Preclinical pharmacology of glucosylceramide synthase inhibitor venglustat in a GBA-related synucleinopathy model

Mutations in GBA, the gene encoding the lysosomal enzyme glucocerebrosidase (GCase), represent the greatest genetic risk factor for developing synucleinopathies including Parkinson’s disease (PD). Additionally, PD patients harboring a mutant GBA allele present with an earlier disease onset and an accelerated disease progression of both motor and non-motor symptoms. Preclinical studies in mouse models of synucleinopathy suggest that modulation of the sphingolipid metabolism pathway via inhibition of glucosylceramide synthase (GCS) using a CNS-penetrant small molecule may be a potential treatment for synucleinopathies. Here, we aim to alleviate the lipid storage burden by inhibiting the de novo synthesis of the primary glycosphingolipid substrate of GCase, glucosylceramide (GlcCer). We have previously shown that systemic GCS inhibition reduced GlcCer and glucosylsphingosine (GlcSph) accumulation, slowed α-synuclein buildup in the hippocampus, and improved cognitive deficits. Here, we studied the efficacy of a brain-penetrant clinical candidate GCS inhibitor, venglustat, in mouse models of GBA-related synucleinopathy, including a heterozygous Gba mouse model which more closely replicates the typical GBA-PD patient genotype. Collectively, these data support the rationale for modulation of GCase-related sphingolipid metabolism as a therapeutic strategy for treating GBA-related synucleinopathies.

Synucleinopathies, including Parkinson's disease (PD) and dementia with Lewy bodies (DLB), are a group of progressive neurodegenerative diseases characterized histopathologically by the aberrant α-synuclein accumulation throughout the central nervous system (CNS) 1,2 . Mutations in the GBA gene are a known genetic risk factor for developing a synucleinopathy 3,4 . GBA encodes for the lysosomal hydrolase glucocerebrosidase (GCase), which catabolizes the glycosphingolipids glucosylceramide (GlcCer) and glucosylsphingosine (GlcSph) into ceramide and sphingosine, respectively. Homozygous and compound heterozygous GBA mutations cause the rare lysosomal storage disorder Gaucher disease, which is characterized by reduced GCase activity and subsequent lysosomal GlcCer and GlcSph accumulation in the brain and viscera 5 .
The link between synucleinopathies and GBA mutations is supported by a growing body of data highlighting that Gaucher disease patients and heterozygous GBA mutation carriers are at an increased risk for developing PD and dementia with Lewy bodies. Additionally, PD patients carrying GBA mutations (GBA-PD) present with earlier onset of disease and faster cognitive decline compared to non-GBA mutation carriers [6][7][8][9][10][11][12] . Interestingly, sporadic PD patients can also present with decreased GCase activity in cerebrospinal fluid (CSF) and brain [13][14][15][16][17][18] . Furthermore, age-related and progressive reduction in GCase activity has been demonstrated in the substantia nigra and putamen of post-mortem brain homogenates from unaffected individuals 19 . Reduced expression of GCase protein has been shown in several cortical areas of the brain that are impacted by α-synuclein deposition and are associated with reduced lysosomal chaperone-mediated autophagy and ceramide levels 18,20 . These observations are supported by data from numerous preclinical studies performed in in vitro and in mouse models of Gaucher disease and synucleinopathy, which highlight the link between diminished GCase activity and

Results
Sustained GCS inhibition reduced glycosphingolipid levels in the plasma and brain of GBA-related synucleinopathy mice. Given that GBA haploinsufficiency significantly increases the risk for developing a synucleinopathy, we evaluated the effects of GCS inhibition on disease pathology in a GBArelated synucleinopathy mouse model (Gba D409V/WT ). Gba D409V/WT mice carry one mutant copy of the murine Gba gene and exhibit multiple features of synucleinopathy, including accumulation of α-synuclein inclusions and reduced GCase activity in the CNS 22 . To evaluate the effects of sustained GCS inhibition on disease pathologies, we performed a long-term GCS inhibitor administration study, in which Gba D409V/WT mice were fed GZ667161 (Fig. 1A,B), a brain-penetrant GCS inhibitor and an analog of the clinical candidate compound venglustat with similar GCS inhibitory activity (Suppl. Table 1). Gba D409V/WT mice were administered GZ667161 in pelleted diet (0.033% wt/wt) starting at 4 weeks of age until they were 10 months old (Fig. 