Homozygous and compound heterozygous rare variants in VPS13C contribute to Lewy body diseases

Dementia with Lewy bodies (DLB) and Parkinson’s disease (PD) are clinically, pathologically and etiologically overlapping disorders. They are included in the Lewy body disease (LBD) continuum characterized by α-synuclein-positive Lewy body pathology in neurons. Homozygous PTC mutations in Vacuolar Protein Sorting 13 homolog C gene (VPS13C) are associated with early-onset PD. Whole genome sequencing of two affected siblings and healthy parents of family A with onset age below 50 years and DLB pathology conrmed at autopsy. Targeted resequencing of the VPS13C coding region in 844 LBD patients and 664 control individuals. Phasing cis-trans location of the compound heterozygous VPS13C variants. Analysis of VPS13C protein expression in lymphoblast cells and brain lysates. Immunohistochemistry and live cell labeling in HeLa and SH-SY5Y cells overexpressing wild type or mutant VPS13C. LBD, and when present bi-allelic in patients, these combinations have variable effects on expression and functioning of VPS13C. Apart from the recessive PTC mutations, some combinations of rare missense mutations might mimic recessive inheritance and explain part of the sporadic LBD patients. We propose that homozygous or compound heterozygous rare missense variants in VPS13C reducing VPS13C protein expression can contribute to risk of LBD. Hematoxylin-Eosin, Klüver-Barrera myelin Immunohistochemistry, performed with against β-amyloid hyperphosphorylated tau (AT8), ubiquitin, TDP-43, FUS, p62 and α-synuclein. P9 Long-read ONT cDNA sequencing, the missense variants in P8 and P9. the VPS13C c.448 + 7A G splice site variant is present downstream of exon 6 of VPS13C isoform 2 amplify VPS13C amplicons, exon 6 or 7 specic for isoform 2, from cDNA derived of lymphoblast cells, preventing phasing of c.448 + 7A > G/p.Ala1687Val and evaluation of the effect of c.448 + 7A > G on splicing. We conrmed cis conguration of the compound heterozygous alleles in 3 patient and in control recessive excluded carriers from further in this


Methods
Whole genome sequencing of two affected siblings and healthy parents of family A with onset age below 50 years and DLB pathology con rmed at autopsy. Targeted resequencing of the VPS13C coding region in 844 LBD patients and 664 control individuals. Phasing cis-trans location of the compound heterozygous VPS13C variants. Analysis of VPS13C protein expression in lymphoblast cells and brain lysates. Immunohistochemistry and live cell labeling in HeLa and SH-SY5Y cells overexpressing wild type or mutant VPS13C.

Results
In the two affected siblings of family A, we identi ed compound heterozygous rare variants located in trans in VPS13C Rare VPS13C variants (MAF ≤ 1%), are signi cantly associated (p = 0.0233) in the LBD patient cohort using optimized sequence kernel association test (SKAT-O). We identi ed 11 carriers of compound heterozygous variants and 1 carrier of homozygous variants in VPS13C. Trans location of the variants in compound heterozygous carriers was con rmed in 4 and cis location in 3 patients. The frequency of patient carriers with bi-allelic variants mimicking recessive inheritance is 1.07% (9/844). In post-mortem brain tissue of two unrelated DLB carriers of compound trans heterozygous variants, we observed a reduction of VPS13C protein expression. Overexpressing of wild type or mutant VPS13C, in HeLa and SH-SY5Y cells, demonstrated that the mutations abolish the endosomal/lysosomal localization of VPS13C.

Conclusions
Our data indicate that rare alleles are associated with LBD, and when present bi-allelic in patients, these combinations have variable effects on expression and functioning of VPS13C. Apart from the recessive PTC mutations, some combinations of rare missense mutations might mimic recessive inheritance and explain part of the sporadic LBD patients. We propose that homozygous or compound heterozygous rare missense variants in VPS13C reducing VPS13C protein expression can contribute to risk of LBD.

Belgian patient and control cohorts
Members of the Belgian BELNEU consortium recruited patients at neurological centers associated with university or general hospitals in Belgium [26,27]. The LBD cohort comprised 844 LBD patients with mean age at onset age (AAO) of 62.9 ± 11.8 years (31.5% female; 20.0% with positive familial history). In the LBD cohort, 233 patients were diagnosed with DLB (mean AAO 70.2 ± 10.2 years, range 34-88; 33.0% females; 24.0% with positive familial history), of which 159 patients (68.2%) had a clinical diagnosis and 74 patients (31.8%) a neuropathological diagnosis of DLB. The remaining 611 patients in the LBD cohort had a clinical diagnosis of PD (mean AAO 60.6 ± 11.3 years, range 24-88; 31.1% females and 18.5% with positive familial history). Patients had a clinical examination by a neurologist and neuroimaging. We collected medical history of patients and family members. A positive family history of disease was given if at least one rst-degree relative was affected. DLB patients were diagnosed in accordance with the established criteria for possible, probable or pathological DLB [7,28], and PD patients according to the NINDS diagnostic criteria for PD [29]. The PD cohort was genetically pro led for the 5 major PD genes (SNCA, LRRK2, PARK2, PINK1 and PARK7) by means of Sanger sequencing for simple mutations and multiplex amplicon quanti cation (MAQ, Agilent, Multiplicom, Niel, Belgium), quantitative real-time PCR or multiplex ligation-dependent probe ampli cation (MLPA) [30] for copy number variants [31].
A geographically matched control cohort consisted of 664 individuals with a mean age at inclusion (AAI) of 72.0 ± 9.4 years (range 34-88) and comprised 60.5% females. Control individuals were recruited among healthy partners of patients visiting a memory clinic, and negative for neurological or psychiatric antecedents or neurological complaints, or community-recruited individuals scoring >25 on a Montreal Cognitive Assessment (MoCA) [32] with a negative individual or familial history of neurodegenerative or psychiatric diseases.

