Glycerol 3-phosphate acyltransferase exacerbates α-synuclein-induced toxicity by increasing lipid peroxidation

Abstract

Although multiple cellular pathways have been implicated in α-Synuclein (α-syn)-associated Parkinson’s disease (PD), the role of lipid metabolism remains elusive. In this study, we identify Drosophila mino, which encodes the mitochondrial isoform of the lipid synthesis enzyme glycerol 3-phosphate acyltransferase (GPAT), as a potent modifier of α-syn. Silencing the expression of mino significantly suppresses α-syn-induced PD phenotypes in Drosophila, including dopaminergic neuronal loss and locomotion defects as well as circadian rhythm-related activities, whereas mino overexpression yields opposite effects. Mechanistically, we find that mino modulates the levels of mitochondrial reactive oxygen species and lipid peroxidation. Importantly, treatment of α-syn-expressing flies with FSG67, a GPAT inhibitor of glycerol 3-phosphate acyltransferase, reproduces the benefits of mino knockdown. FSG67 also inhibits α-syn aggregation and lipid peroxidation in mouse primary neurons treated with α-syn preformed fibrils. Our study elucidates an important factor contributing to α-syn toxicity and offers a therapeutic direction for PD.

The role of lipid metabolism in Parkinson's disease is poorly understood. Here, Ren and co-authors identify a key enzyme in this process, mitochondrial glycerol 3-phosphate acyltransferase, as a critical driver of α-synuclein toxicity.

Introduction

Parkinson’s disease (PD) is a prevalent neurodegenerative disease that is characterized clinically by motoric deficits, including bradykinesia, tremors, and rigidity. The principal neuropathology underlying PD is the loss of dopaminergic (DA) neurons in the substantia nigra pars compacta of the midbrain, which results in the disruption of the nigrostriatal pathway leading to movement disorders. Accompanying the DA neurodegeneration is the presence of intraneuronal inclusions known as Lewy bodies (LBs), which are enriched with α-synuclein (α-syn)^1,2. Rare mutations and copy number variations in the α-syn-encoding gene SNCA, such as A30P and A53T, are causative for monogenic forms of PD^3–5. Genome-wide association studies have established an association between genetic variations in the SNCA gene and the risk for PD⁶. However, the effects of these genetic variations are modest compared to other PD risk genes, like GBA1 or LRRK2⁷, potentially attributable to the polygenic mode of inheritance and interactions with genetic modifiers of SNCA^8–10. SNCA modifiers are linked to diverse biological pathways, including mitochondrial function, antioxidant defense, autophagy, protein degradation, vesicle trafficking, neuroinflammation, cytoskeleton dynamics, and lipid metabolism.

Lipid dyshomeostasis is increasingly recognized as a key player in PD pathogenesis. LBs are not merely aggregates of α-syn but contain abundant lipids from damaged organelles, which implicate lipid dyshomeostasis as a mechanism in LB biogenesis and PD pathogenesis^11,12. Several lipid metabolism genes are linked to PD, including PLA2G6, GBA, LRRK2, Parkin, SMPD1, SCARB2, DGKQ, and SYNJ1^13–15. Lipidomic irregularities in the polyunsaturated fatty acids, glycerophosphatidylcholine, diacylglycerol, and triglycerides have been observed in the serum and brain tissues of PD patients^14,16–18.

The N-terminal domain of α-synuclein has affinity for lipid membranes, and studies have shown that mutations A30P, E46K, and G51D in this region can alter the protein’s oligomeric state^19–21. α-syn deficiency disturbs brain lipid metabolism, as evidenced by changes in palmitate uptake and neutral lipid content. This disruption is associated with alterations in the esterification of arachidonic acid and docosahexaenoic acid into phospholipids^22–24. Excessive α-syn induces the accumulation of lipid droplets in yeast²⁵, and promotes the synthesis of polyunsaturated ether phospholipids in induced pluripotent stem cell-derived human midbrain dopaminergic neurons, leading to lipid peroxidation and increased sensitivity to ferroptosis, a form of cell death marked by iron-dependent accumulation of lipid peroxides^26,27.

Lipid-metabolizing enzymes affect α-syn aggregation and associated neurotoxicity. For example, stearoyl-CoA desaturase (SCD) regulates the unsaturation level of fatty acids and affects α-syn oligomeric composition in yeast, cells, and transgenic mice^28–32; fatty acid CoA synthetase long-chain family member 4 (ACSL4) regulates fatty acids partitioning into the fatty acyl-CoA pool and influences lipid peroxidation and ferroptosis that are linked to α-syn neurotoxicity^26,33,34. Lysophosphatidic acid acyltransferase (LPAAT), which converts lysophospholipids to phospholipids and helps to reconstitute phosphatidylcholine, is upregulated in postmortem PD brains and cultured neurons overexpressing SNCA E35K/E46K/E61K mutants³⁵. The lipase LIPE regulates the equilibrium between α-syn tetramer and monomer, which affects α-syn neurotoxicity³⁶. Several PD models reveal that modulating fatty acid and lipid metabolism genes, including LPCAT, SCD, and LIPE, may reverse α-syn aggregation and associated phenotypes^30,35,36. YTX-7739, an SCD inhibitor, has progressed to phase 1 clinical trials²⁹. Uncovering the lipid involvement in PD pathogenesis is therefore critical, as it paves the way for therapeutic strategies.

In this study, we used transgenic Drosophila overexpressing α-syn as a model for genetic screening, and identify mino, which encodes the mitochondrial isoform of the lipid synthesis enzyme glycerol 3-phosphate acyltransferase (GPAT), as a modifier of α-syn. We find that neuronal silencing of mino mitigates α-syn-associated PD-like phenotypes in Drosophila, including dopaminergic neuronal loss and locomotion defects, whereas mino overexpression yields opposite effects. Neuronal silencing of mino in Drosophila PD model is linked to reduced levels of mitochondrial reactive oxygen species, membrane lipid peroxidation, and less α-syn higher-order oligomers in the brain. Knocking down other GPAT-encoding genes Gpat4, or CG15450 reveals a common role of GPAT in modulating α-syn neurotoxicity. Treatment of α-syn-expressing flies with FSG67, a GPAT inhibitor, reproduces the benefits of mino/GPAT knockdown. FSG67 inhibits α-syn fibril aggregation and lipid peroxidation in mouse primary neurons treated with α-syn preformed fibrils. Taken together, our study elucidates an important factor/pathway contributing to α-syn toxicity and offers a novel therapeutic direction for PD.

Results

Drosophila screen identifies mino RNAi as a suppressor of α-syn neurotoxicity

To identify modifiers of α-syn, we performed a genetic screen in a Drosophila model of synucleinopathy, where we overexpressed human wild-type α-syn in the photoreceptor (PR) neurons R1-R6. Overexpression of α-syn in PR neurons induced vacuole formation in the lamina layer, where PR neurons form synapses with downstream lamina monopolar neurons³⁷. To identify candidate genes for screening, we utilized a deltaSVM model to predict single-nucleotide polymorphisms (SNPs) with differential transcription factor binding among 83 human PD risk loci from the Parkinson’s disease genome-wide association study locus browser (v1.9)^38,39. Based on the location of SNPs, we manually selected 1–3 nearest genes for each locus. Subsequently, we ordered available RNAi or dCas9/gRNA-VPR overexpression lines from the Bloomington Drosophila Stock Center (BDSC). These lines targeted the Drosophila orthologues of the selected human genes, which were identified using the DIOPT Ortholog Finder⁴⁰. Finally, we crossed 107 lines targeting 92 genes with the Drosophila synucleinopathy model to examine whether they can influence α-syn-induced vacuole formation in the lamina⁴¹ (Supplementary Data 1).

From 107 RNAi and overexpression fly lines, we identified five RNAi lines (targeting Ggamma1, mino, nonC, RpII215, or Src64B) and nine overexpression lines (targeting Acbp1, Bap170, bnl, CG5214, CG7461, Psn, Pu, SdhC, and wg) that suppressed vacuole formation in α-syn flies by more than 2-fold at an adjusted P value of 0.01 (Fig. 1a). We subsequently selected mino, CG5214, CG7461, SdhC with roles in energy and lipid metabolism, to test if their RNAi or overexpression could mitigate the climbing declines in a pan-neuronal α-syn (GAL4) fly model which expresses α-syn (nSyb-GAL4 > SNCA) (Supplementary Fig. 1A). mino RNAi alleviates the climbing phenotype in α-syn flies for up to 8 weeks post-eclosion (Fig. 1b and Supplementary Fig. 1A, B). This was also observed in another pan-neuronal α-syn (QF2) fly model, which expresses higher levels of α-syn by nSyb-QF2 > SNCA and shows accelerated decline of climbing abilities (Fig. 1c, d). Overexpressing mino using an upstream activating sequence (UAS) insertion line mino^EY00734 exacerbates α-syn-induced locomotor impairments (Fig. 1d). Analysis of brain-extracted mRNA showed a 50% reduction and a fourfold increase of mino transcripts by mino RNAi and overexpression, respectively (Supplementary Figs. 1B and  2K). These results suggest a role for mino in modulating α-syn neurotoxicity.

