Abstract

   Parkinson’s Disease (PD) is driven by pathological aggregates of
   alpha-synuclein (αSyn), whose formation is facilitated by impaired
   glycosphingolipid metabolism via acidic glucocerebrosidase (GCase). We
   investigated glucosylceramide (GlcCer) accumulation in human, mouse,
   and cellular PD models. Lipidomic analyses revealed elevated plasma
   GlcCer, especially GlcCer24:1, and a shift in phosphatidylcholine (PC)
   species in PD patients. PD patient skin fibroblasts accumulated more
   GlcCer under lysosomal stress. GlcCer and sulfatides (SHexCer) were
   increased in Pink1^−/−SNCA^A53T PD mouse brains, and HT22 neurons
   exposed to preformed αSyn fibrils accumulated GlcCer and ceramides.
   GlcCer24:1 enhanced fibril toxicity, but had no direct or indirect
   effect on G-protein coupled receptors. RNAseq of GlcCer24:1-treated
   dorsal root ganglion neurons showed upregulation of glycolipid response
   genes, similar to pathogen-related signaling. These data indicate
   extracellular GlcCer is elevated in PD and triggers innate immune
   responses in sensory neurons.

   Subject terms: Parkinson's disease, High-throughput screening,
   Transgenic organisms, Metabolomics, Next-generation sequencing,
   Microscopy, Diagnostic markers

Introduction

   The pathology of PD is caused by deposits of oligomeric forms of
   alpha-synuclein (gene SNCA, αSyn)^[48]1,[49]2. Formation of αSyn
   fibrils is precipitated by glucosylceramides or their
   glucosylsphingosine metabolites^[50]3–[51]5, which per se can form
   twisted ribbon, amyloid-like structures^[52]6. Oligomers and fibrils of
   αSyn spread from cell to cell via extracellular vesicles^[53]7 or
   directly via tunneling nanotubes^[54]8 or prion-like cell-to-cell
   transfer^[55]9. The primary and probably only pathway for removal is
   via autophagolysosomal degradation^[56]10–[57]13. If it fails and
   lysosomes are overloaded with αSyn^[58]14, or if the lysosomal pH and
   calcium gradients decline under the load of accumulated lipids,
   lysosomal membranes become leaky^[59]15. The resulting release of
   proteases into the cytoplasm leads to autoproteolysis if it is not
   prevented by lysophagy, i.e. autophagy of damaged
   lysosomes^[60]15,[61]16. Lysophagy can prevent cell death to some
   extent but eventually leads into a vicious cycle unless overloaded or
   damaged lysosomes are expelled as exosomes for temporary relief^[62]17,
   but it results in further spreading of αSyn^[63]18,[64]19. Accumulation
   of αSyn in lysosomes interferes with the catalytic activity of
   lysosomal enzymes, in particular glucocerebrosidase 1 (gene GBA1,
   GCase), which is required for degradation of glucosylceramides
   (GlcCer). In turn, GCase malfunctioning exacerbates deficits of
   αSyn-degradation^[65]3,[66]20–[67]22. Accordingly, mutations in genes
   involved in glycosphingolipid transport such as ATP10B^[68]23 or
   degradation such as VPS35^[69]24,[70]25 increase the risk and severity
   of PD. Particularly mutant GBA1 accelerates αSyn
   pathology^[71]26,[72]27. It is still not completely understood how
   heterozygous PD-associated GBA1 mutations lead to a strong increase of
   the risk for developing PD^[73]28,[74]29 and the severity of PD, in
   particular PD-associated dementia. Homozygous “loss-of-function” GBA1
   mutations are causative for Gaucher disease^[75]30. Such GBA1
   mutations, but not the most frequent PD-associated E326K mutation, lead
   to protein misfolding and hence early degradation, or failure of
   transport and delivery to the lysosomes^[76]31, which depends on the
   shuttle protein, lysosome integrated membrane protein 2
   (LIMP2)^[77]32,[78]33. However, misfolding or substantial loss of GCase
   protein expression or loss of catalytic activity is mostly not seen in
   heterozygous carriers (as in PD), and mechanistically, the
   bidirectional pathology of dysfunctional GCase with αSyn is only
   partially understood^[79]34. It was hypothesized that mutant GCase
   directly precipitates αSyn oligomerization, that accumulation of GCase
   substrates, mostly GlcCer and other glycosphingolipids, drive αSyn
   oligomerization, that GCase deficiency results in a loss of lysosomal
   function and capacity for clearance and turnover of αSyn, that
   increased oligomeric αSyn in the lysosomes impairs membrane targeting
   and hence, activity of GCase, and that GCase substrates in turn shield
   αSyn oligomers from degradation^[80]21,[81]35, all finally leading to
   lysosomal leakage and impairment of mitochondria^[82]36. High levels of
   mutant misfolded GCase protein may also evoke an excessive
   unfolded-protein-response in the endoplasmic reticulum^[83]37 or
   saturation of the ubiquitin–proteasome pathway^[84]38, and defective
   generation of extracellular vesicles^[85]39. It is likely that multiple
   mechanisms are involved in the bidirectional GCase-αSyn pathology in
   PD. Dopaminergic neurons of the substantia nigra are particularly
   vulnerable to αSyn oligomers, and their death manifests as typical PD
   motor function deficits. Even earlier, but often unrecognized, sensory
   and autonomic neurons succumb to αSyn and lipid overload, manifesting
   in sensory loss, chronic muscular, neuropathic and visceral pain and
   headache^[86]40,[87]41. We have shown that sporadic PD patients have
   increased plasma concentrations of glucosylceramides, which was
   associated with a loss of thermal perception, mechanical
   hypersensitivity in Quantitative Sensory Tests, and high pain
   ratings^[88]42. The results suggested that accumulation of
   glucosylceramides may be common phenomenon in PD, not restricted to the
   about 5–15% PD patients carrying a pathogenic GBA1 mutation^[89]28.

   To obtain further insight into the pattern of glycosphingolipids and
   their impact on other lipids we used a back-translational approach and
   assessed lipidomic patterns in human PD plasma versus controls, patient
   derived primary fibroblasts, brain tissue of PD mice carrying a double
   PD-causative Pink1 deletion plus SNCA mutation (SNCA A53T) and in HT22
   hippocampal neurons stimulated with preformed αSyn fibrils, and then we
   investigated effects of GlcCer in vitro on i) G-protein coupled
   receptors (GPCRs), ii) handling of preformed aSyn fibrils and iii)
   transcriptomic changes in vulnerable primary neurons.

   GlcCer or HexCer were increased in all PD species/sites that were
   examined but GlcCer did not affect the uptake of αSyn fibrils or
   function of GPCRs, the latter suggested to be a target of
   glucosylsphingosines (GlcSph). Instead, exposure of primary sensory
   neurons to GlcCer 24:1 resulted in a transcriptomic switch reminiscent
   of cell responses to pathogen-derived glycolipids and suggests that
   extracellular GlcCer may mimic pathogenic threats. Since sensory
   neurons are a site of early premotor manifestations of PD and a source
   of αSyn spreading the observed mechanism may be relevant to the
   progression of the disease and potentially susceptible to therapeutic
   intervention.

Results

Increased plasma glucosylceramides in PD patients

   In previous studies, we^[90]42 and others^[91]43,[92]44 have observed
   increased plasma concentrations of ceramides and glucosylceramides or
   glucosylsphingosines in patients with idiopathic PD who were not
   carrying GBA1 mutations that are associated with mal-folding or
   dysfunction of the GBA1 gene product, lysosomal glucocerebrosidase
   alpha (GCase). To gain further insight into PD-associated ceramide
   pathology we explored and compared lipidomic patterns in plasma (Fig.
   [93]1) and primary fibroblasts of PD patients (Fig. [94]2), in brain
   tissue of PD-mice (Fig. [95]3), and in HT22 mouse hippocampal neurons
   loaded with αSyn preformed fibrils (Figs. [96]4 and [97]5).

Fig. 1. Plasma lipidomic analyses in PD patients versus healthy controls
(HC).