Ethical assurances
The ethics committee of the University Hospital Antwerp and University of Antwerp approved the clinical and genetic research protocols. Collection of biological samples was in accordance with the written informed consent signed by the participant and/or their legal guardian.
Whole genome sequencing Short-read paired-end WGS of two affected siblings of family A (Fig. 1a), subsequent read alignment to the human reference genome (GRCh37/hg19) and base and variant calling were performed by Complete Genomics TM Inc [33]. To annotate and select genetic variants in the WGS data we used GenomeComb [34]. High quality variants were selected based on a sequence coverage of at least 20x, a variant call score (Complete Genomics Inc.) of ≥60 dB, i.e. a probability of ≥10 6 :1 describing the likelihood of the variant call compared to the second most likely call, and genomic location outside repeat regions marked as simple repeats or micro satellites by RepeatMasker v3.0 [35]. Novel or rare (minor allele frequency (MAF) < 1%) variants in the 1000 Genome Project database [36] and below 25% in our in-house next generation sequencing database of Belgian patients with distinct neurological disorders), nonsynonymous coding and splice site variants were selected. Finally, we focused on variants in line with autosomal recessive inheritance, i.e. homozygous or compound heterozygous variants. We slected homozygous variants and multiple compound heterozygous variants in the same gene.

Targeted resequencing of VPS13C
We performed PCR ampli cation of all 86 coding exons and anking splice sites of VPS13C by amplicontarget PCR ampli cation (MASTR technology; Agilent). Amplicons were uniquely tagged; based on the Nextera XT shotgun library preparation protocol (Illumina), containing sample-speci c indices [37]. Libraries (n=384) were pooled and sequenced in one run on the MiSeq platform using the MiSeq V3 chemistry, generating paired-end sequence reads of 300 nucleotides (Illumina). After sample demultiplexing, we mapped sequence reads using the Burrows-Wheeler Aligner (BWA) [38,39] to a minigenome, combining the target sequences extracted from the human genome, reference sequence hg19.
Sequence variants were called using GATKv3.5 HaplotypeCaller [40,41] and variants annotated using GenomeComb [34]. We used the Combined Annotation Dependent Depletion (CADD_Phred) score, to predict the deleterious impact of non-synonymous variants in VPS13C, whereby a score above 20 represents the top 1% most deleterious variants in the genome [42]. Alternative splicing of VPS13C produces four different transcripts, ubiquitously expressed. Coding variants are numbered relative to the translation initiation codon in the largest VPS13C transcript (GenBank Accession Number NM_020821.2. Amino acid changes, numbered according to the largest VPS13C isoform (GenPept Accession Number NP_065872.1). Sequencing reads, visualized with the Integrative Genomics Viewer (IGV) [43] [44]. Target regions were PCR ampli ed from genomic DNA and subsequently Sanger sequencing using the BigDye® Terminator Cycle Sequencing kit v3.1 (Applied Biosystems) on an ABI3730 automated sequencer (Applied Biosystems). Sanger sequences were analyzed using Seqman (DNASTAR) and NovoSNP software [45].
Allele-speci c PCR to determine allele-phase con guration We used allele-speci c PCR ampli cation to determine cis/trans con guration of two VPS13C variants present in the same exon. For each mutation, an allele-speci c and a wild-type primer were designed in combination with a general second primer using the online Primer3 software [44] to amplify both the wildtype and mutant allele separately (  [44]. Native barcoding and adapter ligation were carried out using the Native Barcoding Expansion 1-12 (EXP-NBD104; Oxford Nanopore Technologies, ONT) in conjunction with the Ligation Sequencing Kit (SQK-LSK109; ONT).
VPS13C amplicons from 11 individual samples were pooled equimolar and sequenced on a MinION, using a single FLO-MIN106 ow cell (ONT). In total 900Mb of data was generated (~650K reads). Base calling and barcode de-multiplexing of the raw data was performed with Guppy (v.3.2.2). Further analysis was performed with a pipeline integrated in GenomeComb [34]. Alignment to the hg38 reference sequence [48] was performed with minimap2 using the splicedhq preset [49]. Samtools was used for the conversion to BAM and sorting [50], which enabled visualization with IGV [43]. Single nucleotide variant calling and haplotyping was performed using longshot (v0.4.0) [51].

