GBA1 deficiency negatively affects physiological α-synuclein tetramers and related multimers

To explore whether GBA1 deficiency could affect the formation of α-syn tetramers and related multimers, we generated human neuroblastoma SH-SY5Y cell lines with depletion of GBA1 using the CRISPR/Cas9 system without any desirable off-target effects (SI Appendix, Fig. S1A). These GBA1 KO cells lead to no detectable levels of GBA1 protein, RNA and enzymatic activity, and the subsequent accumulation of α-syn in both whole-cell lysates and lysosome-enriched fractions (SI Appendix, Fig. S1). Consistent with prior studies using GBA1-depleted cells (20, 21), depletion of GBA1 results in accumulation of lysosomal associated membrane protein 1 (LAMP-1) but not the endoplasmic reticulum (ER) stress marker, 78 kDa glucose-regulated protein (GRP78), suggesting that depletion of GBA1 contributes to lysosomal dysfunction but not to ER stress (SI Appendix, Fig. S1B). Since it is difficult to distinguish between the α-syn monomer and helical tetramers using native gel electrophoresis (22), we employed an intact cell cross-linking method that allows detection of the α-syn tetramers with 1 mM disuccinimidyl glutarate (DSG), a membrane-permeable cross-linker, as previously described (11, 12). Since detergents for cell lysis could profoundly affect destabilization of the tetramers, GBA1 KO cells were lysed by sonication as previously described (23). In the GBA1 KO cells, the cross-linked α-syn tetramers (60 kDa) and related multimers (80 and 100 kDa) are decreased, whereas the amount of free monomer is increased compared with control cells (Fig. 1 A and B and SI Appendix, Fig. S2). Similar results are observed in primary cortical neurons carrying the heterozygous L444P GBA1 mutation (SI Appendix, Fig. S3), the most common GBA1 mutation in neuronopathic Gaucher disease (GD) (14, 15). α-Syn 60-, 80-, and 100-kDa species (tetramers/multimers) detected in cells are different from the soluble toxic oligomer species, because soluble oligomeric species of α-syn could only be detected in cells overexpressing α-syn (SI Appendix, Fig. S4). To confirm cross-linking efficiency, dimerization of DJ-1 was assessed in DSG-treated lysates as a positive control (23, 24). DJ-1 is efficiently associated into a homodimer. Unexpectedly, dimeric and monomeric DJ-1 forms are increased, whereas the ratio of DJ-1 dimer to monomer is not altered in GBA1 KO cells compared with control cells (Fig. 1C and SI Appendix, Fig. S2).

GBA1 deficiency primarily leads to pathological accumulation of GSLs in the lysosome due to loss of GBA1 enzymatic activity (18, 20). Importantly, lipid-binding motifs of α-syn protein are thought to be critical for the formation of tetramers and related multimers (11). To examine whether accumulation of GSLs due to GBA1 deficiency influences levels of α-syn tetramers, we assessed the relationship between intracellular GlcCer levels and α-syn tetramer levels in the presence or absence of miglustat, an inhibitor of GlcCer synthase (25). Consistent with previous studies in GBA1-deficient cells (18), GBA1 KO cells have a 1.73-fold and 2.37-fold increase in intracellular GlcCer levels as assessed by immunofluorescence analysis for anti-GlcCer antibody and quantitative lipidomics analysis, whereas treatment with 100 μM miglustat results in a reduction in the levels of intracellular GlcCer to near-normal status (Fig. 2 A and B and SI Appendix, Fig. S5 A and B). Total ceramide levels are not significantly but slightly decreased in GBA1 KO cells (SI Appendix, Fig. S5 C and D), whereas its species composition shows a similar pattern in both control and GBA1 KO cells treated with or without miglustat (SI Appendix, Fig. S5 E and F). Importantly, the GBA1 deficiency-induced reduction of tetramers and related multimers is significantly recovered by treatment with miglustat in GBA1 KO cells (Fig. 2 C and D), but the ratio of DJ-1 dimer to monomer is not altered (Fig. 2E). Together, these data suggest that GSL accumulation due to GBA1 deficiency contributes to the destabilization of α-syn tetramers and related multimers.

To extend our findings, we generated hDA neurons carrying a GBA1 mutation. To this end, PD patient skin fibroblasts harboring the GBA1 mutation (N370S/WT GBA1-PD) were obtained from Coriell (SI Appendix, Table S1) and reprogrammed using standard techniques (26). Immunostaining analysis with control and GBA1-PD iPSC lines shows expression of alkaline phosphatase (AP) and pluripotency protein markers (SI Appendix, Fig. S6A). No differences are observed in the levels of protein and RNA of pluripotency markers between all iPSC and human embryonic stem cells (H9) (SI Appendix, Fig. S6 B–D). Furthermore, we confirmed normal euploid karyotypes (SI Appendix, Fig. S6E) and the formation of teratomas (SI Appendix, Fig. S6F). Importantly, all iPSC-derived hDA neurons show a similar differentiation potential (SI Appendix, Fig. S7 A–D) and comparable protein and RNA levels of neuronal and hDA markers including tyrosine hydroxylase (TH), TUJ1, MAP2, PITX3, Nurr1, and FOXA2 (SI Appendix, Fig. S7 E–G). Notably, no significant differences are observed in dopamine release (SI Appendix, Fig. S8A) and electrophysiological properties among hDA neurons differentiated from control and GBA1-PD iPSC lines (SI Appendix, Fig. S8 B–H).

