Unexpected phenotypic and molecular changes of combined glucocerebrosidase and acid sphingomyelinase deficiency

There is compelling evidence of an excessive burden of lysosomal disease gene variants and lysosomal dysfunction in Parkinson's disease (PD) (Robak et al., 2017; Wallings et al., 2019). Bi-allelic mutations in glucocerebrosidase 1 (GBA1; encoding glucocerebrosidase or GCase) cause Gaucher's disease, a lysosomal storage disorder (LSD), whereas heterozygous mutations are the most common and strongest genetic risk factor for sporadic PD, with a prevalence of ∼5-20% depending on the population investigated (Siebert et al., 2014; Orr-Urtreger et al., 2009; Neumann et al., 2009). PD patients also exhibit reduced GCase activity in different tissues, including the brain (Gegg et al., 2012), regardless of their GBA1 mutation status (Gegg et al., 2012; Parnetti et al., 2017; Atashrazm et al., 2018).

In a similar fashion to GBA1 mutations, homozygous mutations in SMPD1 (encoding acid sphingomyelinase or ASM) also cause an LSD, in this case, Niemann–Pick disease, whereas heterozygous SMPD1 variants are associated with increased risk of sporadic PD (Alcalay et al., 2019; Gan-Or et al., 2013; Foo et al., 2013; Mao et al., 2017; Dagan et al., 2015). Owing to their rarity within the PD population, the functional significance of these SMPD1 variants is still not completely understood (Alcalay et al., 2019; Usenko et al., 2022). However, a reduction of ASM activity is correlated with an earlier age of disease onset in PD as well as in other synucleinopathies, including dementia with Lewy bodies and multiple system atrophy (Alcalay et al., 2019; Usenko et al., 2022). Both GBA1 and SMPD1 encode lysosomal enzymes that converge on ceramide metabolism (Fig. 1). Therefore, an additive interaction between these two enzymes is biologically plausible but awaits experimental confirmation. We hypothesised that ASM deficiency could worsen the functional consequences of GCase deficiency, aggravating phenotypes that could potentially lead to, or enhance, neurodegeneration.

Zebrafish (Danio rerio) are an attractive vertebrate model to study the biological effect of both monogenic PD genes and genetic risk factors for PD (Flinn et al., 2013; Larbalestier et al., 2022), including gba1 deficiency (Keatinge et al., 2015). We had previously characterised a gba1 mutant zebrafish line (gba1^−/−) (Keatinge et al., 2015) and demonstrated its usefulness to study gene-gene interactions (Watson et al., 2019). gba1^−/− zebrafish faithfully model key features of GCase deficiency or Gaucher's disease, including Gaucher cell accumulation, marked inflammation with microglial infiltration, mitochondrial dysfunction and neurodegeneration (Keatinge et al., 2015). gba1^−/− larvae develop normally but gba1^−/− juvenile zebrafish then rapidly deteriorate from 10-12 weeks onwards and die between 12 and 14 weeks.

As expected, combined GCase and ASM deficiency acted synergistically on key sphingolipid metabolites in the gba1^−/−;smpd1^−/− double-mutant zebrafish. However, instead of a worsening of phenotypes, we unexpectedly observed markedly prolonged survival and conventional swimming behaviour in gba1^−/−;smpd1^−/− mutants compared to the behaviour of the gba1^−/− (single) mutant zebrafish. RNA sequencing (RNAseq)-based pathway analysis confirmed the restoration of neuronal health in gba1^−/−;smpd1^−/− mutants compared to that in the gba1^−/− mutants. Mechanistic experiments identified a rescue effect of combined GCase and ASM deficiency on the function of mitochondrial Complexes I and IV in gba1^−/−;smpd1^−/− mutants compared to the marked but distinct mitochondrial dysfunction in gba1^−/− or smpd1^−/− single mutant zebrafish. The mitochondrial rescue led to an abrogation of oxidative membrane damage, further reflecting the overall restorative effect of ASM deficiency on neuronal health in GCase deficiency. Our study highlights the need of functional, mechanistic validation for the interaction of any putative genetic/enzymatic risk factors for human diseases in suitable model systems, rather than readily assuming an additive effect.