Splice-site variant analysis on cDNA
Lymphoblast cells were treated with 100 μg/ml cycloheximide (Sigma-Aldrich BVBA) or equal amounts of dimethyl sulfoxide (Fisher Scienti c) for 4 h. The RNeasy procedure (RNeasy Mini Kit; Qiagen) was used to isolate and purify total RNA from lymphoblast cells as recommended by the manufacturer. Subsequently, total RNA was treated with DNase (Turbo DNase Kit; Ambion). cDNA synthesis of total RNA was performed primed with random hexamer primers using the SuperScript® III First-Strand Synthesis System for RT-PCR (Invitrogen), followed by a RNase H digestion to remove the RNA template from the cDNA:RNA hybrid molecule. To investigate the effect of VPS13C c.4166-8C>A on splicing, primers were designed using the online Primer3 software (Table S8, ) [44]. Exon skipping of exon 38 due to VPS13C c.4166-8C>A was investigated with a F primer located in exon 37 and a R primer located in exon 39. VPS13C cDNA was ampli ed in a total volume of 15 µl. Fragments were analyzed on a 2 % agarose gel, and sized using TrackIt™ 100 bp DNA Ladder (Invitrogen). Additionally, the VPS13C amplicons were puri ed and Sanger sequenced (as described above
The construct was transformed into BL21(DE3) competent E. coli (New England Biolabs) and expressed for 3h, at 28°C, to allow proper folding of the protein. Expressed recombinant protein was puri ed with Ni-NTA resin (Qiagen) followed by gel ltration chromatography using an ÄKTA™ pure system equipped with a HiLoad 26/600 Superdex 75 pg column (GE Healthcare). Purity of recombinant protein was assessed by Coomassie Brilliant blue staining. New Zealand white rabbits were immunized with different amounts of puri ed recombinant protein, and the serum was subsequently tested by Western blot for VPS13C immunoreactivity, using VPS13C knockout HeLa cell extracts as a negative control (Fig. S12). Polyclonal antibodies from the sera displaying the highest titer were a nity puri ed against the recombinant protein immobilized on NHS-activated Sepharose® 4 Fast Flow (GE Healthcare) according to the manufacturer's instructions. Antibodies were eluted from the column with 100mM Glycine pH 2.7 and immediately neutralized with 1M Tris-HCl pH 9.0. The eluted fraction was dialyzed against PBS for 48h, with three daily buffer changes. Finally, puri ed antibodies were concentrated to 1 mg/mL using a Vivaspin 15R, 30 000 MWCO hydrosart (Sartorius AG), and frozen in aliquots containing 50% glycerol.

Western blotting
Protein analysis performed in lymphoblast cells and brain lysates of patient carriers of >1 VPS13C variants and of family A. Cells and brain tissue were lysed in modi ed radioimmune precipitation buffer (RIPA: 1% sodium dodecyl sulfate, 150 mM NaCl, 0.5% Na-Doc, 1% NP-40, 50 mM Tris-HCL; pH, 8.0) supplemented with protease and phosphatase inhibitor (2x Complete Protease inhibitor cocktail and 1x Phosphostop phosphatase inhibitor cocktail; both from Roche). Lysates were sonicated on ice, cleared at 20,000 g for 15 min at 4°C, and supernatants were collected for immunoblotting. Protein concentrations were measured using a BCA Protein Assay Kit (Pierce™; Thermo Fisher Scienti c). Equal amounts of proteins were separated on a 3-8% Tris-Acetate gel (Life Technologies) and electro-blotted onto a polyvinylidene di uoride membrane (Hybond P; Amersham Biosciences). Membranes were blocked in 5% skimmed milk in PBS and probed with primary antibodies against VPS13C (polyclonal rabbit anti-human VPS13C (1:1,000; Novus Biologicals; NBP1-94043), polyclonal rabbit anti-human VPS13C (1:1,000; Novus Biologicals; NBP1-94044) or polyclonal rabbit anti-human (1:2,000; generated in house as described above)) and GAPDH (monoclonal mouse anti-human GAPDH (1:20,000; Genetex; GTX627408)). Immunodetection was performed with speci c secondary antibodies conjugated to horse-radish peroxidase in combination with the ECL prime chemiluminescent detection system (GE Healthcare) or the WesternBright Sirius detection system (Isogen Life Science). The band intensities were determined by quantifying the mean pixel grey values in a rectangular region of interest using Image Quant TL software (GE Healthcare Life Sciences) and subsequently normalized to GAPDH.

Preparation of cDNA constructs
The codon optimized coding sequence of wild type (WT) human VPS13C was purchased in a Gateway® compatible pDONR vector from GeneArt TM Gene synthesis (pDONR-VPS13C WT ; Life Technologies). The VPS13C missense variants p.Trp395Cys and p.Ala444Pro were introduced into the pDONR-VPS13C WT vector with QuickChange Site-Directed Mutagenesis. Sequence veri ed mutant entry clones were subcloned into the Gateway® pcDNA™-DEST40 Vector with a C-terminal V5-6x His tag (Thermo Fisher Scienti c) or into an in-house developed Gateway®-compatible pCR3 Vector with a C-terminal EmGFPtag. Sequences were veri ed by DNA sequencing and primers used for cloning are listed in Table S8 ().

