Inhibition or genetic reduction of ASAH1/acid ceramidase restore α-synuclein clearance in mutant GBA1 dopamine neurons from Parkinson’s patients

Abstract

Bi-allelic mutations in GBA1, a gene that encodes the lysosomal enzyme β-glucocerebrosidase (GCase), cause Gaucher disease (GD). Although GD carriers do not exhibit clinical manifestations, GBA1 mutations are the highest risk factor for Parkinson’s disease (PD) in GD patients and carriers of the disease [1–5]. GCase breaks down glucosylceramide (GluCer), a sphingolipid that accumulates in GD. GluCer is deacylated by the lysosomal enzyme acid ceramidase (ACDase) to glucosylsphingosine (GluSph) [6–8]. GluSph is neurotoxic and accumulates to high levels in neuronopathic GD brains [9, 10]. However, whether this metabolic pathway involving ACDase plays a role in GBA1-associated PD (GBA1/PD) is not known. In this report we used induced pluripotent stem cells (hiPSCs) from PD patients harboring heterozygote GBA1 mutations to examine the role of ACDase in promoting α-synuclein accumulation and aggregation, a hallmark of PD. Compared to isogenic controls, hiPSC-derived PD dopamine (DA) neurons had elevated levels of pathogenic α-synuclein species. There was also reduced nuclear localization of transcription factor EB (TFEB), impaired autophagy, and decreased levels of cathepsin D (CathD), a lysosomal protease involved in α-synuclein degradation [11]. Treatment of the mutant DA neurons with a number of different ACDase inhibitors, or CRISPR/Cas9 knockdown (KD) of the ASAH1 gene, reversed all the phenotypic abnormalities of the mutant DA neurons. We conclude that in GBA1/PD-DA neurons, ACDase contributes to deregulation of key nodes of the autophagy/lysosomal pathway (ALP) involved in α-synuclein clearance. Our results suggest that ACDase is a potential therapeutic target for treating GBA1-associated PD.

Graphical Abstract

Graphical Abstract.

Introduction

Mild mutations in GBA1 cause type 1 Gaucher disease (GD), which is characterized by the appearance of lipid-engorged macrophages [12] due to the inability of these cells to hydrolyze GluCer on the surface of phagocytosed RBCs [13–16]. These Gaucher macrophages lodge in the bone marrow, liver and spleen, leading to hepato-splenomegaly, anemia, thrombocytopenia, and bone manifestations [13, 17]. Enzyme replacement therapy (ERT), and substrate reduction therapy (SRT) with GluCer synthase inhibitors reverse these abnormalities, although skeletal symptoms are more refractory to treatment [18]. In contrast to mild GD mutations, severe GD mutations cause in addition to visceral abnormalities, neuronopathy that is fatal before age 2 (type 2 GD), or chronic (type 3 GD) [19, 20]. GD carriers do not exhibit clinical abnormalities, but these individuals and type 1 GD patients have a 5%–20% increased risk of developing Parkinson’s disease (PD) and Lewy Body Dementia (LBD) [21, 22]. While in types 2 and 3 GD there is neuronal loss in wide areas of the brain, for reasons that are not understood, in GBA1-associated PD (GBA1/PD), the substantia nigra pars compacta is predominantly affected, with progressive loss of dopamine (DA) neurons affecting motor and non-motor functions [23–25]. GBA1/PD exhibits earlier onset and more severe symptoms compared with idiopathic PD [4, 26].

It is well known that sphingolipid accumulation plays a major role as a driver of CNS pathology in neuronopathic GD (nGD). In type 2 GD, there is up to a 500-fold elevation in the neurotoxic sphingolipid GluSph [8, 27, 28]. Although types 1/2 GD and GBA1/PD have different clinical characteristics, the common denominator is GCase deficiency [29], suggesting the possibility that glucosphingolipid alterations may also play a role in GBA1/PD [29–32]. Brains of patients with GBA1/PD and idiopathic PD have lower GCase activity, which has been associated with reduced clearance of α-synuclein aggregates, further supporting the idea that abnormal increases in glucosphingolipids may lead to PD [33–35]. There is also substantial evidence that in addition to loss of function, gain-of-function mechanisms leading to GD and GBA1/PD play an important role in these disorders. These include disruption of proteostatic and transport mechanisms by the mutant GCase protein that are essential for neuronal survival, including endoplasmic reticulum-associated degradation (ERAD) and the unfolded protein response (UPR) [23, 36–38].

GCase deficiency has pleiotropic effects deregulating homeostatic functions in many cell types including macrophages, lymphoid and other hematopoietic cells, microglia, osteoblasts, and neuronal cells [39–45]. Mutations in GBA1 have been shown to interfere with vesicle trafficking during autophagy, autophagic flux, mitochondrial function, and calcium homeostasis [43, 46]. In addition to the deleterious effects of GBA1 mutations on lysosomal and mitochondrial functions, GCase deficiency has been shown to deregulate developmental networks, including the Wnt/β-catenin pathway in neurons and osteoblasts [42, 47–52].