   [98]Fig. 1
   [99]Open in a new tab

   Patients (n = 16 f, 34 m) and controls (n = 25 f, 25 m) were
   age-matched <60-70+ years old at the time of blood sampling.
   Demographic details are shown in Supplementary Table [100]1. A Volcano
   plots show the Log2(fold change) on the x-axis versus the negative
   Log10 of the t-test P-value on the Y-axis. Increased lipids in PD are
   in red, reduced in blue. Lipids with statistically significant P-value
   but below the threshold for fold change are in orange. The Volcano
   plots reveals changes in FA, HexCer and PC which are detailed in (B–D).
   Please note that hexosylceramides (HexCer) represent GlcCer and GalCer.
   Hex2Cer mostly LacCer. B Fatty acids (FA) female and male PD patients
   and controls. C Hexosylceramides obtained by targeted and untargeted
   (TOF) lipidomic analyses. D Phospahtidylcholines (PC). The line is the
   mean, the whiskers show the SD. Each scatter is a subject. Data were
   submitted to 2-way ANOVA and subsequent posthoc analysis for each lipid
   species using a adjustment of alpha according to Sidak. *P < 0.05,
   **P < 0.01, ***P < 0.001, ****P < 0.0001. CAR carnitines, CER
   ceramides, CE cholesterol ester, DG diglycerides, FA fatty acids,
   HexCer hexosylceramides, LPC lysophosphatidylcholines, LPE
   lysophosphatidylethanolamines, LPG lysophosphatidylglycerols, LPI
   lysophosphatidylinositols, PC phosphatidylcholines, PE
   phosphatidylethanolamines, PD phosphatidylglycerols, PI
   phosphatidylinositols, SM sphingomyelins, ST sterols, TG triglycerides,
   UbiQ ubiquitin, –O ether bound.

Fig. 2. Ceramides and hexosylceramides in PD and HC primary fibroblasts upon
stimulation with pimozide.

   [101]Fig. 2
   [102]Open in a new tab

   Primary fibroblast cultures were obtained from 3 mm skin biopsies of
   the lower leg from n = 12 (9 f, 3 m) HC and n = 13 (4 f, 9 m) PD
   patients. Sphingolipids were analyzed by targeted LC-MS/MS from 2.5
   ×105 primary human fibroblasts per sample, and concentrations in ng/ml
   were auto-scaled to have an common mean and variance of 1 (Z-score).
   Sub-confluent cultures were stimulated with 12.5 µM pimozide (PIM
   12.5 µM = 1/5^th of EC50) or vehicle (1:10000 DMSO) and harvested at
   24 h. A Box/scatter plots of sphingolipid z-scores. Pimozide treatment
   raised Cer 24:1 and GlcCer’s and reduced sphingosine (SPH d18:1)
   predominantly in PD-PHF. The box shows the interquartile range, the
   line is the median, whiskers show minimum to maximum, scatters are
   individual subjects. Statistics: 2-way ANOVA and subsequent t-tests
   with adjustment of alpha according to Sidak for the between subject
   factor (4 groups) *P < 0.05, **P < 0.01, ***P < 0.001. B Score plots of
   a canonical discrimination analyses using sphingolipid concentrations
   as input. Clusters of vehicle treated PHF are overlapping, but pimozide
   treated PD-PHF differ from HC-PHF. C Paired analysis of for the most
   abundant GlcCer 16:0 and GlcCer24:1. PD fibroblasts show a stronger PIM
   evoked increase. D Violin plots reveal stronger variance of
   sphingolipid levels in PHF of PD patients than controls. Data as in
   (A).

Fig. 3. Ceramides and hexosylceramides in PD mouse brain.

   [103]Fig. 3
   [104]Open in a new tab

   A Ceramides and hexosylceramides in brain tissue (cortex, subcortex,
   midbrain) in Pink1^−/−SNCA^A53T and Sv129FVB wildtype control mice (age
   50–60 weeks) analyzed by targeted LC-MS/MS analysis. Each scatter is
   one mouse. Data were submitted to 2-way ANOVA using the between subject
   factor “genotype” and the within subject factor “ceramide”, followed by
   posthoc t-tests with adjustment of alpha according to Sidak for
   genotype. *P < 0.05, ****P < 0.0001. B Ceramides and hexosylceramides
   in brain tissue (cortex, subcortex, midbrain) in Pink1^−/−SNCA^A53T and
   Sv129FVB wildtype control mice (age 50–60 weeks) analyzed by untargeted
   UHPLC-MS/MS lipidomic screen. Data were analyzed using 2-way ANOVA with
   the between subject factor “genotype” and the within subject factor
   “ceramide”, followed by posthoc analysis for each lipid using the false
   discovery rate for adjustment of alpha. Heatmaps of top 50 regulated
   lipids are presented in Supplementary Fig [105]S3.

Fig. 4. Preformed alpha synuclein fibrils (αSyn-PFF) increase ceramides in
HT22 cells.

   [106]Fig. 4
   [107]Open in a new tab

   A Ceramide and hexosylceramide concentrations in picograms per 100,000
   HT22 mouse hippocampal neurons as assessed by targeted LC-MS/MS
   analysis. Cells were seeded in the absence or presence of PFF at
   2 µg/ml or 6 µg/ml in culture flasks and grown for 48 h. Control cells
   were treated with the respective volume of DMSO or were left untreated
   (naïve HT22). Each group consisted in 8 replicates. Data were submitted
   to 2-way ANOVA and subsequent posthoc t-test using an adjustment of
   alpha according to Sidak for the between subject factor “treatment”
   (i.e. 4 groups). *P < 0.05, **P < 0.01. High PFF increased all
   ceramides. B Heatmap of top 50 regulated lipids in HT22 mouse
   hippocampal neurons as assessed by UHPLC-MS/MS lipidomic analysis.
   Cells were treated as explained in (A). AUC/IS values were square root
   transformed and auto-scaled to have a common mean and variance of 1
   (z-scores). Lipids were clustered according to Euclidean distance
   metrics using the Ward method. C Principal Component Analysis of HT22
   lipidomic data showing a 3D-score scatter plot of the first three PC.
   Each scatter shows one sample. The circles are 90% confidence
   elipsoids.

Fig. 5. Cellular uptake of preformed alpha synuclein fibrils (αSyn-PFF) in
HT22 cells.

   [108]Fig. 5
   [109]Open in a new tab

   Immunofluorescent studies show the uptake of AF555-labeled αSyn
   preformed fibrils in HT22 mouse hippocampal neurons. Pretreatment with
   GlcCer24:1 (1 µM) had no significant impact on the PFF-AF555
   fluorescence uptake in HT22 cells (pink). Wheat germ agglutin WGA-AF488
   (yellow) is used as counterstain of plasma membranes (early time point)
   and endolysosomes (late time point). Hoechst-33342 is used as nuclear
   stain. Additional zoom-in images are shown in Supplementary Fig.
   [110]S5.

   Plasma lipidomic studies of 50 PD patients versus 50 age-matched
   healthy controls (HC) revealed increased hexosylceramides in PD,
   particularly GlcCer24:1 (Fig. [111]1A, Supplementary Fig. [112]S1),
   stronger in male than female PD patients (Fig. [113]1C), whereas a
   number of mono- and poly-unsaturated fatty acids were reduced without
   differences between sexes (Fig. [114]1B). In addition,
   phosphatidylcholine species (PC) shifted from long chain
   poly-unsaturated PC species to shorter mono-unsaturated PCs (Fig.
   [115]1D) without overall change of summed PCs, which agrees with a
   recent study where a set of serum lipids including PC species allowed
   for a better prediction of PD progression/outcome than age, sex and
   risk gene LRRK2 mutation status^[116]45. PCs are connected with GlcCer
   via the phospholipid translocase ATP10 (ATP10A, 10B and 10D), which
   flips PC in exchange with GlcCer in the plasma membrane, and
   contributes thereby to membrane dynamics^[117]23,[118]46,[119]47, and
   is considered as PD risk gene^[120]23. A recent study shows that ATP10
   rather than GCase accounts for increased GlcCer in PD^[121]48.

   Plasma lipidomic data and metadata are available at
   [122]https://www.ebi.ac.uk/biostudies with the accession number
   S-BSST1880.
   [123]https://www.ebi.ac.uk/biostudies/studies/S-BSST1880?key=a0dc6f53-0
   365-410a-96e3-d83e51a04282.

Increased glucosylceramides in PD primary fibroblasts upon stimulation with
pimozide

   GlcCer are generated in the ER, from there transported and inserted
   into the plasma membrane and membranes of organelles, and they are
   degraded in the lysosomes^[124]49. Considering the autophagolysosomal
   pathophysiology of PD the most likely mechanism underlying increased
   GlcCer in plasma is mal-degradation in the
   lysosomes^[125]12,[126]14,[127]50,[128]51. To study this aspect we used
   primary patient derived fibroblasts that were stimulated with
   pimozide^[129]52.

