notable in individuals of African and African admixed ancestries. Over the last two decades, we have witnessed a revolution in the research area of complex genetic diseases. In the PD field, large-scale genome-wide association studies in the European, Asian, and Latin American populations have identified multiple risk loci associated with disease. These include 78 loci and 90 independent signals associated with PD risk in the European population, nine replicated loci and two novel ancestry-specific signals in the Asian population, and a total of 11 novel loci recently nominated through multi-ancestry GWAS efforts. Nevertheless, the African and African admixed populations remain completely unexplored in the context of PD genetics.

Added Value of this Study

To address the lack of diversity in our research field, this study aimed to conduct the first genome-wide assessment of PD genetics in the African and African admixed populations. Here, we identified a genetic risk factor linked to PD etiology, dissected ancestry-specific differences in risk and age at onset, characterized known genetic risk factors, and highlighted the utility of the African and African admixed risk haplotype substructure for future fine-mapping efforts.

Implications of all the Available Evidence

We nominate a novel signal impacting GBA1 as the major genetic risk factor for PD in the African and African admixed populations. The present study could inform future GBA1 clinical trials, improving patient stratification. In this regard, genetic testing can help to design trials likely to provide meaningful and actionable answers. We identified a novel disease mechanism via expression changes consistent with decreased GBA1 activity levels. This novel mechanism may hold promise for future efficient RNA-based therapeutic strategies such as antisense oligonucleotid es or short interfering RNAs aimed at preventing and decreasing disease risk. This work represents a valuable resource in an underserved population, supporting pioneering research within the Global Parkinson’s Genetics Program (GP2) and beyond. Deciphering causal and genetic risk factors in all these ancestries will help determine whether interventions, potential targets for disease modifying treatment, and prevention strategies that are being studied in the European populations are relevant to the African and African admixed populations.