We initially hypothesised that deficiency of both GCase and ASM enzymes may synergise, leading to a further aggravation of gba1-deficient phenotypes in vivo. To address this, we identified a single smpd1 orthologue in zebrafish (ENSDARG00000076121) with 59% shared identity to the human SMPD1 gene at both the DNA and the protein level. CRISPR/Cas9 technology was used to generate a stable smpd1 mutant line (smpd1^−/−). The selected mutant allele contained a 5 bp deletion and 136 bp insertion within exon 3, resulting in a frame shift and the generation of a premature stop codon (Fig. 2A; Fig. S1). Enzymatic activity of ASM in smpd1^−/− at 5 days post fertilisation (dpf) was reduced by 93% (P=0.006, Fig. 2B). The large reduction in ASM enzymatic activity resulted in a significant increase of key glycolipid substrates in the smpd1^−/− larvae already at 5 dpf (Fig. 2C). The smpd1^+/− line was crossed with the gba1^+/− line to generate gba1^+/−;smpd1^+/−. The latter were subsequently in-crossed to generate double mutants, single mutants and wild-type (WT) controls for all subsequent experiments. At each in-cross, larvae were genotyped at 3 dpf and then raised in genotype-specific tanks. Every genotype was present in its expected Mendelian ratio (1/16) during genotyping at the larval stages (Table S1), but only the WT, gba1^−/−, smpd1^−/− and gba1^−/−;smpd1^−/− lines were raised for experiments.

We had previously reported marked sphingolipid accumulation in gba1^−/− zebrafish (Keatinge et al., 2015). We hypothesised that combined (enzymatic) GCase and ASM deficiency would synergistically increase distinct sphingolipid subtypes. Using mass spectrometry, a comprehensive panel of glycolipid substrates was analysed in the brains of gba1^−/− and smpd1^−/− single mutant as well as in gba1^−/−;smpd1^−/− double mutant zebrafish and WT controls at 12 weeks of age. As expected, a marked additive effect of combined GCase and ASM deficiency was observed for glucosylceramide levels (the direct substrate of GCase) (Fig. 3A). This was likely due to both a block in glucosylceramide catabolism and metabolic compensation in the flux of sphingolipid generation. Combined GCase and ASM deficiency also resulted in an additive effect on lactosylceramide, ceramide and sphinganine levels (Fig. 3B-D). Sphingosine levels were increased in gba1^−/−;smpd1^−/− compared to those in WT, reflecting an increase compared to those in gba1^−/− but not compared to those in smpd1^−/− (Fig. 3E). Unexpectedly, there was no synergistic effect in sphingomyelin levels in the gba1^−/−;smpd1^−/− double mutants (Fig. 3F).

The inflammation markers chitotriosidase (CHIT1) and β-hexosaminidase (heterodimer or homodimer of HEXA and/or HEXB) are markedly increased in the serum of Gaucher's disease patients and used as biomarkers to monitor disease activity (Grabowski, 2012). We previously observed a marked increase in chitotriosidase and β-hexosaminidase activity in gba1^−/− zebrafish brain tissue at 12 weeks (Keatinge et al., 2015). As key GCase substrates were synergistically increased in gba1^−/−;smpd1^−/− double-mutant zebrafish, we investigated whether combined GCase and ASM inactivation may also result in a further increase of chitotriosidase and β-hexosaminidase activity. Unexpectedly, gba1^−/−;smpd1^−/− double-mutant zebrafish displayed a similar increase in chitotriosidase and β-hexosaminidase activity compared to that seen for gba1^−/− zebrafish (Fig. S2). Furthermore, we had previously detected Gaucher cell invasion in the central nervous system of end-stage gba1^−/− zebrafish (Keatinge et al., 2015). Analyses of the retinas across all genotypes demonstrated similar Gaucher cell invasion in the double mutants (Fig. 4), with both gba1^−/− and gba1^−/−;smpd1^−/− mutants showing an ∼50% increase in cells positive for 4C4 (which marks microglia) compared to WT. These 4C4-positive cells in both gba1^−/− and gba1^−/−;smpd1^−/− mutants also tended to be larger and rounder, indicative of the Gaucher cells we had previously described. These data suggest persistent yet unaltered neuroinflammatory states in the double mutants despite a marked synergistic increase in sphingolipid metabolites.