Fig. 1. Drosophila genetic screen identifies mino RNAi as a suppressor of α-syn neurotoxicity.

a Plot of fold change (FC) of the size of α-syn-induced lamina vacuoles vs. BH-adjusted P value of two-sided nonparametric two-sample Mann–Whitney U test of genetic screen for α-syn modifiers using EGFP RNAi as control, using the Matlab function ranksum. The RNAi or overexpression (OE) with log2(FC) < −1 and P < 0.01 were labeled. Each data point was generated by the Matlab function plot(log2(FC), –log10(p),’.’) using FC and P for each group. b, d Startle-induced climbing assay of flies incubated at 25 °C, expressing pan-neuronal α-syn either by nSyb-GAL4 > SNCA (b) or nSyb-QF2 > SNCA (d). c Western blot and quantification of relative α-syn protein levels in the head of flies incubated at 29 °C for 3 weeks, with genotypes nSyb-GAL4 > SNCA, nSyb-QF2 > SNCA, and respective controls by western blot. e, g Immunostaining of PAM cluster TH+ neurons (e, g) and pixel classification of TH+ pixels using Labkit (e) in flies nSyb-GAL4 > SNCA incubated at 25 °C for 32 days or 49 days (e), or in flies nSyb-QF2 > SNCA incubated at 29 °C for 21 days (g). Scale bar = 20 μm. f, h Quantification of PAM cluster TH+ neurons in flies with nSyb-GAL4 > SNCA (f) or nSyb-QF2 > SNCA (h). b, d Each data point indicates the mean climbing height of 1 vial containing 20–30 flies, measured repeatedly across timepoints. c Each data point and each lane indicates one pool of dissected fly brains. f, h Each data point indicates one fly. Number of vials (n) per group is 13, 14, 12, 14, 14, 14, 14, 14 for 8 time points in (b), respectively, and 12, 20, 16, 19 for 4 time points in (d), respectively. Number of pools of dissected fly brains (n) per group 3 for (c). Number of flies (n) per group is 50, 48, 57, 40, 20, 17, 50, 29 for (e, f); 71, 39, 35, 50, 64, 55 for (g, h). Data in (c) were analyzed using a two-sided T test. Data in (b, d, f, h) at each timepoint were analyzed with one-way ANOVA followed by Tukey’s and Dunnett’s multiple comparisons test. The group comparisons in (b, d) were analyzed using two-way ANOVA followed by Tukey’s multiple comparisons test. The asterisks in (b) indicate the comparisons between α-syn with mino RNAi and α-syn with EGFP RNAi. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Source data are provided as a Source Data file. A single experiment with biological replicates was performed.

Pan-neuronal overexpression of α-syn is toxic to DA neurons⁴². Among the DA neuron clusters in the fly brain, the protocerebral anterior medial (PAM) cluster contains more than 100 tyrosine hydroxylase-positive (TH+ ) DA neurons and is involved in startle-induced negative geotaxis^43,44. We utilized the Labkit pixel classification to quantify the PAM TH+ neurons (Supplementary Fig. 1C)⁴⁵. α-syn (GAL4) model exhibits a less severe phenotype with a loss of PAM TH+ neurons evident at day 49 but not at day 32, compared to control flies (nSyb > EGFP RNAi) (Fig. 1e, f). In contrast, the α-syn (QF2) model shows an early toxicity with neuronal loss at day 32 compared to control flies (nSyb-QF2, nSyb > EGFP RNAi) (Supplementary Fig. 1D). When housed at a higher temperature of 29 °C, the reduction of PAM TH+ neurons in α-syn (QF2) flies occurs earlier at day 21 (Fig. 1g, h). Although varying α-syn expression levels or temperatures are associated with differences in the timing of neuronal loss, mino RNAi mitigates PAM TH+ neuronal loss in both models (Fig. 1g, h). In contrast, mino overexpression exacerbates the loss of PAM TH+ neurons in α-syn (QF2) flies (Fig. 1g, h). The protocerebral posterior medial 1/2 (PPM1/2), posterior lateral protocerebrum (PPL1), and dorsal medial (DM) clusters did not show DA neuron loss in α-syn (QF2) flies at day 21 (Supplementary Fig. 1E). Although we observed a lower number of TH+ cells in α-syn (QF2) flies upon eclosion, the effect of mino on α-syn-associated neurodegeneration is likely manifested during the post-developmental stage (Supplementary Fig. 1F). α-syn (QF2) model was used for all the subsequent studies.

Several α-syn PD models have been reported with disruptions in circadian rhythm-associated locomotor behavior^46–48. Similarly, we observed that α-syn induces an accelerated decline in daily activities, as shown by the disappearance of the morning peak and reduced amplitude at the evening peak (Fig. 2a–c). This decline is alleviated in the presence of neuronal mino RNAi (Fig. 2a–c). Although mino RNAi restored the evening peak’s average intensity, it did not lead to recovery of the morning peak (Supplementary Fig. 2A, B). In the absence of α-syn, neither mino RNAi nor its overexpression affects the daily activities (Fig. 2a–c).

Fig. 2. mino RNAi suppresses the α-syn-induced decline of daily locomotor activity.

a Mean actograms of flies incubated at 25 °C in 12-h light/dark cycles, recorded using the Drosophila activities monitor (DAM) for consecutive 10 days. The periods of light and dark are indicated using open or filled rectangles on the top, respectively. b Quantification of daily total locomotor activities for each fly of (a). c Heatmap of average hourly locomotor activities for each hour of the day across the 10 days, with individual flies as one column. a, b Measurements were taken from distinct flies for each timepoint with each data point as one fly, and were measured repeatedly across timepoints. The number of flies (n) per group is 4 for (a, b). Data in (b) were analyzed with two-way ANOVA followed by Tukey’s multiple comparisons test. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Source data are provided as a Source Data file. A single experiment with biological replicates was performed.

Neuron–glia metabolic crosstalk is associated with energy homeostasis and neurodegeneration⁴⁹. We further explored whether the changes of mino in glia also affect the daily locomotor activities of α-syn flies. Glia-specific mino RNAi did not show obvious α-syn-modifying effects, whereas simultaneous mino RNAi expression in neurons and glia mitigates the decline in daily locomotor activities in α-syn flies (Supplementary Fig. 2C–F). From mino insertion mutants, we identified mino^f04927, which shows a reduction of mino transcripts level to 16% and 50% of control flies in homozygote and heterozygote adult, respectively (Supplementary Fig. 2G, H, J). mino^f04927 flies exhibit increased transcription of transposons flamenco-2, 1731, and 42AB (Supplementary Fig. 2H), in agreement with its role in regulating primary transposon silencing in germline⁵⁰, as well as reduced PAM TH+ neurons compared to control flies (Supplementary Fig. 2I, L), suggesting an adverse effect of systematic mino insufficiency. We integrated mino^f04927 into α-syn model and found that α-syn flies with homozygous mino^f04927 did not show significant differences of PAM TH+ neurons compared to α-syn flies (Supplementary Fig. 2I, L). However, mino^f04927/+ mitigates the PAM TH+ neuronal loss and the decline of daily locomotor activities in α-syn flies (Supplementary Fig. 2N, O), suggesting that a limited systematic insufficiency of mino may be protective against α-syn toxicity.

Taken together, our results identified mino as a modifier of α-syn in two Drosophila models of PD.

mino modulates α-syn-induced lipid peroxidation and associated cell death

mino encodes glycerol 3-phosphate acyltransferase which catalyzes the acylation of acyl-CoA to glycerol-3-phosphate contributing to the synthesis of phospholipids and triglycerides⁵¹. Previous studies have shown that α-syn oligomerization and toxicity could be enhanced by the interaction with phospholipid membranes, and that α-syn alters the lipid profiles or is associated with lipid peroxidation^{17,26,27,52,53}. Thus, we examined lipid peroxidation in α-syn fly brains and if mino could affect it. To analyze lipid peroxidation levels of neuronal membranes using BODIPY C11, we labeled neuronal membranes with a cell membrane-localized infrared fluorescent protein (mIFP)⁵⁴. Although we separated the imaging sessions for BODIPY C11 and mIFP to minimize bleed-through, we found that BODIPY C11-stained brains without the mIFP transgene have emission in the infrared channel when excited by 640 nm laser; therefore, the presence of mIFP fluorescence was unable to differentiate between neuronal or glial membranes if co-stained with BODIPY C11 (Supplementary Fig. 3A, B). When flies were cultured at 25 °C for 3 weeks, the membrane lipid peroxidation levels in the brain optic lobe were not affected by α-syn (Supplementary Fig. 3C, D). However, the lipid peroxidation levels of α-syn fly brains were further decreased by mino RNAi and increased by mino overexpression (Supplementary Fig. 3C, D). When flies were raised at 29 °C for 3 weeks, α-syn caused an increase in membrane lipid peroxidation levels, which were reduced by mino RNAi and increased by mino overexpression (Fig. 3a, b). In the absence of α-syn, neither mino RNAi nor its overexpression affects the lipid peroxidation levels (Fig. 3a, b). Thus, the effect of mino on lipid peroxidation levels is dependent on α-syn.

Fig. 3. mino modulates α-syn-associated lipid peroxidation and cell death.

a BODIPY C11 staining for lipid peroxidation in the optic lobe of flies incubated at 29 °C for 3 weeks. Oxidized and reduced C11 were imaged using 488 nm laser and 561 nm laser, respectively. The pixel-wise ratio between oxidized C11 and reduced C11 in the membrane was shown as a pseudo-coloring heatmap. Scale bar = 3 μm. b Quantification of lipid peroxidation index using the mean ratio of oxidized to reduced C11 for (a). c TUNEL staining of the optic lobe of flies incubated at 29 °C for 3 weeks. Scale bar = 40 μm. d Quantification of TUNEL-positive spots in (c). e Detection of cleaved Dcp-1 and ATPCL using western blot for dissected fly brains. Each data point in (b) indicates the lipid peroxidation level of one fly, averaged from four images. Each data point in (d) indicates one fly. Each data point in (e) indicates one pool of flies. Number of flies (n) per group is 35, 30, 33, 31, 30, 32 for (b); 29, 28, 25, 28, 23, 31, 35, 21, 26, 22, 22, 28 for (d). The number of pools (n) per group is 1 for (e), except 3 for ATPCL quantification. Data in (b, d) were analyzed with one-way ANOVA followed by Tukey’s and Dunnett’s multiple comparisons test. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Source data are provided as a Source Data file. A single experiment with biological replicates was performed.