1C). Control Gba D409V/WT mice were given vehicle rodent chow for the same duration. At the time of sacrifice, mean drug exposures were 223 ± 81 ng Timeline of GZ667161 administration. GBA-related synucleinopathy mice (Gba D409V/WT ) were fed GZ667161 as described in the Methods and untreated littermates were fed a control diet without the small molecule compound. (D) GZ667161 reduces brain and plasma GlcCer and brain glucosylsphingosine (GlcSph), an accumulating glycosphingolipid in this mouse model, following 9 months of treatment. Plasma GlcSph levels are not shown as they are below the lower limit of quantification. GZ667161 concentration in plasma and brain are shown below the x-axis. Results are represented as mean ± SEM, N ≥ 15. Unpaired t-test; ****p < 0.0001. (E-G) Brain sections from Gba D409V/WT mice were immunostained for proteinase K-resistant α-synuclein (E), ubiquitin (F), and tau (G) aggregates. GCS inhibition by GZ667161 results in an insignificant numerical decrease in proteinase K-resistant α-synuclein and tau hippocampal aggregates and a significant reduction in ubiquitin aggregates compared to littermate controls. The representative images (right) show proteinase K-resistant α-synuclein immunoreactivity (E; red), ubiquitin immunoreactivity (F; green), and tau (G; green) of control and GZ667161 treated Gba D409V/WT mice at 10 months of age. DAPI-stained cell nuclei are shown in blue. One-way ANOVA followed by Tukey's post-hoc analysis. All data represent the mean ± SEM, with N = 10-15 per group; *p < 0.05. www.nature.com/scientificreports/ of drug per mL of plasma and 237 ± 72 ng drug per gram of wet brain homogenate (Fig. 1D). GlcCer levels were significantly reduced in Gba D409V/WT mice following GZ667161 administration. After 9 months of treatment, GlcCer levels in plasma were reduced to 8 ± 6% relative to controls (p < 0.0001) while GlcCer levels in the brain were reduced to 58 ± 2% relative to controls (p < 0.0001; Fig. 1D). The levels of GlcSph in the brains of control and treated mice were below the limit of quantitation and were therefore not compared (Suppl. Table 2) 22 . Together, these data demonstrated that sustained GCS inhibition effectively reduced GCase-related glycosphingolipid substrates in the CNS of GBA-related synucleinopathy mice.

Scientific
Long-term GCS inhibition reduced hippocampal protein aggregates in GBA-related synucleinopathy mice. Aberrant protein misfolding has been implicated in synucleinopathies including PD and DLB 31 , a pathological phenotype which was recapitulated in mutant Gba mouse models. To that effect, we have previously shown progressive CNS accumulation of protein inclusions composed of tau, ubiquitin, and proteinase K-resistant α-synuclein in Gaucher-related synucleinopathy mice (Gba D409V/D409V ) 22 . These protein inclusions were reduced following long-term GCS inhibition 27 . Here, we applied the same therapeutic agent to the GBA-related synucleinopathy mice (Gba D409V/WT ). While untreated Gba D409V/WT mice accumulated some protein aggregates in the brain, mice treated for 9 months with GZ667161 exhibited a decrease in proteinase K-resistant α-synuclein, tau, and ubiquitin aggregates (p = 0.2, 0.3, and 0.03, respectively; Fig. 1E-G). However, due to the modest levels of inclusions that accumulated in these mice even by 10 months of age, the reduction in protein aggregates following GZ667161 treatment was only statistically significant for the ubiquitin aggregates (p = 0.03). Regardless, the beneficial outcomes achieved here further support the rationale for modulating sphingolipid metabolism homeostasis as a means of relieving aberrant protein aggregation in the CNS.
Venglustat, a brain-penetrant GCS inhibitor, reduced glucosylceramide in mouse models of GBA-and Gaucher-related synucleinopathy. Following our evaluation of GCS inhibitors using GZ667161 to mediate disease pathology in GBA-related synucleinopathy mice (Gba D409V/WT ), we sought to evaluate the efficacy of the clinical candidate GCS inhibitor compound, venglustat. Venglustat is currently in development as a novel GCS inhibitor in multiple phase 2 and phase 3 clinical trials for several diseases in which the sphingolipid metabolic pathway at the node of GlcCer/GCS is pathophysiologically implicated. Based on results from phase 1 clinical studies and nonclinical studies evaluating the safety and tolerability, pharmacokinetic, and pharmacodynamic profiles, oral administration of venglustat was generally well-tolerated with minimal side effects and off-target pharmacological effects (Suppl. Tables 3 and 4) 32 .
