Putative second hit rare genetic variants in families with seemingly GBA-associated Parkinson’s disease

Rare variants in the beta-glucocerebrosidase gene (GBA1) are common genetic risk factors for alpha synucleinopathy, which often manifests clinically as GBA-associated Parkinson’s disease (GBA-PD). Clinically, GBA-PD closely mimics idiopathic PD, but it may present at a younger age and often aggregates in families. Most carriers of GBA variants are, however, asymptomatic. Moreover, symptomatic PD patients without GBA variant have been reported in families with seemingly GBA-PD. These observations obscure the link between GBA variants and PD pathogenesis and point towards a role for unidentified additional genetic and/or environmental risk factors or second hits in GBA-PD. In this study, we explored whether rare genetic variants may be additional risk factors for PD in two families segregating the PD-associated GBA1 variants c.115+1G>A (ClinVar ID: 93445) and p.L444P (ClinVar ID: 4288). Our analysis identified rare genetic variants of the HSP70 co-chaperone DnaJ homolog subfamily B member 6 (DNAJB6) and lysosomal protein prosaposin (PSAP) as additional factors possibly influencing PD risk in the two families. In comparison to the wild-type proteins, variant DNAJB6 and PSAP proteins show altered functions in the context of cellular alpha-synuclein homeostasis when expressed in reporter cells. Furthermore, the segregation pattern of the rare variants in the genes encoding DNAJB6 and PSAP indicated a possible association with PD in the respective families. The occurrence of second hits or additional PD cosegregating rare variants has important implications for genetic counseling in PD families with GBA1 variant carriers and for the selection of PD patients for GBA targeted treatments.


INTRODUCTION
Rare variants in the gene GBA1, which encodes betaglucocerebrosidase (GBA, OMIM: #606463; EC 3.2.1.45), are considered as risk factors for developing alpha synucleinopathy manifesting clinically as Parkinson's disease (PD) 1,2 . The GBAassociated PD (GBA-PD) often aggregates in families but the connection between GBA1 variants and PD pathogenesis remains ambiguous. Contributing to this ambiguity is an unknown molecular mechanism by which GBA variants translate into cellular alpha-synuclein mishandling. Genetic observations such as the occurrence of non-symptomatic GBA1 variant carriers (e.g. incomplete penetrance) and a high rate of PD symptomatic noncarriers (e.g. phenocopies of mostly unknown origin) in seemingly GBA-PD families further obscure the link between GBA and PD [3][4][5] . From a genetic point of view, reduced penetrance and phenocopies in GBA-PD families can be explained by interactions between GBA1 variants and unidentified additional genetic and/or environmental risk factors or second hits. For instance, additional risk factors may influence cellular alpha-synuclein homeostasis concurrently with GBA1 variants in an epistatic manner, thereby modifying the risk of alpha synucleinopathy attributed to GBA variants and hence their penetrance. Alternatively, additional risk factors may act independently of GBA variants and increase the PD risk non-epistatically of the GBA1 locus, thereby resulting in the occurrence of phenocopies in GBA-PD families.
The identification of additional factors such as second genetic hits can improve the clinical management of GBA-PD on many fronts. For instance, this knowledge will enable genetic risk management, counseling, and clinical surveillance in seemingly GBA-PD families, which is otherwise difficult to achieve solely based on GBA1 variants. Furthermore, GBA has recently become a prominent target for therapeutic development. Drugs such as GZ/SAR402671 (ClinicalTrials.gov Identifier: NCT02906020) and Ambroxol (ClinicalTrials.gov Identifier: NCT02941822 and NCT02914366) have been promoted for randomized interventional trials in PD patients with GBA1 variants. These drugs aim to augment lysosomal transport and function in GBA1 variant carriers thereby restoring healthy lysosomal function, which is required for cellular alpha-synuclein handling [6][7][8] . A major concern is that additional risk factors in GBA1 variant carriers may affect cellular alpha-synuclein homeostasis and thereby may confound the outcome of these trials. 1 Massively parallel sequencing, when combined with familial genetic studies, has the potential to reveal high-risk rare genetic variation in complex neurodegenerative disorders that show a substantially heritable component 9 . Application of sequencing technologies in PD genetics has recently led to the discovery of additional potentially pathogenic rare genetic variation in PD patients carrying primary pathogenic mutations, thus providing evidence for oligogenic inheritance and pathway interactions in PD 10,11 . In this study, we uncovered rare variants in DNAJB6 and PSAP as putative second hits in two PD families segregating GBA1 mutations c.115+1G>A and p.L444P, respectively.