Since reduction in GBA1 protein levels and activity are a feature of sporadic PD and GBA1-PD (27), we determined GBA1 protein levels and enzyme activity and α-syn levels in PD patient fibroblasts, iPSC, neuronal precursors, and hDA neurons. GBA1 protein levels and enzyme activity are profoundly reduced in all of these cells with the GBA1 N370S mutation compared with control-derived cells (Fig. 3 A and B and SI Appendix, Figs. S9 and S10A). These results are similar to the finding in postmortem PD brains carrying a heterozygous GBA1 mutation and hDA neurons differentiated from iPSC harboring a heterozygous GBA1 mutation (16, 28). As described previously (17), α-syn protein is only detected in iPSC and neuronal precursors, and the protein levels are higher in GBA1-PD iPSC and neuronal precursors compared with control (SI Appendix, Fig. S9 D, E, G, and H). Interestingly, the level of α-syn protein is markedly higher in hDA neurons differentiated from GBA1-PD iPSC compared with hDA neurons differentiated from the controls (Fig. 3A and SI Appendix, Fig. S10B), whereas the level of TH protein is not changed in all hDA neurons (Fig. 3A and SI Appendix, Fig. S10C), suggesting that GBA1 N370S perturbs α-syn processing (Fig. 3A and SI Appendix, Fig. S10B). Intriguingly, α-syn protein is not detected in fibroblasts (SI Appendix, Fig. S9A). Next, GSL levels were assessed via liquid chromatography mass spectrometry (LC-MS) using hDA neurons at 60 d in vitro (DIV 60) (29). The analysis shows a 1.87-fold increase in GBA1–PD1-1 and a 2.08-fold increase in GBA1-PD2 compared with control hDA neurons (Fig. 3C and SI Appendix, Fig. S11A). Interestingly, total ceramide levels are slightly decreased in both GBA1–PD1-1 and GBA1-PD2 compared with controls (Fig. 3D and SI Appendix, Fig. S11B), whereas ceramide species composition shows a similar pattern among these hDA neurons (SI Appendix, Fig. S11 C and D). Since accumulated GlcCer levels were observed, we sought to determine whether the accumulation of GlcCer alters the ratio of α-syn tetramers and related multimers to free monomers. Strikingly, DSG cross-linking analysis using hDA neurons reveals that the ratios of α-syn tetramers and related multimers to free monomers are significantly decreased in all GBA1-PD–derived hDA neurons, suggesting that destabilization of the tetramers and related multimers is correlated with the levels of GlcCer (Fig. 3 E and F). No difference is observed in the ratio of DJ-1 dimer to monomer (Fig. 3 E and G) and in the ratio of TH tetramers to monomers between GBA1-PD and control hDA neurons (Fig. 3E).

Since our findings demonstrate that GlcCer accumulation destabilizes the α-syn tetramers and related multimers, which subsequently leads to accumulation of the free monomer of α-syn in GBA1-PD hDA neurons, we next sought to determine whether miglustat could increase α-syn tetramers and related multimers in GBA1-PD hDA neurons through reducing GlcCer synthesis. For effective inhibition of GlcCer synthesis, hDA neurons were treated with 100 μM miglustat as the concentration did not affect the GBA1 protein levels and enzyme activity as well as TH protein levels (SI Appendix, Fig. S12). DSG–cross-linking analysis shows that treatment with miglustat significantly reverses the destabilization of the α-syn tetramers and related multimers due to GBA1 deficiency-mediated GSL accumulation as assessed by Western blot analysis (Fig. 4 A and B). Similarly, augmentation of GBA1 protein via lenti-viral transduction also recovers the ratio of α-syn tetramers and related multimers to monomers (Fig. 4 D and E). Interestingly, the increases in DJ-1 level due to GBA1 deficiency are decreased in GBA1-overexpressed GBA1-PD hDA neurons (Fig. 4D) but not miglustat-treated GBA1-PD hDA neurons (Fig. 4A). On the other hand, the ratio of DJ-1 dimer to monomer is similar in both miglustat-treated and GBA1-overexpressed GBA1-PD hDA neurons compared with control hDA neurons (Fig. 4 C and F). No differences in TH protein levels and the ratio of TH tetramers to monomer are observed in all groups (Fig. 4 A and D).