We investigated whether mino’s effects on α-syn-associated lipid peroxidation correlate with changes in cell death in the optic lobe. As a positive control, we treated flies with H₂O₂, which induces positive TUNEL signals (Fig. 3n). TUNEL-labeled cell death in the optic lobe of α-syn fly brains was reduced by mino RNAi and increased by mino overexpression (Fig. 3c, d). Neither mino RNAi nor its overexpression alone induces TUNEL-labeled cell death in optic lobes (Fig. 3c, d). We did not observe positive staining of cleaved caspase-3 or cleaved Dcp-1 in α-syn fly brains (Fig. 3e and Supplementary Fig. 3E, H), whereas the hs-hid (heat shock promoter-head involution defective) third instar larva brain shows positive staining after heat shock (Supplementary Fig. 3F, G). Similarly, we observed the cleavage of apoptosis reporters Apoliner⁵⁵ and hPARP1-Venus⁵⁶ in hs-hid larva brains upon heat shock but not in α-syn fly brains (Supplementary Fig. 3I–L). The cleavage of hPARP1-Venus occurs only in one brain among 50 brains without α-syn examined (Supplementary Fig. 3M). Given the limitations of these approaches, we cannot completely exclude the possibility of apoptosis in α-syn fly brains.

Neuronal α-syn causes LDs accumulation in glia

Lipid droplets (LDs) accumulation is a strategy by which cells manage stress, including protection against lipid peroxidation⁵⁷. We examined if LDs accumulate in fly brains with pan-neuronal α-syn using BODIPY 493/503. Replacing Triton X−100 by saponin during the fixation of cultured cells reportedly could stabilize the LDs for staining⁵⁸. In contrast to cultured cells, we observed that 0.1% Triton X−100 in the fixative could resolve the LDs fluorescence from the background signals in fly brains (Supplementary Fig. 4A). In 3-week-old control fly brains without α-syn, we observed massive LDs in central brains but not in optic lobe (Fig. 4a). However, in fly brains with pan-neuronal α-syn, the optic lobe also accumulates massive LDs (Fig. 4a). We quantified the time-dependent accumulation of LDs within the optic lobe cortex from day 1 to week 7 (Fig. 4b, c). In flies with and without α-syn, LDs could be observed at day 1 post-eclosion (Fig. 4b, c), but were reduced after week 1 (Fig. 4b), which is similar to the findings of LD dynamics in fly retina⁵⁹. In flies without α-syn, no massive LDs were observed up to week 6 examined (Fig. 4b, c). However, in fly brains with α-syn, LDs showed time-dependent accumulation from week 1 to 6 examined (Fig. 4b, c). To determine if LDs were localized in neurons or glia, we expressed EGFP-tagged TAG lipase bmm (UAS-bmm.EGFP) either in glia or neurons of fly brains and co-stained the LDs with LipidTOX. In α-syn fly brains, LDs co-localizes with glial bmm but not with neuronal bmm (Fig. 4d). Genomic-tagged bmm.GFP could be detected surrounding the peri-brain fat bodies and at variable portions of LDs in the optic lobe of the day 1 fly brain (Supplementary Fig. 4B). We expressed GFP either in glia or neurons in fly brains with α-syn and found that LDs localize outside neuronal GFP but are enclosed within glial GFP (Supplementary Fig. 4C).

Fig. 4. Neuronal α-syn causes LDs accumulation in glia.

a Images of z-stack projections of LDs in the optic lobe and central brain of fly brains incubated at 29 °C for 3 weeks, stained by BODIPY 493/503. Scale bar = 40 μm. b Z-stack projections of cortex LDs in optic lobe stained by BODIPY 493/503 at multiple timepoints. Scale bar = 20 μm. c Quantification of LDs for (b). d Imaging of the optic lobe for bmm.GFP and LDs by LipidTOX in flies incubated at 29 °C for 3 weeks. Scale bar = 10 μm. e, g Z-stack projections of cortex LDs stained by BODIPY 493/503 in the optic lobe of flies incubated at 29 °C for 3 weeks. Scale bar = 20 μm. f, h Quantification of LDs for (e, g), respectively. Number of flies (n) per group is 3 for (a); 8, 6, 6, 7, 5, 6, 7 for multiple time points in (c) (nSyb-GAL4); 8, 6, 6, 7, 6, 6, 7 for multiple time points in (c) (nSyb > mino RNAi); 8, 7, 8, 8, 8, 8, 8 for multiple time points in (c) (nSyb-QF2 > SNCA, nSyb-GAL4); 8, 8, 8, 8, 8, 8, 8 for multiple time points in (c) (nSyb-QF2 > SNCA, nSyb > mino RNAi); 10 for (d); 37, 36, 37, 40, 39, 36, 39, 42 for (e, f); 30, 30, 27, 32, 27, 31, 20, 26 for (g, h). Data in (f, h) were analyzed with one-way ANOVA followed by Tukey’s and Dunnett’s multiple comparisons test. The group comparisons in (c) were analyzed using two-way ANOVA followed by Tukey’s multiple comparisons test. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Source data are provided as a Source Data file. A single experiment with biological replicates was performed.

To determine if the LDs were coated by Drosophila Perilipin (encoded by Lsd-1 and Lsd-2), using genomic-tagged Lsd-1.GFP we found that Lsd-1.GFP was detectable enclosing the peri-brain fat bodies, but not on the LDs of α-syn brains or the central brain (Supplementary Fig. 4D), which confirmed the tissue specificity of Lsd-1⁶⁰. From the public fly brain or head RNA sequencing datasets^61–63, we found that fly brain preferentially expressed Lsd-2 and that Lsd-2 is mainly expressed in glia and head sensory neurons but little in brain central nervous system (CNS) neurons (Supplementary Fig. 4E–I). We overexpressed Lsd-2 or triglyceride lipase bmm either in glia or neurons and found that glial bmm expression, but not neuronal expression, suppressed the α-syn-associated LDs accumulation (Fig. 4e, f). On the other hand, glial but not neuronal Lsd-2 overexpression increased the α-syn-associated LDs (Fig. 4g, h). These results confirmed the glial localization of LDs. The presence of LDs in fly photoreceptor neuron expressing α-syn has been reported⁶⁰. Fly head or brain RNA-sequencing datasets^61–63 show that in addition to Lsd-2, mdy, the fly brain-expressing homolog of DGAT catalyzing the last step of triglycerides synthesis, are expressed in glia and head sensory neurons but little in CNS neurons (Supplementary Fig. 4E–I). These data align with the localization of LDs in glia of fly brains and in α-syn-expressing photoreceptor neurons in retina⁶⁰.

LDs have been associated with an antioxidant role that retards α-syn neuronal toxicity^59,64. We examined if modulating glial LDs affect neuronal α-syn-associated neurodegeneration. The α-syn-associated decrease of PAM TH+ neurons was not affected by bmm or Lsd-2 overexpression in glia or neurons (Fig. 4i, j). The TUNEL-labeled cell death in the optic lobe of α-syn fly brains was not affected by Lsd-2 overexpression in glia or neurons, or by bmm overexpression in glia. However, bmm overexpression in neurons mildly increased the cell death in α-syn fly brains (Fig. 4k–m).

Taken together, our results indicate that fly brains with pan-neuronal α-syn accumulate LDs predominantly in glia. Due to differences in the localization between glial LDs and neuronal α-syn, we were unable to address the possibility that LDs with specific compositions, or in different cellular contexts, could potentially modify α-syn aggregation or toxicity in ways not captured by our assays.

TCA cycle metabolites isocitrate and citrate increased in α-syn fly brains

Glia Lsd-2 overexpression causes increased LDs in pan-neuronal α-syn fly brains in contrast to control brains without α-syn (Fig. 4g, h). Given the reported pathway that fatty acids synthesized in neurons are transported to glia and stored as LDs^59,65, we reasoned that α-syn fly brains may have increased de novo synthesis of fatty acids, possibly from citrate (Fig. 5a). In the tricarboxylic acid (TCA) cycle, citrate is converted to α-ketoglutarate, which is further converted to succinyl-CoA by α-ketoglutarate dehydrogenase (OGDH) complex, whose dysfunction or deficiency is implicated in PD^66–68. A blockage of OGDH may cause the build-up of metabolites in preceding steps (Fig. 5a). To address these possibilities, we conducted a short-chain fatty acids (SCF) metabolomics assay on dissected 3-week-old fly brains (10 brains pooled, n = 1, no technical or biological replicates) with or without pan-neuronal α-syn (Fig. 5b and Supplementary Data 2, Supplementary Data 6). Among the TCA cycle metabolites, isocitrate showed a tenfold increase, and citrate showed a fivefold increase in α-syn fly brains compared to brains without α-syn, whereas the increase is less than 1.5-fold for malate, fumarate, and succinate. Lactate showed a threefold increase, consistent with the increased lactate dehydrogenase level in α-syn-expressing fly brains (Fig. 5a, b). Glycolate, which was reported as TH+ neuron-protective metabolite associated with DJ-1⁶⁹, is reduced to 36% compared to brains without α-syn.

Fig. 5. Profiling of TCA cycle metabolites and LDs accumulation in fly brains.

a Schematic illustration of the glycolysis, TCA cycle, and de novo lipid synthesis pathways. The fold changes of metabolites are indicated in red. Green arrows indicate the de novo fatty acid synthesis and lipogenesis. Purple arrows indicate the malate-aspartate shuttle. Gray-filled boxes indicate the transporters on the mitochondria's inner membrane. b Short-chain fatty acids profiling in dissected fly brains with pan-neuronal α-syn or not, incubated at 29 °C for 3 weeks. The numbers indicate the average metabolite (pmol) per fly brain from the pooled samples of ten brains. The number of pools (n) per group is 1. N.D., not detected or had concentrations below the lower limit of quantification. N.A. not available. c Z-stack projections of the cortex of the optic lobe of flies incubated at 29 °C for 3 weeks, stained by BODIPY 493/503. Scale bar = 20 μm. d Quantification of mean intensity of BODIPY 493/503 for (c). e Z-stack projections of cortex LDs stained by BODIPY 493/503 in the optic lobe of flies incubated at 29 °C for 3 weeks. Scale bar = 20 μm. f Quantification of LDs for (e). Each data point indicates one fly. Number of flies (n) per group is 19, 16, 19, 17, 21, 19 for (c, d); 19, 16, 19, 17, 21, 19, 19, 21, 22, 25, 27, 24 for (e, f). Data in (d) were analyzed with one-way ANOVA followed by Tukey’s and Dunnett’s multiple comparisons test. Data in (f) were analyzed two-sided with Brown–Forsythe and Welch’s ANOVA followed by Dunnett’s T3 multiple comparisons test. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Glu glutamate, Asp aspartate, MPC mitochondrial pyruvate carrier, CTP citrate transport protein, AGC aspartate glutamate carriers, MKA malate α-ketoglutarate antiporter, GOT1 glutamate oxaloacetate transaminase 1, GOT2 glutamate oxaloacetate transaminase 2, MDH1 malate dehydrogenase 1, MDH2 malate dehydrogenase 2. Source data are provided as a Source Data file. A single experiment with biological replicates was performed.