Clinical findings
Family A comprised four PD subjects, with a mean age of 58.5 years (range , mean age at onset of 46.5 years (range [40][41][42][43][44][45][46][47][48][49][50][51], and a mean disease duration of 12.5 years. Depression was noted in all PD patients of the family A. Patients with longer disease duration presented with cognitive decline (Supplementary Table 1 Table 1). The PD patients in family B were treated with carbidopa-levodopa (Sinemet) with a good response.
Identification and familial segregation of rare genetic variants in family A and B The contribution of rare genetic variation to PD risk in family A and B was explored by whole-genome sequencing followed by rare variant filtering and segregation analysis. The summary of study design and data analysis is shown in Supplementary Fig. 1. Genome sequencing was performed for three PD patients in family A and one PD patient in family B (Fig. 1A, C; asterisks). Sequencing was also performed for one control individual from each family (Fig. 1A, C; asterisks), which helped to reduce the final list of candidate rare variants. The list of candidate rare variants identified in the two families is shown in Table 1.
A GBA1 splice-site variant (rs104886460-G/A; c.115+1G>A; ClinVar ID: 93445) was identified in the WGS data of family A. Segregation testing of this variant in family A revealed four PD subjects as heterozygous carriers of c.115+1G>A and in addition three asymptomatic carriers (Fig. 1A, symbols marked with thick lines; IV:6, V:4, and V:5). The younger asymptomatic carriers V:4 and V:5 (aged 41 and 40 years, respectively) may still develop PD or may already exhibit prodromal features of the disease as suggested previously 12,13 . We therefore assessed the three asymptomatic GBA1; c.115+1G>A carriers for olfactory (odor identification ability), visual and cognitive functions (Mini-Mental State Examination score), and depression (HAM-D score), which were all normal (data not shown). The occurrence of an asymptomatic carrier (IV:6, age 65 years) considerably older than the average age of PD onset in this family (i.e 46.5 years) suggests that the PD penetrance of the heterozygous variant GBA1; c.115+1G>A in family A is incomplete.
Another GBA1 gene variant (rs421016-T/C; c.1448T>C which encodes p.L444P, also called p.L483P; ClinVar ID: 4288), was identified by WGS of a PD patient (Fig. 1C, subject III:1) from family B. Segregation analysis of the p.L444P variant in family B revealed two additional PD patients as heterozygous carriers (Fig. 1C, subjects II:2 and II:4). No asymptomatic carrier of p.L444P was observed in this family. The PD patient III:4 in family B (Fig. 2C, male, black symbol with a white circle) did not segregate the GBA; p.L444P variant. He presented with action tremor, slight rigidity, and slowness of the movement in his upper right limb at the age of 49, which affected his daily activities, including writing, dressing, and his job as a restaurant worker. His motor symptoms worsened over a span of 2 years. Based on his medical history, a review of his signs and symptoms, and a neurological and physical examination he was formally diagnosed with PD at the age of 51 years (Supplementary Table 1). Due to his non-carrier status for the p.L444P variant in GBA1, which otherwise segregated with PD in family B, the patient III:4 was considered a phenocopy PD patient of unknown origin.
Exonic Rare variants c.A577G (p.T193A) in the DNAJB6 gene (NM_058246.4, rs770053224_A/G) and c.A470G (p.N157S) in the PSAP gene (NM_002778.4, rs756379007_T/C) segregated with PD in family A and B, respectively. All four PD patients (IV:1, IV:4, IV:5, and V:1) of family A were heterozygous carriers for the DNAJB6; p.T193A and GBA1; c.115+1G>A variants whereas the three nonpenetrant carriers of GBA1; c.115+1G>A in family A (IV:6, V:4, and V:5) lacked this concurrence. This indicates that the DNAJB6; p.T193A variant contributes to PD risk in this family (Fig. 1A). In family B, segregation of PSAP; p.N157S in all four PD patients including the phenocopy PD patient III:4 ( Fig. 1C, black symbol with a white circle) indicated that the rare genetic variation in the PSAP gene is an independent risk factor for PD in this family which complies with a previous report 10 .
