Lack of evidence for genetic association of saposins A, B, C and D with Parkinson’s disease

We read with much interest the paper by Oji et al. (2020), which suggested an association between mutations in the saposin D region of PSAP and Parkinson’s disease. In their paper, the authors showed that variants in saposin D only partially segregate with Parkinson’s disease in three families of Japanese origin. For example, in Family 1, at least three carriers of the PSAP p.Q453P variant did not have Parkinson’s disease, including one of the parents who was an obligate carrier. The authors also report an association of six common intronic variants within the saposin D region of the gene when comparing Japanese Parkinson’s disease patients and East Asian control subjects. Among these six variants, rs4747203 and rs885828 remained significant also when comparing the combined Japanese-Taiwanese Parkinson’s disease cohorts with East Asian controls. These two variants are in perfect linkage disequilibrium (LD, r² = 1) in East Asians and represent the same allele. The authors further demonstrated, using different models, that PSAP mutations may affect prosaposin localization in the lysosome, autophagy, progressive motor decline in mice, and that patient-derived fibroblasts and dopaminergic neurons had α-synuclein accumulation.

PSAP encodes prosaposin, which is later cleaved into four active saposins: Saposin A, B, C and D. Saposins A–D are lysosomal hydrolases required for the degradation of sphingolipids (Kishimoto et al., 1992; Lieberman, 2011), and mutations in saposins may lead to accumulation of these sphingolipids. Saposins A, B, C and in some case reports, saposin D (Diaz-Font et al., 2005; Radha Rama Devi et al., 2019), have been associated with specific lysosomal storage disorders (Kishimoto et al., 1992). In particular, saposin C activates glucocerebrosidase, a lysosomal enzyme encoded by GBA. Biallelic mutations in GBA and in rare cases in PSAP cause Gaucher disease, a lysosomal storage disorder. GBA mutations are also present in up to 20% of patients with Parkinson’s disease. However, a previous study suggested that PSAP variants are not associated with Parkinson’s disease. In this study it was shown that the specific saposin regions A–D, except for a nominal association of the saposin C region (with two variants identified, rs747170456 and rs745951723), were not associated with Parkinson’s disease (Ouled Amar Bencheikh et al., 2018).

To examine whether variants in saposin D or the other saposins are associated with Parkinson’s disease, we analysed large cohorts of Parkinson’s disease patients and controls (details on the specific cohorts are provided in Supplementary Table 1). These cohorts included a total of 4132 Parkinson’s disease patients and 4470 control subjects with full sequencing data on PSAP, and were used for analysing the burden of rare variants in PSAP and saposins A–D separately. We further analysed data on PSAP variants from the recent Parkinson’s disease genome-wide association study (GWAS) on 37 688 cases, 18 618 UK Biobank proxy-cases (i.e. subjects with a first degree relative with Parkinson’s disease) and ~1.4 million controls, all Europeans (Nalls et al., 2019). To examine whether variants in the PSAP locus are associated with age at onset (AAO) of Parkinson’s disease, we analysed data from a recent GWAS on the AAO of Parkinson’s disease, with data on 28 568 patients (Blauwendraat et al., 2019).

The cohorts analysed at the NIH (Supplementary Table 1) were captured using TruSeq DNA or TruSeq DNA PCRFree sample preparation kits, followed by whole genome sequencing (WGS). The cohorts analysed at McGill University were captured using molecular inversion probes (MIPs), which specifically target the regions of interest, as we previously described (Ross et al., 2016). A complete pipeline for MIPs design and a supplementary table including the MIPs used for the sequencing are available in Ouled Amar Bencheikh et al. (2018), and are also available upon request. In both NIH and McGill cohorts, variants were filtered for minor allele frequency < 0.01, genotype missingness (<0.05 in NIH cohorts; <0.10 in McGill cohorts) and Hardy-Weinberg equilibrium (P ≤ 1 × 10⁻⁶). In addition, in the NIH cohorts, quality control for mean sequence depth (<25×) and contamination rate (>3%) of the samples was performed. In the McGill cohorts, separate analyses with different depth of coverage thresholds (15×, 30× and 50×) were performed. Our study has power >80% to detect statistically significant (P < 0.05) rare variants frequency difference of 1% in one group versus 2% in the second group (patients or controls), and power of 100% to identify the effects of rs4747203 and rs885828 with P < 0.05 as reported by Oji et al. (2020). We tested the association of rare variants in saposin D and the other saposins with Parkinson’s disease using optimized sequence Kernel association tests (SKAT-O). These burden tests were first done on all rare variants, followed by analyses only with variants that are potentially functional (non-synonymous, splice site, frameshift and stop variants).