   Patient derived primary fibroblasts are considered a viable model for
   studying neurodegenerative changes due to their metabolic and
   biochemical relationships with neurons^[130]53. In particular, the
   phenotype of skin fibroblasts of late-onset sporadic PD subjects
   recapitulated features of iPSC derived midbrain dopamine neurons from
   the same patient, including PD-associated changes in morphology,
   mitochondrial function, and autophagy^[131]54. Fibroblasts of PD
   patients express about 5-10 fold more αSyn^[132]55 and are used for
   early diagnosis of prodromal symptoms^[133]56. Further, sphingolipid
   changes found in PD fibroblasts precipitated the aggregation of
   αSyn^[134]57. Primary patient derived fibroblasts were therefore chosen
   to further study GlcCer alterations.

   To mimic PD-like lysosomal dysfunctionality, primary human fibroblasts
   (PHF) of 13 PD patients and 12 HC were stimulated with pimozide, which
   is a lysosome-tropic, antipsychotic drug. As a weak base (pKa 8.36)
   pimozide is trapped in lysosomes but does not cause major lysosomal
   alkalinization and therefore does not cause complete disruption of
   lysosomal functions like chloroquine^[135]52. At high concentrations,
   its accumulation in lysosomes can cause lysosomal membrane
   permeabilization and lysophagy^[136]15 but at lower concentrations it
   mainly interferes with lysosomal (sphingo)-lipid metabolism that
   manifests in an increase of ceramides and GlcCer in cell
   lysates^[137]15,[138]52. It was therefore an ideal stimulus to assess
   subtle differences of lysosomal functions in PD versus control
   fibroblasts.

   The IC50 in primary human fibroblasts (PHF) of pimozide-evoked cell
   death was 90-100 µM (Supplementary Fig. [139]S2). To assess effects on
   lipid homeostasis, PHF were stimulated for 24 h with 1/8^th of the IC50
   (i.e. 12.5 µM). At baseline, there was no difference between PD and HC
   fibroblasts in any of the analyzed sphingolipids (Fig. [140]2).
   Pimozide increased ceramides and reduced sphingoid base sphingolipids
   (sphingosine, S1P) in both PD and HC fibroblasts in line with previous
   observations in tumor cells^[141]15, but the pimozide-evoked increase
   of glucosylceramides was stronger in PD fibroblasts (Fig. [142]2A, C).
   Canonical discrimination analyses using all sphingolipids as input was
   able to distinguish PD-PIM versus HC-PIM fibroblasts (Fig. [143]2B).
   Further, violin plots of Z-transformed sphingolipids revealed higher
   variability of PD fibroblasts as compared with HC fibroblasts (Fig.
   [144]2D). The results suggest higher vulnerability of PD-fibroblasts
   towards a pimozide-evoked raise of lysosomal pH possibly because PD
   fibroblasts have a higher lysosomal burden of αSyn^[145]55 and
   therefore lower capacity to adjust lysosomal hydrolase activity.

   Fibroblast sphingolipid raw data and metadata are available at
   [146]https://www.ebi.ac.uk/biostudies with the accession number
   S-BSST1881.
   [147]https://www.ebi.ac.uk/biostudies/studies/S-BSST1881?key=4e8ae53f-1
   7f0-4179-80ff-40392e418270.

Increased ceramides and hexosylceramides in PD mouse brain

   Plasma hexosylceramides are maintained and replenished from peripheral
   and central sources. The relative contribution is unknown. Elevated
   plasma levels in PD patients suggested disease-associated metabolic
   changes, but plasma levels represent the extracellular compartment and
   may not necessarily directly reflect brain tissue. Therefore, we used
   PD-mice to get direct insight into brain ceramides. As explained in the
   description of the PD mouse model, we opted for Pink1^−/−SNCA^A53T mice
   because they combine features of human PD including a premotor disease
   with sensory symptoms and development of spontaneous motor deficits.

   Targeted (Fig. [148]3A) and untargeted (Fig. [149]3B) lipidomic studies
   revealed increased ceramides in the brain of double mutant
   Pink1^−/−SNCA^A53T mice compared with wildtype Sv129-FVB control mice,
   in agreement with previous ceramide studies in Pink1^−/− mouse
   brain^[150]58. Importantly, Pink1^−/−SNCA^A53T mice had no clinical
   symptoms of PD motor disease at the time of tissue sampling at ~12
   months of age. Ceramides, hexosylceramides (GlcCer or GalCer) and
   sulfatides (SHexCer) were increased in Pink1^−/−SNCA^A53T brains
   whereas acetyl-HexCer (AHexCer) were decreased (Fig. [151]3B).
   Lipidomic analyses further revealed increased diacylglycerols (DG) and
   lysophosphatidylethanolamines (LPE) in Pink1^−/−SNCA^A53T brains
   (Supplementary Fig. [152]3). DGs are normally low in the brain, and
   high brain DGs have been suggested to indicate CNS disease^[153]59.
   LPEs are contained in Lewy bodies and likely precipitate αSyn
   aggregation, depending on C-chain length and saturation^[154]43.
   Lipidomic heat maps of candidate lipids also show a switch of PC
   species not equal but corresponding to the patterns in human plasma,
   mainly a decrease of long-chain PC-O and increase of shorter PCs,
   mostly with low saturation.

   Brain sphingolipid and lipidomic and metabolomic raw data and metadata
   are available at available at [155]https://www.ebi.ac.uk/biostudies
   with the accession number S-BSST1888.
   [156]https://www.ebi.ac.uk/biostudies/studies/S-BSST1888?key=6230dc37-d
   998-4eca-bf0f-3502f7164655.

Increased ceramides in HT22 neurons upon ingestion of preformed αSyn fibrils

   Mutant GBA1 and αSyn mutually increase neuronal damage likely
   converging on lysosomal collapse^[157]22,[158]50 and mitochondrial
   toxicity^[159]36. To mimic the αSyn part, HT22 mouse hippocampal
   neurons were exposed to active preformed αSyn fibrils, which were
   supplied via the culture medium. Using confocal live imaging microscopy
   it was confirmed that HT22 cells ingested PFF, thus creating PD-like
   HT22. HT22 cell were chosen as a model because of their
   well-established sensitivity to PD-inducing pro-oxidative agents,
   expression of dopamine receptors, clathrin endocytosis machinery and
   the observed reliable uptake of αSyn-PFF.

   Lipidomic analyses of such PD-like HT22 showed that the exposure and
   ingestion of αSyn fibrils increased ceramides and hexosylceramides. The
   higher dose of 6 μg/ml caused stronger changes than the low dose of
   2 µg/ml (Fig. [160]4A). αSyn fibrils alone had no effect on cell
   viability as assessed by WST-1 assays (Supplementary Fig. [161]4A, B),
   nor did it consistently affect polar metabolites. At high
   concentration, αSyn-PFF treatment resulted in an increase of LDH
   activity in supernatants in some cultures (Supplementary Fig.
   [162]S4C), suggesting damage of the cell membrane (cytosolic LDH) or
   lysosomal exocytosis (lysosomal LDH). Lipidomic analyses (Fig. [163]4B,
   C) revealed that ceramides were among the top upregulated lipids
   exposed to αSyn PFF. In agreement with mouse brain and human plasma,
   PFF ingestion was associated with an increase of short-chain PC but
   decrease of long-chain PC (Supplementary Fig. [164]S4C). In addition,
   heatmap analyses showed a strong increase of phosphatidylinositol (PI)
   species which had not been observed in mouse brain or human plasma.
   Score plots of a principal component analysis showed a distinction of
   PFF-treated versus not-treated cells, but no PFF dose-dependent
   difference.

   HT22 lipidomic and metabolomic data and metadata are available at
   [165]https://www.ebi.ac.uk/biostudies with the accession number
   S-BSST1897.
   [166]https://www.ebi.ac.uk/biostudies/studies/S-BSST1897?key=d2a20a65-a
   88f-44a4-9654-672a72f8a9ac.