The PD cosegregating variants in DNAJB6 and PSAP genes identified in this study are listed with extremely low frequencies (MAF < 0.0001) in the ethnically matched controls in GnomAD v2.1.1 (ref. 15 ) and were classified as the 1% most deleterious substitutions in the genome (CADD > 20) 16 (Table 1). No other additional candidate rare variant segregated completely with PD in the two families.
Functional impact of the rare variant in DNAJB6 DNAJB6, a member of the HSP40 family of J-protein cochaperones, acts as an aggregation suppressor preventing Table 1. Candidate rare variants identified through genome sequencing and analysis of family A and B. fibrillation of several aggregation-prone proteins 17,18 . Steady-state levels and cellular distribution profiles of the wild-type DNAJB6 and the p.T193A variant proteins expressed in HEK293 cells were not significantly different (data not shown). DNAJB6 dimer-/ oligomerization was also not affected by p.T193A substitution ( Supplementary Fig. 2), suggesting that p.T193A substitution exerts its influence by altering the cellular function of DNAJB6 protein. The p.T193A substitution in DNAJB6 protein identified in the current study is located in a highly conserved serine/ threonine-rich motif (i.e SxSTST at residues 190 and 192-195), which lines the concave side of the DNAJB6 dimer near the peptide-binding cleft ( Fig. 2A, B). This structural conformation is necessary for cellular functions of the DNAJB6 protein 19,20 , which includes alpha-synuclein homeostasis 21 . We expressed wild-type DNAJB6 or p.T193A variant containing DNAJB6 proteins in an alpha-synuclein reporter cell line lacking endogenous DNAJB6 (see "Methods"). Diffuse cytoplasmic distribution of alphasynuclein-dsRed conjugate was observed in the majority of cells in case of wild-type DNAJB6 protein expression whereas the expression of DNAJB6; p.T193A only partially restored the diffuse cytoplasmic alpha-synuclein distribution. A significantly higher number of cells expressing DNAJB6; p.T193A variant displayed a focal cytoplasmic distribution of alpha-synuclein-dsRed compared to wild-type DNAJB6. Collectively, these data indicate altered cellular homeostasis of alpha-synuclein as a direct consequence of p.T193A variant expression in HEK293 cells (Fig. 2C, D).
Functional impact of the rare variant in PSAP Prosaposin (PSAP) is a lysosomal protein, which in its native form or processed peptides (e.g. Saposins A-D and several prosaptides) plays distinct cellular roles 22 . The PD cosegregating PSAP variant p.N157S identified in family B (Fig. 2C, D) is located in the Intersaposin A-B region of the PSAP protein (Fig. 3A). Rare potentially pathogenic variants in other regions of PSAP protein have previously been reported to increase PD risk in a GBAdependent manner 10,[23][24][25] . To determine the functional impact of the p.N157S variant, we expressed wild type or p.N157S variant PSAP as Flag or EGFP fusion proteins in the HEK293 cells. EGFPtagged PSAP; p.N157S displayed a lower co-localization with perinuclear lysosomal LAMP1-positive puncta compared to wildtype PSAP (Fig. 3C, D). The reduced lysosomal localization of the p.N157S variant may be due to an overall reduction in protein levels 48 h post-transfection (Fig. 3D, top panel), indicating that variant PSAP protein is less stable than that of wild-type protein in HEK293 cells. It has previously been shown that the intersaposin A-B region in the PSAP protein is necessary for cleavage of native PSAP into biologically active Saposin peptides 26 (Fig. 3A, Saposin A-D). PSAP is a functional partner of lysosomal Cathepsin D (CTSD)   Supplementary Fig. 3C. All blots are derived from the same experiment and gels/blots were processed in parallel. E Representative images showing LAMP1-positive intracellular inclusions decorated with alpha-synuclein-EGFP. Cells were transfected with three plasmids: a plasmid-expressing