The marked additive effect of combined GCase and ASM deficiency on sphingolipid levels led us to hypothesise that ASM deficiency would further worsen the motor phenotype and shorten survival in gba1^−/−;smpd1^−/− double-mutant zebrafish. Unexpectedly, genetic inactivation of ASM led to a complete rescue of this behaviour in the gba1^−/−;smpd1^−/− double-mutant zebrafish [Movie 1 (WT), Movie 2 (smpd1^−/−), Movie 3 (gba1^−/−) and Movie 4 (gba1^−/−;smpd1^−/−)]. Importantly, disease-free survival, in which animals could consistently maintain buoyancy, was also markedly increased by 22% in gba1^−/−;smpd1^−/− double-mutant zebrafish compared to that in gba1^−/− zebrafish (median survival of 102 dpf in gba1^−/− and 125  dpf in gba1^−/−;smpd1^−/−, P=0.0055; Fig. 5A). Despite not exhibiting the same barrel rolling phenotype as the gba1^−/− mutants, and also being able to maintain their buoyancy, the gba1^−/−;smpd1^−/− double mutants would ultimately be found unresponsive at the bottom of the tank and were thus culled for humane reasons. We also raised smpd1^−/− mutants to determine their lifespan, but never encountered a decrease in viability compared to that of WT zebrafish, even up to the age of 18 months (data not shown).

We next applied RNAseq-based pathway analysis to further elucidate the underlying mechanisms of the observed rescue effect. The differential gene expression analysis in all four genotypes (WT, gba1^−/− and smpd1^−/− single mutants, and gba1^−/−;smpd1^−/− double mutants) identified a total of 512 genes that were dysregulated in gba1^−/− but rescued in gba1^−/−;smpd1^−/−. Amongst these, 219 genes were downregulated and 293 genes were upregulated in gba1^−/− compared to wild-type and gba1^−/−;smpd1^−/− [adjusted P-value≤0.05, |log₂(fold change)|≥1]. We next employed ClusterProfiler analysis on Gene Ontology (GO) categories to identify functionally relevant pathways within the rescued gene sets. Key neuronal pathways including the GO terms for synaptic signalling, chemical synaptic transmission and calcium ion-regulated exocytosis were markedly downregulated in gba1^−/− but normalised in gba1^−/−;smpd1^−/− (Fig. 5B). This suggests that key aspects of neuronal function were restored in the gba1^−/−;smpd1^−/− double mutants.

We also observed an enrichment of upregulated genes in gba1^−/−compared to those in gba1^−/−;smpd1^−/− in a broad range of GO terms, the top five of which are thought to regulate muscle function. However, as our RNAseq analysis was carried out on brain tissue, we consider these changes to be of limited relevance (Fig. S3). Upregulation of the inflammatory signature in gba1^−/− was retained in gba1^−/−;smpd1^−/− but not further enhanced (data not shown).

As both GCase and ASM are lysosomal hydrolases, we specifically focused on the effect of isolated GCase deficiency in gba1^−/− compared to the effect of combined GCase and ASM deficiency in gba1^−/−;smpd1^−/− on lysosomal transcriptomic pathways. Gene set enrichment analysis led to the identification of 27 leading-edge, dysregulated lysosomal genes, which account for the enrichment signal of the pathway. The expression of these 27 lysosomal genes was increased in gba1^−/− compared to wild-type and gba1^−/−;smpd1^−/− (Fig. 5C; Table S2). Amongst these 27 genes, acid hydrolases contributed the most. Cathepsin L.1 (ctsl.1), involved in the initiation of protein degradation, ranked as the top-rescued gene. The apparent normalisation of lysosomal gene expression profiles in gba1^−/−;smpd1^−/− was in contrast to the observed marked increase in a wide range of sphingolipid levels in gba1^−/−;smpd1^−/− compared to the sphingolipid levels in gba1^−/− or smpd1^−/− single mutants (see above).