In the analyses of both McGill (with 15×, 30× and 50× thresholds) and NIH cohorts, SKAT-O results showed no association between PSAP variants and Parkinson’s disease. Furthermore, when we separately analysed saposins A–D, no association was found (Table 1). Of note, no known pathogenic saposin D mutations were found in either McGill or NIH cohorts. To examine the association of common variants in the PSAP locus with Parkinson’s disease in individuals of European origin, and specifically the association of rs4747203 and rs885828, we extracted data from the most recent GWASs on Parkinson’s disease risk and age at onset (Blauwendraat et al., 2019; Nalls et al., 2019). No PSAP variants passed the significance threshold for an association with risk or AAO of Parkinson’s disease (Fig. 1). In particular, the two SNPs identified by Oji et al. as significant both in the Japanese and in the combined Taiwanese-Japanese cohorts, rs4747203 and rs885828 (Oji et al., 2020), showed lack of association with risk of Parkinson’s disease [P = 0.7577, odds ratio (OR) = 1.0032, 95% confidence interval (CI) = 0.9828–1.0239 and P = 0.7288, OR = 0.996, 95% CI = 0.976–1.017, respectively] or with its AAO [P = 0.0164, b = 0.2367, standard error (SE) = 0.0986 and P = 0.01679, b = −0.2358, SE = 0.0986, respectively]. These P-values are not corrected for multiple comparisons and are not statistically significant after Bonferroni correction. Of note, these two SNPs are in perfect linkage disequilibrium (r² = 1, D′ = 1) both in East Asian and European populations, implicating that they represent the same allele (https://ldlink.nci.nih.gov). Since Oji et al. compared Japanese Parkinson’s disease patient population to East Asian controls, which include multiple ethnicities, we further compared the frequencies of these two variants with Japanese controls from a previously published GWAS (Satake et al., 2009, data also found in Oji et al., 2020  Supplementary Table 8). The alternative allele frequency of rs885828 (which is in perfect LD with rs4747203) was 0.7 in healthy Japanese control subjects, very similar to the frequency in Japanese Parkinson’s disease patients reported in the Oji et al. study, 0.72 (P = 0.1, χ² = 2.732).

Table 1.

Association of rare saposin variants with Parkinson’s disease

	P-value
	SKAT-O McGill All	SKAT-O McGill Func	SKAT-O NIH All	SKAT-O NIH Func
Sap A	0.40	0.39	0.34	NA
Sap B	0.13	0.15	0.79	NA
Sap C	0.76	0.76	NA	NA
Sap D	0.72	0.42	0.47	NA
PSAP	0.19	0.16	0.42	0.67

Func = functional variants, including non-synonymous, frame-shift, splice site and stop variants; NA = not applicable, not enough variants for analysis; Sap = saposin; SKAT-O = Optimized Sequential Kernel Association Test.

Figure 1.

Zoom plots of common genetic variants in PSAP and their association with Parkinson’s disease. (A) Zoom plot of variants within PSAP and their association with risk of Parkinson’s disease. Data are taken from the recent Parkinson’s disease genome-wide association study (GWAS), which included 37 688 cases, 18 618 UK Biobank proxy-cases and ~1.4 million controls (Nalls et al., 2019). None of the variants in PSAP were associated with risk of Parkinson’s disease. (B) Zoom plot of variants within PSAP and their association with age at onset (AAO) of Parkinson’s disease. Data are from the recent GWAS on the AAO of Parkinson’s disease, which included 28 568 patients (Blauwendraat et al., 2019). None of the variants in PSAP were associated with AAO of Parkinson’s disease.

In conclusion, our findings do not support an association between either rare or common variants in saposin D or any of the other saposins, and Parkinson’s disease. We cannot completely rule out the possibility of very rare pathogenic saposin D variants associated Parkinson’s disease, although we did not find an association between functional variants and Parkinson’s disease. It is possible that differences in the populations tested could have led to the differences between our findings and those reported by Oji et al. However, the current results suggest that PSAP variants play either a very minor role, or more likely, no role at all in Parkinson’s disease.

Data availability

The data that support the findings of this study are available from the corresponding author, upon reasonable request.