EGFP-tagged aggregation-prone alpha-synuclein protein (see "Methods"), a plasmid-expressing mTurquoise2-LAMP1 expression protein as a lysosomal marker, and a plasmid-expressing wild type or p.PN157S variant containing PSAP. Arrowheads in the merged panel point to the EGFP-PSAP puncta in the perinuclear region overlapping with LAMP1-mTurquoise2 as well as alpha-syuclein-EGFP puncta indicating lysosomal aggregation. F The quantification of lysosomal alpha-synuclein-EGFP aggregates in transfected cells is shown (n = 3 transfections). Statistical analysis was performed by ANOVA. Graphs represent mean and standard deviation, **P < 0.01. and both proteins are mutually dependent for lysosomal trafficking. Moreover, PSAP-CTSD interaction is necessary for stimulation of the CTSD proteolytic activity, which mediates the processing of PSAP at the intersaposin A-B region [26][27][28][29] . Computational predictions indicated that p.N157S substitution in PSAP introduces a novel phosphorylation site (LESSKI, AA 153-158) in the intersaposin A-B region near a predicted CTSD-specific site (PSAP residues 176-179, LLL) (Fig. 3A, B and Supplementary Fig.  3A). We tested the influence of p.N157S substitution on the PSAP-CTSD interaction. Fluorescence resonance energy transfer (FRET) analysis of co-expressed PSAP and CTSD as FRET capable fusion proteins indicated that the p.N157S substitution reduces PSAP-CTSD interaction in the LAMP1+lysosomal perinuclear compartment (Supplementary Fig. 3B). Thus, reduced PSAP synthesis and reduced interaction with CTSD were identified as functional consequences of the p.N157S substitution. This may influence cellular alpha-synuclein homeostasis [30][31][32][33] considering the role of PSAP in lysosomal CTSD delivery and activation 27,28 . We, therefore, tested the influence of the PSAP; p.N157S expression on lysosomal handling of alpha-synuclein in heterologous cells. To this end, we co-expressed mCherry-tagged PSAP or PSAP; p.N157S with an aggregation-prone alpha-synuclein-EGFP (see "Methods"). The number of LAMP1-positive alphasynuclein-EGFP decorated inclusions per cell as well as the number of cells with such inclusions was higher when coexpressing PSAP; p.N157S than when co-expressing wild-type PSAP (Fig. 3E, F). Collectively, these results indicate that expression of p.N157S variant PSAP in HEK293 cells results in a lysosomal defect, which affects alpha-synuclein handling.

DISCUSSION
We here report rare genetic variants in the HSP70 chaperone DNAJB6 and the lysosomal protein PSAP as likely second hits in addition to the GBA1 gene variants in two PD families. The segregation pattern of additional rare variants in both families (Fig. 1A, C) and their ability to manipulate cellular alpha-synuclein handling when expressed in reporter cells (Figs. 2C, D and 3E, F) lead to the hypothesis that they contribute to the PD pathophysiology.
The precise mechanism by which GBA1 variants increase the risk for developing alpha synucleinopathy in PD remains unknown. Cellular alpha-synuclein is intrinsically prone to aggregation and toxic fibrillation 34,35 . Thus, a major portion of alpha-synuclein is delivered to lysosomes for degradation via non-vesicular (chaperone-mediated) or vesicular (micro-and macro-autophagy) pathways 36,37 . It was proposed that GBA1 variants result in an abnormal lysosomal glycosphingolipid environment that leads to an impaired autophagy-lysosomal pathway and promotes alphasynuclein aggregation 38 . However, genetic observations such as reduced PD penetrance 5 of GBA1 variants in familial cases, the occurrence of phenocopies in GBA-PD families 3,4 , and comparable age-specific PD risk in homo-and heterozygous GBA1 variant carriers 39 indicate that GBA1 variants alone do not cause alpha synucleinopathy or PD in the heterozygous GBA1 variant carriers. This suggests that additional risk influence cellular alpha-synuclein homeostasis in GBA-PD.