Transfections
HeLa or SH-SY5Y cells were transfected using either X-tremeGENE™ 9 DNA Transfection Reagents (Sigma Aldrich), Lipofectamine LTX plus (Invitrogen), both according to manufacturer's protocol, or Polyethylenimine (PEI) according to an in house optimized protocol. Brie y, cells were seeded in a 6-well plate at 2.5x10 5 cells per well, 24 hours before transfection. On the day of transfection, cell medium was replaced by medium without antibiotics. 1.44µg plasmid DNA was diluted in 115µl Opti-MEM (Life-Technologies) and in parallel, 7.21µl PEI was diluted 115µl Opti-MEM. The diluted PEI was added to the DNA and mixed gently by vortexing. After 10 min of incubation at room temperature, the solution was added to the cells. All cells were evaluated 48h post-transfection.
Immunohistochemistry and live cell labeling

Image acquisition and analysis
Images were taken with a Zeiss LSM700 confocal microscope using either a 63x/1.40 Plan-Apochromat or a 40x/1.30 Plan-Neo uar objective. Filters, dichroics and scanning modes were set to exclude crosstalk between the different uorescence channels, pixel sizes were set according to the Nyquist sampling theorem and z-stacks comprising entire cells, acquired at optimal step sizes. To visualize co-localization of VPS13C with different organelle markers, Fiji software generated line-intensity plot pro les [52]. For quanti cation of VPS13C mislocalization, two researchers who were blind for the genotype visually scored cells.

Identi cation of compound heterozygous VPS13C missense variants in family A
In family A (Fig. 1a), genomic DNA was available of two affected siblings and their unaffected parents.
The index patient, P1 (II.2. Fig. 1a), rst presented with non-uent aphasia at age 42. A clinical neurological examination revealed non-uent aphasia, extrapyramidal signs consisting of hypomania, bradykinesia, gait disturbances, cogwheel rigidity and resting tremor (Table S1). Frontal disinhibition signs such as glabella re ex, snout re ex and palmomental re exes were present. Later in the disease course, myoclonus was present and behavioral symptoms became more apparent. The disease progressed rapidly and the patient died at the age of 54 years. Patient P1 (II.2, Fig. 1a) received a neuropathological con rmation of diffuse LBD, neocortical type. The affected sibling of patient P1 (II.1, Fig. 1a), presented at age 41 years with initial symptoms of anxiety and depression, combined with word nding di culties and an episode of visual hallucinations and delusion. The patient received a clinical diagnosis of unspeci ed dementia and died at the age of 47 years. There were no other familial antecedents of early-onset neurodegenerative brain diseases. Both parents were unaffected at advanced ages (> 80 years), showing that segregation of the disease in family A is consistent with autosomal recessive inheritance (Fig.1a).
We obtained WGS data of the two affected siblings of family A (Fig. 1a) Fig. 1a; Table 1). The multiple heterozygous variants in WISP3, RNF6 and CLLU1 were all in cis con guration in one parent and absent in the other parent. The VPS13C p.Trp395Cys and p.Ala444Pro alleles were absent in the control cohort of 664 individuals and belong to the 1% most deleterious substitutions in the human genome, indicated by a CADD_Phred score above 20 (Table 1) [42].
Additional VPS13C rare variants in the LBD families and patient cohort Targeted resequencing of the 86 coding exons and splice site junctions of VPS13C in the LBD patient and control cohorts identi ed 71 non-synonymous coding and splice site rare variants with a MAF ≤1% in VPS13C (Table S3, Fig. S1). All VPS13C variants were present in isoform 2, the largest VPS13C transcript (NM_020821.2), containing exons 6 and 7, and the main splice variant in brain, suggesting brain-speci c gene functions [53]. To study a potential role of these rare non-synonymous or splice VPS13C variants in LBD, we performed a SKAT-O analysis on the variants listed in  Fig. S2c). Further, we observed 10 patients and 7 controls with rare compound heterozygous variants in VPS13C (Table 1, Table S5) We phased the compound heterozygous VPS13C variants to identify carriers with trans con guration in line with recessive inheritance. Besides patient P1 (Family A, Fig. 1a), we identi ed 3 additional patients, P3, P4 and P5, carrying trans compound heterozygous variants in VPS13C (Table 1). Genotyping the two VPS13C variants and haplotype sharing analysis in relatives revealed that the child of patient P3 (Family B, Fig. 1a) carried one allele, p.Thr1218Ala, con rming that the compound p.Thr1218Ala/p.Ile2789Thr variants are in trans con guration in the affected parent (Fig. 1a. Fig. S2b). In the P4 patient, the compound p.Met2711Ile/p.Ile2789Thr variants are located in exon 61 and allele-speci c PCR analysis con rmed trans con guration (Fig. S3). In the P5 patient, long-read cDNA sequencing demonstrated trans con guration for p.Ala1687Val/p.Ser2904Leu and different haplotypes. In addition, we identi ed 3 patients and 5 controls with compound heterozygous variants in cis con guration (Table S5, Table S6).
Of the patient P10 in Family C (Fig. 1a), the two children were negative for both VPS13C variants, p.Ser963Gly/p.Ser2026Phe, present in the compound heterozygous parent. The p.Lys171Glu/c.4166-8C>A variants are present in patient P11 and three controls C3, C4 and C5. The relative high frequency in the Belgian cohorts of this VPS13C variant combination (4/1506; 0.27% observed; 0.00027% expected) ( Table S5, Table S6), we suspected that the two variants were located in cis in all four carriers. We were able to con rm the cis con guration by haplotype sharing analysis showing a common haplotype shared by all carriers (Fig. S2c). In the control C6, the VPS13C p.Met2764Ile/p.Val2765Leu variants are located in cis based on their joint occurrence in sequencing reads obtained via targeted resequencing of VPS13C (Fig. S4). Long-read cDNA sequencing in patient P12 and control C7 con rmed a shared haplotype and cis con guration for p.Ser963Gly/p.Ser2026Phe (P12) and p.Ser2282Phe/p.Arg3176Gly (C7).
To summarize, we identi ed 1 patient with homozygous and 11 patients with compound heterozygous rare variants in VPS13C. Four patients have a con rmed trans and 3 patients con rmed cis con guration of the compound variants in VPS13C. The 4 remaining patients have an unknown variant con guration. The frequency of patient carriers with bi-allelic variants, including the 4 non-phased patients is 1.07% (9/844) ( Table 1). All their compound mutated alleles are clustering in VPS13α involved in lipid transport, the putative WD40 involved in late endosomal/lysosomal localization or the pleckstrin homology involved in lipid droplet binding, domain of VPS13C (Fig. 1b) [54]. The clinical characteristics of all 9 patient carriers are summarized in Table S1. All were negative for mutations in the major PD genes and other genes associated with neurodegenerative brain diseases (Table S4), except for the PD patient P9, with the VPS13C p.Ile2789Thr/p.Ile3726Val un-phased alleles, who also carries the LRRK2 p.R1441C mutation. Of the 7 compound heterozygous carriers in the control group, 5 controls have a con rmed cis con guration of their VPS13C variants while the 2 remaining controls had no con rmed phasing (2/664; 0.30%).
Effect of VPS13C splice site variants on mRNA splicing.
Apart from the missense variants we also identi ed three splice site variants, c.448+7A>G, c.4056+3A>C and c.4166-8C>A, each partnering with a missense mutation, in control C2, patient P8 and in patient P11 and controls C3, C4, and C5 (Table1 , Table S5, Table S6). The c.4056+3A>C variant in the PD patient P8 was absent in 664 controls and predicted to affect the canonical splice donor site of exon 36 by three in silico splicing prediction programs (Table S7). Of patient P8 we had no lymphoblast cells or brain to evaluate the actual effect on mRNA splicing. The splicing predictions for c.4166-8C>A and c.448+7A>G were inconsistent between prediction programs (Table S7). We did not observed exon skipping of c.4166-8C>A when treating the lymphoblast cells of the 4 carriers with cycloheximide before isolating total mRNA (Fig. S5).