To confirm that administration of venglustat formulated in a pelleted diet would produce adequate drug exposure and/or target engagement in plasma, brain, and CSF, we performed a 2-week treatment study in Gba D409V/WT mice. Treated mice were fed venglustat formulated in a pelleted diet (0.03% wt/wt) and untreated littermates were fed a control diet for two weeks ( Fig. 2A,B, C left). At necropsy, plasma, brain, and CSF were collected to evaluate drug quantities and/or glycosphingolipid levels. As a meaningful change in the accumulation of aggregated proteins in the CNS was unlikely to occur within such a short time frame, morphological assessment of hippocampal inclusions was not included as a study endpoint. CSF collected from Gba D409V/WT mice was reserved for glycosphingolipid level analysis, as the very limited volumes of CSF obtained from these mice precluded us from also measuring drug concentrations in this biofluid. Adequate levels of exposure of this brain-penetrant GCS inhibitor was confirmed in plasma (1616 ± 221 ng of venglustat per mL of plasma) and brains (901 ± 110 ng of venglustat per g of brain lysate) of treated animals (Fig. 2C, left). Total GlcCer levels were quantified via mass spectrometry. Though Gba D409V/WT mice did not accumulate GlcCer levels beyond that of their WT counterparts, animals treated with venglustat showed reduced levels of this glycosphingolipid in plasma, brain, and CSF (14 ± 2%, 68 ± 1% and 37 ± 3% of control Gba D409V/WT , respectively; p < 0.0001, N = 12) relative to untreated animals ( Fig. 2C, left).
While we observed putatively beneficial outcomes using GCS inhibitors in Gba D409V/WT mice, we next opted to evaluate the effects of long-term venglustat treatment in a more severe Gba model (Gaucher-related synucleinopathy mice, Gba D409V/D409V ). These mice provide a more robust platform for evaluating the potential pharmacological benefits of GCS inhibitors. In addition to accumulating greater amounts of hippocampal protein aggregates and GlcSph, these mice develop cognitive impairment starting at approximately 4 months of age 22 . Our previous studies had validated GCS inhibition as a therapeutic strategy for synucleinopathies in Gba D409V/D409V mice by demonstrating a significant improvement in CNS pathology and the associated memory deficit following treatment 27 . However, these results were achieved using GZ667161 and not the compound which has or is currently being evaluated in clinical trials for several disease indications, including GBA-PD 33 . As such, we sought to evaluate the effects of GCS inhibition in Gba D409V/D409V mice using the clinical candidate compound, venglustat.
We first confirmed that the presence of an additional copy of mutant Gba did not negatively affect compound distribution or target engagement with a short-term treatment experiment. Here, four-month-old Gba D409V/D409V mice were treated with venglustat in pelleted rodent chow (0.03% wt/wt) for two weeks prior to necropsy, where plasma, brain and CSF were collected. As a meaningful change in the accumulation of aggregated proteins in the CNS was unlikely to occur within such a short time frame, morphological assessment of hippocampal inclusions was not included as a study endpoint. The limited volumes of CSF collected from mice was reserved for measuring drug concentration. In treated Gba D409V/D409V mice, venglustat concentrations were 1706 ± 51 ng/ mL in plasma, 1184 ± 56 ng/g in the brain, and 56 ± 3 ng/mL in CSF (CSF data not shown) (Fig. 2C, right). Total GlcCer levels were quantified via mass spectrometry. As expected, Gba D409V/D409V mice treated with venglustat had reduced levels of GlcCer in plasma and brain (10 ± 1% and 76 ± 2% of control Gba D409V/D409V , respectively; p < 0.0001, N ≥ 44) relative to untreated animals (Fig. 2C, right). Collectively, these data confirmed CNS exposure and GCS inhibition activity of venglustat in a mouse model of Gaucher-related synucleinopathy. www.nature.com/scientificreports/ Venglustat reduced glycosphingolipids in Gaucher-related synucleinopathy mice. To evaluate the effects of sustained GCS inhibition in this more robust synucleinopathy model, we performed a long-term treatment study using the pelleted chow formulation of venglustat in Gaucher-related synucleinopathy mice starting at 4 weeks of age (Fig. 3A). A control group of littermates was fed vehicle rodent chow. Brain