α-synuclein is a 14 KDa protein that is involved in neurotransmitter release by synaptic vesicles, and its pathologic aggregation into high MW complexes in Lewy bodies is a hallmark of synucleinopathies including PD and LBD [53, 54]. This leads to neuronal death, in particular in the substantia nigra pars compacta. The factors that destabilize the normal conformation of this aggregation-prone, prion-like protein are not well understood, and there is much effort in identifying treatments that can reverse the aggregation or increase the clearance of pathogenic α-synuclein species. Lysosomes regulate vital metabolic functions, remove damaged organelles, recycle nutrients through autophagy, and clear α-synuclein aggregates [55–57]. A major regulator of these lysosomal functions is mTORC1 [58]. This is a multi-subunit complex on the lysosomal surface that acts as an energy and nutrient sensor. mTORC1 regulates anabolism, catabolism, and critical functions of the ALP. mTOR, a Ser/Thr kinase within this complex regulates the steady-state phosphorylation of transcription factor EB (TFEB), the master regulator of lysosomal and autophagy genes [59–63]. mTOR phosphorylation of TFEB retains this transcription factor in the cytoplasm, whereas TFEB dephosphorylation results in its transport to the nucleus [64]. In the nucleus, TFEB induces the expression of genes involved in lysosomal biogenesis, autophagy, and it upregulates the expression of hydrolases including cathepsins, which degrade α-synuclein aggregates [65–67]. We previously reported that in neurons from types 2 and 3 GD, there was mTOR kinase hyperactivation and a concomitant decrease in TFEB levels and activity, resulting in autophagy defects [68]. These abnormalities were prevented by incubation of the mutant neurons with mTOR inhibitors. The abnormal phenotype of neurons from neuronopathic GD was directly related to the accumulation of GluCer and GluSph, as incubation of these cells with inhibitors of GluCer synthase or ACDase prevented mTOR hyperactivation, and rescued lysosomal biogenesis and autophagy [69, 70]. Furthermore, incubation of WT neurons with GluSph recapitulated the mTOR hyperactivation and lysosomal abnormalities of mutant neurons, which were prevented by co-incubation of GluSph with mTOR inhibitors in both, neuronopathic GD and GBA1/PD models [69, 71]. These reports suggest that this lipid is a key metabolite capable of interfering with mTORC1-dependent lysosomal homeostasis, leading to pathogenic changes in neurons.

A key question that arises is whether these findings from GD studies are applicable to the mechanisms leading to GBA1/PD. Recent studies appear to support overlapping pathogenic mechanisms between GD and GBA1/PD. Using hiPSC-derived DA neurons from GBA1/PD patients and isogenic controls, Kumar et al. reported that in these cells, there was also mTOR hyperactivation, autophagic flux abnormalities, and increased levels of pathogenic α-synuclein species, and that inhibitors of GluCer synthase and ACDase rescued the mutant phenotype [71]. Mubariz et al reported similar findings [72]. About 7% of patients with PD exhibit GBA1 mutations, but not all individuals with these mutations develop PD [73]. The low penetration of GBA1 mutations in type 1 GD and in GD heterozygotes results in a low (10%) incidence of GBA1/PD [74]. Thus, in addition to GCase deficiency, genetic background, age, environmental and other epigenetic factors, are likely to be involved in the onset and progression of PD [4, 75]. Nevertheless, the use of a stringent isogenic hiPSC system where the only variable was a heterozygote GBA1 mutation suggests a role of GCase deficiency in the abnormal phenotype of GBA1/PD neurons [43, 71]. While both, GluCer synthase and ACDase inhibitors reversed the mutant phenotype of GBA1/PD neurons, ACDase inhibition by carmofur, a potent inhibitor of this enzyme, was sufficient to rescue the mutant phenotype, suggesting that ACDase activity was an important determinant in the GBA1 pathogenic cascade [71]. A role of ACDase in GD and GBA1/PD had also been reported in in vitro and animal models including mice and Zebrafish [6, 76, 77].

Given the importance of ACDase as a potential therapeutic target for synucleinopathies linked to GBA1 mutations, we undertook a more extensive study of the consequences of pharmacological and genetic inhibition of ACDase/ASAH1 in DA neurons from GBA1/PD patients. We report that in addition to carmofur, 3 other different ACDase inhibitors, as well as ASAH1 knockdown (KD), reversed the phenotypic alterations we observed in the mutant GBA1/PD neurons compared to isogenic controls. Our results strongly suggest that ACDase/ASAH1 is a potential therapeutic target for the prevention, amelioration or treatment of GBA1/PD.

Results

In the present study we used hiPSC-derived midbrain DA neurons (GBA1/PD-DA) from three different PD patients with heterozygote GBA1 mutations, namely RecNciI/WT, L444P/WT, N370S/WT, and the corresponding gene-edited WT/WT isogenic controls. These cells were then differentiated into DA neurons and stained with antibodies to the DA marker Tyrosine hydroxylase (TH) and the neuronal marker Tuj1 (Fig. S1A–C) as described in M&M. The differentiation efficiency to DA neurons for all 3 mutant genotypes and isogenic controls was about 70–80% (Fig. S1A–C). As shown in Fig. S1D, the mutant RecNciI/WT and L444P/WT DA neurons had about a 40%–50% reduction in the levels of GCase enzymatic activity compared to the corresponding isogenic controls, whereas N370S/WT neurons had about 70% activity compared to controls. The higher GCase activity in the N370S mutant is consistent with the milder severity of this mutation.

ACDase inhibitors lower ⍺-synuclein aggregation in GBA1/PD-DA neurons

To examine the effect of ACDase inhibitors on α-synuclein aggregation, RecNciI/WT, L444P/WT, and N370S/WT PD-DA neurons, along with their respective isogenic gene-corrected (GC) controls, were differentiated into DA neurons, and the levels of phospho-α-synuclein/S129 (α-synuclein/pS129) monomer and oligomer were analyzed by immunoblot using antibodies to α-synuclein/pS129. As shown in Fig. 1A, D, G, the mutant GBA1/PD-DA neurons had elevated levels of monomeric and oligomeric α-synuclein/pS129 compared to isogenic controls. After initial dose optimization, the 3 mutant DA lines were treated with the ACDase inhibitors 22 m (10 μM), ARN14988 (10 μM), and LCL521 (20 μM) during the last 10 days of DA differentiation. WB analysis showed that there were significant increases in the levels of high MW oligomers and monomers of α-synuclein/pS129 in RecNciI/WT (Fig. 1A–C), L444P/WT (Fig. 1D–F), and N370S/WT PD (Fig. 1G–I) DA neurons, and that 22 m treatment caused a reduction in these α-synuclein/pS129 species compared to untreated mutant GBA1/WT lines. Similar results were observed with two other ACDase inhibitors, ARN14988 (Fig. S2a) and LCL521 (Fig. S2b).