Reduced expression for VPS13C mutant alleles
To determine the effect VPS13C missense variants on transcript and protein expression we used qPCR and Western blot analysis of lymphoblast cells of DLB patient P1 (Family A, Fig 1a, II.2) and unaffected parents (Fig. 2a-c). No lymphoblast cells were available of the affected sib of P1 (Fig. 1a, II.1). We did not observe signi cant difference in VPS13C transcript expression for patient P1 and parents, compared to controls negative for rare VPS13C variants (Fig. 2a). However, endogenous VPS13C protein expression were reduced ~40% in the parent carrying the p.Ala444Pro allele (p =0.0618) and ~70% in the parent carrying the p.Trp395Cys allele (p = 0.0024). In the patient P1, carrying both VPS13C alleles p.Trp395Cys/p.Ala444Pro there is 90% reduction (p = 0.0002), compared to the control individuals ( Fig.   2b and 2c). We also investigated the protein expression in all patient and control carriers of compound heterozygous VPS13C variants and lymphoblast cells available. We observed a reduction in VPS13C expression in lymphoblast cells of patient P2, homozygous for p.Ala444Pro, patient P5 trans compound heterozygous for p.Ala1687Val/p.Ser2904Leu and patient P7 un-phased compound heterozygous for p.Thr766Ala/p.Leu1846Ser, compared to control individuals (Fig. 2d, e). None of the control carriers showed a reduction in VPS13C protein expression (Fig. 2d, e). We observed the largest reductions in patients P1 and P2, indicating that the p.Trp395Cys and p.Ala444Pro mutated alleles had the strongest effect compared to the other VPS13C alleles in lymphoblast cells (Fig. 2d). Brain tissue was available of patient carriers P1 and P3, and two unrelated control individuals, negative for rare VPS13C variants (MAF≤1%). In all studied brain tissues and brain regions (prefrontal cortex, temporal cortex, cerebellar cortex, hippocampus, substantia nigra and nucleus caudatus), VPS13C protein expression was abnormally reduced in patients compared to control individuals, with almost no VPS13C protein levels in patient P1 (Fig. 2f) of Family A (Fig. 1a), the discovery family we used for gene identi cation.