ATP citrate lyase, encoded by ACLY in humans and ATPCL in flies, catalyzes the conversion of citrate to oxaloacetate and acetyl-CoA, which links carbohydrate to lipid metabolism⁷⁰ (Fig. 5a). A reported proteomics analysis of α-syn-expressing fly head showed a 26-fold decrease of ATPCL isoform E2QCF1 and a 1.5-fold increase of isoform Q7KN85⁷¹. E2QCF1 differs from Q7KN85 only by the absence of phosphorylation-regulated peptide ALSPTAAKPIKLPPISA, a region also differs between fly ATPCL and human ACLY (Supplementary Fig. 5A). We detected the expression of isoform Q7KN85 but not E2QCF1 in the mRNA of fly brains (Supplementary Fig. 5B). The mRNA and protein levels of ATPCL in α-syn fly brains are comparable to those without α-syn (Fig. 3e and Supplementary Fig. 5B–D).

Next, we ask if knockdown or overexpression of neuronal mino affect lipid staining in fly brains. Although the massive LDs were not observed from the BODIPY 493/503-stained background signal in fly brains without α-syn, neuronal mino overexpression was associated with an increase in the background signal in the optic lobe cortex, whereas neuronal mino RNAi was associated with a reduction of the background signal (Fig. 5c, d). On the other hand, in fly brains with α-syn and BODIPY 493/503-stained LDs, neuronal mino knockdown caused an increase in LDs levels, whereas neuronal mino overexpression did not affect LDs levels (Fig. 5e, f). These mino-associated changes of lipid staining suggest a possibility that neuronal mino may affect the membrane lipid level in normal fly brains and LDs levels in α-syn fly brains, likely involves the neuron–glia transport and deposition of fatty acids into lipids which deserves further investigation.

We conducted lipidomic analyses of fly brains to explore the possible lipid differences associated with α-syn and mino (Supplementary Fig. 6 and Supplementary Data 3 and 6). Phospholipids constitute the largest proportion (40.05%) of total fly brain lipids (Supplementary Fig. 6A). The percentage of phospholipids increased with α-syn expression (log₂(FC) = 0.15). This α-syn effect seems to be mitigated in the presence of mino RNAi (log₂(FC) = 0.07) and enhanced when mino is overexpressed (log₂(FC) = 0.17) (Supplementary Fig. 6A). Phosphatidylcholine (PC) and phosphatidylethanolamine (PE), the two most abundant species comprising over 95% of total brain phospholipids, exhibited similar trends in response to α-syn expression and mino RNAi (Supplementary Fig. 6B). PC and PE in α-syn-expressing brains showed increased presence of monounsaturated fatty acids (MUFAs) and decreased presence of polyunsaturated fatty acids (PUFAs) (Supplementary Fig. 6C, D). α-syn fly brains showed increased transcripts of Pcyt1 and Pect, which encode the rate-limiting steps of PC and PE synthesis, and increased Desat2, which encodes stearoyl-CoA 9-desaturase, a homolog of human α-syn toxicity modifier stearoyl-CoA desaturase (SCD)⁷² (Supplementary Fig. 6E). Three lipid species, including Cer-NS (d16:1/24:0), TG (50:4)-FA18:3, and TG (54:2)-FA18:2, were detected in α-syn fly brains but not in control brains without α-syn (Supplementary Fig. 6F).

Taken together, our data demonstrates the increase of TCA cycle metabolites in α-syn fly brains, and that neuronal mino perturbations affect brain lipid staining in both normal and α-syn fly brains.

A common role of GPAT as modulators of α-syn neurotoxicity

To address if other GPAT members⁵¹ have a role in modulating α-syn neurotoxicity, we extended our RNAi targeting to Gpat4, and CG15450, two other Drosophila homologs of GPAT. mino is predicted to reside on the mitochondria, whereas Gpat4, and CG15450 are predicted to reside on the endoplasmic reticulum⁷³ (Fig. 6a). Similar to mino RNAi, neuronal knockdown of Gpat4 and CG15450 suppressed the decline of climbing abilities (Fig. 6b), mitigated the reduction in daily locomotor activities (Fig. 6c, d), attenuated the decline of PAM TH+ neurons in α-syn flies, and showed a similar effect in brain lipid staining as mino RNAi (Figs. 6e, f and 5c–f). Similarly, the Gpat4^f0898/+ heterozygote that exhibits a 34% reduction of Gpat4 transcript level compared to w¹¹¹⁸ mitigated PAM TH+ neuron loss and locomotor declines in α-syn flies (Supplementary Fig. 2M–Q). α-syn flies treated with the commercial GPAT inhibitor FSG67⁷⁴ showed improvements in daily locomotor activities and PAM TH+ neuron loss (Fig. 6g–i and Supplementary Fig. 7A). In contrast, RNAi of Gnpat, an enzyme predicted to reside on the peroxisome for ether lipid synthesis, did not affect the locomotion activities of α-syn flies (Fig. 6a–d). Despite the observation that Gnpat RNAi mitigated the loss of DA neurons in the PAM cluster (Fig. 6e), these results highlight the broad effects of GPAT activity in modulating α-syn neurotoxicity.

Fig. 6. mino represents GPAT activity in modulating α-syn neurotoxicity.

a Predicted subcellular localization of mino, Gpat4, CG15450, and Gnpat using MULocDeep. b Startle-induced climbing assay of flies incubated at 25 °C, with neuronal RNAi of mino, Gpat4, CG15450, or Gnpat in fly brains with pan-neuronal α-syn or not. c Mean actograms of flies incubated at 25 °C in 12-h light/dark cycles for 20 consecutive days. The periods of light and dark are indicated using open or filled rectangles on the top, respectively. d Quantification of daily total locomotor activities for each fly of (c). f, g Immunostaining of PAM cluster TH+ neurons in flies incubated at 29 °C for 21 days. Scale bar = 20 μm. e, h Quantification of PAM cluster TH+ neurons for (f, g), respectively. g, h FSG67 was added to fly food at 1 mM. i Quantification of daily total locomotor activities for flies incubated at 25 °C and fed with FSG67 or not. b Each data point indicates the mean climbing height of 1 vial containing 20–30 flies, measured repeatedly across timepoints. d, i Measurements were taken from distinct flies for each timepoint, with each data point as one fly, and were measured repeatedly across timepoints. Each data point indicates one fly in (e, h). Number of flies (n) per group is 16 for (b) (Syb-QF2, nSyb-GAL4); 8 for (b) (Syb-QF2, nSyb-GAL4 > gene RNAi); 16, 8, 8 for (b) (day 3, 6, 12, for all the groups containing Syb-QF2 > SNCA); 4 for (c, d); 46, 38, 40, 26, 39, 54, 54, 52, 49, 47 for (e); 65, 62, 66 for (h); 7 for (i). Data in (b) at each timepoint and data in (e, h) were analyzed with one-way ANOVA followed by Tukey’s and Dunnett’s multiple comparisons test. The group comparisons in (b, d, i) were analyzed using two-way ANOVA followed by Tukey’s and Dunnett’s multiple comparisons test. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Source data are provided as a Source Data file. A single experiment with biological replicates was performed.

We have previously shown that α-syn expression in fly brains is associated with glycolysis reprogramming and increased levels of Lactate dehydrogenase (Ldh)⁴¹. GPAT RNAi did not change the Ldh levels of α-syn flies (Fig. 7a, b) or the total α-syn or pS129 levels. Thus, mino does not seem to function upstream of α-syn.

Fig. 7. GPAT RNAi suppresses α-syn higher-order oligomer formation.

a Western blot of α-syn, Pgi, Ldh, and pS129 in dissected fly brains of flies incubated at 29 °C for 3 weeks. b Quantification of relative protein levels in (a). c Western blot of α-syn higher-order oligomers formation in dissected fly brains expressing pan-neuronal α-syn, with flies incubated at 29 °C for 3 weeks. A 3D heatmap of α-syn was shown on the right. d Quantification of the ratio between α-syn oligomers and monomers, or the percentage of monomers and oligomers to total α-syn in (c). e α-syn flies were fed with food supplemented with 1 mM FSG67 or not, incubated at 25 °C for 3 weeks before oligomer analysis. Each data point in (b, d, e) and each lane in (a, c) indicates one pool of dissected fly brains. Number of pools of dissected fly brains (n) per group is 4 for α-syn and Ldh, 2 for Pgi, and 3 for pS129/α-syn in (b); 5 for (d); 4 for (e). Data in (b) were analyzed with one-way ANOVA followed by Tukey’s and Dunnett’s multiple comparisons test. Data in (d, e) were analyzed with two-way ANOVA followed by Fisher’s LSD test. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Source data are provided as a Source Data file. A single experiment with biological replicates was performed.