We had previously observed marked mitochondrial dysfunction in gba1^−/−. We therefore also focussed on the analysis of mitochondrial genes involved in the oxidative phosphorylation pathway. This leading-edge mitochondrial gene subset included 16 genes encoding the subunits of the Complexes I, II, IV and V in the mitochondrial electron transport chain (Table S3). Interestingly, gene set enrichment analysis showed an upregulation of this mitochondrial gene subset in gba1^−/−, presumably as a compensatory mechanism to the impaired function of the mitochondrial respiratory chain, but showed similar mitochondrial gene expression levels in WT and gba1^−/−;smpd1^−/− (Fig. 5D).

We next compared the mitochondrial respiratory chain function across all four genotypes to further determine whether the normalised gene expression levels for oxidative phosphorylation-related genes would be reflected in normalised mitochondrial function. Complex I activity was reduced by 65% in smpd1^−/− compared to WT levels (P=0.0198, Fig. 6A) but restored to 92% of WT levels in gba1^−/−;smpd1^−/− (P=0.0445, Fig. 6A). Complex II activity was not significantly altered in any of the genotypes (Fig. 6B). Complex III activity in gba1^−/− was reduced by 45% compared to WT levels (P=0.0091, Fig. 6C) as previously observed (Keatinge et al., 2015). Complex III activity in the gba1^−/−;smpd1^−/− double-mutant zebrafish was reduced by only 9% compared to WT levels; however, this did not reach significance compared to the levels observed in gba1^−/− (P=0.1688). Complex IV activity was unchanged in smpd1^−/− compared to WT but reduced by 40% in gba1^−/− compared to WT, as previously reported (P=0.0491, Fig. 6D). Remarkably, there was a marked improvement of complex IV activity in gba1^−/−;smpd1^−/− with an increase in activity of 69% compared to that in gba1^−/− (P= 0.0005, Fig. 6D). Thus, there was rescue of mitochondrial respiratory chain function by mitochondrial Complexes I and IV, in which ASM deficiency normalised mitochondrial Complex IV function in gba1^−/− and GCase deficiency normalised mitochondrial Complex I function in smpd1^−/−. Malfunction of the mitochondrial respiratory chain can result in oxidative stress and subsequent lipid peroxidation. We therefore investigated whether the observed rescue in the activity of mitochondrial Complexes I and IV resulted in reduced oxidative stress-related damage. Mitochondrial lipid peroxidation was increased in whole gba1^−/− adult fish by 63% above WT levels (P= 0.0214, Fig. 6E). As predicted, lipid peroxidation levels were reduced by 70% in gba1^−/−;smpd1^−/− double mutants compared to gba1^−/− and thus effectively normalised (P=0.0094, Fig. 6E).

Biochemically, GCase and ASM both play a key role in sphingolipid metabolism (Hannun and Obeid, 2008; Quinville et al., 2021). Unexpectedly, we observed a rescue of motor behaviour and a marked prolongation of life expectancy following combined GCase and ASM deficiency, despite clear evidence of an additive effect on the intracellular level of key sphingolipids and their metabolites. The remarkable rescue effect of mitochondrial Complexes I and IV in gba1^−/−;smpd1^−/− on behaviour and survival suggests a central role of mitochondrial dysfunction in GCase deficiency. The profound normalisation of neuronal function in gba1^−/−;smpd1^−/−, as indicated in our RNAseq-based pathway analysis, is in keeping with the observation of the rescued motor phenotype. The normalisation of intracellular homeostasis is also reflected by the normalisation of both lysosomal and mitochondrial transcriptional pathways.