Figure 1.

ACDase inhibitor 22 m lowers ⍺-synuclein aggregation in GBA1/PD-DA neurons. RecNciI/WT, L444P/WT, and N370S/WT PD-DA neurons were either left untreated or incubated with the ACDase inhibitor 22 m (10 μM) during the last 10 days of DA differentiation as indicated. (A) Representative immunoblot showing ⍺-SYN-pS129 oligomer (top panel) and monomer (middle panel) in RecNciI/WT and isogenic gene-corrected (GC) control neurons. (B and C), quantitation of immunoblots: (B) high MW and (C) low MW ⍺-SYN-pS129 in RecNciI/WT and isogenic GC control DA neurons, normalized to actin. (D) Representative immunoblot showing ⍺-SYN-pS129 oligomer (top panel) and monomer (middle panel) in L444P/WT and isogenic GC control DA neurons. (E and F), quantitation of immunoblots: (E) high MW and (F) low MW ⍺-SYN-pS129 in L444P/WT and isogenic GC control DA neurons, normalized to actin. (G) Representative immunoblot showing ⍺-SYN-pS129 oligomer (top panel) and monomer (middle panel) in N370S/WT and isogenic GC control DA neurons. (H and I), quantitation of immunoblots: (H) high MW and (I) low MW ⍺-SYN-pS129 in N370S/WT and isogenic GC control DA neurons, normalized to actin. All plots represent the results from three independent experiments (n = 3). Error bars represent mean ± SEM. p-values were determined using unpaired Welch’s t test. Asterisks indicate the level of statistical significance: ^*P ≤ 0.05, ^**P ≤ 0.01, ^***P ≤ 0.001, ^****P ≤ 0.0001. Numbers at the right of immunoblots refer to MW markers.

We conclude from these results that the ACDase inhibitors 22 m, ARN14988, and LCL521 prevent aggregation and lower the levels of α-synuclein/pS129 in GBA1/PD-DA neurons. These results are similar to those previously reported using carmofur as an ACDase inhibitor [71]. The fact that 3 other different ACDase inhibitors were all effective in reducing levels of pathogenic α-synuclein species, should alleviate concerns that the previous results obtained with carmofur and the present results are due to off-target effects.

CRISPR/Cas9 knockdown of ASAH1 in heterozygote GBA1/PD-DA neurons prevents the formation of ⍺-synuclein aggregates

We then sought genetic evidence that the results obtained by pharmacological inhibition of ACDase were indeed due to a role of this protein in the elevation of pathogenic α-synuclein species. To this end, we used CRISPR/Cas9 to knock down (KD) ASAH1, the gene that encodes ACDase, in the RecNciI/WT, L444P/WT, and N370S/WT hiPSCs lines, followed by differentiation to DA neurons. WB analysis using anti-ACDase antibodies showed that ASAH1 KD resulted in ~ 50% reduction in ACDase in DA neurons derived from all three mutant GBA1/PD lines (Fig. S3A–F). Interestingly, we were unable to isolate clones with a reduction in ACDase greater than 40%–50%, suggesting that this enzyme may be essential for survival of the hiPSC clones used for CRISPR/Cas9 gene-editing. WB analysis revealed a significant reduction in α-synuclein/pS129 high MW oligomers and monomers in KD RecNciI/WT (Fig. 2A–C), KD L444P/WT (Fig. 2D–F), and KD N370S/WT (Fig. 2G–I) DA neurons, compared to the parental lines before KD.

Figure 2.

ASAH1 knockdown reduces ⍺-synuclein aggregation in GBA1/PD-DA neurons. RecNciI/WT, L444P/WT, and N370S/WT hiPSCs were subjected to ASAH1 gene knockdown (KD) using the CRISPR/Cas9 system. The KD hiPSC, parental, and GC control lines were then differentiated into DA neurons. (A) Representative immunoblot of ⍺-SYN-pS129 oligomer (top panel) and monomer (middle panel) in RecNciI/WT-ASAH1 KD, parental RecNciI/WT, and isogenic GC control DA neurons. (B and C), quantitation of immunoblots: (B) high MW and (C) low MW ⍺-SYN-pS129, normalized to actin. (D) Representative immunoblot of ⍺-SYN-pS129 oligomer (top panel) and monomer (middle panel) in L444P/WT-ASAH1 KD, parental L444P/WT, and isogenic GC control DA neurons. (E and F), quantitation of immunoblots: (E) high MW and (F) low MW ⍺-SYN-pS129, normalized to actin. (G) Representative immunoblot of ⍺-SYN-pS129 oligomer (top panel) and monomer (middle panel) in N370S/WT-ASAH1 KD, parental N370S/WT, and isogenic GC control DA neurons. (H and I), quantitation of immunoblots: (H) high MW and (I) low MW ⍺-SYN-pS129, normalized to actin. All plots represent the results from three independent experiments (n = 3). Error bars represent mean ± SEM. p-values were determined using unpaired Welch’s t test. Asterisks indicate the level of statistical significance: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.