Subcellular localization of VPS13C
We transfected HeLa cells with wild type or mutant VPS13C variants (p.Trp395Cys or p.Ala444Pro) and investigated the effect of the missense variants on the subcellular localization of the protein. Wild type VPS13C localized to small organelles, whereas mutant VPS13C alleles, p.Trp395Cys or p.Ala444Pro, localized at larger cytosolic structures in the majority of cells (Fig. 3a-b). Triple immunostaining of V5 (VPS13C constructs with a C-terminal V5-6x His tag), with markers for late endosomes (Rab7) and lysosomes (Lamp1), revealed a late endosomal/lysosomal localization of wild type VPS13C. However, the late endosomal/lysosomal localization of VPS13C was lost when the VPS13C mutant alleles were present (Fig. 3c), which was independently con rmed with GFP-tagged VPS13C constructs and other markers (CD63, late endosomes; Lysotracker, lysosome; Fig. S6, S7). Moreover, we could con rm this subcellular localization of wild type and mislocalization of mutant VPS13C at late endosomes and lysosomes in human neuroblastoma SH-SY5Y cells (Fig. S8). Immunostaining in HeLa cells with markers for the ER (PDI; Fig. S9), cis-and medial-Golgi (Giantin; Fig. S10) and trans-Golgi (TGN46; Fig. S11) demonstrated no co-localization of wild type and mutant VPS13C with these organelles.
Pathological phenotype of compound heterozygous missense mutation carriers Patient P1 died at age 54, and we obtained and autopsy with 12 hours post mortem delay (PMD). We observed moderate frontotemporal atrophy with the superior temporal gyrus affected more than the medial temporal gyrus (Fig. 4a). In the midbrain, the zona compacta of the substantia nigra was very pale. The substantia nigra showed severe neuronal loss, most explicit in the lateral part of the zona compacta. Using the rating scheme for cerebrovascular lesions of Deramecourt and colleagues, no more than grade 2 of vascular pathological alterations could be scored [55]. The lateral occipitotemporal gyrus of patient P1 showed microspongiotic changes in the cortex (Fig 4c). The hippocampus and parahippocampal gyrus were affected with a moderate number of neuro brillary tangles, neuritic threads and dystrophic neurites (Fig. 4d), whereas other brain structures did not present tau pathology. 4G8 staining to detect β-amyloid pathology and TDP-43 and FUS staining for frontotemporal dementia (FTD) pathology were all negative. The α-synuclein staining showed an abundance (grade 3) of Lewy bodies and Lewy neurites in the frontal cortices, temporal neocortex, hippocampus, parahippocampal gyrus, amygdala, and in the pigmented nuclei of the mesencephalon, pons and medulla oblongata (Fig. 4e, i-k). Rare Lewy body pathology was found in the occipital cortex, neostriatum, hypothalamus. Based on our neuropathological ndings, the patient received a diagnosis of diffuse Lewy body disease, neocortical type [7,56].
Patient P3 died at age 64 and autopsy brain was obtained 8 hours following death. We observed severe atrophy of the temporal lobe while the parietal and occipital lobes were less severely affected (Fig. 4b).
Ventricular dilation was severe and most explicit in the temporal horn. The pars compacta of the substantia nigra was markedly thin and more rostrally, completely depigmented. The locus caeruleus in the pons was also severely depigmented. Histochemistry showed severe neuronal loss in the frontal and temporal cortices, and to a lesser extent in the parietal cortex with spongiosis and astrocytic gliosis (Fig.  4f). There was a severe atrophy of the hippocampus, amygdala and parahippocampal gyrus. The superior temporal gyrus was severely gliotic, with a thinning of the cortex to 1.5mm. The dorsomedial formation of the thalamus was affected with neuronal cell loss and gliotic changes and severe neuronal loss was observed in the substantia nigra. Neuro brillary tangles, neuropil threads and dystrophic neurites is found in every cortical sample examined, as well as the thalamus, neostriatum, corpus mamillare, medial geniculate body, substantia nigra, and reticular formation of mesencephalon, pons and medulla oblongata (Fig. 4g). The lesion load was compatible with that of stage VI (Braak & Braak) [57], and with a stage B3 of Montine et al. [56]. Classical and diffuse senile plaques were present in frontal, temporal, parietal and striatal cortices, as well as in the thalamus, putamen, medial geniculate body, corpus mamillare and in the molecular layer of the cerebellar cortex (Fig 4h). Cerebral amyloid angiopathy was mild [58] and the β-amyloid pathology was compatible with Phase 5 of Thal et al. [59], whereas the load of classical senile plaques was severe, compatible with CERAD stage 3 [60]. These ndings are compatible with AD neuropathological changes A3B3C3 [56]. Both the TDP-43 and FUS stainings were negative. α-synuclein staining showed a moderate amount of Lewy bodies in the hippocampus and parahippocampal gyrus, and a severe amount in the amygdala. Sparse Lewy bodies were found in prefrontal cortex, the substantia nigra, the pons and the medulla oblongata (Fig. 4l-m).
Based on these observations, patient P3 received a neuropathological diagnosis of LBD, amygdala predominant type, and of AD neuropathological changes A3B3C3 [7,56].

