Coding Variation in GBA Explains the Majority of the SYT11-GBA Parkinson’s Disease GWAS Locus

A landmark discovery in the genetics of Parkinson’s disease (PD) was the determination that mutations in the gene encoding glucocerebrosidase (GBA) convey risk for PD and dementia with Lewy bodies.^1,2 Mutations in GBA are also causing the autosomal-recessive disorder Gaucher’s disease (GD). GD is a lysosomal storage disorder that presents with a variety of features, including parkinsonism. Although it has been clear that some GD-causing GBA mutations such as p.N409S are risk factors for PD, there has been some controversy regarding GBA mutations that do not cause GD and the risk for PD. Resolving this issue and the effect of such mutations on, for example, glucocerebrosidase activity will have important implications for our understanding of the pathobiology of PD.

We present here a large genetic assessment of GBA using NeuroX genotyping of 6249 PD cases and 6032 controls.³ In total, we detected 11 GBA coding variants. No large deletions or duplications at the GBA locus were identified. As expected p.N409S and p.E365K were enriched in cases compared with controls (Table 1). Unfortunately, another relatively common disease associated, L483P, could not be genotyped.¹ Interestingly, variant p.T408M, reported as uncertain pathological significance, was also significantly overrepresented in cases. In total, the majority of identified GBA variants were more common in cases compared with controls, the exception being p.K13R, which is also reported as benign (Table 1). Although some individuals were carrying multiple GBA mutations, no high linkage disequilibrium patterns were identified between coding variants (Supplementary Results).

Table 1.

GBA variant frequencies in cases and controls

Amino acid	GRCh38 (bp)	rs Number	Clinvar	HGMD pathogenicity	Cases				Controls				ExAC (EUR)
					M/M	M/W	W/W	MAF	M/M	M/W	W/W	MAF	M/M	MAF	OR	P
K13R (K-27R)	155240707	rs150466109	Benign	DM?	0	0	6249	0	0	2	6030	0.00017	0	0.00021	NA	0.999
C62Y (C23Y)	155240008	rs145888253	na	na	0	1	6248	0.00008	0	0	6032	0	0	0.00001	NA	1
E150K (E111K)	155239622	none	na	DM?	0	1	6248	0.00008	0	0	6031	0	0	0	NA	0.9993
F255Y (F216Y)	155237576	rs74500255	Likely	Pathogenic	0	2	6247	0.00016	0	1	6031	0.00008	0	0	NA	0.5827
F298L (F259L)	155237446	none	na	pathogenic	0	1	6248	0.00008	0	0	5990	0	0	0	NA	0.9993
A348V (A309V)	155236426	rs78396650	Pathogenic	Pathogenic	0	5	6244	0.00040	0	0	6032	0	0	0	NA	0.9984
E365K (E326K)	155236376	rs2230288	Conflicting interpretations	DM?	4	256	5989	0.02158	0	121	5911	0.01013	7	0.01196	2.037	1.64E-09
T408M (T369M)	155236246	rs75548401	Conflicting interpretations	DM?	2	169	6078	0.01404	0	107	5925	0.00895	5	0.00976	1.53	0.001188
N409S (N370S)	155235843	rs76763715	Likely pathogenic	Pathogenic	0	87	6161	0.00701	0	41	5991	0.00341	1	0.00363	1.882	0.002014
D419N/Y	155235814	none	na	Pathogenic	0	2	6247	0.00016	0	0	6032	0	0	0	NA	0.9991
(D380N/Y) E427K (E388K)	155235790	rs149171124	Uncertain significance	DM?	0	10	6239	0.00080	0	5	6027	0.00041	0	0.00031	NA	0.2157
na	155390201	rs71628662	na	na	1	187	6059	0.01513	0	88	5935	0.007305	na	na	2.07	1.324E-07
na	154925709	rs114138760	na	na	1	166	6082	0.01344	0	107	5925	0.008869	na	na	1.52	0.0001853

Indicated are uncorrected P values (Bonferroni correction for 0.05 lies at > 0.0045) and odds ratios. Tests were carried out in plink using logistic regression with covariates age, sex, first 20 principal components based on genome-wide genotype data, and sample provenance.

Na, not applicable; bp, base pair; W, wild type; M, mutant; MAF, minor allele frequency; EUR, European; ExAC, Exorne Aggregation Consortium (http://exac.broadinstitute.org/); OR, odds ratio; HGMD, Human Gene Mutation Database.

In previous PD genome-wide association studies (GWAS), a locus approximately 150 kb from GBA was implicated as a risk factor for PD (nominated as SYT11-GBA).⁴ Analysis of the non-coding variant rs71628662, previously implicated, revealed a strong association for disease (Table 1). Previous analyses suggested that this was independent of known GD-causing mutations, and the mechanism of action at this locus was therefore unclear. Conditioned analysis of rs71628662 on each of the identified coding variants revealed that the association at p. E365K completely accounted for the previously identified GWAS signal (conditional p.E365K rs71628662, P = 0.6234). Within the current data set, the linkage disequilibrium between p.E365K and rs7168662 is very high (D′ = 0.987, r² = 0.84).

Previously, another noncoding variant (rs114138760) was also identified as an independent signal.⁵ In the NeuroX data set, this variant also showed association with disease (P = 0.0001853) and remained significant following adjustment for the primary GWAS variant (P = 0.0001511). In addition, this signal also remained significant following adjustment for p.E365K, p.N409S, and p.T408M (P > 0.007146), and of the 15 rs114138760 risk variant carriers, none carry p.L483P.

In summary, this work affirms that functional alleles within GBA (p.E365K and p.T408M) that are insufficient to cause GD variants are robustly associated with PD. Further, we have shown that a previously reported GWAS signal at this locus can be largely explained by protein coding variability within GBA. These data are consistent with p.E365K being the functional effector allele underlying the original GWA locus (GBA-SYT11 locus), as suggested previously.⁶ The results from this current study suggest that efforts directed at replacing or augmenting glucocerebrosidase activity already under way for PD may have wider-ranging applicability than just for individuals with GD-associated mutations in glucocerebrosidase.