The phospholipid membrane plays a role in α-syn oligomerization^75,76. We investigated whether GPAT affects α-syn oligomerization levels in α-syn fly brains. α-syn expressed in fly brains is mainly detected as monomers^42,77. Cross-linkers dimethyl dithiobispropionimidate (DTBP)⁷⁸ could stabilize the α-syn oligomers for detection (Supplementary Fig. 8A). With 4 mM DTBP treatment, we found that GPAT RNAi, but not Gnpat RNAi, suppressed the formation of higher-order α-syn oligomers (Fig. 7c, d). GPAT RNAi specifically reduced α-syn tetramers and higher-order oligomers, without impacting α-syn dimers or trimers, either in comparison to their respective monomers or to total α-syn levels (Fig. 7c, d). We did not detect proteinase K-resistant α-syn aggregates in α-syn fly brains (Supplementary Fig. 8B).

α-syn oligomers or aggregates were associated with the accumulation of mitochondrial reactive oxygen species (ROS), including H₂O₂⁷⁵. We expressed the H₂O₂ sensor mito-roGFP2-Orp1 in photoreceptor neurons to assess transient mitochondrial H₂O₂ levels⁷⁹. mino RNAi or overexpression did not affect the mitochondrial H₂O₂ index in photoreceptor neurons without α-syn (Fig. 8a, b). In α-syn-expressing photoreceptor neurons, mino RNAi was associated with a reduction, whereas mino overexpression was associated with an increase, of the H₂O₂ index (Fig. 8a, b). MitoTimer is a mitochondrial ROS-sensitive sensor reflecting the cumulative redox history of mitochondria by shifting the emission fluorescence from green to red⁸⁰. We verified the mitochondrial localization of pan-neuronally expressed MitoTimer (Supplementary Fig. 8C). α-syn expression in fly brains accelerated MitoTimer aging in the optic lobe, and this effect was mitigated by mino RNAi but enhanced by mino overexpression (Fig. 8c, d). We noted a mild suppression of MitoTimer aging by mino RNAi in the absence of α-syn (Fig. 8c, d). GPAT RNAi showed similar effects as mino in affecting MitoTimer aging (Fig. 8e, f).

Fig. 8. GPAT RNAi suppresses mitochondrial ROS accumulation and mitochondrial aging in α-syn fly brains.

a Imaging of mitochondrial H₂O₂ sensor in photoreceptor axons of flies incubated at 25 °C for 2 weeks, expressing H₂O₂ sensor Rh1-GAL4 > Mito-roGFP2-Orp1, and expressing α-syn (Rh1 > SNCA) or not. The H₂O₂ increased as the merged color shift from green to yellow and red. Scale bar = 5 μm. b Quantification of the ratio between mean intensity in 405 nm and 488 nm channels. Treatment with 20 mM H₂O₂ for 5 min was used as a positive control. c, e, g Imaging of pan-neuronal MitoTimer in the cortex of optic lobe reflecting the maturation of mitochondria from green to red color with time (c), or at day 5 of flies incubated at 25 °C (e), or fed with FSG67 or not and incubated at 22 °C or 25 °C for 1 day, 1 week, and 2 weeks (g). Scale bar = 10 μm (c, g) and 5 μm (e). d, f, h Quantification of MitoTimer aging index using the ratio of mean intensity of red to green channels of MitoTimer in (c, e, g), respectively. Each data point indicates one fly. Number of flies (n) per group is 21, 21, 21, 27, 33, 31, 22 for (b); 33, 16, 20, 26, 28 for multiple time points in (d) (nSyb-QF2, nSyb-GAL4); 27, 21, 16, 24, 29 for multiple time points in (d) (nSyb-QF2, nSyb > mino RNAi); 38, 18, 19, 29, 23 for multiple time points in (d) (nSyb-QF2 > SNCA, nSyb-GAL4); 29, 23, 22, 33, 26 for multiple time points in (d) (nSyb-QF2 > SNCA, nSyb > mino RNAi); 32, 27, 30, 30, 29, 26, 29, 29, 57, 50, 36, 41, 48, 40 for (f); 31, 25, 31, 29, 33, 26; (31, 25 shared day 1 data), 31, 32, 34, 30, 31, 31, 30, 29 for (h). Data in (d) at each timepoint and data in (b, f, h) were analyzed with one-way ANOVA followed by Tukey’s and Dunnett’s multiple comparisons test. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Source data are provided as a Source Data file. A single experiment with biological replicates was performed.

α-syn flies fed with 1 mM FSG67 for 3 weeks at 25 °C showed a reduction of tetramers and higher-order oligomers in comparison to monomers (Fig. 7e and Supplementary Fig. 8D). FSG67 also caused a suppression of MitoTimer aging in α-syn flies incubated at 22 °C (Fig. 8g, h). However, this suppression was lost when flies were incubated at 25 °C (Fig. 8g, h).

Taken together, our results demonstrate GPAT as a modulator of α-syn neurotoxicity by affecting the α-syn oligomerization and mitochondrial ROS accumulation.

Pharmacological inhibition of GPAT suppresses α-syn-induced lipid peroxidation in mouse primary neurons treated with α-syn preformed fibrils (PFF)

We investigated the effect of GPAT on α-syn aggregation in primary mouse neurons treated with α-syn preformed fibrils (PFF). PFF treatment induced the formation of fibers and aggregates positive for α-syn phosphorylation at serine 129 (pS129) (Fig. 9a). Co-treatment of neurons with PFF and FSG67 attenuated the levels of pS129 signals (Fig. 9a, b). PFF-induced pS129 fiber and aggregates partially colocalized with the staining of lipid peroxidation product 4-HNE, and that PFF-treated neurons exhibited elevated levels of staining of 4-HNE and MDA (Fig. 9a, c–e). 4-HNE and MDA staining in PFF-treated neurons were ameliorated by FSG67 co-treatment (Fig. 9a, c–e). BODIPY C11 staining showed elevated lipid peroxidation levels in PFF-treated neurons, which were mitigated when co-treated with FSG67 (Fig. 9f, g). Collectively, our results suggest that GPAT is a relevant player in α-syn-induced cytotoxicity in the mammalian context.

Fig. 9. FSG67 suppresses α-syn aggregation and lipid peroxidation in neurons transfected with PFF.

a Mouse primary neurons were transfected with α-syn monomers or PFF with 75 μM FSG67 for 2 weeks and stained with antibody against phosphorylated α-syn pS129 and lipid peroxidation product 4-HNE. Scale bar = 30 μm. b, c Quantifications of pS129 and 4-HNE intensity sum. For pS129 (b), each data point is the mean of technical replicates from one animal (a total of three independent experiments using three animals). Number of animals (n) per group is 3. d Mouse primary neurons were transfected with PFF with 0, 75, or 150 μM FSG67 for 2 weeks and stained with antibody against pS129 and the lipid peroxidation product MDA. Scale bar = 30 μm. e Quantifications of MDA intensity sum. For 4-HNE (c) and MDA (e), all data points are technical replicates ( > 30) from only one animal. Number of animals (n) per group is 1. f, g Mouse primary neurons were transfected with PFF with 0, 75, or 150 μM FSG67 for 3 weeks and live-stained with BODIPY C11 (f) for the lipid peroxidation analysis (g). Each data point is the mean of technical replicates ( > 48) from one animal (total three independent experiments using three animals). The number of animals (n) per group is 3. Scale bar = 40 μm. Data were analyzed with one-way ANOVA followed by Tukey’s and Dunnett’s multiple comparisons test. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Source data are provided as a Source Data file.

Discussion

In this study, we identified mino/GPAT as genetic modifiers of α-syn-induced neurotoxicity. Using Drosophila or a primary mouse neuronal model of synucleinopathy, we demonstrated that pharmacological or genetic reduction of neuronal mino/GPAT mitigates α-syn-associated pathological phenotypes, including locomotor decline, TH+ neuron loss, mitochondrial ROS accumulation, and lipid peroxidation. We also uncovered the upregulation of glycolysis and blockage of the OGDH-catalyzed step in the TCA cycle in α-syn fly brains, which is associated with citrate accumulation and massive LDs formation in the optic lobe, where LDs were seldom observed. These insights support a role for energy and lipid metabolism in the pathogenesis of PD and therein offer a therapeutic angle for PD. In cultured PD cells, which were bathed in a high volume of culture medium for efficient metabolite exchange, OGDH suppression specifically decreased TCA cycle metabolites downstream of OGDH, whereas citrate levels remained unchanged⁶⁷. In contrast, the metabolite changes observed in native fly brains, where neurons are intricately wrapped by glia and exchange is more complex, characterize a PD-like environment in a more physiologically relevant context.

Our results also point to a hypothesis that neuronal mino may influence the fate of fatty acids by modulating their contribution between glial LD storage and membrane lipid synthesis. This hypothesis aligns with our findings that neuronal mino knockdown enhances glial LD accumulation in α-syn fly brains, whereas mino overexpression alters background lipid staining levels in fly brains without α-syn, possibly reflecting changes in brain membrane lipids. Future in vitro studies on the effect of GPAT on neuronal membrane lipid changes and flux of fatty acids from neurons to culture medium, and studies tracing the flux of fatty acids to glia in neuron–glia co-culture, will be helpful to test this hypothesis.

α-syn comprises an amphipathic N-terminal region, a hydrophobic non-amyloid-component (NAC) domain, and an acidic C-terminal domain. The N-terminal with multiple lipid membrane binding motifs underlie its interaction with lipid membranes^19,81. α-syn aggregation is enhanced in the presence of phospholipid surfaces^82–84, and its toxicity can also be altered by the saturation level of fatty acids²⁸. Using Drosophila PD models exhibiting mild or severe disease phenotypes, we demonstrated that α-syn neurotoxicity can be modulated by the lipid synthesis enzyme mino. Neuronal RNAi against mino ameliorates α-syn-associated pathological phenotypes, including the decline in locomotor activities and the loss of DA neurons, while elevated mino expression exacerbates these phenotypes. The beneficial effects of mino downregulation are similarly observed in α-syn flies harboring RNAi targeting the predicted differentially subcellular-localized Gpat4, and CG15450, suggesting a common role of GPAT in modulating α-syn neurotoxicity. Fluorescence-tagged transgenic mino expressed in the female reproductive system is known to colocalize with both mitochondria and ER⁵⁰. The lipid exchange between mitochondria and ER may be related to the subcellular independence of the modifier role of GPAT⁸⁵.

