African ancestry neurodegeneration risk variant disrupts an intronic branchpoint in GBA1

Abstract

Recently, an African ancestry-specific Parkinson disease (PD) risk signal was identified at the gene encoding glucocerebrosidase (GBA1). This variant ( rs3115534 -G) is carried by ~50% of West African PD cases and imparts a dose-dependent increase in risk for disease. The risk variant has varied frequencies across African ancestry groups but is almost absent in European and Asian ancestry populations. GBA1 is a gene of high clinical and therapeutic interest. Damaging biallelic protein-coding variants cause Gaucher disease and monoallelic variants confer risk for PD and dementia with Lewy bodies, likely by reducing the function of glucocerebrosidase. Interestingly, the African ancestry-specific GBA1 risk variant is a noncoding variant, suggesting a different mechanism of action. Using full-length RNA transcript sequencing, we identified partial intron 8 expression in risk variant carriers (G) but not in nonvariant carriers (T). Antibodies targeting the N terminus of glucocerebrosidase showed that this intron-retained isoform is likely not protein coding and subsequent proteomics did not identify a shorter protein isoform, suggesting that the disease mechanism is RNA based. Clustered regularly interspaced short palindromic repeats editing of the reported index variant ( rs3115534 ) revealed that this is the sequence alteration responsible for driving the production of these transcripts containing intron 8. Follow-up analysis of this variant showed that it is in a key intronic branchpoint sequence and, therefore, has important implications in splicing and disease. In addition, when measuring glucocerebrosidase activity, we identified a dose-dependent reduction in risk variant carriers. Overall, we report the functional effect of a GBA1 noncoding risk variant, which acts by interfering with the splicing of functional GBA1 transcripts, resulting in reduced protein levels and reduced glucocerebrosidase activity. This understanding reveals a potential therapeutic target in an underserved and underrepresented population.

Fig. 1. Overview of the African ancestry PD GBA1 GWAS locus.

a, LocusZoom plot showing rs3115534 as index variant (purple diamond) and located in intron 8 of GBA1. b, rs3115534 as eQTL for GBA1 RNA expression from Kachuri et al. showing increased GBA1 expression with G risk genotypes. Genotype counts: GG, n = 22; GT, n = 185; TT, n = 537. c, rs3115534 as pQTL for GBA1 protein expression from Surapaneni et al. showing decreased protein levels with G risk genotypes. Genotype counts: GG, n = 16; GT, n = 134; TT, n = 317. For all box plots, the center line represents the median, edges of the box represent the first and third quartiles and ends of bars represent the maximum and minimum (not including outliers).

Fig. 2. GBA1 intron 8 expression is correlated with rs3115534 genotype.

a, ONT long-read RNAseq of eight LCLs shows a consistent pattern where the rs3115534-G risk allele is associated with intron 8 expression and absent in homozygous T (nonrisk allele) individuals generated using IGV. b,c, Quantification of intron 8 expression is significantly associated with the G allele in both the 40-bp region before exon 9 (b) and the full intron 8 (c) (linear regression, P < 0.05). d,e, No significant differences were identified in the two neighboring exons 8 (d) and 9 (e). Coverage for all panels was normalized by dividing the mean depth by the total number of mapped reads per million (Methods). In b–e, a linear regression was run with GG + GT in one group versus TT. For all panels, genotype counts are as follows: GG, n = 1; GT, n = 4; TT, n = 3. Error bars represent the s.d. for all panels with the center at the mean. The unadjusted r² is displayed.

Fig. 3. Increased intron 8 expression across datasets in G allele carriers.

a, Intron 8 coverage from human frontal cortex sequenced with ONT (n = 8). Intron 8 expression is only present in G allele carriers but does not reach statistical significance (P = 0.281) likely because of a smaller sample size. b, Intron 8 coverage from LCLs (n = 18) sequenced with Illumina. Expression is significantly associated with the G allele (P = 3.88 × 10⁻⁵). c, Intron 8 expression from human frontal cortex sequenced with Illumina (n = 92). Expression is significantly associated with the G allele (P = 2.76 × 10⁻¹⁵). d, Intron 8 coverage from LCLs in the 1000 Genomes Project cohort (n = 88). Expression is significantly associated with the G allele (P = 4.08 × 10⁻³¹). e, Intron 8 coverage from blood in the AMP PD cohort (n = 148). Expression is significantly associated with the G allele (P = 8.05 × 10⁻⁷). f, CRISPR editing of LCLs showed that the rs3115534-G risk allele is significantly associated with intron 8 expression. Here, coverage is shown for the full allelic series of ND22789, a TT line originally CRISPR edited through to a GG line. Coverage for all panels normalized by dividing the mean regional depth by the total number of mapped reads per million (Methods). In a,b,e, a linear regression was run with GG + GT in one group versus TT. In c,d, a linear regression was run with GG, GT and TT in separate groups. Error bars represent the s.d. for all panels with the center at the mean. The unadjusted r² is displayed.

Fig. 4. The GBA1 intronic rs3115534 variant acts as a splicing branchpoint.

The causal variant rs3115534 (highlighted in gray on the top and in red dashed box on the bottom) in intron 8 is located in the key splicing branchpoint according to Branchpointer. When rs3115534 is in a nonrisk state (T), on the antisense strand, the A allele functions as a branch site for the spliceosome, whereas, in the risk state (G), on the antisense strand, the C allele disrupts this branch site resulting in abnormal splicing.

Fig. 5. Measuring GCase activity across GBA1 genotypes.

a, GCase activity was measured across rs3115534GBA1 genotypes showing a significant dose G-allele-dependent reduction across genotypes (P = 0.029). A linear regression was run with GG, GT and TT in separate groups. The unadjusted r² is displayed. b, When measuring GCase activity across multiple heterozygous GBA1 genotypes, it appears that rs3115534-GT and rs3115534-GG reduce GCase activity to similar levels to heterozygous PD risk variants (p.E365K and p.T408M) but remain higher than heterozygous GBA1 Gaucher disease-causing variants (GBA1_mild, such as p.N409S, and GBA1_severe, such as p.L483P). Note that a was extracted from b and the same data are included in both. Group counts are as follows: non-PD, n = 97; LRRK2_G2019S, n = 95; PD_idiopathic, n = 122; rs3115534_het, n = 99; rs3115534_hom, n = 9; PD_risk, n = 99; GBA1_mild, n = 90; GBA1_severe, n = 99. Error bars represent the s.d. for all panels.

Fig. 6. Suggested variant-to-function hypothesis of rs3115534.

rs3115534-T is transcribed to pre-mRNA as A, a highly conserved branchpoint nucleotide. However, rs3115534-G, which confers elevated PD risk, is instead transcribed to C, causing the observed branchpoint disruption (rs3115534, denoted as red star). This single-nucleotide change in intron 8 impacts splicing by disrupting the normal binding from the adenosine branchpoint nucleotide and the 5′ splice site (GU). Subsequently, an alternative branchpoint is used, uncovering an alternative 3′ splice site upstream of the normal splice site immediately proximal to exon 9 and resulting in partial and complete intron retention and fewer functional GBA1 mRNA transcripts. Downstream, abnormal splicing of GBA1 leads to reduced GCase protein and subsequent lower GCase activity, which is a known pathomechanism of PD and DLB. ss, splice site. Generated with BioRender.com.