DNAJB6 is a member of the HSP40 family of J-protein cochaperones. Rare genetic variants in several members of this family have previously been associated with PD [40][41][42][43][44] . Our data support that a p.T193A substitution, which is located in a conserved serine-threonine-rich SxSTST motif (residues 190 and 192-195) of DNAJB6, affects its anti-aggregation properties with respect to cellar alpha-synuclein homeostasis as was described previously for other aggregation-prone proteins 19,20 . In our study, DNAJB6; p.T193A and GBA1; c.115+1G>A variants were found to segregate concurrently only in the PD patients of family A (Fig. 1A,  B). Individuals lacking this concurrence did not show cardinal motor or prodromal signs of PD (Fig. 1A, subjects IV:6, IV:7, V:4, and V:5). The DNAJB6; p.T193A variant when expressed in HEK293 cells promotes alpha-synuclein aggregation (Fig. 2C, D). Taken together, our results indicate that DNAJB6; p.T193A is a rare genetic variant, which apparently acts independently in relation to GBA, to enhance PD risk in the GBA1; c.115+1G>A variant carriers of the family A.
PSAP is a multifunctional protein, which is secreted in its native form as ligand for the PD-associated neuroprotective receptor GPR37 (refs. 45,46 ). PSAP is intracellularly cleaved by lysosomal proteases such as Cathepsin D (CTSD) 26 into mature monosaposins. One of the cleavage products, namely Saposin C, is essential cofactor for lysosomal GBA 24,25 . Short peptides derived from PSAP (e.g. Prosaptides) have protective effects on dopaminergic neurons in vitro and in models of PD in vivo 47,48 . At the molecular level, PSAP protein engages in reciprocal functional interaction with CTSD and this interaction affects lysosomal trafficking and function of both PSAP and CTSD [27][28][29] . Interestingly, PSAP mRNA is upregulated in the substantia nigra of PD patients 49 . This upregulation may be a protective mechanism against PD, given the neuroprotective functions of PSAP and our study, which indicates that a dysfunctional PSAP contributes to PD.
In family B, three PD patients were concurrent carriers of GBA; p.L444P and PSAP; p.N157S variants. However, patient III:4 carried only the PSAP; p.N157S variant. This patient was clinically indistinguishable from the three PD patients with both variants, indicating a similar underlying pathogenic mechanism (Supplementary Table 1, Family B). Indeed both GBA and PSAP proteins interact at the molecular level 24,25 . PD pathogenesis due to GBA; p.L444P variant involves accumulation of alpha-synuclein aggregates 50,51 . The PSAP; p.N157S variant when expressed in HEK293 cells also promoted alpha-synuclein aggregation (Fig. 3E, F). The exact molecular mechanism by which PSAP; p.N157S variant promotes alpha-synuclein is currently not known. One possibility is a reduced molecular interaction between PSAP; p.N157S variant and CTSD ( Supplementary Fig. 3B), which may decrease lysosomal CTSD function 27 . Since CTSD is the main lysosomal protease clearing alpha-synuclein 52,53 , the PSAP; p.N157S variant may promote alpha-synuclein aggregation and PD risk independently of GBA; p.L444P variant. Thus, the PSAP; p.N157S variant may contribute to PD risk in all four PD patients of the family B independent of the GBA; p.L444P variant. We thus considered the subject III:4 of the family B to be a genetic phenocopy lacking the GBA; p.L444P variant.
Finally, we focused on rare genetic variation as additional risk factors for GBA-PD in a family-based design. Variants in the PSAP and DNAJB6 genes were selected from the list of background WGS candidates primarily based on segregation testing and the potential roles of DNAJB6 and PSAP proteins in the cellular alpha-synuclein handling. In this study, we tested in heterologous cells (HEK293 cells) whether the identified variants have disrupted functions compared to the wild-type DNAJB6 and PSAP proteins. Further investigation on patient-derived material such as iPSCs is required to definitely link the identified rare genetic variants in DNAJB6 and PSAP with PD pathogenesis. The prioritization of DNAJB6 and PSAP in this study does not ignore the role of other rare variants in the list of candidates ( Table 1) that may in addition play a role in PD pathogenesis. For instance, a potential pathogenic rare variant in the Tenascin receptor gene (TNR: p.N180H; ClinVAR ID 224863) was found in family A (Table 1). This finding is an independent replication of a recent exome sequencing study involving familial PD patients from a distant population 54 . In our study, the TNR variant was detected in three of the four PD patients of family A but in none of the controls, which indicates that the variant may have a pathogenic effect in the segregating individuals. In addition, we did not consider common genetic variation at the loci associated with cellular alpha-synuclein homeostasis. Large-scale genetic association studies are required to uncover such interactions. Despite these limitations, findings from our study contribute to a deeper molecular understanding of GBA-PD. Such understanding will not only be helpful for the clinicians when providing genetic counseling to seemingly GBA-PD patients and families but also for selecting patients for novel interventional therapies aiming to treat PD by augmenting lysosomal GBA function.