No significant increase of PFF ingestion in the presence of GlcCer24:1

   It has been suggested that GBA1 deficiency exaggerates the exosomal
   spreading of αSyn^[167]60,[168]61 likely because the release of
   exosomes which are enriched in GlcCer^[169]62 is a compensatory
   mechanisms to get rid of excess GlcCer. Spreading also requires
   subsequent uptake in recipient cells. Therefore we studied in a series
   of confocal live imaging experiments if extracellular GlcCer 24:1
   enhanced the uptake or distribution of αSyn PFF in HT22 hippocampal
   neurons. The plasma membrane was visualized with fluorescent wheat germ
   agglutinin (WGA-AF488). During the observation time, WGA was taken up
   and moved to the lysosomes (Supplementary Fig. [170]S5 WGA co-staining
   with LysoID®). Hence, in the first images, WGA is localized mostly at
   the outer membrane (left panel in Fig. [171]5), whereas it
   co-segregates with αSyn PFF in the lysosomes at late images (right
   panel, Supplementary Fig. [172]S5 co-staining with LysoID®). The middle
   panel is a time point in between. GlcCer 24:1 treatment had no overt
   effect on the quantity of ingested αSyn PFF but the cell size was more
   variable in GlcCer 24:1 plus PFF-treated cell cultures and some cells
   had thin cell-to-cell contacts (Fig. [173]5 bottom row left image,
   Supplementary Fig. [174]S5 Zoom-in) and WST-1 assays showed that the
   combination of GlcCer 24:1 plus αSyn PFF for 48 h increased the
   percentage of non-viable cells (Supplementary Fig. [175]S4).

No activation of G-protein coupled receptors by GlcCer 24:1 and 18:1

   High levels of GlcCer in extracellular fluids (plasma) suggest that
   GlcCer get into contact with the outer membrane, membrane receptors,
   transporters or channels. We have previously observed that GlcCer in
   culture medium increased stimulated calcium fluxes in a fraction of
   neurons, but not per se without additional stimulus^[176]63. GlcCer are
   important components of membrane microdomains and hence involved in the
   regulation of membrane receptor insertion and function^[177]64,[178]65.
   Considering the importance of dopamine receptors in PD and previous
   reports showing activation of serotonin receptors with GlcSph we
   assessed putative direct effects of the top PD-associated GlcCer
   candidates in human plasma, GlcCer18:1 and GlcCer24:1, on G-protein
   coupled receptors via beta-arrestin screening and dynamic mass
   redistribution (DMR) assays. Compared with the positive control
   (carbachol, muscarinic GPCR), GlcCer18:1 and GlcCer24:1 had no effect
   on any of the screened GPCRs (Fig. [179]6A, B; Supplementary Fig.
   [180]S6 showing replicates) at a reasonable concentration range up to
   10 µM. At 10 µM, GlcCer24:1 increased beta-arrestin binding up to
   3-fold in HTLA cells with heterologous expression of OPRL1 (opioid
   related receptor, nociceptin/orphanin FQ). The result was reproducible
   in quadruple replicates (Supplementary Fig. [181]S6) but still weak
   compared with the positive controls.

Fig. 6. GlcCer18:1 and GlcCer24:1 do not activate heterologous expressed
G-protein coupled receptors.

   [182]Fig. 6
   [183]Open in a new tab

   A Beta arrestin-based screening of GPCR activity in a heterologous
   expression model in COS cells upon stimulation with GlcCer 18:1 versus
   vehicle at 1, 5 and 10 µM. The heatmap shows the mean of four
   replicates for candidate GPCR selected from an initial screen (330
   GPCRs, 8 neg and 4 × 2 pos controls). Details in Supplementary Fig.
   [184]1. B As in A but stimulation with GlcCer 24:1 C: Dynamic Mass
   Redistribution (DMR) analysis of candidate GPCRs stimulated with GlcCer
   18:1 or GlcCer 24:1 at 1 µM. DMR measures a ligand-induced shift in
   resonant wavelength in picometers. Three candidate GPCR were tested
   with GlcCer 18:1 (upper left) and 14 with GlcCer 24:1. Carbachol
   activation of muscarinic receptor CHRM1 was used as positive control
   (right Y-axis). MCHR1 melanin-concentrating hormone receptor 1, FFA3
   free fatty acid receptor 3 (GPR41), CXCR7 CXC-type chemokine receptor
   7, GPR4 G-protein coupled receptor 4, pH-sensing, GPRC5A orphan GPCR,
   alias: retinoic acid-induced protein 3 (RAI3), GPR77 C5a anaphylatoxin
   chemotactic receptor, GPR82 and GPR85 both orphan, P2RY8 P2Y
   purinoceptor, NPSR1 neuropeptide S receptor 1, GRM5 metabotropic
   glutamate receptor, GPPR gastrin releasing peptide receptor, GPR123
   adhesion GPCR, GPR143 ocular albinism type 1 (OA1), receptor for
   tyrosine, L-dopa and dopamine, BDKRB2 bradykinin receptor B2, AVPR2
   arginine vasopressin receptor 2, ADRA1B alpha-1B adrenoreceptor.

GlcCer 24:1 causes inflammatory transcriptional response in primary mouse
neurons

   The loss of motor functions in PD is often preceded with non-motor
   symptoms originating from manifestations of the disease in the
   peripheral somatosensory and autonomous nervous
   systems^[185]66,[186]67. The premotor phase may last for years and
   causes (among others) PD-associated sensory neuropathy^[187]68 with
   sensory loss and pain^[188]69–[189]71, and restless leg
   syndrome^[190]72. In “body-first” PD, αSyn spreading originates from
   the peripheral nervous system^[191]73–[192]75. Somatosensory neurons
   are therefore in many cases the site of the disease onset where the
   progression may still be responsive to disease-modifying intervention.
   We have shown previously that sensory loss and pain is associated with
   high GlcCer in plasma (PD patients) or DRGs and sciatic nerve tissue
   (mice) and that treatment of dorsal root ganglia (DRG) neurons with
   GlcCer 24:1 leads to an exaggerated calcium influx upon
   stimulation^[193]42,[194]63. Therefore, we used adult DRG neurons to
   gain further insight into the effects of GlcCer using mRNA sequencing
   to reveal transcriptomic changes that occur when DRG neurons are
   exposed to GlcCer 24:1 (Fig. [195]7).

Fig. 7. GlcCer24:1 of primary neurons triggers upregulation of genes involved
in GO “response to glycolipid”.

   [196]Fig. 7
   [197]Open in a new tab

   A Volcano plots of regulation of mRNA expression assessed by RNAseq in
   primary sensory neurons of the dorsal root ganglia from adult mice
   treated with 1 µM GlcCer24:1 versus vehicle. The x-axis shows the
   Log2(Fold change) of TMM-normalized counts. The y-axis shows the
   negative logarithm of the t-test P-value. The data are from n = 4
   cultures per condition. Upregulated genes are shown in red,
   downregulated in blue. B Candidate genes were selected according to the
   FDR-adjusted P-value and are sorted according to abundance and
   presented as TMM-normalized reads (trimmed mean of m values). C
   Candidate genes were submitted to GO and pathway enrichment analysis
   using DAVID, STRING and Panther consistently showing an enrichment of
   genes involved in “response to glycolipid”.

   RNAseq revealed an upregulation of a number of membrane-associated
   genes (GO CC: membrane) and overrepresentation of genes involved in the
   “response to glycolipids” (GO BP or similar terms such as “response to
   LPS”, “response to pathogens”). The results suggest that extracellular
   GlcCer can induce inflammatory responses similar to those induced by
   pathogens. The results agree with a recent study showing that
   glucosylceramide accumulation due to GCase deficiency caused macrophage
   activation and neuroinflammation^[198]76. Candidate genes, gene
   descriptions and GO studies are presented in a [199]Supplementary Excel
   file. RNAseq data have been deposited as GEO dataset with the
   provisional accession number GSE262573.

   To review GEO accession GSE262573.

   Enter token clytuucuxjupzmj into the box.