exposure after 8 months of administration was 617 ± 41 ng/g (Fig. 3B), while plasma exposure was 944 ± 53 ng/mL (Fig. 3B). As anticipated, mice treated with venglustat for 8 months had significantly reduced levels of Glc-Cer relative to untreated animals, with 73 ± 2% remaining in the brain and 13 ± 1% remaining in the plasma compared to controls (Fig. 3B, left; p < 0.0001, N ≥ 17 per group; Suppl. Table 5). GlcSph, the accumulating glycosphingolipid substrate in this model, was even more substantially reduced in the CNS following venglustat administration, with 63 ± 2% remaining in the brain relative to untreated animals (Fig. 3B, right; p < 0.0001, N ≥ 17 per group; Suppl. Table 5). The reduction in plasma GlcSph was similar to that of GlcCer, with 17 ± 1% remaining compared to untreated animals (Fig. 3B, right; p < 0.0001, N ≥ 17 per group; Suppl. Table 5). These data are consistent with our previously reported reductions in glycosphingolipids in Gba D409V/D409V mice treated with GZ667161 27 and demonstrated the target engagement capability of the clinical candidate GCS inhibitor venglustat in this mouse model of Gaucher-related synucleinopathy.  Four-month-old mice were orally administered venglustat, as described in the Methods, for two weeks prior to euthanasia and tissue collection. Littermates were fed the same rodent chow without the small molecule drug. Venglustat administration in GBA-related synucleinopathy mice (Gba D409V/WT ) significantly reduced glucosylceramide (GlcCer) in the plasma, brain, and cerebrospinal fluid (CSF) compared to untreated control mice (N = 6-12; left). Venglustat also significantly reduces plasma and brain GlcCer in treated Gaucher-related synucleinopathy mice (Gba D409V/D409V ) compared to controls (N = 44-50; right), confirming target engagement. Limited CSF volumes precluded us from measuring GlcCer in Gaucher-related synucleinopathy mice. Plasma and brain exposure to venglustat was confirmed in both mouse models and is shown below the x-axes. CSF exposure was not evaluated in GBA-related synucleinopathy mice due to limited sample volumes. Drug concentration in CSF was 56 ± 3 ng/mL (N = 36; data not shown) in Gba D409V/D409V mice. Results are represented as the mean ± SEM. Unpaired t-test; ****p < 0.0001. www.nature.com/scientificreports/ www.nature.com/scientificreports/ α-synuclein, ubiquitin, and tau in the CNS 22,23 . In the present study, abnormal accumulations of these protein aggregates were absent in 9-month-old wild-type (WT) controls and 4-week-old Gba D409V/D409V untreated baseline mice. In contrast, 9-month-old Gba D409V/D409V controls had significant levels of hippocampal inclusions (Fig. 3C-E). Venglustat treatment reduced the amount of proteinase K-resistant α-synuclein, ubiquitin, and tau inclusions in Gba D409V/D409V mice after 9 months of treatment (P = 0.009, 0.004, and < 0.0001, respectively; Fig. 3C-E). Collectively, these data support the hypothesis that restoration of lipid metabolism homeostasis by venglustat in mutant Gaucher mice slows or halts the accumulation of protein aggregates in the CNS. These findings are significant, as they further validated modulation of glycosphingolipid metabolism as a diseasemodifying approach for treating Gaucher-related synucleinopathies.

Scientific Reports
Venglustat administration prevented the development of the hippocampal-related memory deficit in Gaucher-related synucleinopathy mice. Mounting evidence implicates GBA mutations as a significant contributor to faster cognitive decline in patients with synucleinopathies, including PD 9-12 . In addition to the lipidopathy and proteinopathy phenotypes, Gaucher-related synucleinopathy mice also present with cognitive impairment 22,23 . To determine the effects of GCS inhibition by venglustat on cognition, hippocampalrelated memory was evaluated with a novel object recognition test following 7 months of treatment. While WT mice preferentially investigated the novel object, untreated Gba D409V/D409V mice showed no such preference, indicative of a memory deficit (Fig. 3F). Long-term treatment of Gba D409V/D409V mice with venglustat attenuated these memory deficits as demonstrated by the preferential investigation of the novel object (P < 0.0001; Fig. 3F). These data demonstrated the ability of venglustat to improve this aberrant behavioral phenotype in a mouse model of Gaucher-related synucleinopathy.