Study participants and clinical evaluation
Our study followed the tenets of the Declaration of Helsinki. The Institutional Review Board of the Institute for biomedical and genetic engineering (IBGE, Islamabad, Pakistan) provided the approval (Ref. IBGE/ SARK04/1201/2012). Written informed consent was obtained from all participants.
Medical records of the neurology department of Shaheed Zulfiqar Ali Bhutto Medical University in Islamabad were screened rigorously to identify PD patients with a positive family history (defined as having at least one first-degree relative with PD who was also registered in the same clinic). We identified two families (family A and B) with four PD patients each. There was no history of consanguinity in the two families. After obtaining written informed consent, members of family A and B underwent a detailed neurologic examination by at least one experienced movement disorders specialist. The diagnosis of the eight PD patients was verified based on the UK Parkinson's disease society brain bank 55 and Gelb criteria 56 . Clinical data were collected, which comprised assessment of motor disability scores (Unified Parkinson Disease Rating Motor subscale III, UPDRS-III, total score 108) 57 , PD severity assessment (Hoehn and Yahr scale, H-Y, stages 1-5) 58 , cognitive impairment rating (Mini-Mental State Examination, MMSE, cut-off 24/30) 59 , depression rating (7-item version of the Hamilton Depression Rating Scale; scores 0-8: normal, 9-16: mild, 17-24: moderate and ≥25: severe) 60 and an assessment of olfactory and visual functions.

Genome sequencing and analysis
Genomic DNA was isolated from peripheral blood cells using a QIAamp DNA Blood Mini Kit. DNA libraries were prepared using the TruSeq Nano DNA PCR-free kit (Illumina San Diego CA USA). Paired-end sequencing (2 × 150 bp) was performed at >30× coverage per sample (Illumina HiSeq X TEN and HiSeq2500, Illumina San Diego CA). Raw reads were aligned to the human reference genome (version build GRCh37, version hs37d5) using bwa mem (version 0.7.8) with minimum score threshold set to zero [-T 0] and remaining settings left at default values 61 . Duplicates were marked using biobambam (version 0.0.148) 62 . SNVs were called using SAMtools (version 0.1.19) 63 and short indels were called using Platypus (version 0.7.4) 64 . Functional classifications of the variants were done using ANNOVAR 65 with gene model definitions from Gencode (version v19).
Variant filtering, candidate selection, and Sanger sequencing Quality-filtered variants (QUAL score >20) with minimum coverage >10× were carried forward for filtering and candidate selection. A stringent filtering approach was used to prioritize rare pathogenic variants likely representing candidates worthy of functional validation. We only considered variants with minor allele frequency MAF < 0.01 in ExAC (version 0.3) since high-risk rare variants with large effect sizes tend to aggregate in families. All functional rare variants, including nonsynonymous SNVs, frameshift, and non-frameshift indels, stop-gain/loss and splice-site SNVs or indels were considered. Rare functional variants that were identified only in the control individuals but were absent in one or more of the affected individuals of the family were removed. Deleteriousness prediction was carried out using a set of prediction tools, including CADD 16 , PolyPhen-2 (ref. 66 ), SIFT 67 , LRT 68 , Mutation Taster 69 , FATHMM 70 , PROVEAN 71 , and MetaSVM and MetaLR 72 . Benign variants were filtered out using the mutation significance cut-off (MSC) tool, a gene-level-specific cut-off for CADD/PolyPhen-2/SIFT scores as described previously 73,74 . Variants in genes with high brain expression in the GTEx Portal 75 were prioritized. The SmallPedigree-WGS module of the Canvas program (version 1.40.0.1613) (https://academic.oup.com/bioinformatics/article/32/ 15/2375/1743834) was applied to the WGS data of two families separately to detect CNVs.