Discussion

   Lewy bodies are primarily composed of α-synuclein fibrils intertwined
   with lipids. Mutations in genes responsible for lipid degradation or
   transport, such as GBA1, ATP13a2 and ATP10b, have been identified as
   factors that increase the risk of PD^[200]23,[201]77. Although
   mechanistically complex, lipidomic research suggests that an
   accumulation of glycosylated sphingolipids may enhance the
   pathogenicity of αSyn^[202]3,[203]4,[204]78 by precipitating the
   formation of αSyn fibrils^[205]6. In the present study, we show that
   glucosylceramides are increased in the plasma of PD patients, in the
   brain of PD mice, in primary human fibroblasts of PD patients treated
   with pimozide and in HT22 hippocampal neurons after ingestion of
   preformed αSyn fibrils, together strongly suggesting that αSyn
   interferes with metabolic homeostasis and/or release of
   hexosylceramides. The predominant candidate was GlcCer 24:1. Except
   AHexCer in the mouse brain, all GlcCer (HexCer) species were increased
   in human PD, or PD-models. However, GlcCer 24:1 (and GlcCer 18:1 in
   some experiments) had little effect on cell viability, morphology,
   ingestion of αSyn, or GPCR activation which agrees with a recent study
   that shows high plasma GlcCer in association with PD but not in a
   causative manner. The authors concluded that GlcCer per se was not
   pathogenic^[206]48 but changes of GlcCer need to be considered in the
   context of more complex changes of lipids that are connected via
   metabolic and transport pathways. In particular, we and several
   previous studies observed changes in a number of phosphatidylcholine
   (PC) species^[207]43,[208]79–[209]82 whose specific functions are
   presently mostly elusive, but determine the secondary structure of αSyn
   oligomers^[210]83,[211]84 and are intricately linked to endolysosomal
   transfer and membrane insertion of GlcCer via
   flippases^[212]77,[213]79,[214]85. Nonetheless, alterations of HexCer
   are the most robust finding in PD and models of PD and our RNAseq
   studies have shown that GlcCer 24:1 treatment of primary neurons
   elicited a pro-inflammatory response. In support, a recent study in a
   PD fly model showed that GlcCer caused cellular immune
   activation^[215]76.

   Considering the close association of glucosylceramides with the risk
   and the course of PD we asked how elevated glucosylceramides might
   contribute to an unfavorable course of PD. We considered three putative
   mechanisms: (i) by activation or inhibition of GPCR signaling (based on
   studies with hexosylsphingosines^[216]86,[217]87), (ii) by facilitating
   αSyn uptake and (iii) by causing a glycolipid-like immune response. We
   found no evidence for the first two hypotheses but some support for the
   third.

   In β-arrestin-based GPCR screening assays, GlcCer 24:1 induced a
   dose-dependent increase of β-arrestin binding in OPRL1 positive clones
   (nociception/orphanin-FQ) up to 3-fold. However, the effect was weak
   compared with the positive control carbachol, which increased the
   β-arrestin signal up to 50-fold upon binding to muscarinic choline
   receptors (CHRM1). In DMR-based GPCR screening, GlcCer had no effect.
   We infer that GlcCer 24:1 (and 18:1) do not activate or inhibit GPCRs
   at concentrations which might be reached in extracellular fluids.
   Importantly, GlcCer 24:1 or 18:1 had no direct effect on dopaminergic
   receptors which could have been a putative disease-relevant site of
   action. It is of note that glucosylsphingosines (GlcSph), which are the
   deacylated (lyso-) form of glucosylceramides, were found to activate
   serotonin receptors^[218]88 and that glucosylsphingolipids may
   indirectly affect GPCR by changes of membrane composition and lipid
   rafts^[219]89.

   Both, GlcCer and GlcSph accumulate in brain and peripheral organs in
   Gaucher disease^[220]90, but GlcSph are preferred as “biomarkers” in
   Gaucher disease^[221]91 because GBA1 deficient macrophages (Gaucher
   cells) were reported to accumulate predominantly the
   lyso-glycosphingolipid^[222]92, which however does not confer more or
   less toxicity to the acylated or deacylated version for neurons in the
   context of neuronal type Gaucher disease and PD, but might explain why
   GlcSph has gained more attention^[223]93–[224]95 and is analyzed for
   newborn Gaucher screening^[225]96,[226]97. But GlcCer and gangliosides
   rather than GlcSph were found to precipitate αSyn oligomerization in
   acidic environments in vitro^[227]98.

   αSyn monomers normally bind to the surface of synaptic
   vesicles^[228]99,[229]100. Their oligomerization properties enable
   vesicle tethering but αSyn is not essential for synaptic vesicle
   cycling and release^[230]101. However, excess or mutant or
   phosphorylated αSyn adopts beta sheet rich conformations which are
   prone to oligomerization, further aggregation and fibril formation and
   spread like prion-like proteins^[231]9,[232]102,[233]103. αSyn uses
   several means of cell-to-cell spreading^[234]104,[235]105 including
   transsynaptic transfer from neuron to neuron, traveling through
   tunneling nanotubes^[236]8,[237]51,[238]106, packaging into and release
   of extracellular vesicles which can reach far distant target cells and
   are taken up endocytosis^[239]7,[240]107. Further, αSyn can be expelled
   via lysosomal exocytosis likely together with GlcCer thereby creating a
   hostile microenvironment^[241]108. GlcCer are also components of
   exosomal membranes^[242]109 and may assist in packaging of αSyn into
   exosomes^[243]110, and define the target cells^[244]111. Our experiment
   created a neuronal microenvironment with excess free non-exosome-packed
   αSyn fibrils (plus/minus GlcCer24:1) and high extracellular GlcCer, a
   situation that occurs when αSyn fibrils or aggregates are expelled from
   cells via exocytosis or lysosomal exocytosis. Because of the excess of
   PFF, all neurons ingested the fibrils, and our experiment did not allow
   for live-observation of αSyn-PFF spreading from one population of
   neurons to another. In extracellular fluids, GlcCer likely bind to
   albumin or other proteins, but the relative contribution of exosomal,
   albumin/protein-bound and free GlcCer is unknown. Under the
   experimental conditions of this study, GlcCer did neither increase the
   uptake of αSyn-PFF nor the accumulation in lysosomes. However, results
   may differ in mixed cultures that include immune cells and glia, and/or
   in compartmentalized cultures with localized preformed fibril (PFF)
   seeds.

   In myeloid immune cells, GlcCer have been described as ligands of
   Mincle (Clec4e)^[245]112,[246]113, which is an inducible C-type lectin
   and innate immune receptor of macrophages, neutrophils and dendritic
   cells that senses glycolipids of the mycobacterial cell wall (trehalose
   6,6’-dimycolate, known as cord factor)^[247]114,[248]115 and other
   glycolipid species derived from pathogens or damaged and dead
   cells^[249]116–[250]118. Mincle (Clec4e) was weakly expressed in our
   primary sensory neuron cultures (RNAseq studies). For comparison,
   Clec4e total read counts ranged from 11 to 70, whereas total read
   counts of marker genes for DRG neurons were in the range of 100–4000,
   e.g. TRPA1 ~100, TRPV1 ~300, Scn10a (Nav1.8) ~∼800, TAC1 (Substance P)
   ~4000, and relevant lipid-sensitive GPCRs were e.g. LPAR5 ~∼15, PTGER4
   (EP4) ~∼50, CNR1 (CB1) ~∼300, S1PR1 ~∼70 and S1PR3 ~∼600. Hence, Clec4e
   expression was low but still in a range where its activation may result
   in measurable effects. In line with this reasoning we found that
   treatment of primary sensory neurons with GlcCer 24:1 resulted in a
   transcriptional response similar to that observed by stimulation of
   cells with bacterial glycolipids. Gene ontology analyses of upregulated
   genes showed an enrichment of terms like “toll-like receptor
   signaling”, “positive regulation of innate immune response”, “pattern
   recognition receptor signaling”. The data suggest that high
   extracellular GlcCer may induce an innate immune response, which is
   adequate for the defence against mycobacterium but may contribute to
   neuroinflammation if it were to occur in the brain. The latter is
   supported by a study where GlcCer stimulated microglia phagocytosed
   living neurons^[251]119. We used primary sensory neurons because they
   are a site of early PD manifestations and sensitive to GlcCer but it
   needs to be assessed in future studies if GlcCer elicits similar innate
   immune responses in neurons of the CNS.