Discussion
Mounting evidence has strengthened the link between synucleinopathies, GBA mutations, and the glucocerebrosidase pathway and supports targeting the sphingolipid pathway as a potential therapeutic strategy for synucleinopathies. Mutations in GBA are the most significant risk factor for the development of synucleinopathies, acting to decrease the age of onset and accelerate disease progression 3,4,34,35 . Emerging evidence for reduced GCase activity in the CSF of idiopathic PD patients suggests that dysfunction in sphingolipid metabolism may even play a role in synucleinopathies beyond the genetically defined population (Parnetti et al., 2017). The present study supports the hypothesis that modulation of the sphingolipid pathway via inhibition of de novo GlcCer synthesis by brain-penetrant GCS inhibitors can ameliorate the proteinopathy and cognitive impairment phenotypes associated with synucleinopathies. Notably, these outcomes were achieved using venglustat, a brain-penetrant GCS inhibitor developed for clinical investigations. The complexities of Parkinson's disease pathogenesis cannot be replicated in a single preclinical model. In this study, we employed mouse models of synucleinopathy which vary in Gba mutation copy number and resultant GCase activity and exhibit a spectrum of synucleinopathy phenotypes 22 . Gaucher-related synucleinopathy mice (Gba D409V/D409V ) have greatly reduced GCase activity (~ 25% of wild-type) 22 relative to GBA-related synucleinopathy mice (Gba D409V/WT ) (~ 50 of wild-type) 22 . As a consequence, they accumulate GlcSph and protein aggregates in the CNS to a greater extent and additionally develop a cognitive defect, a clinically relevant phenotype. While we observed beneficial outcomes in both Gba mouse models, the more robust behavioral and pathological phenotypes exhibited in the homozygous model allowed us to better evaluate the effects of venglustat on disease pathology. Of note, sustained administration of venglustat in Gba D409V/D409V mice resulted in a significant improvement in memory impairment. These data are consistent with our previous work, which demonstrated an amelioration of cognitive dysfunction and even a reversal of the cognitive decline phenotype, following long-term administration of a tool compound GCS inhibitor (GZ667161) in Gba D409V/D409V mice 27 . Importantly, the present study demonstrates the efficacy of the clinical candidate compound venglustat in ameliorating the aberrant cognitive phenotype in this mouse model of Gaucher-related synucleinopathy.
There are some technical limitations to evaluating substrate reduction therapy efficacy in the GBA-related synucleinopathy mouse model. Plasma GlcSph levels in Gba D409V/WT mice could not be quantitated due to mass spectrometry assay sensitivity limits. Likewise, brain GlcSph levels were only detected in 4 out of 15 treated and 3 out of 10 control Gba D409V/WT mice (Suppl. Table 1), limiting our statistical analysis. Future assay optimization experiments may improve GlcSph detection, expanding the dynamic range and permitting evaluation of subtle but potentially meaningful differences between treated and untreated heterozygous animals. A similar drawback exists for our assessment of insoluble CNS proteins, as Gba D409V/WT mice accumulate fewer proteinaceous aggregates than Gba D409V/D409V mice. A longer treatment duration and/or more sensitive aggregate detection methods may enable better differentiation between treated and untreated heterozygous animals. In addition, the lack of neurodegeneration in this mouse model precludes the assessment of whether GCS inhibition may provide neuroprotection.