Commercially available M13-tailed primer pairs (Table 1; Thermo Fisher Scientific) were used for the amplification of candidate genomic regions. The genomic segment covering the position of p.L444P variant in GBA1 (rs421016 position) was amplified using forward (TGGGTGCGTAACTTTGTC GACAG) and reverse (CCACAGCAGGATCCTTGATGGTAA) primers, as described previously 76 . The genomic segment covering a novel stop-gain variant in the ABCA7 gene (corresponding to p.Q1062X) was amplified using M13-tailed forward (TACTACCTGACGCTGGTGAAGG) and reverse (AGTTAAGGCACAGCCACCCCACTG) primers (M13 sequence not shown). Thermocycling conditions used were as follows: 3 min at 94°C, 30 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 45 s, and a final extension for 5 min at 72°C. PCR products were resolved on 1.5% agarose gel and purified using the MinElute Gel Extraction Kit (Qiagen, Germany). Bidirectional Sanger sequencing was performed at StarSEQ GmbH, Mainz, Germany.
In silico analysis and protein structure predictions Online tools PROSITE My Domains 77 , ClustalOmega 78 , and Jalview 79 were used for drawing protein domain features, sequence alignment, and visualization, respectively. N-glycosylation sites were predicted using the GlycoEP and NetNGlyc web-servers 80,81 . Analysis of post-translational modification sites was performed with the Eukaryotic Linear Motif webserver 82 . Three-dimensional structure of human PSAP protein (Uniprot ID: P07602) was predicted using the I-TASSER web-server 83 . The ModLoop web-server 84 was used to repair broken peptide bonds. Previously, Sun et al. 85 showed that Cathepsin D (CTSD) specific sites are leucine rich. Based on this observation, we scanned the PSAP sequence, which predicted residues 176-179 (LLL) in the intersaposin region A-B of the PSAP as a potential CTSD-specific site. The three-dimensional structure of the human DNAJB6 homodimer was described previously by Söderberg et al. 19  Plasmids, cell culture, and transfection Full-length cDNA encoding wild-type human PSAP cloned in pCMV3 with a C-terminal Flag tag was obtained from Sino Biological (HG16224-CF) and subcloned into pEGFP-N1 with HindIII and AgeI to obtain pEGFP-N1-PSAP (WT) plasmid. A fragment of PSAP cDNA encoding PSAP amino acid residues 1-320 encoding N157S variant was synthesized via GeneArt synthesis (Thermo Fisher Scientific) and inserted in pEGFP-N1-PSAP (WT) via HindIII and EcoRI, replacing the corresponding wild-type sequence to generate the pEGFP-N1-PSAP (N157S) plasmid. PSAP cDNA sequences encoding the wild type and the p.N157S variant were similarly cloned in pmCherry-N1. The cDNA sequence encoding human Cathepsin D (CTSD) was amplified from a human brain cDNA library and inserted into pmCherry-N1 using HindIII and AgeI restriction sites. The pcDNA3.1 plasmid-expressing wild-type DNAJB6 as a fluorescent fusion protein was previously published 86 and was kindly provided by Professor Conrad Chris Weihl (Washington University School of Medicine). A wild-type DNAJB6 cDNA clone was modified by PCR to introduce the c.A577G/p.T193A variant in the coding sequence using the following primers: CAAATCGA-TATCAGCTTCAAC (forward primer; variant position is underlined) and CAACAGATGGCTGGCAACTAG (reverse primer). An alpha-synuclein-dsRed aggregation reporter cell line lacking endogenous DNAJB6 (ref. 21 ) was kindly provided by Dr. Cristian Hansen, Lund University, Sweden. A bicistronic plasmid encoding aggregation-prone alpha-synuclein (A53T) cDNA tagged with PDZ-binding motif (HSTTRV) of neuroligin-1 and PDZ1 domain of synaptic scaffolding molecule (S-SCAM) fused to mCherry has previously been described 87 and was kindly provided by Professor Björn H. Falkenburger (University Hospital Carl Gustav Carus at the Dresden University of Technology, Germany). This plasmid was modified to replace mCherry with the EGFP coding sequence. A plasmid encoding Lamp1-mTurquoise2 was obtained from Addgene (Plasmid # 98828). Sanger sequencing verified the sequence integrity of all the above-mentioned plasmids.