   C-type lectin activation leads to spleen tyrosine kinase (Syk) mediated
   downstream signaling involving phospholipase C, protein kinase C, MAP
   kinases and nuclear factor kappa B (NFκB). Considering that the most
   frequent loss-of-function mutation, GBA1 N370S, that causes Gaucher
   disease confers resistance to tuberculosis in a zebrafish
   model^[252]120 it is tempting to speculate that GlcCer-mediated
   Mincle/Clec4e activation is protective against pathogens but possibly
   harmful if it sustains a microglial attack of neurons. It has been
   shown that Syk inhibitors reduce neuroinflammation, but so far only in
   models of stroke or vascular brain injury^[253]121, not yet for Gaucher
   disease or GBA1-associated Parkinson’s disease. It is of note that a
   recent study suggested to add a Mincle activating adjuvant to
   αSyn-peptides to enhance the efficacy of active PD
   vaccination^[254]122. The vaccine consisted in a glucan-αSyn conjugate,
   i.e. not a lipid-conjugate but the strategy reduced fibril propagation
   in an in vivo synuclein-seeding-spreading model. Hence, although the
   molecular target of GlcCer remains unidentified its immune-stimulating
   effect might be beneficial if it could be harnessed to enhance fibril
   elimination.

   The present study has limitations. It focused on GlcCer in
   non-GBA1-associated PD, but it would have been valuable to include some
   confirmed GBA1 mutant patients and GBA1-mutant mice for comparison as
   “positive” control, particularly because we did not find quantifiable
   glucosylsphingosine species in our lipidomic analysis, which are
   supposed to be biomarkers for Gaucher disease. Although the frequency
   of PD-associated GBA1 mutations is low in Germany, it is a limitation
   that our patients were not genotyped for GBA1 or other PD-associated
   genes that affect lipid metabolism or transport. Hippocampal mouse HT22
   neurons and primary mouse DRG neurons are valuable models for PD
   research but human CNS neurons may respond differently to PFF or GlcCer
   which needs to be assessed in human neurons or organoids in future
   studies.

   In summary, we show that extracellular GlcCer is increased in PD and in
   PD models likely originating from defective lysosomal functions and the
   data suggest that extracellular GlcCer may induce an innate immune
   response in peripheral sensory neurons. Such immune activation may
   sustain a proinflammatory state but may also attract glia and immune
   cells to remove extracellular αSyn.

Methods

PD patients versus age-matched healthy controls

   PD patients, diagnosed using ICD10 diagnostic criteria for Parkinson’s
   Disease, were consecutively recruited from the Movement Disorders
   Department of Neurology of the University Hospital in Frankfurt am
   Main, Germany as described^[255]42. Aged healthy control subjects were
   consecutively recruited from spouse and friends of PD patients and from
   the outpatient stroke unit of the University Hospital in Frankfurt am
   Main. Demographic data of patients and controls have been shown in
   detail in ref. ^[256]42. PD disease severity was assessed according to
   the Hoehn Yahr rating scale in “ON” periods for patients who
   experienced substantial disease fluctuations (16 out of 50 patients).
   Patients were not routinely genotyped for PD-associated GBA1 mutations
   because the frequency in Germany is low (5-10%). Informed written
   consent was obtained from all subjects. The study was approved by the
   Institutional Ethics Committee of the University Hospital of Frankfurt
   (Approval #458/16, date 13.01.2017, amendment 01.10.2018). Data
   acquisition and blood sample collection adhered to the Declaration of
   Helsinki. Demographic data are presented in Supplementary Table [257]1.

Skin biopsy and primary fibroblast culture of PD patients and controls

   After local anesthesia by intradermal injection of lidocaine a 3 mm
   skin punch biopsy was obtained from the skin of the lower leg 10–15 cm
   above the ankle. The biopsy was placed in ice-cold DMEM medium for
   transport. Culture dishes (6-well plates) were prepared by coating the
   surfaces with FBS and dried under the clean bench. The biopsy was
   placed in a culture dish within 1 h and cautiously covered with
   fibroblast medium (low glucose DMEM (100 mg/l), 10% FBS, 1x MEM
   Non-Essential Amino Acids, 2 mM glutamine, 1% PenStrep) and fibroblasts
   were allowed to grow for 3–4 days. Fibroblasts were harvested with
   trypsin and sub-cultured in T75 flasks in fibroblast medium or in DMEM
   high glucose (5 g/l), 10% FBS, 1% PenStrep. Targeted lipidomic analyses
   were obtained from naïve cultures and upon stimulation with pimozide
   versus vehicle. WST-1 assays and immunofluorescence analyses were used
   to assess EC50 values of pimozide in PHF.

   For lipidomic analyses, primary human fibroblast (PHF) cultures were
   stimulated with 1/8^th of the EC50 of pimozide (i.e. 12.5 µM; Merck,
   #573110-100MG) or vehicle (DMSO 1:10000) for 24 h, in two rounds, each
   consisting in 25 PHF cultures (13 PD, 12 HC cultures) of matched HC and
   PD patients. Sub-confluent cultures were split and plated on 10 cm
   culture dishes, each two for DMSO (Control) and Pimozide. Cells were
   allowed to attach and grow for 24 h, before treatments were added and
   cells grown for another 24 h. Cells were washed 3 × in 1× PBS,
   harvested by trypsinization, pelleted, washed, and adjusted to 250,000
   cells per cell pellet which was frozen at −80° until lipid analyses.

   Demographic data of PFB donors are presented in Supplementary Table
   [258]2.

Pimozide IC50 in primary human fibroblasts

   WST-1 and sulforhodamine B (SRB) cell viability/cell death assays were
   used to define stimulating but non-toxic pimozide concentrations and
   conditions. PHF were seeded in 100 µl medium/well of a 96 well plate,
   5000 per well and allowed to grow for 24 h. The medium was replaced
   with full medium containing Pimozide at concentrations ranging from 1
   to 256 µM (log2 scale 1–8 µM) or 1:10000 DMSO (=vehicle). The next day,
   10 µl WST-1 reagent (Merck #11644807001) was added to each well and
   cells were incubated for 60 min at 37 °C. Formation of the formazan
   dye, which relies on the metabolic activity was then measured via its
   absorbance in a multi-well spectrophotometer (TECAN Infinite F200 Pro,
   RRID: SCR_020543) at 405 nm and 620 nm (reference). Subsequently, the
   reaction was stopped by adding 20 µl 50% trichloroacetic acid, TCA
   (w/v) to a final concentration of 10% TCA for 1 h at 4 °C. Wells were
   washed with distilled water 7-times and were dried. To quantify cell
   mass, fixed cells were stained by adding 100 µl 0.4% (w/v)
   sulforhodamine B (SRB) in 1% acetic acid for 30 min at room temperature
   under shaking. Wells were washed with 1% acetic acid 5-times and cells
   were lysed by adding 250 µl 10 mM Tris/HCl pH 10.5 for 10 min at room
   temperature. Absorbance of SRB at 540 nm was measured with TECAN plate
   reader. The percentage of viable cells compared with DMSO was plotted
   versus the log2 concentration and fitted with a inhibitory sigmoidal
   Emax model in GraphPad Prism 9 or 10 (RRID:SCR_002798).

   The survival of HT22 cells in the presence of GlcCer24:1 or vehicle and
   AF555-labeled αSyn-PFF was tested using the WST-1 reagent (Merck
   #11644807001) as described above. HT22 cells were seeded in 96 well
   plates (4000 cells/well) and treated with 1 µM GlcCer24:1, with or
   without αSyn-PFF. After 48 h of incubation, 10 µl of the WST-1 reagent
   was added to the cells and after a 60 min incubation at 37 °C the
   absorbance was measured at 450 nm and 620 nm (reference) in a
   multi-well spectrophotometer as above. Subsequently, the SRB assay was
   performed.

Double mutant Pink1^−/−SNCA^A53T Parkinson-model mice

   Homozygous Pink1^−/− plus SNCA^A53T double mutant mice and Pink1^−/−
   single mutant mice were generated as described in refs.
   ^[259]63,[260]123 by crossing Pink1^−/− mice (MGI:3850370) with PrPmtA
   mice expressing human mutant SNCA A53T (mtA) under the prion promoter
   (PrP) (MGI:3723258) and then, interbreeding the littermates. Mice are
   available from Jackson lab as cryopreserved sperm (FVB;129-Pink1^tm1Aub
   X Tg(Prnp-SNCA*A53T)^AAub/J; Strain #:017678). Double mutant mice are
   referred to as Pink1^−/−SNCA^A53T or in short, Pink1SNCA.