Mitochondrial dysfunction is a key feature of both familial and sporadic PD, as well as LSDs, as the mitochondrial network and lysosomal system are also known to be tightly interlinked (Kim et al., 2021; Magalhaes et al., 2018; Plotegher et al., 2019; Plotegher and Duchen, 2017a,b; Raimundo et al., 2016). ASM activity must also be tightly controlled, as either too much or too little has been shown to negatively affect mitochondrial function, depending on the cell type, tissue or experimental paradigm under study (Niu et al., 2022; Gillmore et al., 2022; Novgorodov et al., 2019). However, when focussed on neuronal health, ASM inhibition consistently ameliorates phenotypes in the context of neuronal loss (Novgorodov et al., 2019; Lee et al., 2014; Hagemann et al., 2020). A plausible rescue mechanism of mitochondrial function in the double-knockout (DKO) zebrafish could be glutamate/calcium signalling. Neuronal GCase deficiency in vitro has been shown to sensitise mitochondria to physiological levels of glutamate (Plotegher et al., 2019). This leads to pathological responses in calcium signalling and downstream mitochondrial dysfunction (Plotegher et al., 2019). Conversely, ASM deficiency in vitro has been shown to cause a decreased vulnerability to glutamate-linked excitotoxicity in neurons (Yu et al., 2000). This was not only linked to a decrease in intracellular calcium levels, but also to a decrease in oxidative stress (Yu et al., 2000). Inhibiting ASM function in primary oligodendrocyte culture can also rescue glutamate-induced mitochondrial dysfunction (Novgorodov et al., 2019). Of note, nine genes involved in calcium ion-regulated exocytosis were downregulated in gba1^−/− single mutants but subsequently normalised in the DKO mutant; namely, syt2a, syt7a, cplx3a, cadpsa, snap47, cacna1hb, rims1b, cbarpb and napbb.

An alternative mechanism could be intracellular redistribution of the sphingolipid profile in double mutants. We only detected a synergistic increase in sphingolipid levels in the DKO mutant compared to the gba1^−/− single mutant, but not a normalisation. However, this does not exclude an effect on subcellular localisation of a specific sphingolipid metabolite that may underpin the observed rescue mechanism. Sphingolipid signalling is vitally important for many varied intercellular and intracellular processes (Hannun and Obeid, 2008). Sphingolipid signalling must remain highly compartmentalised due to its pleiotropic effects (Canals and Clarke, 2022; Ivanova, 2020; Piccinini et al., 2010). However, due to technical reasons, we used bulk brain tissue for our metabolite analysis, which would therefore not allow for the detection intercellular and intracellular sphingolipid differences. Future work should involve metabolic analyses of sphingolipids separated by cell type and by specific cellular fractions to produce a spatial understanding of the distinct sphingolipid metabolism and distribution across the different gba1^−/− and smpd1^−/− genotypes. Furthermore, whole-body analyses using recently developed techniques to monitor in situ glucosylceramide generation would give novel insights into the glycolipid dysregulation in our double mutants (Katzy et al., 2022).

Intriguingly, deficiency of another LSD gene, asah1b, which functions on a separate arm of the ceramide pathway to smpd1 (Fig. 1), also ameliorates gba1-deficient phenotypes in vivo and in vitro. Biallelic ASAH1 mutations cause the LSD Farber disease in humans. By developing a DKO zebrafish for gba1^−/− and asah1b^−/−, Lelieveld et al. (2022) demonstrated that asah1b deficiency also led to a rescue of behavioural and neuronal phenotypes in a similar manner to that of our DKO gba1^−/−;smpd1^−/− zebrafish. In keeping with our own data, the rescue effect observed was not due to an amelioration of neuroinflammation, as DKO zebrafish retained the upregulation of tnfb, il1b and apoeb exhibited by gba1^−/− (Lelieveld et al., 2022). Similarly, Kim et al. (2018) demonstrated that pharmacological inhibition of ASAH1 led to a significant reduction in GBA1-linked cellular phenotypes, including accumulation of ubiquitinated proteins and α-synuclein in dopaminergic neuronal cultures derived from PD GBA1^+/− patients.