We conclude from these results that ASAH1 KD significantly lowers the levels of α-synuclein/pS129 in GBA1/PD-DA neurons harboring different mutations in GBA1. The results obtained by genetic reduction of ASAH1 are in agreement with those obtained by pharmacological inhibition of ACDase, providing strong support to the idea that ACDase plays a key role in the elevation of α-synuclein in GBA1/PD-DA neurons.

ACDase inhibition and ASAH1 knockdown in GBA1/PD-DA neurons increase nuclear localization of TFEB

TFEB is a major regulator of lysosomal function through transcriptional activation of the Coordinated Lysosomal Expression and Regulation (CLEAR) network of lysosomal and autophagy genes [78]. TFEB phosphorylation by mTOR and other kinases retains this transcription factor in the cytoplasm, whereas its dephosphorylation results in translocation to the nucleus, where it induces its target genes [79]. We have previously shown that in neuronopathic GD neurons, there was a reduction in TFEB levels and activity, and a concomitant decrease in expression of its target genes, leading to lysosomal dysfunction and autophagy block [67]. To investigate the effect of ACDase inhibition on a key ALP regulator involved in α-synuclein clearance, we examined the subcellular localization of TFEB in GBA1/PD-DA neurons before and after treatment with ACDase inhibitors. Confocal microscopy analysis of RecNciI/WT DA neurons showed a considerable reduction in MFI of nuclear TFEB compared to the isogenic control (Fig. 3A), and treatment of the mutant cells with ARN14988, 22 m, and LCL521 caused a significant increase in the nuclear localization of TFEB compared to untreated cells (Fig. 3A and B).

Figure 3.

ACDase inhibition and ASAH1 KD in RecNciI/WT GBA1/PD-DA neurons increases the nuclear localization of TFEB. (A and B) RecNciI/WT GBA1/PD-DA neurons were either left untreated or incubated with the ACDase inhibitors ARN19988, 22 m, or LCL521 during the last 10 days of DA differentiation as indicated. (A) Confocal microscopy analysis using antibodies to neuron-specific class βIII-tubulin and TFEB; nuclei were stained with DAPI. (B), quantitation of immunofluorescence staining showing the percentage of βIII-tubulin-positive neurons with nuclear TFEB localization. p-values were determined using one-way ANOVA followed by Bonferroni’s multiple comparisons test. (C and D) RecNciI/WT hiPSCs were subjected to ASAH1 gene knockdown using the CRISPR/Cas9 system. The KD hiPSC lines, parental, and isogenic GC controls were then differentiated to DA neurons. (C) Confocal microscopy analysis using antibodies to βIII-tubulin and TFEB; nuclei were stained with DAPI. (D), quantitation of immunofluorescence staining showing the percentage of βIII-tubulin-positive neurons with nuclear TFEB localization. Plots represent the results from three independent experiments (n = 3). p-values were determined using unpaired Welch’s t test. n = 3 independent experiments. Error bars represent mean ± SEM. Asterisks indicate the level of statistical significance: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. Scale bar, 50 μm.

We then determined the effect of ASAH1 KD on the subcellular localization of TFEB by confocal microscopy analysis of parental vs. KD RecNciI DA neurons. As shown in Fig. 3C and D, there was a 50% reduction in TFEB nuclear localization in RecNciI/WT neurons compared to isogenic WT/WT neurons, and ASAH1 KD of the mutant neurons resulted in a significant rescue of TFEB nuclear localization compared to the parental control.

Analysis of the effect of ACDase inhibition and ASAH1 KD in the two other GBA1/PD mutants gave similar results. In L444P neurons there was about a 50% reduction in the nuclear localization of TFEB compared to the isogenic control, and treatment of the mutant cells with ARN14988, 22 m, and LCL521 caused a significant increase in the nuclear localization of TFEB compared to the untreated L444P/WT neurons. (Fig. S4A and B). As shown in Fig. S4C and D, ASAH1 KD of L444P/WT neurons partially rescued TFEB nuclear localization. In N370S/WT neurons there was also considerably lower nuclear localization of TFEB, which was significantly increased by both, ACDase inhibition (Fig. S5A and B), and ASAH1 KD (Fig. S5C and D).

We conclude from these results that in RecNciI/WT, L444P/WT and N370S/WT DA neurons, there was considerable reduction in the nuclear localization of TFEB compared to the corresponding isogenic controls and that both, ACDase inhibition and ASAH1 KD nearly rescued the nuclear localization of TFEB.

ACDase inhibition and ASAH1 KD rescue CathD expression

CathD is a lysosomal enzyme involved in the degradation and clearance of α-synuclein, and its expression is regulated by TFEB [11]. It has been previously reported that in neuronopathic GD neurons and in GBA1/PD-DA neurons with heterozygote mutations in GBA1, there is a decrease in the expression of Cathepsins and other hydrolases [67, 72]. Confocal microscopy analysis showed that RecNciI/WT DA neurons had about a 60% reduction in CathD levels compared to isogenic controls (Fig. 4A and B), and that treatment with ARN14988, 22 m, and LCL521 caused an increase in the expression of this hydrolase. As shown in Fig. 4C and D, KD of ASAH1 resulted in a considerable increase in CathD expression compared to the parental line. CathD expression in L444P/WT (Fig. S6A and B) and in N370S/WT neurons (Fig. S7A and B) was about 60% lower than in the corresponding isogenic controls. ACDase inhibition in L444P/WT (Fig. S6A and B) and N370S/WT neurons (Fig. S7A and B) partially restored CathD expression. Similar results were obtained after ASAH1 KD of L444P/WT (Fig. S6C and D) and N370S/WT (Fig. S7C and D) neurons.

Figure 4.