Our study indicates that RNAi targeting GPAT affects the formation of higher-order oligomers of α-syn. The detection of α-syn oligomers in Drosophila is less observed compared to mammalian cells^42,77,86. In vivo crosslinking with DTBP⁷⁸ enhances the sensitivity of detecting α-syn oligomeric species. While RNAi against GPAT did not result in a change in α-syn protein levels, changes in α-syn from higher-order oligomers to lower-order oligomers and monomers were observed. This was associated with a decrease in α-syn neurotoxicity.

Elevated α-syn levels are known to trigger caspase activation and apoptosis in various PD models^87,88. Recent observations of enhanced lipid peroxidation in neurons containing SNCA gene triplication suggest sensitivity to ferroptosis²⁶. In the enzymatic ferroptotic pathway, lipoxygenase (LOX) catalyzes iron-dependent oxidation of polyunsaturated PE-AA (arachidonoyl) and PE-AdA (adrenoyl)⁸⁹. Mitochondria, the primary ROS-generating organelle, contribute to the non-enzymatic pathway of ferroptosis via the Fenton reaction⁹⁰. We detected accelerated mitochondria cumulative redox levels in α-syn fly brains. Drosophila has a distinct lipidomic composition lacking polyunsaturated fatty acids including AA and AdA, may utilize linoleic acid (18:2n-6) and linolenic acid (18:3n-3) for ROS-induced non-enzymatic lipid peroxidation on ER and mitochondria^91,92. Linoleic acid in choline glycerophospholipids is prone to forming hydroperoxide during peroxidation in human erythrocyte membrane⁹³. 4-HNE has been observed in fat bodies, neurons, and muscles of aging flies and was further enhanced by paraquat feeding⁹⁴. Although our results favor a model by which α-syn contributes to neuronal death through lipid peroxidation⁹⁵, we cannot rule out apoptosis as a driver of neuronal death, as the TUNEL assay lacks specificity in distinguishing between apoptosis, necrosis, and autolytic cell death⁹⁶ and TUNEL-positive nuclei are observed in cells undergoing ferroptosis⁹⁷. Moreover, despite the absence of cleaved caspase-3 staining and the lack of positive apoptotic signals of two reporters in our study, a previous study using hPARP-based apoptosis reporter has detected apoptosis in pan-neuronal α-syn-expressing fly brains⁴². It is possible that the differences in the α-syn expression levels may determine the extent to which the apoptosis and ferroptosis-driven neurodegeneration⁹⁸. While our data point towards a potential role for ferroptosis, additional experiments, including the use of specific ferroptosis modulators, will be necessary to elucidate the mechanisms underlying neuronal death.

Our study characterized age-dependent glial LDs accumulation in the optic lobe of α-syn fly brains. The glial localization of LDs in α-syn fly brains differs from Girard et al.’s findings, where the α-syn coating the LDs in PR neurons resists proteinase K digestion⁶⁰. Recently, it was reported that age-dependent LDs accumulation in the glia of the fly central brain is linked to mitochondrial dysfunction and senescence⁹⁹. Although previous studies suggest that GPAT knockdown reduces lipid accumulation as LDs and/or membrane lipids^100–102, we found that neuronal mino RNAi increases the LDs accumulation in glia, whereas neuronal mino overexpression does not. This suggests the complexity of the effect of GPAT on lipid levels in the context of neuron–glia interaction.

Finally, we demonstrated in both the α-syn fly as well as primary mouse neuronal models that pharmacological inhibition of mino/GPAT activity through FSG67 treatment promotes beneficial outcomes that illustrate the transferability of our findings from non-mammalian to mammalian contexts. Notably, current disease-modifying experimental therapeutics for PD aimed at α-syn were mostly focused on reducing its aggregation or enhancing its degradation. Our approach targeting a lipid synthesis pathway that α-syn acts on, offers a therapeutic angle that may augment the anti-aggregation and/or pro-degradation approaches in mitigating α-syn-induced neurotoxicity. Interestingly, FSG67 was originally developed as a drug to treat obesity and diabetes¹⁰³. Given the optimism surrounding the use of GLP agonists to treat PD, which was similarly developed to treat diabetes and subsequently to manage weight loss, it is interesting to postulate a converged pathway between FSG67 and GLP agonists that emphasizes the need to better understand the role of lipid aberrations in the pathogenesis of PD.

In conclusion, our research not only reveals a role of glycerol 3-phosphate acyltransferase in modulating α-syn aggregation and its neurotoxic effects but also highlights its potential as a target for mitigating the progression of PD.

Methods

Mouse primary neuron culture and staining

Mice-related studies were approved by and conformed to the guidelines of the Institutional Animal Care Committee (IACUC) of the Nanyang Technological University Lee Kong Chian School of Medicine (IACUC Protocol A20037). C57BL/6 mice were housed in groups of two to five in individual ventilated cages (IVC) under specific pathogen-free conditions in a temperature-controlled environment (23 °C) with 70% humidity and a 12-h light/dark cycle (lights on at 7:00 a.m.). Food and water were provided ad libitum, and all procedures adhered to humane care guidelines. Embryonic day 17.5 mouse fetuses without gender identification from pregnant female C57BL/6 were obtained. Sex was not factored into the experimental design due to the technical impracticability of determining fetal sex at the embryonic stage. A total of three pregnant mice were used, one for each of the three repeats of the experiment. The cortices were isolated and dissociated with papain (LS003119, Worthington) for 30 min at 37 °C. Papain was removed, and the cortices were washed with DMEM/FBS and spun down at 200×g for 5 min. Cells were titrated, spun down, and resuspended in Neurobasal medium supplemented with B27 and GlutaMAX (Gibco). In total, 80,000 cells/well were plated. At day 7, mice primary cortical neurons were treated with 1 μg/ml α-syn monomers (SPR-323, StressMarq) or 1 μg/ml synuclein PFF (SPR-322, StressMarq) with or without 75 μM or 150 μM FSG (HY−112489, MedChemExpress LLC) for 2 weeks. Following treatment, cells were fixed in 4% PFA/4% sucrose/PBS for 30 min at room temperature, then washed three times with PBS. Subsequently, cells were blocked in 0.2 M glycine, 0.1 mg/ml saponin, and 30 mg/ml BSA in PBS for 45 min. Primary antibody staining was carried out in antibody diluent (0.1 mg/ml saponin and 1 mg/ml BSA in PBS) for 1 h at room temperature using anti-Malondialdehyde antibody (Abcam, 11E3, ab243066, 1:500), anti-4-Hydroxynonenal antibody (Abcam, HNEJ-2, ab48506, 1:500), and anti-α-syn phospho S129 (Abcam, ab51253, 1:500). After washing three times with PBS, cells were incubated with secondary antibodies in antibody diluent for 1 h, followed by four washes with PBS. Finally, cells were stained with BODIPY 493/503 (Thermo Fisher Scientific, D-3922, 1:1000) and 1 μg/ml DAPI, mounted, and imaged using an FV3000 confocal microscope (Olympus). The 4-HNE and MDA signals were quantified by thresholding the image and then measuring the intensity sum of the area above the threshold on the image using ImageJ/Fiji. For the lipid peroxidation assay, 5 mM BODIPY C11 stock solution was added to the culture medium of a glass-bottom culture dish in 1:1000 dilutions and incubated for 0.5 h, then directly imaged in the green and red channels, respectively. The ratio of mean intensity in the cell was analyzed using ImageJ/Fiji.

Drosophila stocks

Flies were grown at 25 °C with a standard yeast-corn meal-dextrose diet. Only male flies were collected for experiments. Aging experiments were conducted at 25 °C, unless otherwise stated. Adult male flies were gathered in groups of 20–30 per vial for aging. During aging, the food was changed every 2 days for the initial 2 weeks and subsequently replaced daily from the 3rd week onward. All the Drosophila stocks used in this study, except for Rh1-GAL4, could be purchased from BDSC. There are no newly generated fly stocks from this study. The sources of Drosophila stocks are listed in Supplementary Table 1. mino^f04927 is sourced from Exelixis, Inc. mino^f04927, mino^EY00734 and Gpat4^f06898 mutants were outcrossed to w¹¹¹⁸. The nSyb-QF2 > SNCA stock was deposited to BDSC (RRID: BDSC_600605) by Koh Tong-Wey. The reference to Rh1-GAL4 is provided within the Supplementary Information.