Immunoblotting, cell imaging, and quantification
For immunoblotting, HEK293 cells transfected with plasmids expressing Flag-tagged proteins were washed and lysed. SDS-PAGE was performed with 10% acrylamide gels. Proteins were transferred to polyvinylidene difluoride (PVDF) membrane, which was blocked in 10% skimmed milk for 1 h. The membranes were incubated with mouse anti-Flag antibody (1:1000, F1804, Sigma) overnight at 4°C, followed by incubation with peroxidase-linked secondary antibody (1:10,000, P0161, Dako) for 1 h. Protein bands were detected on a ChemiDoc XRS + imaging system (Bio-Rad). After Flag tag detection, the PVDF membrane was stripped and betaactin expression was detected on the same membrane using a rabbit antiactin antibody (1:1000, ab8227, Abcam).
For cell imaging, cells transfected with appropriate plasmids were washed three times with phosphate buffer saline (PBS) solution and were fixed in 4% paraformaldehyde for 10 min. Nuclei were stained with 4′,6′diamidino-2-phenylindole (Sigma, St Louis, MO, USA). Coverslips were mounted onto slides with ProLong Gold anti-fade mountant (Thermo Fisher Scientific). Images were obtained with a ×63/1.4 oil objective on a TCS SP5 confocal imaging system (Leica Microsystems, Heidelberg GmbH). For the analysis of lysosomal localization of PSAP variants, Fiji software was used to quantify the integrated intensity of EGFP fluorescence (i.e PSAP-EGFP fusion) overlapping with turquoise signal (i.e mTurquoise2-LAMP1 fusion). At least 30 cells for each condition from triplicate coverslips were analyzed. Quantification of alpha-synuclein aggregation was performed as described previously 21,87 . Briefly, whole-field images were obtained from three coverslips for each condition. Alpha-synuclein aggregation was defined as a cellular region with markedly increased corresponding immunofluorescence signal compared with surrounding cytoplasmic areas of the same cell. An investigator blinded to the experimental conditions classified at least 100 cells per coverslip as positive or negative for cytoplasmic alpha-synuclein aggregates (Fig. 2), or cells with no, single, or multiple lysosomal aggregates (Fig. 3). All experiments were repeated at least three times.

FRET analysis
For the FRET experiments, cells were transfected with plasmids expressing EGFP and mCherry as a donor-acceptor FRET pair alone as controls or as fusion proteins. A construct expressing EGFP and mCherry as a tandem dimer separated by a spacer was used as FRET standard. Twenty-four hours after transfection, cells were fixed and mounted as described above. FRET measurements were performed by using the FRET acceptor photobleaching wizard in the Leica Application Suite on a Leica scanning confocal microscope (TCS SP5 Leica Microsystems, Heidelberg GmbH, Germany). FRET measurements were performed for at least three independent transfections for each condition. FRET efficiency was defined as FRET eff = (donor post − donor pre )/donor post , where donor post describes the fluorescence intensity after, and donor pre the intensity before acceptor photobleaching. FRET eff was calculated by the FRET wizard of the Leica Application Suite.

Statistical analysis
We used QtiPlot software (version 4.8.7) for statistical analysis and data presentation. Graphs represent mean ± standard deviation. The statistical tests and number of independent experiments for each analysis are noted in the figure legends. P < 0.05 was considered significant.
MAÖ and RCW gratefully acknowledge the support of the Klaus Tschira Foundation and the European Union's Horizon 2020 Framework Programme for Research and Innovation under the Specific Grant Agreement Nos. 720270 and 785907 (Human Brain Project SGA1 and SGA2). A.U. and S. N. gratefully acknowledge the support from the German Academic Exchange Service (DAAD Scholarship) and International Research Support Initiative Program of the higher education commission of Pakistan respectively.

Reporting summary
Further information on experimental design is available in the Nature Research Reporting Summary linked to this paper.

DATA AVAILABILITY
The datasets generated during and/or analyzed during the current study are available in the EGA repository under the following accessions: Study: EGAS00001004777 and Dataset: EGAD00001006561.