ACDase inhibition and ASAH1 KD in RecNciI/WT GBA1/PD-DA neurons increases the expression levels of CathD. (A and B) RecNciI/WT GBA1/PD-DA neurons were either left untreated or incubated with the ACDase inhibitors ARN19988, 22 m, or LCL521 during the last 10 days of DA differentiation as indicated. (A) Confocal microscopy analysis using antibodies to βIII-tubulin and CathD; nuclei were stained with DAPI. (B), quantitation of immunofluorescence staining showing the mean fluorescence intensity of CathD within βIII-tubulin-positive neurons. p-values were determined using one-way ANOVA followed by Bonferroni’s multiple comparisons test. (C and D) RecNciI/WT hiPSCs were subjected to ASAH1 gene knockdown using the CRISPR/Cas9 system. The KD hiPSC lines, parental, and isogenic GC controls were then differentiated to DA neurons. (C) Confocal microscopy analysis using antibodies to βIII-tubulin and CathD; nuclei were stained with DAPI. (D), quantitation of immunofluorescence staining showing the mean fluorescence intensity of CathD within βIII-tubulin-positive neurons. p-values were determined using unpaired Welch’s t test. Asterisks indicate the level of statistical significance: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. Scale bar, 50 μm.

We conclude from these results that in GBA1/PD-DA neurons there is a considerable reduction in CathD levels, and that ACDase inhibition increases this hydrolase expression, albeit with different efficiencies, with ARN14988, and the brain-penetrant 22 m being more effective than LCL521. In addition, KD of ASAH1 in all 3 lines of GBA1/PD-DA neurons also showed considerable increases in the levels of CathD.

Comparison of the effect of ACDase inhibition on TFEB nuclear localization and CathD expression in RecNciI/WT, L444P/WT, and N370S/WT DA neurons

As shown in Fig. S8A, the effect of ACDase inhibition on TFEB nuclear localization by 22 m (10 μM), ARN14988 (10 μM), and LCL521 (20 μM) was similar across the 3 different GBA1 mutations. A similar analysis of CathD rescue by ACDase inhibition showed no notable differences among different GBA1 mutations with ARN14988 and 22 m. However, treatment with LCL521 led to a significant restoration of CathD levels in N370S/WT neurons compared to RecNciI/WT and L444P/WT neurons (Fig. S8B).

ACDase inhibition and ASAH1 KD restore autophagic flux in GBA1/PD-DA neurons

We previously showed that neuronopathic GD neurons and heterozygote GBA1/PD-DA neurons have autophagic defects, as determined by analysis of autophagy markers [69, 71]. SQSTM1/p62 is an adaptor protein that brings protein aggregates to the autophagosome and is degraded by lysosomes during autophagy, whereas the accumulation of p62 reflects a block in autophagy flux [80]. WB analysis showed that GBA1/PD-DA neurons from RecNciI/WT (Fig. 5A and D), L444P/WT (Fig. 5B and E), and N370S/WT (Fig. 5C and F) had elevated levels of p62 compared to isogenic controls, and that the three ACDase inhibitors caused a significant reduction in p62. In addition, KD of ASAH1 in DA neurons from RecNciI/WT (Fig. 5G and J), L444P/WT (Fig. 5H and K), and N370S/WT (Fig. 5I and L) brought down p62 levels to those in isogenic controls.

Figure 5.

ACDase inhibition and ASAH1 KD rescue autophagy in GBA1/PD-DA neurons. RecNciI/WT, L444P/WT, and N370S/WT PD-DA neurons were either left untreated or incubated with the ACDase inhibitors ARN19988, 22 m, or LCL521 during the last 10 days of DA differentiation. Cell lysates from mutant and isogenic GC controls were analyzed by WB for levels of P62. (A–C) representative immunoblots of P62 in RecNciI/WT (A), L444P/WT (B), and N370S/WT (C) DA neurons. (D–F), quantitation of immunoblots showing P62 normalized to actin in RecNciI/WT (D), L444P/WT (E), and N370S/WT (F) DA neurons. Graphs represent the results from three independent experiments (n = 3). p-values were determined using one-way ANOVA followed by Bonferroni’s multiple comparisons test. (G) Cell lysates from KD RecNciI/WT, parental RecNciI/WT, and isogenic GC control DA neurons were analyzed by WB using antibodies to P62. (H) Immunoblots of P62 in KD L444P/WT, parental L444P/WT, and GC control DA neurons. (I) Immunoblots of P62 in KD N370S/WT, parental N370S/WT, and GC DA neurons. Quantitation of the corresponding immunoblots is shown in J, K, and L. graphs represent the results from three independent experiments (n = 3). p-values were determined using unpaired Welch’s t test. Error bars represent mean ± SEM. Asterisks indicate the level of statistical significance: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.

We conclude from these results that GBA1 mutations in GBA1/PD-DA neurons interfere with normal autophagy, and that ACDase is part of the mechanism that deregulates the ALP in the mutant neurons.

Discussion

In this report we used an isogenic hiPSC system to evaluate ACDase as a potential therapeutic target for GBA1-associated PD. To this end, we analyzed the effect of pharmacological and genetic inhibition of ACDase in reverting the phenotypic abnormalities of dopamine neurons derived from PD patients with heterozygote mutations in GBA1. Our results showed that ACDase inhibition or ASAH1 knockdown resulted in increased clearance of neurotoxic α-synuclein species, through rescue of TFEB-dependent lysosomal functions involved in the degradation of this aggregation-prone protein.