Drosophila brain immunostaining

Larvae or adult fly brains were dissected in PBS, fixed in 3.7% paraformaldehyde in PBS with 0.1% Triton X−100 on ice for 0.5 h, and subsequently fixed at room temperature for an additional 0.5 h. For LDs staining, room temperature fixation was extended to 1 h. Following fixation, brains were rinsed in PBS with 0.4% Triton X−100 (PBST) and subjected to primary antibody staining for 1 day at 4 °C, then washed in PBST and incubated in secondary antibodies for 1 day at 4 °C, followed by washing with PBST. In the TUNEL assay, after removal of the buffer, the TUNEL reaction mixture (Roche, 12156792910) was added, incubated in a 37 °C water bath for 1 h, and briefly washed. Subsequently, brains were stained in PBST with 4’,6-diamidino-2-phenylindole (DAPI) for 3 h, briefly washed, and mounted with Prolong Gold (Thermo Fisher Scientific) on bridged slides, positioned as a line with the anterior side facing upward and both sides padded with double layers of Scotch transparent tape (except for a single layer for TH staining in the PAM cluster). For positive control of apoptosis, third instar larvae of w¹¹¹⁸ or hs-hid were heat-shocked in a 37 °C water bath for 2 h and allowed to recover for 5 h before dissection. For Apoliner, the heat shock is 1 h without recovery. Imaging was performed using an FV3000 confocal microscope with consistent image settings for samples within each experiment and analyzed using Bitplane Imaris 8.4 or ImageJ/Fiji. TUNEL analysis was performed using the Imaris Spots module. The primary antibodies used include rabbit anti-Tyrosine Hydroxylase TH (Pel Freeze, P40101−150, Lot ajo391o, 1:500), rabbit anti-Cleaved Caspase-3 (Asp175) (Cell Signaling Technology, 9661, Lot 43, 1:500), rabbit anti-Cleaved Caspase-3 (Asp175) (5A1E) (Cell Signaling Technology, 9664, Lot 22, 1:500), mouse anti-repo (DSHB, 8D12, Lot 3/10/22, 1:500), mouse anti-α-tubulin (DSHB, 12G10, Lot 1/13/22, 1:5000), rabbit anti-Cleaved Drosophila Dcp-1 (Asp216) (Cell Signaling Technology, 9578, Lot 5, 1:500), rabbit anti-cleaved PARP (Abcam, ab2317, Lot 559599, 1:500), mouse anti-GFP (Roche, 11814460001, Lot 14442000-30021090, 1:500), rabbit anti-GFP (Thermo Fisher Scientific, A6455, Lot 2415918, 1:500), mouse anti-ATP5A (Abcam, ab14748, Lot 2101063327, 1:500).

Drosophila brain lipid droplet imaging

Adult fly brains were dissected in PBS, fixed in 3.7% paraformaldehyde in PBS with 0.1% Triton X-100 on ice for 0.5 h, and subsequently fixed at room temperature for an additional 1 h. Following fixation, brains were rinsed in PBS with 0.1% Triton X-100 (PBST) with DAPI for 1 h and then stored at 4 °C. Before imaging, brains were stained in PBS with 0.1% Triton X-100 and BODIPY 493/503 (Thermo Fisher Scientific, D-3922, 1:1000) or LipidTOX (Thermo Fisher Scientific, H34476, 1:1000) for 0.5 h in room temperature with continuous rolling, then briefly washed with PBS and mounted using 9:1 mixture of glycerol and 1.5 M Tris-HCl, pH 8.8 between a gap formed by double layers of Scotch transparent tape on glass slide. Z-stacks aimed at the center of the optic lobe are imaged immediately within 1 h using FV3000. BODIPY 493/503 was excited using 488 nm laser with 0.01% laser power and collected at 500–550 nm emission light. To quantify the volume of LDs, representative sample images were concatenated along the T-axis and trained using Labkit (Fiji) to generate a classifier of the LDs. The classifier was used to automatically predict the LDs of all samples in a batch and generate binary masks. The volume of LDs per sample was analyzed using Bitplane Imaris 8.4.

Drosophila PAM TH+ neuron quantification

The TH+ neurons within a single PAM cluster were imaged using FV3000 with 512×512 resolution and 0.34 µm slice interval for both TH and DAPI channels, generating z-stacks. Labkit training and TH+ region prediction were performed on a computer with an NVIDIA Quadro P6000. Twenty representative z-stacks, encompassing a spectrum of TH signals and backgrounds, were vertically merged into a unified stack using Fiji/ImageJ. In Labkit (Fiji), the stack underwent more than 100 cycles of “scribbling-predicting-correcting” to train the TH+ pixel classifier against backgrounds. The resulting classifier file with a size of 379 megabytes was applied consistently for subsequent TH+ neuron analysis. For TH+ region identification, the classifier file was loaded into Labkit to batch process a folder of z-stack TH images. The defects, including noise and vacuoles on the Labkit-predicted mask images, were further refined using a customized Matlab (Mathworks) code “Correct_Labkit_TH.m” (version 1.0). The source code, instructions for installation and use, description of code dependencies, and the test data are available in Supplementary Data and also at https://github.com/rmd13/FlyBrain_PAM_TH_DAPI_Labkit/. The code sequentially processes each slice of a Labkit-generated 0/1 mask image. It smooths the mask using a Gaussian smoothing filter with a sigma of 1.5 to enhance the image quality. Following smoothing, a threshold of 0.25 is applied to the resulting image; thus, only the pixels above this threshold are labeled as a mask. The mask slice is then dilated using a square-shaped structuring element with a width of 5 pixels to fill in small gaps within the mask. Additionally, the code identifies and fills larger holes that still exist within the mask, provided these holes have an area of less than 200 pixels. Finally, the mask is eroded using the same square-shaped structuring element with a width of 5 pixels. The processed mask images were imported into Bitplane Imaris 8.4.1 as a new channel alongside the raw images. TH+ surfaces and DAPI spots were created using Imaris in the mask and DAPI channel, respectively. The number of TH+ neurons in the PAM cluster was counted as DAPI spots located within the TH+ surface by running the Bitplane Imaris 8.4’s surface distance transformation extension outside the TH+ surfaces.

Drosophila brain lipid peroxidation assay

Fly brains expressing pan-neuronal infrared fluorescent protein (mIFP) were dissected in PBS and transferred into Eppendorf tube 1 containing 100 μl of PBS. In Eppendorf tube 2, 0.5 μl of 5 mM BODIPY C11 DMSO solution (Thermo Fisher Scientific, D3861) and 0.5 μl of 1% saponin (Sigma, S7900) were added into 350 μl PBS (B), mixed, and transferred to tube 1, which was then rolled at room temperature for 30 min. Following the rolling process, brains were briefly washed in PBS and mounted with PBS on a bridged slide using two layers of Scotch transparent tape, creating a 5 mm gap. Thick nail polish was used to secure the four corners of the coverslip, and additional PBS was added to fill any gaps if necessary. BODIPY C11 was immediately imaged in FV3000 using a ×100 objective lens. The brain optic lobe was located and live-imaged using a 640 nm laser to determine the position of labeled neuronal membranes. Subsequently, images were acquired in the green channel (488 nm laser, 0.04% laser power, 515–545 nm emission) and red channel (561 nm laser, 0.02% laser power, 575–605 nm emission), respectively. Four to five images were acquired for each brain. For each group, the brains were imaged within 1 h. The lipid peroxidation index was analyzed as the ratio of the mean intensity of membrane area in the green channel to the mean intensity in the red channel using ImageJ/Fiji. For each brain, the lipid peroxidation index from four to five images was averaged. A pseudo-colored image was generated by creating a ratio channel and masking out the unstained area in Bitplane Imaris 8.4.

Drosophila daily locomotor activity assay

Fly food, extracted from a vial, was trimmed into two semi-circles with a height of 5 mm and then placed on tissue paper for a 0.5-h drying period. To administer FSG67 (HY-112489, MedChemExpress LLC), 10 µl of 0, 1, 2.5, or 5 mM FSG67 dissolved in PBS was evenly spread on the fly food until absorption. One end of a clean glass tube for locomotor activity assay was inserted down vertically into the food and capped. Anesthetized flies were individually transferred into each tube, and then the tube was blocked with a sponge on the other end. For the locomotor activity assay, these loaded glass tubes were then affixed to the DAM2 Drosophila activity monitors (TriKinetics). The locomotor activities were recorded at a frequency of 10 s over a period of one to three weeks, maintaining a 12-h light/dark cycle of LED illumination controlled via an Arduino-controlled relay by the same computer. The food was replaced every three to five days. Actograms captured during this period were analyzed using the ActogramJ plugin¹⁰⁴ in ImageJ/Fiji. Daily locomotor activities were further analyzed and plotted using Microsoft Office Excel (2010) and GraphPad Prism 8.3.0. For the other assay of flies using FSG67, the loaded tubes were placed on trays inside the incubator.

Drosophila brain short-chain fatty acids and lipidomic assay

For each genotype for short-chain fatty acids or lipidomic assay, ten fly brains per genotype were dissected in cold PBS and pooled into an Eppendorf tube on ice with 50 μl PBS. Then PBS was carefully removed under a stereomicroscope, and tubes were frozen at −80 °C. The tubes were sent to the Singapore Phenome Center (SPC) on a cryobox pre-chilled at −80 °C overnight. The total number of sample pools (n), i.e., tubes, is 1 for nSyb-QF2 and 1 for nSyb-QF2 > SNCA, with nSyb-QF2 as the control. There are no technical replicates and no biological replicates. Short-chain fatty acids and lipidomic assays were performed, and data were returned as Excel files. The detailed methods are provided in Supplementary Methods and Supplementary Table 3. The spectral data are provided in Supplementary Data 6.