For practical reasons, we modelled combined enzymatic deficiency using homozygous, and not heterozygous, mutants for gba1 and smpd1, intrinsically modelling LSDs and not PD. However, the unexpected nature of our results demonstrates the need for future characterisation of combined partial LSD gene deficiencies in the wider context of PD.

All larval and adult zebrafish were housed at the University of Sheffield; experimental procedures were in accordance with the UK Home Office Animals (Scientific Procedures) Act 1986 (project license PPL 70/8437, held by O.B.). Adult zebrafish were housed at a density of 20 per tank, on a cycle of 14 h of light, 10 h of dark. Adults and embryos were kept at a constant temperature of 28°C.

The gba1^−/− mutant lines was generated using TALEN technology (Keatinge et al., 2015). The smpd1^−/− mutant line was generated by the CRISPR/Cas9 method as previously described (Keatinge et al., 2021; Hruscha et al., 2013). The following ultramer template was used: 5′-AAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAACGGATTGAGGCTTGTGTCTCCCTATAGTGAGTCGTATTACGC-3′. The smpd1^−/− line was genotyped using the following primers: F, 5′-AGCCGTGGTGGTTTCTACAG-3′, and R, 5′-CCTTCTCTCCCTTGTTCTCG-3′. The smpd1^−/− line was crossed to gba1^+/− to generate double-heterozygous individuals. These were subsequently in-crossed to generate double mutants, single mutants and WT controls. At each in-cross, larvae were genotyped at 3 dpf by larval tail biopsy as previously described (Wilkinson et al., 2013). Each genotype was raised in genotype-specific tanks at a density of 10-15 fish per tank. All individuals were re-genotyped at 10 weeks post fertilisation. For survival curves, animals were culled for humane reasons when they could no longer maintain consistent buoyancy.

WT and mutant fish were fixed in 4% paraformaldehyde overnight at 4°C before removal of the eye, which was incubated in 30% sucrose in PBS overnight at 4°C. Eyes were then embedded in OCT compound (Tissue Tek O.C.T.; Sakura, 4583) and cryosectioned at 20 µm (Leica). Slides were rehydrated in PBS before blocking for 1 h at room temperature using 150 µl blocking solution (1% sheep serum, 5% bovine serum albumin, 0.3% Triton X-100 and 0.1% Tween-20 in PBS). Slides were then incubated with 150 µl primary antibody solution (4C4; a generous gift from Noemie Hamilton, Department of Biology, University of York, York, UK; 1:50 dilution in block solution; antibody registry ID: AB_10013752) overnight at 4°C. After incubation, slides were washed three times in PBS for 20 min followed by incubation with secondary antibody (Alexa Fluor 647 anti-mouse IgG; Invitrogen, A-21235) for 2 h. Slides were then washed in PBS three times for 20 min, before adding Fluoroshield with DAPI (Sigma-Aldrich, F6057-20ML) and applying a glass coverslip. Slides were imaged on a Zeiss LSM 900 confocal microscope using a 40× water-immersion objective.

ASM activity was determined using homogenates prepared as follows: tubes containing 20 embryos (5 dpf) were sonicated in 500 μl MilliQ water and centrifuged (3400 g). Then, 20 μl of supernatant was incubated with the substrate HMU-PC (6-hexadecanoylamino-4-methylumbelliferyl-phosphorylcholine; 0.66 mM, Moscerdam Substrates, The Netherlands) at pH 5.2 and 37°C for 2 h. Fluorescence intensity was measured at 415 nm (excitation) and 460 nm (emission) using a plate reader (Perkin Elmer, LS55). Lysosomal and mitochondrial enzyme assays as well as mass spectrometry were undertaken as previously described (Keatinge et al., 2015). Enzyme assays were performed on brain homogenates at 12 weeks post fertilisation at a concentration of 1 mg/ml and at 28°C. Chitotriosidase, β-galactosidase and β-hexosaminidase activity was measured using 4-methylumbelliferyl-β-N,N′,N″-triacetyl-chitotriose (Sigma-Aldrich), 4-methylumbelliferyl-galactopyranoside (Sigma-Aldrich) and methylumbelliferyl-2-acetamido2-deoxy-β-gluco-pyranoside (Sigma-Aldrich), respectively, all dissolved in the respective McIlvaine citrate–phosphate buffer.