We then determined whether reduced free monomers of α-syn by miglustat treatment rescues the pathology induced by preformed fibrils (PFF) (30). To this end, control and GBA1-PD hDA neurons at DIV 60 were treated with 100 μM miglustat 3 d before PFF treatment (5 μg/mL). Ten days after PFF treatment, the levels of phospho-Ser129 α-syn (p-α-syn), a marker for pathologic α-syn, are significantly increased in GBA1-PD hDA neurons with vehicle treatment, while treatment of miglustat significantly decreases the levels of p-α-syn in GBA1-PD hDA neurons as assessed by immunofluorescence analysis with anti–p-α-syn antibody (Fig. 5 A and B). For further analysis of the levels of p-α-syn, the control and GBA1-PD hDA neurons were fractionated into Triton X-100 (TX)-soluble and -insoluble fractions. TH levels are decreased in both control and GBA1-PD hDA neurons with PFF treatment in TX-soluble fractions. Notably, the reduction in TH levels are greater in GBA1-PD hDA neurons treated with PFF compared with that in control hDA neurons treated with PFF (Fig. 5C and SI Appendix, Fig. S10E). Importantly, the reduction in TH levels is partially but significantly recovered in both control and GBA1-PD hDA neurons by treatment of miglustat in the TX-soluble fraction (Fig. 5C and SI Appendix, Fig. S10E). However, endogenous α-syn levels are unchanged and p-α-syn levels are barely detectable in the TX-soluble fraction (Fig. 5C and SI Appendix, Fig. S10D). Importantly, PFF treatment leads to accumulation of abnormal α-syn aggregate species (Fig. 5D and SI Appendix, Fig. S10F) and an increase in the level of p-α-syn (Fig. 5D and SI Appendix, Fig. S10G) in the TX-insoluble fraction in both control and GBA1-PD hDA neurons, whereas miglustat reduces the accumulation of the TX-insoluble α-syn aggregate species (Fig. 5D and SI Appendix, Fig. S10F) and p-α-syn (Fig. 5D and SI Appendix, Fig. S10G). On the other hand, the accumulation in α-syn aggregate species and p-α-syn levels in the TX-insoluble fraction is greater in GBA1-PD hDA neurons treated with PFF compared with those in PFF-treated control hDA neurons. Next, we monitored neuronal toxicity induced by PFF treatment in hDA neurons through AlamarBlue and lactate dehydrogenase (LDH) assays. PFF treatment for 2 wk induces neuronal toxicity in control and GBA1-PD hDA neurons. There is a greater neuronal toxicity in GBA1-PD hDA neurons treated with PFF compared with that in PFF-treated control hDA neurons (Fig. 5 E and F). Importantly, the enhanced PFF-induced neuronal toxicity in the GBA1-PD hDA neurons is significantly protected by treatment of miglustat (Fig. 5 E and F). Similar results are observed in hDA neurons augmenting GBA1 protein via lentiviral transduction (SI Appendix, Fig. S13). Taken together, these data indicate that lowering pathogenic GSL accumulation stabilizes α-syn tetramers and related multimers with a decrease in free monomers, which is protective.

GBA1 KO cells, mouse primary cortical neurons, and hDA neurons were suspended in PBS (pH 8.0) followed by three washes with PBS (pH 8.0). Washed cells were incubated in 1 mM DSG for 30 min at 37 °C with 500 rpm shaking in Eppendorf ThermoMixer C. This reaction was quenched with 20 mM Tris (pH 7.4) for 15 min at room temperature. Complete protease inhibitor mixture (Roche) was added to DSG–cross-linked cells followed by sonication for 15 s (1 s pulse on/off) at 10% amplitude (Branson Digital sonifier). After sonication, the cell lysates were centrifuged at 100,000 × g at 4 °C for 1 h. The generation of GBA1 KO cells, primary cortical neuronal cultures, and detailed biochemical studies are described in SI Appendix, SI Materials and Methods.

Six iPSC lines were generated from the skin fibroblast of control and heterozygous N370S GBA1-PD obtained from the Coriell Institute for Medical Research (SI Appendix, Table S1). All experiments using human stem cells were monitored by The Johns Hopkins University Institutional Stem Cell Research Oversight Committee. Further characterization and differentiation procedures are described in SI Appendix, SI Materials and Methods.

Student’s t test was used in GBA1 KO characterization. One-way ANOVA with Tukey’s post hoc test was used for immunostaining quantification of GlcCer, the ratio of multimers to monomer in hDA neurons and miglustat-treated GBA1 KO cells, GBA1 enzymatic activity, dopamine release, and lipidomics. Two-tailed t test was used for the comparison of recording data of electrophysiology. Two-way ANOVA with Bonferroni post hoc test was used for immunoblotting quantification of hDA neuron differentiation, the ratio of multimers to monomer in miglustat-treated and LV-GBA1–transduced hDA neurons, p-α-syn staining quantification, and cell death assay. Assessments with P < 0.05 were considered significant. Statistical calculations were performed with GraphPad Prism software, Version 5.0 (www.graphpad.com).