Discussion
VPS13C is a risk gene for PD patients identi ed in a meta-analysis of genome-wide association studies (GWAS) with an estimated odds ratio of 1.1 [16,61,62]. However, Lesage and colleagues identi ed homozygous and compound heterozygous PTC mutations in VPS13C in patients with a distinct form of early-onset parkinsonism, characterized by rapid and severe disease progression and early cognitive decline (Fig. 1b) [18]. The presence of Lewy bodies in the brainstem, limbic system and many cortical areas in one of the PD patients was reminiscent of diffuse LBD [18]. Later, a diagnostic WES data analysis of 80 persons with PD symptoms at early age and a negative family history, identi ed compound heterozygous variants affecting canonical splice sites in VPS13C in one early-onset PD patient (Fig. 1b) [19]. This patient presented overall with milder motor symptoms and disease progression, but presented with a rapid deterioration of cognitive functioning [19]. Shortly after, a large homozygous deletion comprising 50 exons of VPS13C was identi ed by WGS data of an sporadic patient with sensorimotor polyneuropathy and early-onset parkinsonism (Fig. 1b) [20]. This patient presented with normal cognitive functioning and a milder disease severity. Taken together, autosomal recessive loss of function mutations in VPS13C are rare and associated with early-onset PD with a high probability of cognitive deterioration and suggestive diffuse LBD pathology.
In contrast to the loss-of-function (LOF) mutations due to PTC mutations or deletions, we report for the rst time that some combination of rare missense variants (MAF ≤ 1%) in VPS13C can also contribute to DLB and PD by loss-of-function mechanism and as such mimicking recessive inheritance. Our ndings in family A triggered our interest in the role of rare missense variants in VPS13C and risk for LBD. In this family of healthy parents with two siblings affected with DLB at age early 41-42, we aimed at identifying the underlying genetic etiology using WGS. Taking that the pedigree suggested a potential recessive inheritance pattern, we searched for multiple rare variants in the same gene. The results indicated that rare compound heterozygous missense variants could explain the genetics in family A, with the patients combining 2 rare variants, p.Trp395Cys/p.Ala444Pro, inherited from their parents.
To expand our ndings in family A, we screened the complete VPS13C coding region for multiple rare variants in the Belgian LBD cohort comprising DLB and PD patients and in controls. We observed a signi cant association (p = 0.0233) of rare (MAF ≤ 1%) variants and LBD patients. The full burden test including all genetic variants in VPS13C was not signi cant (p = 0.175). VPS13C is likely to contain deleterious, protective and null variants. Recently, a VPS13C haplotype including the common (MAF > 1%) variants p.R153H-p.I398I-p.I1132V-p.Q2376Q was associated with a reduced risk for PD (p = 0.0052, odds ratio = 0.48, 95% con dence interval = 0.28-0.82) [63]. The same study, which included 1.567 late-onset PD patients, did not observe a statistically signi cant burden of rare (MAF ≤ 1%) VPS13C variants in PD, neither bi-allelic, non-synonymous coding and splice site variants in VPS13C in patients [63]. However, this study only included compound heterozygous variants with a MAF < 0.1% in late-onset patients, while we included variants with a MAF < 1% in both early-and late-onset patients. An independent study, investigating 4476 sporadic PD patients (mean AAO 60 years) and 5140 healthy control individuals reported a signi cant association (p = 0.002296) of rare (MAF ≤ 1%) VPS13C variants in PD [64].
In our Belgian LBD cohort, we identi ed 5 patients (4 DLB and 1 PD) with homozygous or trans compound heterozygous rare variants in VPS13C, including the patient of family A. We also observed 4 patients (1 DLB and 3 PD) with compound heterozygous variants in VPS13C with unknown phase (9/844; 1.07%). In the control cohort, we observed two carriers (2/664; 0.30%) of compound heterozygous VPS13C variants of unknown phase (Table S6). We were unable to phase the VPS13C alleles in the 4 patients and 2 controls ( Table 1, Table S6). DNA of relatives for segregation or biomaterials for RNA isolation were not available for the carriers P8, P9 and C1. Long-read ONT cDNA sequencing, could not call the missense variants in P8 and P9. In C2, the VPS13C c.448 + 7A > G splice site variant is present downstream of exon 6 of VPS13C isoform 2 (NM_020821). We were not able to amplify VPS13C amplicons, containing exon 6 or 7 speci c for isoform 2, from cDNA derived of lymphoblast cells, preventing phasing of c.448 + 7A > G/p.Ala1687Val and evaluation of the effect of c.448 + 7A > G on splicing. We con rmed cis con guration of the compound heterozygous alleles in 3 patient and in 5 control carriers. This cis con guration is not tting recessive inheritance and we excluded the carriers from further analyses in this study.
DLB patients P1 with compound heterozygous variants p.Trp395Cys/p.Ala444Pro, P2 with homozygous p.Ala444Pro and P3 with p.Thr1218Ala/p.Ile2789Thr presented with a severe disease progression (Table   S1). Neuropathological examination of patient carriers P1 and P3 indicated Lewy bodies in multiple brain regions ( Fig. 4i-n). Since there was also extensive AD pathology in patient P3, there is a small likelihood that the pathological ndings are associated with a DLB clinical syndrome. Nevertheless, both our clinical and neuropathological data of VPS13C mutation carriers supports the phenotype of most VPS13C patient carriers reported to date, including an early-onset age, severe disease progression and the co-occurrence of parkinsonism and dementia (Table 1, Table S1, ) [18][19][20].