There is a wealth of data supporting the hypothesis of a bidirectional pathogenic feedback loop between GCase and α-synuclein. Decreased GCase activity and the subsequent accumulation of undegraded glycosphingolipids (GSLs) can promote changes in membrane lipid homeostasis and the misprocessing of α-synuclein. In turn, increased intracellular accumulation of α-synuclein may prevent proper GCase trafficking from the endoplasmic reticulum to the Golgi apparatus and lysosomes. The resulting decrease in GCase activity in the lysosome can result in the accumulation of GSLs and exacerbate α-synuclein oligomerization 21 . In addition, reduced GCase activity may increase α-synuclein cell-to-cell transmission 24 and reduce chaperone-mediated autophagy 20  www.nature.com/scientificreports/ patients 30 . This hypothesis is further supported by the present studies, in which the reduction of GCase-related GSLs (by venglustat or GZ667161) yielded a numerical decrease in insoluble α-synuclein levels in the CNS of Gba mouse models. These data are consistent with our previous study, wherein GCS inhibitors (GZ667161) reduced the accumulation of these proteinaceous inclusions in the brains of Gba D409V/D409V mice 27 . Similar beneficial outcomes have been demonstrated following AAV-mediated overexpression of human GCase in the CNS of Gba mutant homozygous mice; providing further validation that modulating GCase-related sphingolipid metabolism may be a viable therapeutic strategy for improving the proteinopathy and pathological phenotypes in synucleinopathies 23 . Clinical trials have been completed or are currently underway to evaluate venglustat, an orally available, highly selective, brain-penetrant, small molecule GCS inhibitor, in several patient populations including chronic neuronopathic Gaucher disease (Gaucher disease type 3; NCT02843035), Fabry disease (NCT02228460), GM2 gangliosidosis (NCT04221451), autosomal-dominant polycystic kidney disease (NCT04705051), and early stage PD patients harboring heterozygous GBA mutations (NCT02906020), as a potential disease-modifying therapeutic. Despite the strength of these and other preclinical data, GCS inhibition by venglustat in early stage GBA-PD patients did not alter disease progression following one year of administration. Though clinical development for this indication has been halted, venglustat is still being pursued as a potential disease-modifying therapy in the additional rare disease indications (https:// www. sanofi. com/ en/ media-room/ press-relea ses/ 2021/ 2021-02-05-07-30-00). These differences between the clinical and preclinical results warrants further investigation to better understand the role of sphingolipids in Parkinson's disease pathophysiology beyond their relationship with α-synuclein aggregation. Ongoing efforts to identify genetic modifiers of GBA-PD risk and age at onset, via genome-wide association studies, may aid in our understanding of these results 36 . Furthermore, such efforts may help to improve the design of future clinical studies by narrowing the selection of GBA-PD patients carrying particular variants, including those that are implicated in lysosomal dysfunction. Certainly, additional studies are required to determine the full impact of lysosomal dysfunction on PD pathogenesis. Yet, the ability to alter the course of disease through restoration of lipid metabolism indicate that the GCase-associated sphingolipid pathway is a key target for the development of transformative therapies for synucleinopathies.

Materials and methods
Animals. The Institutional Animal Care and Use Committee (IACUC) at Sanofi approved all procedures. All experiments were performed in accordance with relevant guidelines and regulations as stipulated by the IACUC. Animals were individually housed under light:dark (12:12 h) cycles and provided with food and water ad libitum. Environmental enrichment was provided in the form of corn cob bedding, nestlets, and cardboard nest boxes. Dietary enrichment was not provided to any groups. Mouse models of Gaucher disease, including Gaucher-related synucleinopathy (Gba D409V/D409V ) and GBA-related synucleinopathy (Gba D409V/WT ), harbor a point mutation at residue 409 in the mouse glucocerebrosidase (Gba) gene 22 on a C57Bl/6 background. Age-matched C57Bl/6 wild-type (WT) mice, used as controls, were either littermates to Gba D409V/WT mice or were purchased (Charles River Laboratories) for use in Gba D409V/D409V studies. All behavioral testing was performed during the animals' light cycle, between the hours of 8 AM and 4 PM. All methods and data are reported in accordance with ARRIVE guidelines (https:// arriv eguid elines. org) and recommendations.