Severe bi-allelic mutations in GBA1 cause neuronopathic GD, but mild mutations such a N370S do not cause neurological manifestations. However, GBA1 mutations are the highest risk factor for PD, and about 7% of all PD patients have mutations in this gene, which are associated with earlier disease onset and increased severity compared to idiopathic PD [73]. While GBA1 mutations increase the risk of PD by 5%–20%, only about 10%–30% of these individuals develop PD [81]. Thus, in contrast to the strong neurological phenotype caused by severe bi-allelic GBA1 mutations, heterozygote GBA1 mutations are insufficient to cause GBA1/PD, and genetic background, age, environmental and other epigenetic factors are important determinants of this disorder. Because of the low penetrance of mono-allelic GBA1 mutations, in order to study the specific contribution of GCase deficiency to development of GBA1/PD, we used an isogenic system in which the only variable is the presence or absence of a mono-allelic GBA1 mutation. The hiPSC lines we used in this communication harbor frequent GBA1 mutant genotypes: a mild N370S mutation, and the severe L444P and RecNciI mutations.

The hallmark of PD and LBD is the presence of α-synuclein aggregates that accumulate in Lewy bodies [82]. The aggregation of this 14 kDa protein leads to neuronal death, in particular in the substantia nigra pars compacta [53, 54]. α-synuclein is largely cleared by cathepsin proteases in the lysosome [11]. Under homeostatic conditions, mTOR kinase phosphorylates TFEB, preventing its migration to the nucleus. Starvation conditions or mTOR inhibition prevent TFEB phosphorylation, enabling its migration to the nucleus [64]. In the nucleus, TFEB upregulates autophagic and lysosomal genes, inducing autophagy and the expression of lysosomal cathepsins, which are both required for the degradation of α-synuclein [11, 79]. In GBA1/PD-DA neurons, heterozygote mutations in GBA1 caused hyperactivation of mTOR, impaired autophagic flux, decreased TFEB levels in the nucleus, and lowered cathepsin D expression [67, 68]. This phenotype correlated with increased α-synuclein aggregation in the mutant GBA1/PD neurons. Our analysis showed the involvement of ACDase in all of these alterations. It was previously shown that inhibition of ACDase by carmofur in an animal model of GBA1-associated PD decreased levels of oligomeric α-synuclein in brain, and that in HEK293-FT cells deficient in GCase and in GBA1/PD dopamine neurons, carmofur decreased oxidized α-synuclein [6, 71, 83]. We previously showed that treatment of iPSC-derived GBA1/PD-DA neurons with carmofur decreased levels of α-synuclein to those in isogenic controls. These previous reports suggested that ACDase was involved in the formation of toxic α-synuclein species.

As off-target effects of carmofur cannot be ruled out, we sought additional evidence that ACDase contributes to the pathogenesis of GBA1/PD. The results reported here using 3 additional unrelated ACDase inhibitors, and CRISPR/Cas9 KD of ASAH1, showed that inhibition or genetic reduction of this enzyme results in a significant increase in α-synuclein clearance. This was likely the result of improved autophagy through prevention of mTORC1 hyperactivation in the mutant neurons. Decreased mTOR kinase activity enables TFEB transport to the nucleus, which results in higher levels of expression of CathD, one of the target genes upregulated by TFEB. Our present results lend support to the idea that ACDase plays a key role in GBA1-associated neurodegeneration, further suggesting that this enzyme may be a therapeutic target for alternative or combination SRT.

The ACDase inhibitors 22 m, ARN14988 and LCL521 rescued the GBA1/PD phenotype in all three mutant GBA1/PD lines, albeit with different efficiencies depending on the parameters measured. Of the 3 ACDase inhibitors used in this study, 22 m is the only one reported to cross the BBB. 22 m was shown target engagement in in vivo models of Krabbe and Gaucher diseases [84]. Krabbe is a fatal demyelinating disease caused by mutations in GALC, the gene that encodes galactosylceramidase [85]. In Krabbe disease, the toxic lipid that accumulates is galactosylsphingosine (GalSph), which is generated by the action of ACDase on galactosylceramide (GalCer). Treatment of Krabbe (Twitcher) and GD mice with 22 m lowered the levels of GalSph and GluSph in brains of Krabbe and GD mice, respectively [84, 86]. Although these studies did not test whether 22 m would reduce the pathology in Krabbe or GD mice, intraperitoneal injection of the brain-penetrant ACDase inhibitor carmofur into Krabbe mice, decreased the levels of GalSph and extended their lifespan [86], raising the question of whether 22 m may also have similar therapeutic effects in GALC or GBA1 deficiencies. Studies in Zebrafish have provided additional genetic evidence that Asah1/ACDase may be involved in the loss of DA neurons and disruption of motor functions in GBA1/PD. Gba1^−/− Zebrafish exhibited a decreased lifespan, loss of dopaminergic neurons, locomotor abnormalities, and development of an abnormal curved back posture [76]. This phenotype was corrected when Gba1^−/− Zebrafish were crossed with asah1^−/− fish. Whether inhibiting ACDase with 22 m in animal models of GBA1-associated PD would reverse α-synuclein accumulation as carmofur did [77], will be addressed in future studies.