We were unable to isolate sufficient mitochondria from brain tissue to robustly measure lipid peroxidation signals above background levels. However, sufficient mitochondria could be isolated from 3-month-old adult zebrafish bodies to perform the assay robustly. Bodies were homogenised in ice-cold mitochondrial isolation buffer [ice-cold sucrose buffer (0.4 M phosphate buffer pH 7.4, 0.25 M sucrose, 0.15 M KCl, 40 mM KF and 1 mM N-acetyl-cysteine)]. The Abcam lipid peroxidation kit (ab118970) fluorometric assay was used to measure lipid peroxidation according to the manufacturer's instructions. Results were normalised to WT samples.

RNA was prepared from brain tissue of 12-week-old zebrafish. A TRIzol-based protocol was used to extract RNA from the tissue. Briefly, individual brains were homogenised in 250 µl TRI Reagent (Sigma-Aldrich) and incubated at room temperature before adding 50 µl chloroform (Thermo Fisher Scientific). The samples were centrifuged at 13,300 g and the top aqueous phase was collected and transferred to a separate tube. RNA was precipitated from the aqueous phase by mixing with an equal volume of isopropanol (Thermo Fisher Scientific) and centrifugation at 13,300 g. The precipitated RNA was resuspended in DEPC-treated water (Thermo Fisher Scientific), and its concentration and quality were quantified using the Nanodrop 1000 Spectrophotometer. Approximately 750 ng of high-quality total RNA, with an RNA integrity number of 9 or above, was used in the preparation of sequencing libraries using the NEB Ultra II Directional RNA Library Prep Kit (New England Biolabs, E7760), following the polyA mRNA workflow [NEBNext® Poly(A) mRNA Magnetic Isolation Module]. Libraries were individually indexed and pooled for sequencing. Single-end 100 bp sequencing was performed on the Illumina HiSeq 2500 platform using Rapid Run mode with V2 chemistry.

Raw sequencing reads were processed using the bcbio workflow system. The quality of the samples was checked using FastQC and multiqc (Ewels et al., 2016). The salmon tool (v0.9.01) was used to quantify genes from the zebrafish reference transcriptome (Danio_rerio.GRCz11.98.gtf from https://www.ensembl.org/info/data/ftp/index.html) (Patro et al., 2017). The salmon files were then imported into R using the tximport Bioconductor package (Soneson et al., 2016). Unsupervised clustering and principal component analysis with DESeq2 revealed a batch effect corresponding to sample preparation date (Love et al., 2014). Differential expression was performed using DESeq2 incorporating a batch factor into the model. The contrast tested was between gba1^−/−, double mutants and WT; a positive log₂(fold change) indicated higher expression in gba1^−/− single mutants. The ClusterProfiler Bioconductor package was used to identify enriched pathways in upregulated [adjusted P-value less than 0.05 and log₂(fold change)>1] and downregulated genes [adjusted P-value less than 0.05 and log₂(fold change)<−1] (Yu et al., 2012).

The analysis was performed with Gene Set Enrichment Analysis (GSEA) software version 4.0.3. GSEA preranked analysis was used with default settings except for ‘Collapse/Remap to gene symbols’ set to ‘No_Collapse’. A ranked list used for the analysis was calculated with each gene assigned a score based on the adjusted P-value and the log₂(fold change). Zebrafish lysosomal and mitochondrial gene sets were prepared by identifying zebrafish homologues of the genes in human gene sets in Molecular Signatures Database (MSigDB) v7.1.

GraphPad Prism v6 software was used for statistical analysis and all error bars shown denote the mean±s.d. All experiments were performed in biological triplicate unless otherwise stated. All data were analysed with either two-tailed unpaired t-test or two-way ANOVA. Significance in all enzyme activity assays was determined by two-way ANOVA with Tukey's multiple comparison test.

We thank Niemann-Pick UK (NPUK) for their support through this project.