   We used Pink1^−/−SNCA^A53T double mutant mice because unlike other PD
   models, they develop spontaneous PD-like motor symptoms with an onset
   at >15 months of age and a frequency of about 20-30 percent^[261]123
   and they combine the complex PD pathophysiology of pro-oxidative and
   dysfunctional autophagolysosomal pathways^[262]12,[263]124–[264]126.
   Further, in contrast to single mutant mice they develop sensory
   premotor deficits reminiscent of the human premotor phase that
   manifests with a sensory loss and pain^[265]127 and is associated with
   increased plasma GlcCer species^[266]42. The onset of sensory deficits
   in Pink1^−/−SNCA^A53T double mutant is at around 6 months of age and is
   progressive^[267]127. Using 1-year-old mice ensured the presence of
   PD-like pathology (mitochondrial morphology^[268]127, respiratory
   dysfunction^[269]127, synuclein aggregates^[270]123, ceramide
   accumulation^[271]127) but avoided overt serious motor deficits and
   suffering of the animals. Overall, Pink1^−/−SNCA^A53T double mutant
   mice show many features of human PD.

   Mice had free access to water and food, and they were maintained in
   climate-controlled rooms with a 12 h light-dark cycle. The maintenance
   of the breeding colony of Pink1^−/−SNCA^A53T and observation up to
   onset of PD symptoms were approved by the local Ethics Committee for
   animal research (Darmstadt, Germany; FK1032 (12.02.205-11.02.2020) and
   FK1131 (17.02.2020-31.12.2023). The studies adhered to the guidelines
   of the Society of Laboratory Animals (GV-SOLAS) and were in line with
   the European and German regulations for animal research. Mice were
   observed for ~12 months and were euthanized to obtain tissue and plasma
   samples for lipid analyses before onset of overt clinical symptoms
   which occur in about 20 percent of mice >18 months^[272]63,[273]123.

Mice tissue collection: brain and plasma

   Mice were euthanized with carbon dioxide and blood withdrawal by
   cardiac puncture, whereby blood was collected into K3^+ EDTA tubes,
   centrifugated at 1300 × g for 5 min, and plasma was transferred to a
   fresh tube and snap frozen on dry ice or in liquid nitrogen. The brain
   was dissected for lipidomic and proteomic analyses. Cerebellum and
   olfactory bulb were removed, and the brain was cut sagittal. Left and
   right halves were weighed with precision scales and snap frozen on dry
   ice. Samples were stored at −80 °C until analysis.

HT22 cell culture, αSyn-PFF loading, and immunofluorescence analysis

   HT22 is an immortalized mouse hippocampal neuronal cell line which are
   derived from hippocampal neuronal HT4 cells and are widely used as an
   in vitro model of PD because of their high sensitivity to glutamate
   excitotoxicity and PD-inducing toxicants like MPP+
   (1-methyl-4-phenylpyridinium) and rotenone^[274]128,[275]129. They
   express dopamine D4 receptors^[276]130.

   HT22 mouse immortalized hippocampal neurons (RRID:CVCL_0321) were grown
   in High Glucose Dulbecco’s modified eagle medium (DMEM, Gibco)
   containing 10% fetal bovine serum (FBS, Gibco), 2 mM L-glutamine
   (Gibco) and 1x penicillin/streptomycin (Sigma) at 37 °C and in 5% CO[2]
   atmosphere in a humidified cell culture incubator.

   For stimulation of HT22 neurons with pre-formed active aSyn fibrils
   (aSyn-PFF) (A53T Mutant Alpha Synuclein Active Protein, MyBioSource),
   sub-confluent cells were trypsinized, pelleted, and resuspended in full
   medium containing αSyn-PFF and seeded in T75 flasks (Greiner) to
   achieve a final concentration of 2 µg/ml PFF or 6 µg/ml PFF and cell
   density of 0.4 ×106 cells/flask. Control cells were plated in full
   medium without PFF.

   For lipid analyses cells were incubated for 72 h and harvested by
   trypsinization. Cells were washed 3-times with 1× PBS, resuspended in
   1 ml 1× PBS, counted and adjusted to 250,000 cells per sample, finally
   pelleted. The supernatant was discarded and pellets stored at −80 °C
   until analysis. For lipid analysis, 8 replicates were generated for
   each treatment.

   For analysis of cell viability upon exposure with αSyn-PFF, WST-1 and
   LDH activity assays were used. The WST-1 assay is based on the cleavage
   of the water-soluble tetrazolium salt by cellular enzymes to a formazan
   dye, which is quantified by measuring absorbance and correlates with
   the number of metabolically active viable cells. Cells were seed in
   96-well plates (3500 cells/well) and were cultured for 24 h or 48 h
   with/without 2 µg/ml αSyn-PFF. The medium was then replaced with 100 μl
   of fresh medium (per well), and 10 µl WST-1 solution were added to each
   well. After 90 min in the incubator, the absorbance was read on a
   multimode microplate reader (SpectraMax i3X, RRID:SCR_026346) at
   450 nm, and the reference wavelength of 620 nm.

   For analysis of LDH activity, 5 µl cell culture supernatant and NADH
   standards (0–12.5 nmol/well) were pipetted into the wells of a 96 well
   plate. The master reaction mix consisting of assay buffer and substrate
   mix was added to the wells. The absorbance was measured on a microplate
   reader at 450 nm every 5 min until the value of the most active sample
   exceeded the value of the highest standard. The absorbance was
   corrected with the blank, and LDH activity was calculated as described
   in the manufacturer’s protocol (Sigma).

   For live cell immunofluorescence analyses of αSyn-PFF uptake, HT22
   cells were seeded in 8-well cover glass bottom culture slides in full
   medium in the presence of 1 µM GlcCer24:1 (AvantiPolar Lipids #860549)
   or vehicle (2:1 mixture of chloroform and methanol, final dilution
   1:10000) for 48 h. The chambers were coated with Poly-L-lysine (0.01%,
   Sigma) before seeding of 5000 cells per well. Then 2 µg/ml αSyn-PFF
   labeled with a fluorophore (Alexa Fluor 555 Microscale Protein Labeling
   Kit, Invitrogen) were added to the medium and incubated for 24 h. Live
   culture images were obtained on a Leica Stellaris 8 fluorescent
   confocal microscope (RRID:SCR_024660). To visualize the plasma membrane
   and the nucleus, cultures were incubated 15 min before imaging with
   wheat germ agglutinin-Alexa Fluor 488 reagent (WGA, Invitrogen).
   Hoechst 33342 was used as nuclear stain (Thermo Scientific). For
   analysis in FIJI ImageJ, mono-channel images were converted to 8-bit
   images, stacked and arranged in montages. Using the threshold
   default-algorithm, images were converted to binary masks which were
   then submitted to the particle counter for each image, and the
   immunofluorescent (IF) area was used for further quantitative analysis.
   The red-PFF IF area was normalized by the respective Hoechst IF area
   representing the nuclei, and the area ratio was submitted to
   statistical t-test comparisons.

Beta-arrestin screening of G-protein activation by GlcCer24:1

   PRESTO-Tango beta-arrestin assay^[277]131 was used to screen for the
   activation profile of heterologously expressed G-protein coupled
   receptors.

   The assay uses HTLA cells which are HEK293T cells stably expressing a
   luciferase reporter gene and stably expressing a human β-arrestin-2
   fused to Tobacco Etch Virus protease. 5000 HTLA cells were seeded in a
   white, transparent and poly-L-Lysin-coated 384-well plate from
   PerkinElmer. The cells were co-transfected after 6 h with a plasmid
   from the PRESTO-Tango kit (Addgene #1000000068). We used a mixture of
   10 ng plasmid and 0.04 µl Lipofectamine 2000 per well performing the
   transfection using the protocol described by Kroeze et al.^[278]131. We
   used GFP as a transfection control and 100 µM carbachol as an assay
   control at the muscarinic M5 receptor. After 24 h, the medium was
   aspirated and replaced by 45 µl of serum-free medium. The ligand (5 µl)
   was then added at a final concentration of 1, 5 and 10 µM for ~24 h.
   Afterwards, the medium was aspirated, and the cells lysed using 50 µl
   of bright-Glo reagent (Promega) diluted 10X with 1x PBS. After 15 min
   of incubation with the lysis buffer, the luminescence (endpoint,
   1500 ms integration time) was measured using a flexstation 3 device.