Although nGD and GBA1/PD have very different clinical manifestations, GBA1 mutations cause a similar deregulation of the mTORC1/TFEB/ALP cascade, suggesting the possibility of overlapping disease mechanisms between neurological disorders caused by mutant GBA1 [65, 67, 72]. In neuronopathic GD, there is a considerable accumulation of GluCer and its neurotoxic metabolite GluSph in brain [49, 87]. Sphingolipid alterations are also the cause of more than 50 monogenic lysosomal storage diseases, most of which are characterized by neurodegeneration, indicating that neurons are very sensitive to perturbations in sphingolipid homeostasis [88]. A key question in the pathogenesis of GBA1/PD is whether sphingolipid alterations play any significant role in deregulating the ALP cascade that clears α-synuclein aggregates. Although in nGD up to 500-fold elevations in GluSph have been demonstrated, GSL elevations in GBA1/PD have not been easy to detect, and this is reflected in a conflicting literature. In some reports, lipid analysis showed that brains from PD patients and from animal models with heterozygote mutations in GBA1 exhibit GSL elevations [33, 89]. It has also been reported that GCase levels in brains of patients with idiopathic PD are lower than in non-affected individuals. In vitro experiments showing that GluCer and GluSph induce the formation of α-synuclein aggregates suggest that direct interactions between GSLs and this aggregation-prone protein can also lead to the formation of pathogenic α-synuclein species [34, 77]. On the other hand, other investigators reported that there is no evidence of GSL accumulation in PD patients [90]. The difficulty to detect GSL and particularly GluSph elevation in brains from heterozygote GBA1/PD patients may be due to the low levels of GSL alterations that can be expected when only one GBA1 allele is affected, and the other is WT. As GBA1-associated PD takes years to develop, it is possible that minimal but persistent generation of GSLs over long periods of time, may contribute to the onset and progression of GBA1/PD. We previously showed that in nGD neurons, ALP rescue was due to the ability of ACDase inhibitors to prevent the formation of neurotoxic GluSph [69]. Whether the increase in α-synuclein clearance in GBA1/PD-DA neurons by inhibition or KD of ACDase/ASAH1 reported in this study is also due to curtailment of its enzymatic activity will be addressed in future studies.

In the present study, CRISPR/Cas9 reduction of ACDase in GBA1/PD-DA neurons to about 40%–50% of parental levels was sufficient to prevent α-synuclein aggregation, providing a genetic validation of the results obtained by pharmacological inhibition of ACDase. It is important to note that ASAH1 deficiency causes Farber disease (FD), a fatal autosomal lysosomal recessive disorder caused by the accumulation of ceramide in various organs including the brain. A more modest ASAH1 deficiency causes spinal muscular atrophy with progressive myoclonic epilepsy (SMA-PME) [91–94]. SMA-PME presents as a milder disorder of ASAH1 deficiency, but in rare cases SMA-PME can progress to FD. Thus, it appears that FD and SMA-PME represent a spectrum of ASAH1 deficiency rather than two separate disorders [94]. As ASAH1 genotype–phenotype correlations have not been established, it would be interesting to determine if there are residual levels of ACDase in affected individuals, and if this impacts clinical phenotypes. In our studies, ASAH1 reduction of ACDase protein levels to 40%–50% of controls did not result in observable cytotoxicity for DA neurons, and in vivo studies showed no toxicity of 22 m at doses that showed target engagement [84]. These observations may suggest that in clinical studies, appropriate dosing and careful monitoring may help prevent unwanted side effects of pharmacological inhibition of ACDase.

In summary, this study provides pharmacological and genetic evidence implicating ACDase in dysregulation of ALP functions involved in the clearance of α-synuclein. Our results further suggest that ACDase may be a potential therapeutic target for amelioration or treatment of GBA1-associated PD.

Materials and methods

Maintenance and differentiation of hiPSCs to DA neurons

The hiPSC lines from PD patients carrying heterozygous GBA1 mutations (RecNcil/WT, L444P/WT, N370S/WT) and the corresponding gene-corrected WT/WT isogenic hiPSCs controls were maintained and differentiated to DA neurons as previously described [43, 95]. These lines are listed in Supplementary Table 1a. The hiPSCs were cultured using standard protocols as we described [71]. For hiPSC differentiation to DA neurons, single cell suspensions of hiPSCs were cultured on Matrigel-coated plates at a density of 40 000 cells/cm² in serum replacement media containing growth factors and small molecules including FGF8a (100 ng/ml), SHH C25II (100 ng/ml), LDN193189 (100 nM), SB431542 (10 μM), CHIR99021 (3 μM) and Purmorphamine (2 μM) for five days. The next six days, cells were maintained in neurobasal medium containing B27 minus vitamin A, N2 supplement, and LDN193189 and CHIR99021. Single cell suspensions were then seeded at a density of 400 000/cm² on polyornithine- and laminin-coated plates in neurobasal media containing B27 minus Vitamin A, BDNF (20 ng/ml), GDNF (20 ng/ml), TGFβ (1 ng/ml) ascorbic acid (0.2 mM), cAMP (0.5 mM) and DAPT (10 μM), and cultured until maturation for 60 days.

Immunocytochemistry

hiPSC-derived DA neurons were grown on polyornithine- and laminin-coated-glass coverslips for microscopy as follows. Glass coverslips were coated with polyornithine at 100 μg/ml for a minimum of 3 h at room temperature, followed by 3 washes in sterile water. Laminin was added to the precoated glass coverslips at 5 μg/ml in sterile phosphate-buffered saline (PBS) and incubated for a minimum of 8 h in a 5% CO2 incubator. Laminin was removed and DA differentiation was carried out in the coated coverslips. After differentiation, neurons were fixed with 4% paraformaldehyde (Santa Cruz) for 15 min at room temperature followed by 3 washes with 1X PBS. The neurons were permeabilized with 0.3% (Vol/Vol) Triton X-100 for 15 min and blocked for 1 h in 5% (Vol/Vol) normal goat serum in PBS. This was followed by incubation with the indicated primary antibodies diluted in 5% goat serum/PBS overnight at 4°C. Following 3 washes with 1x PBS, cells were incubated with secondary antibody diluted at 1:1000 in 5% goat serum/PBS for 1 h in the dark at room temperature. The secondary antibody was removed, followed by 3 washes with 1x PBS. Cover slips were mounted with DAPI-containing mounting media on glass slides and kept in the dark. After 24 h, mounted cover slips were ready for confocal imaging.