Western blot

An equal number of fly brains were dissected and pooled into PBS on ice. PBS was removed, and brains were homogenized with pestles in brain lysis buffer that contains Tris-HCl, pH 8.0 10 mM, NaCl 200 mM, Triton X-100 1%, EDTA 5 mM, Glycerol 5%, PMSF 0.5 mM, DTT 1 mM, Protease Inhibitor Cocktail (Roche, 11873580001). For α-syn pSer129 blotting, Phosphatase Inhibitor Cocktail 2 (Sigma, P5726) and Phosphatase Inhibitor Cocktail 3 (Sigma, P0044) were added to brain lysis buffer (1:300). For proteinase K treatment, equal number of fly heads were homogenized in lysis buffer (50 mM Tris-HCl pH 7.5, 5 mM EDTA, 0.1% NP40, 1 mM DTT, and 1% protease inhibitor cocktail, incubated for 1 h at 25 °C, and centrifuged at 17,949×g for 1 min. The supernatants were collected and incubated for 30 min at 25 °C with proteinase K (Roche, 03115828001). For sample loading, an equal volume of 2× Laemmli loading buffer (Tris-HCl, pH 6.8, 125 mM, bromophenol blue 0.02%, SDS 4%, glycerol 20%, β-mercaptoethanol 10%) was added to the lysate and heated at 50 °C for 0.5 h. For detection of α-syn oligomers, fly brains were dissected in PBS and cross-linked in the same-day prepared buffer A (HEPES 20 mM pH 8.5, KCl 120 mM, EDTA 2 mM, sucrose 0.25 M, glycerol 1%, Dimethyl dithiobispropionimidate (DTBP) 4 mM, protease inhibitor cocktail (Roche)) on ice for 1 h. Then the reaction was halted by adding 10 mM Tris-HCl pH 8.8 for 10 min on ice. Brains were then homogenized with a pestle, and an equal volume of 2× Laemmli loading buffer was added without β-mercaptoethanol or DTT, followed by heating at 50 °C for 0.5 h. For head lysate but not brain lysate, the samples were further centrifuged at 10,621×g for 2 min, and the supernatant was collected. The samples were loaded into Mini Protean TGX (stain-free) 4–20% gel (Bio-Rad, 4568094, 4568095, 4568096) to run for 100 V 1.5 h. The gel was UV-activated and transferred onto a PVDF membrane using transfer buffer (Tris 25 mM, Glycine 192 mM, 20% methanol) at 100 V for 1.5 h. The membrane was fixed in PBS with 3.7% PFA and 0.025% glutaraldehyde for 0.5 h, stained-free imaged, and blocked in 5% milk in Tris-buffered saline with 0.1% Tween-20 (TBST). Primary antibodies used include mouse anti-α-syn (BD 610787, Lot 6092600, 1:1000), rabbit anti-Gpi (Merck, HPA024305, Lot B118974, 1:1000), rabbit anti-LDH (Thermo Fisher, PA5-26531, Lot XJ3718422A, 1:1000), rabbit anti-ATPCL (Cell signaling, 4332, Lot 2, 1:1000), rabbit anti-pSer129-α-synuclein D1R1R (Cell signaling, 23706, Lot 2, 1:1000), and secondary antibodies include anti-mouse HRP (CST, 7076S, 1:3000) and anti-rabbit HRP (Amersham, NA934-1ML, 1:3000). The HRP was detected using WesternBright ECL and Sirius (Advansta) in ChemiDoc imaging system (Bio-Rad). The membrane was subsequently re-blotted with mouse anti-α-tubulin (DSHB, 12G10, Lot 1/13/22, 1:5000). The image was analyzed using Image Lab (Bio-Rad, 6.0.0) and ImageJ/Fiji (v1.53q), and the results were plotted in GraphPad Prism (8.3.0).

Mito-roGFP2-Orp1 imaging for H₂O₂

Retina expressing mito-roGFP2-Orp1 were dissected in PBS and fixed in PBS with 3.7% PFA, 0.1% Triton X-100, and 20 mM N-Ethylmaleimide (Thermo Fisher Scientific, 23030) for 0.5 h on ice, followed by another 0.5 h at room temperature. The fixed retina was washed in PBST with 5% goat serum overnight at 4 °C and then was mounted with Prolong Gold anti-fade mountant (Thermo Fisher Scientific, P10144). Mito-roGFP2-Orp1 was imaged by exciting with a 405 nm or 488 nm laser, respectively, and collecting 500–530 nm emission light. In all, 405 nm laser was used for the H₂O₂-bound form and 488 nm laser for the H₂O₂-free form of the sensor. The H₂O₂ index was determined by the ratio of the mean intensity of the mitochondria in the two channels using ImageJ/Fiji.

MitoTimer imaging

Brains were dissected in PBS and fixed in PBS with 3.7% PFA and 0.1% Triton X-100 for 0.5 h on ice and an additional 0.5 h at room temperature. Following fixation, the brains were washed in PBST with 5% goat serum and DAPI overnight at 4 °C and then mounted with Prolong Gold anti-fade mountant (Thermo Fisher Scientific, P10144). The green channel (488 nm laser, 500–540 nm emission) and the red channel (561 nm laser, 570–620 nm emission) were imaged on the cortex layer of the optic lobe. The ratio of the mean intensity of MitoTimer between the red and the green channel was analyzed using ImageJ/Fiji.

Drosophila climbing assay

The climbing assay was conducted on a customized workstation composed of a background white foam, side LED light, bottom light, 50 ml glass cylinders, an elastic cushion, and a Nikon digital camera securely affixed to a tripod. A sponge pad was positioned in the inner bottom of the cylinder to minimize potential harm to the flies during the climbing assay. In total, 10–20 flies were flipped into a capped cylinder, tapped to the bottom, and allowed to climb for 5 s, following which an image was promptly captured using the camera. The average climbing height for all the images was analyzed using a customized FlySpotter program (Srinivas Gorur-Shandilya, https://github.com/sg-s/fly-spotter) and graphically represented using GraphPad Prism.

Real-time PCR

Fly brains were dissected in PBS, with trachea sacs and fat bodies detached, and were subsequently transferred into RNAlater (Thermo Fisher Scientific, R0901). Alternatively, heads or bodies were collected. Following removal of excess fluid, RNA was extracted using TRIzol™ Reagent (Thermo Fisher Scientific, 15596026). The mRNA concentrations were measured using NanoDrop 2000 Spectrophotometers, and the concentration was adjusted by adding RNase-free water (Thermo Fisher Scientific, AM9932) to reach an equal concentration across all samples. The mRNA was then treated with dsDNase (Thermo Fisher Scientific, EN0771). Subsequently, cDNA was prepared using the Invitrogen SuperScripII First-Strand Synthesis System for RT-PCR (Thermo Fisher Scientific, 11904-018). qPCR was performed using iQ™ SYBR Green Supermix (Bio-Rad, 1708882) and the LightCycler 96 Instrument (Roche). The data was analyzed using LightCycler 96 software (Roche). RPL32 served as a single internal control for mRNA quantification of experiments not involving α-syn. RPL32 and ed served as double internal controls for mRNA quantification of experiments involving α-syn.

Public RNA-sequencing dataset illustration

The expression levels for genes of interest, as well as cell identities, were extracted from the FlyCellAtlas Drosophila head RNA-sequencing dataset (https://flycellatlas.org/) and additional RNA-sequencing datasets GSE223626, GSE218661^61–63 using HDF5viewer (3.1.4). Raw read counts from FlyCellAtlas were normalized per cell when necessary. The combined expression data was imported into Excel and arranged by cell types, and imported to Matlab (MathWorks R2016a) for visualization using the function imagesc.

Synuclein preformed fibril preparation

In total, 15 μl of 2 mg/ml synuclein PFF (StressMarq SPR-324) aliquoted into screw-capped tubes were sonicated in an ultrasonic bath sonicator for 2.5 h at the following conditions: Sweep, 37 Hz frequency, 60% power. Water was changed at 30-min intervals to maintain the water temperature at 15 °C.

Statistics and reproducibility

For mouse experiments, the biological or technical replicates are detailed in the figure legend. For each experiment with fruit flies, the design involves an experiment using multiple biological sample replicates for each group to ensure robustness of our findings. No statistical method was used to predetermine sample size. No data were excluded from the analyses except that for the α-syn modifier screening where data distributions were not normal in over 90% of groups, only the 25% to 75% percentile of data points were utilized for testing to identify candidate genes. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment, with the exception of collecting data for α-syn modifier genetic screening, short-chain fatty acids, and lipidomic analysis. For fly experiments, every single fly is a biological replicate that contributes to a single data point of the plot, unless otherwise indicated in the figure legend. For each plot or experiment, the number of flies (n) used for multiple groups of genotypes and chemical treatments are not identical between groups, and these numbers are indicated in the figure legend. All the statistics are two-sided. Statistical analyses were conducted using GraphPad Prism 8.3.0 and Matlab (MathWorks R2016a). For the genetic screening, the Anderson-Darling test was applied to assess the non-normal distribution of all the data, followed by the Mann–Whitney U test. The P values were adjusted by the Benjamini–Hochberg method using the Matlab function padj = PVAL_ADJUST (p, ‘BH’), where p is the input vector of all the P values (fakenmc/pval_adjust, version 1.2.0.2, by Nuno Fachada. https://github.com/nunofachada/pval_adjust). Apart from genetic screening, data distribution was not formally tested to be normal or not. For data analysis of multiple groups, one-way ANOVA or two-way ANOVA was used to determine significant differences among multiple groups of interest, unless otherwise indicated in the figure legend. This is followed by Tukey’s post hoc test for comparisons between two groups, or Dunnett’s post hoc test for comparisons between a selected control group such as α-syn alone with other groups such as α-syn with other treatments or genetic manipulations. Data are presented as mean ± SD or a violin plot. Significance levels are *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

M.R. and K.L.L. participated in the design of experiments, data analysis, and interpretation. W.T. and G.G.Y.L. designed and performed the experiments using primary mouse neurons. M.R. and K.L.L. wrote the manuscript with inputs from all the authors. We are grateful to the Singapore Phenome Center (SPC) for their expertise and for providing the methodological details and technical support for the metabolomics and lipidomics analyses.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

The genetic screening data (Supplementary Data 1), raw short-chain fatty acids excel data (Supplementary Data 2), raw lipidomics excel data (Supplementary Data 3), the Matlab code, Labkit classifier file and test images (Supplementary Data 4), the spectral data of lipidomic and short-chain fatty acids (Supplementary Data 6) generated in this study are provided as Supplementary Data. The numeric data used for plotting the graphs in the figures and Supplementary figures, and the raw western blot images are provided in Source Data. The Graphpad Prism pzfx files used to generate the plots also with the statistics comparisons information tabs are provided in Supplementary Data 5. The sources of RNA sequencing datasets for the gene expression profiles in Supplementary Fig. 4e–i are cited and publicly available. Source data are provided with this paper.

Code availability

The Labkit classifier file for PAM cluster neuron prediction and the code for the correction of Labkit-predicted TH+ regions can be downloaded from Github (https://github.com/rmd13/FlyBrain_PAM_TH_DAPI_Labkit). We also provided the code, Labkit classifier file and test images generated in this study in Supplementary Data 4. The code was under MIT license.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68325-3.