Dynamic Mass Redistribution Assay (DMR)

   Label-free dynamic mass redistribution enables real-time detection of
   refractive index alterations on biosensor-coated microplates that
   originate from stimulus-induced changes in the total biomass in
   proximity to the sensor surface. Dynamic mass redistribution monitoring
   was performed with a Corning EPIC BT system. 5000 HEK293 cells were
   seeded in a transparent and poly-L-Lysin precoated 384-well plate from
   Corning. The transfection was done as described for the beta arrestin
   screen. After 48 h, the culture medium was removed. Assay buffer was
   added and cells incubated in assay buffer for 2 before a 1 min baseline
   was recorded. Diluted lipids (final concentration 1 µM) or positive
   control were then added and DMR signals were acquired every 3 s for a
   period of 60 min. The readout is the shift of resonance wavelength over
   time ∆λ(t) shown as response curves. Carbachol effects on muscarinic
   receptors were used as positive control.

Primary DRG neuron culture

   To assess gene regulations under treatment with GlcCer24:1, primary
   sensory neurons were prepared from naïve adult C57BL6 mice which had
   free access to food and water, were kept in climate-controlled rooms
   with a 12 h light-dark cycle. Primary adult neuron-enriched cultures of
   dorsal root ganglia (DRG) of each four mice per group were prepared by
   dissecting DRGs of adult mice into Hank’s balanced salt solution (HBSS,
   Merck), followed by digestion with 5 mg/ml collagenase A (Millipore)
   and 1 mg/ml dispase II (Roche Diagnostics, Germany) before treatment
   with DNase (Sigma, 250 U per sample). Triturated cells were centrifuged
   through a 15% fat-free bovine serum albumin (BSA) solution. Primary
   sensory neurons were seeded on poly-L-lysine and laminin-coated cover
   slips (9–12 cover slips per mouse). Cells were cultured in serum-free
   Neurobasal medium (Gibco) containing 1x B27 supplement, 1x Pen/Strep,
   200 ng/ml nerve growth factor and 2 mM L-glutamine at 37 °C and 5%
   CO[2]and 95% humidity. Neurons were cultured in the presence of 1 µM
   GlcCer24:1 or vehicle (2:1 mixture of chloroform and methanol, final
   dilution 1:10000) and submitted to RNAseq.

RNA sequencing in DRG neurons upon stimulation with GlcCer24:1

   The day after seeding of primary DRG neurons, GlcCer24:1 was added to
   the medium at a final concentration of 1 µM. Control cells received the
   same volume of vehicle. Cells were collected two days after adding
   treatments. Total RNA was isolated using a single-cell RNA extraction
   kit (PicoPure, Thermo Fisher). Quantity of total RNA was determined
   using Invitrogen’s Qubit HS assay and quality was checked on an Agilent
   2100 Bioanalyzer Instrument (RRID:SCR_018043). First-strand cDNA was
   prepared and amplified from 1 ng total RNA using SMART Seq v4 Ultra
   Low-Input RNA Kit for sequencing (TaKaRa) according to the manual
   compatible with Illumina NGS. Barcoded sequencing libraries were
   subsequently prepared using the low-input library preparation kit (Sv4
   PLUS kit, TaKaRa). The Sv4 PLUS kit includes the first stand cDNA
   synthesis using SMART technology and library preparation that
   incorporates enzymatic fragmentation and stem-loop adapters to
   construct high-quality, Illumina-compatible libraries. Concentrations
   were determined using Invitrogen’s Qubit HS assay and fragment size was
   analyzed on Agilent’s 2100 Bioanalyzer on a HS DNA chip. RNA sequencing
   was performed of 4 biological replicates per group using an Illumina
   NextGen 2000 system (RRID:SCR_023614).

   The sequence alignment was done with Qiagen’s CLC genomic workbench (v.
   22, RRID:SCR_011853). Sequence reads were trimmed for adapter sequences
   and low-quality sequences using CLC genomic workbench standard settings
   of RNA sequencing (quality limit 0.05). Sequence reads were aligned to
   the reference genome mm10 provided from UCSC (GRCm38) as template,
   using CLC’s default setting for RNAseq. Read counts extraction and
   normalization (TMM) were done using CLC genomic workbench. The
   expression value unit is the trimmed mean of M-values, TMM (EdgeR
   algorithm). TMM reads were Log2 transformed.

   Differential gene expression was assessed using t-tests and fold change
   using CLC genomic workbench. Genes were filtered for at least 4 valid
   values out of 8 samples to exclude low-expression genes. The P value
   was set at 0.05 and adjusted according to the False Discovery Rate
   (FDR). Hierarchical clustering with Euclidean distance metrics was used
   to assess gene expression patterns. Genes were ranked according to P
   and q value, fold change and abundance. Ranked genes were submitted to
   gene ontology enrichment analysis using DAVID (Database for Annotation,
   Visualization and Integrated Discovery^[279]132; RRID:SCR_001881) and
   ExpressAnalyst ([280]https://www.expressanalyst.ca/)
   (RRID:SCR_025651)^[281]133. The RNAseq data have been deposited via the
   Gene Expression Omnibus (RRID:SCR_005012) as GEO dataset with the
   provisional accession number GSE262573.

Lipidomic and metabolomic analyses (plasma, cells, supernatant)

   Human blood was collected in 9 ml K3 + EDTA tubes, mouse blood in
   500 µl K + EDTA tubes, centrifuged at 1300 × g or 1500 × g and plasma
   was transferred into 1.5 ml tubes and stored at −80 °C until further
   use. Cells were collected by trypsinization, washed, pelleted,
   resuspended in 100 µl counted and adjusted to yield 1.5x 10exp6 cells
   per pellet (experiment-1) or 2.5x 10exp5 cells (experiment-2). Finally,
   the cells were pelleted at high speed and supernatant removed with
   100 µl and subsequently 20 µl pipet. The “dry” pellet was frozen at
   −80 °C until lipid/metabolite extraction. Fibroblasts were homogenized
   as explained below (targeted analyses). Lipidomic and metabolomic
   analysis were conducted applying the same procedure as previously
   described^[282]134. Lipidomic analysis of the human plasma samples used
   a slightly different procedure^[283]135. Targeted sphingolipid analyses
   were done in plasma, cell and tissue extracts essentially as described
   previously^[284]42 using liquid chromatography-electrospray
   ionization-tandem mass spectrometry (LC-ESI-MS/MS), according to
   procedures described in detail in ref. ^[285]136. Further protocol
   details for extraction and analyses of different matrices (cells,
   tissue, plasma) are described in the supplementary methods (Excel
   file).

Statistics

   Group data are presented as mean ± SD or median ± IQR as specified in
   the respective figure legends. Data were analyzed with SPSS 29
   (RRID:SCR_016479) and GraphPad Prism 9 or 10 (RRID:SCR_002798), Origin
   Pro 2024 (RRID:SCR_014212), and MetaboAnalyst 5.0 (RRID:SCR_016723)
   ([286]https://www.metaboanalyst.ca)^[287]137. Bioinformatic analysis of
   “omic” data (RNAseq, lipidomic, metabolomic) is explained in the
   respective paragraphs. Area under the curve (AUC) data of lipidomic and
   metabolomic analyses transformed to square root values to adjust skewed
   distributions. For testing the null-hypothesis that groups were
   identical, two groups were compared with 2-sided, unpaired Student’s t
   tests. Lipidomic and metabolomic data were submitted to 2-way analysis
   of variance (ANOVA) using e.g., the factors “feature” (e.g. lipid,
   metabolite) and ‘group’ (e.g. PD versus HC with/without stimulation).
   In case of significant differences, groups were mutually compared using
   post hoc t-tests according to Šidák or false discovery rate (FDR). The
   meaning of asterisks in figures is explained in the legends.
   ANOVA-simultaneous component analysis (ASCA)^[288]138 was used for
   analysis of multiple covariates on lipidomic data from patients and
   fibroblasts. ASCA is a combination of ANOVA and PCA plus feature
   extraction method for multivariate data to model two major components
   and their interaction. The meta-data for patients included sex, age,
   BMI, telomere length, and sensory loss. For multivariate analyses, data
   were normalized to have a common mean and variance of 1
   (Z-scores = (x − x̄)/SD). Volcano plots were used to assess fold
   differences of lipids versus the negative logarithm (Log10) of the
   t-test P value according to standard procedures. Partial Least Square
   Discriminant analysis (PLS-DA) or canonical discrimination analysis
   were used to assess group membership and variable importance.

Supplementary information

   [289]Supplementary figures and tables^ (2.3MB, pdf)
   [290]Supplementary Excel file^ (28KB, xlsx)
   [291]Supplementary information^ (158.9KB, xlsx)

Acknowledgements