Administration of the glucosylceramide synthase inhibitors: venglustat and tool compound GZ667161. A subset of animals received glucosylceramide synthase inhibitors, venglustat (aka GZ402671) or GZ667161, via pelleted diet at 0.03%-or 0.033%-wt/wt, respectively. For each experiment, sex and siblings were randomly matched for group assignment. Target engagement and exposure confirmation studies included Gba D409V/D409V or Gba D409V/WT mice administered venglustat for two consecutive weeks beginning at approximately 4 months of age. Mice included in sustained GCS inhibition studies were administered either GZ667161 or venglustat upon weaning at ~ 4 weeks of age. Wild-type, baseline, and control groups were fed vehicle rodent chow. GCS inhibitor and vehicle diets were continuously provided to mice until necropsy and tissue collection. CSF collection. Animals were anesthetized via an intraperitoneal injection of a 10:1 Ketamine/Xylazine cocktail prior to being placed into a surgical ear bar rig. After making a midline cut to remove a small patch of skin from the head, the fat and muscle layers were opened using a cautery pen (Thermo Fisher Scientific; Waltham, MA) to reveal the base of the skull and occipital crest. The remaining tissue was then removed to expose the cisterna magna membrane. Using a pulled glass pipette (World Precision Instruments; Sarasota County, FL), the cisterna magna membrane was punctured to allow CSF to flow freely into the pipette via capillary action. After collecting approximately 10-20 uL, CSF was transferred to a clean protein lo-bind tube (Eppendorf; Hamburg, Germany). CSF samples with visible blood contamination were excluded from analyses. Animal perfusion and tissue and blood collection. Prior to whole blood collection, mice were anesthetized via a 200 uL intraperitoneal injection of sodium pentobarbital. Following the loss of response to a footpinch and corneal reflex, approximately 250 uL of whole blood was collected from the retro-orbital sinus using a glass capillary tube into a Microtainer® tube (BD Biosciences; Billerica, MA) containing K2 EDTA anticoagulant. Whole blood samples were collected retro-orbitally and immediately placed on ice. Plasma was isolated after 5 min centrifugation at 8000 RPM at 4 °C. Immediately following blood collection, animals were transcardially perfused with cold phosphate-buffered saline (PBS) at a rate of 18 mL/minute, for two minutes. After cutting the brains sagittally along the midline, the left hemisphere was microdissected into various regions, snap-frozen in liquid nitrogen, and stored at − 80 °C until use 27  Measurement of glycosphingolipid levels. Quantitative analysis of sphingolipids was performed by liquid chromatography and tandem mass spectrometry (LC-MS/MS) 28 . Briefly, brain tissue was homogenized in 10 volumes of water (w/v). Ten microliters of homogenate or plasma was extracted with 1 ml of extraction solution (50:50 acetonitrile/methanol) by protein precipitation. Mouse CSF sphingolipids were extracted by liquid-liquid extraction, as previously described 37  . Sections were coverslipped with aqua-poly/mount (Polysciences; Warrington, PA) and the stratum radiatum external to the CA1 hippocampal cell body layer was imaged with a SPOT camera (SPOT Imaging; Sterling Heights, MI) paired with a Nikon Eclipse E800 fluorescence microscope equipped with a 20 × objective lens, as previously described 22 . Two to three sections were imaged per animal and immunofluorescence was quantitatively measured via threshold fluorescent area on MetaMorph Software (Molecular Devices; San Jose, CA) 22 . All procedures were performed blinded to the treatment or genotype and the percent threshold area is expressed as the mean ± SEM 22 .

Novel object recognition test. Testing was performed in a white Plexiglas open arena box measuring
20 × 20 × 10-high inches. Animals acclimated in the arena for 10 min prior to a 5 min inter-trial-interval in their home cage. Animals were then exposed to two symmetrically placed objects, A and B, for 5 min and the total number of target investigations was recorded. Animals that did not investigate each object a minimum of two times during these 5 min were disqualified from the test and excluded from analysis. After a 24-h retention period (day 2), non-disqualified animals were placed back into the arena and presented with a familiar object, A, and a novel object, C, for 5 min where the total number of target investigations were again recorded. Animals that did not investigate the novel object, C, a minimum of two times were disqualified from the test and excluded from the analysis. All procedures were performed unblinded to the genotype and treatment due to genotypespecific variations in the animals' coat colors and the presence of a color-indicator in the formulated diets. Two animals, each from different treatment group, were tested simultaneously during each trial in separate arenas and scored by an independent investigator. Results are reported as the percentage of target investigations for the novel object during the testing trial on day 2. A target investigation score of 50% indicated no preference for either object.
Statistical analyses. Parametric statistical analyses were performed either by Student's t-test or analysis of variance (ANOVA) followed by Tukey's post-hoc analysis. Nonparametric data were analyzed by Kruskal-Wallis test with Dunn's multiple comparison analysis. Novel object recognition data was analyzed by column statistics and one-sample t-test to compare column means to a hypothetical value of 50% (50% signifies cognitive impairment using this behavioral test). All statistical analyses were performed using GraphPad Prism v4.0 (GraphPad Software; San Diego, CA). A p-value of < 0.05 was considered a statistically significant difference. Sample sizes were determined via a sample size calculation using historical means and standard deviations (obtained from the novel object recognition test), α = 0.05, β = 0.2, and a desired power of 0.80.

Data availability
All data generated and/or analyzed during this study are included in this article and its Supplementary Information files. www.nature.com/scientificreports/