Our genetic data suggest an enrichment of homozygous or compound VPS13C mutation carriers in patients compared to control though not signi cant (p = 0.1258), possibly because of the small numbers in the Belgian cohorts. We estimated the number of subjects needed for an adequate study power, taking into account the low frequency of compound carriers, showed that a minimum of 4200 subjects (~ 2100 per patient/control cohort) would be required to reach 80% power. Replication of PTC mutations and missense variants in VPS13C in larger cohorts is necessary as well as the con rmation of trans con guration of the compound heterozygous variants to identify the carriers that can contribute to recessive inheritance.
The variability in onset age and the presence of potential compound heterozygotes in the control group can be explained by variable penetrance of the VPS13C mutant alleles might be explained by variability of loss of protein expression and functioning of the different mutant VPS13C alleles. In the patients P1 (family A) and P2, with bi-allelic p.Trp395Cys and/or p.Ala444Pro (Table 1, Fig. 2d), was the reduction of VPS13C protein expression in lymphoblast cells the most severe. Both patients developed the disease at very early age (40-42 years) and had a marked severe disease progression (Table S1). In brain tissue of the two patients, P1 and P3 (family B), with con rmed trans compound heterozygous VPS13C variants, VPS13C protein expression was remarkably reduced compared to control individuals, with the strongest reduction observed for patient P1 (Fig. 2f). The VPS13C mutant alleles of DLB patient P3 were present in one younger sibling II.4 (Family B; Fig. 1b). This sibling's age 62 is close to the onset age of the index patient P3 with a current age of 61 indicating that this sibling is still at risk for disease.
The human VPS13 family consists of four proteins, VPS13A/Chorein, VPS13B, VPS13C and VPS13D, with all family members having a strong homology to yeast Vps13. Yeast studies have suggested that Vps13 may have a role in lipid exchange between organelles and showed that yeast mutants lacking Vps13 cause defects in mitochondrial membrane integrity [65,66]. In mammalian cell models, VPS13C was suggested to partially localize to the mitochondrial outer membrane and silencing VPS13C results in a lower mitochondrial membrane potential, mitochondrial fragmentation and increased respiration rates [18]. However, recent research showed that human VPS13C functions as a tether between the ER and late endosomes/lysosomes, and between the ER and lipid droplets, enabling transport of glycerolipids between membranes [54]. We con rmed the localization of wild type VPS13C at late endosomes and lysosomes (Fig. 3, Fig. S6, Fig. S7, Fig. S8). Overexpressing wild type or mutant VPS13C, containing p.Trp395Cys or p.Ala444Pro, in HeLa and SH-SY5Y cells demonstrated that the late endosomal/lysosomal localization of VPS13C is completely lost in mutants (Fig. 3, Fig. S6, Fig. S7, Fig.  S8). Surprisingly, both variants are located in the VPS13α domain and not in the putative WD40 modules responsible for endosomal/lysosomal localization (Fig. 1b). Because these variants are nearby the FFATmotif required for interaction with the ER, they may directly affect the localization to ER-contact sites as well. Overall, the different domains might be required for the protein's structural stability, needed for its localization to ER-late endosome/lysosome and ER-lipid droplet contact sites. Variants within the VPS13α domain may overall negatively affect the stability of the protein thereby affecting its localization. Besides p.Trp395Cys or p.Ala444Pro, we identi ed 12 alleles in compound heterozygous patient carriers mimicking recessive inheritance, including 11 missense and one splice site variants ( Fig. 1b; Table 1), awaiting further functional investigation to estimate their pathogenicity.
Loss-of-function mutations in the other human VPS13 genes are associated with different recessive neurodegenerative or neurodevelopmental disorders. Mutations in VPS13A or Chorein cause a progressive neurodegenerative disorder, chorea-acanthocytosis, primarily characterized by chorea and red blood cell acanthocytosis, but may also include parkinsonism, dystonia frontal and cognitive impairment [67]. VPS13B is associated with Cohen syndrome, a clinically variable syndromic neurodevelopmental disorder with symptoms of microcephaly, intellectual disability and motor delay amongst others other features [68] whereas mutations in VPS13D cause a form of cerebellar ataxia with spasticity, which also appears to lead to mitochondrial dysfunction [69].

Conclusions
Overall, our results indicate that homozygous and compound heterozygous variants in VPS13C are associated with increased risk for LBD. We identi ed missense variants, i.e. p.Trp395Cys and p.Ala444Pro in VPS13C, with loss of functional protein con rming their pathogenicity. Understanding the contribution of the different mutated VPS13C alleles to the genetic etiology of LBD, needs additional research.

Availability of data and materials
All data relevant to this study are included in the research paper or added to the supplementary information. The corresponding author will share additional information upon reasonable request.

Competing interests
None of the authors reported personal competing nancial interests.  compound heterozygous variants with a MAF of ≤ 1%, observed in the families and in the DLB and PD cohorts. The VPS13Cα domain is involved in transport of glycerophospholipids, the putative WD40 modules contain the binding site for late endosomes/lysosomes and the DH-like (DHL)-pleckstrin homology (PH) domains is the lipid droplet-binding region of VPS13C [54]. We showed published PTC mutations and deletions below the VPS13C protein, with at the left rst author name and year of publication [18][19][20].  and diffuse senile plaques (4G8 stain). i-k. α-synuclein pathology of patient P1 in (i) the frontal cortex, (j)