Microscopy and imaging

Immunofluorescence images were captured using a Nikon A1 confocal laser scanning microscope under 20X or 60X oil objectives. The excitation wavelengths used were 405, 488 and 561 nm for blue, green and red fluorophores, respectively. Neuronal images were acquired as Z-stacks or without stacking. Identical pixel acquisition settings were used for all experiments. Further image processing and analysis was done using Fiji or Image J software (https://imagej.nih.gov/ij). Fluorescence intensity of the respective signals was obtained from 15–30 neurons from at least three independent experiments. The mean fluorescence intensity (MFI) was calculated accordingly.

Chemical reagents and treatments

The ACDase inhibitor ARN14988 was purchased from Cayman (Cat# 24284), 22 m was purchased from MCE (Cat# HY-141866), LCL521 dihydrochloride was purchased from MCE (Cat# HY-103593A). Stock solutions of these inhibitors were reconstituted in DMSO. Neuronal cultures were incubated with ARN14988 and 22 m at a final concentration of 10 μM, and with LCL521 at 20 μM during the last 10 days of DA differentiation and were replenished with every media change.

List of reagents

All the reagents used in this study are listed in Supplementary Table 2.

Immunoblot analysis

For immunoblot analysis, differentiated DA neurons were scraped from culture wells on ice into 1X PBS, and collected by centrifugation. Cell lysates were made in a buffer containing 1X PBS, 1% Triton X100 (Vol/Vol) and phosphatase/protease inhibitor mixture (Cell Signaling Technology), followed by centrifugation for 30 min at 4°C, 14000 rpm, The supernatants were collected and kept at −80°C until use. The samples were analyzed on 4%–20% Tris-Glycine gradient SDS-PAGE gels (Thermo Fisher Scientific). Electrophoresis was followed by protein transfer onto nitrocellulose membranes. Membranes were blocked with 5% (WT/Vol) non-fat dry milk in Tris-buffered saline with 1% Tween-20 (TBS-T) and incubated with the indicated primary antibodies for 2–3 h at room temperature or overnight at 4°C. After horseradish peroxidase-conjugated secondary antibody incubation for 1 h at room temperature, the membranes were developed with SuperSignal West Femto Substrate (Thermo Fisher Scientific) and imaged using the Chemidoc system and Imagelab software (BioRad). Densitometry analysis was done using Image J software.

Mean fluorescence intensity quantitation

Confocal Z stacked images were analyzed for mean fluorescent intensity (MFI). For MFI measurements, regions of interest (ROI) were measured from groups of 15–30 neurons from three independent experiments. Manually, ROI were drawn on Tuj1-positive neurons, and MFI was analyzed with Image J software (version 2.0.0-rc-68/1.52e, open-source platform for biological image analysis) for MAC OS X, using the RGB measure function [96].

Cell counting

Images were acquired for TUJ/TH/DAPI staining using a confocal laser-scanning microscope under a 20X or 60X oil objective from three coverslips for each condition. A total of 450–550 cells were counted for each condition using the cell counter function in Image J or directly under the microscope.

CRISPR Cas9 knockdown (KD)

All single guide RNAs (sgRNA) were designed using an online tool from Synthego. The top three sgRNA were selected for the initial KD in hiPSCs. A combination of two sgRNAs (Supplementary Table 1b) gave the highest KD efficiency and were selected for further analysis as described in the text. sgRNA transfection was performed using a Neon transfection system (Thermo Fisher Scientific) according to the manufacturer at 1200 V for 20 milliseconds. Transfected cells were plated on Matrigel-coated plates on mTeSR Plus media containing ROCK inhibitor for 48rs. The transfected hiPSCs were then cultured in media containing 400 ng/ml of puromycin for one week, with a change of media every 48 h Puromycin-resistant cells were collected after 7 days and expanded for the experiments described in the text.

GCase assay

The assay for GCase enzymatic activity in intact cells was carried out as described [97]. DA progenitor cells were plated in 96-well plates at a density of 400 000/cm² for DA differentiation. At day 60 of differentiation, the medium was removed, and neurons were washed with PBS. The assay reaction was started by the addition of 50 μl of 2.5 mM 4-methylumbelliferyl β-D-glucopyranoside (MUG) (Sigma) in 0.2 M acetate buffer (pH 4.0) to each well. Plates were incubated at 37°C for 2 h and the reaction was stopped by the addition of 150 μl of 0.2 M glycine buffer (pH 10.8) to each well. Released 4-methylumbelliferone was measured using a SpectraMax Gemini plate reader (Molecular Device, Sunnyvale, CA) (excitation 365 nm, emission 445 nm). Conduritol B epoxide (CBE) was added at 1 mM to replicate wells for the duration of the assay, to control for non-GCase enzymatic activity.

Quantitation and statistical analysis

Statistical analysis was performed using GraphPad Prism software version 7.0a. The n number indicates the number of independent experiments. Statistical analyses were performed using one-way ANOVA followed by Bonferroni’s multiple comparisons test or using unpaired Welch’s t tests as indicated in the Figure legends. All data sets were selected for analysis. Multiple comparisons of the mean of preselected pairs of data sets were analyzed using Bonferroni post hoc tests. In each figure, asterisks indicate the level of statistical significance: ^*P ≤ 0.05, ^**P ≤ 0.01, ^***P ≤ 0 0.001, ^****P ≤ 0.0001 and ns, nonsignificant. Results are expressed as mean ± Standard error of mean (SEM).

Conflict of interest statement: None declared.

Funding

See Acknowledgements section.

Data availability

All the data used in this study are available upon request.