Acetyl-leucine slows disease progression in lysosomal storage disorders

Abstract Acetyl-dl-leucine is a derivative of the branched chain amino acid leucine. In observational clinical studies, acetyl-dl-leucine improved symptoms of ataxia, in particular in patients with the lysosomal storage disorder, Niemann-Pick disease type C1. Here, we investigated acetyl-dl-leucine and its enantiomers acetyl-l-leucine and acetyl-d-leucine in symptomatic Npc1−/− mice and observed improvement in ataxia with both individual enantiomers and acetyl-dl-leucine. When acetyl-dl-leucine and acetyl-l-leucine were administered pre-symptomatically to Npc1−/− mice, both treatments delayed disease progression and extended life span, whereas acetyl-d-leucine did not. These data are consistent with acetyl-l-leucine being the neuroprotective enantiomer. Altered glucose and antioxidant metabolism were implicated as one of the potential mechanisms of action of the l-enantiomer in Npc1−/− mice. When the standard of care drug miglustat and acetyl-dl-leucine were used in combination significant synergy resulted. In agreement with these pre-clinical data, when Niemann-Pick disease type C1 patients were evaluated after 12 months of acetyl-dl-leucine treatment, rates of disease progression were slowed, with stabilization or improvement in multiple neurological domains. A beneficial effect of acetyl-dl-leucine on gait was also observed in this study in a mouse model of GM2 gangliosidosis (Sandhoff disease) and in Tay-Sachs and Sandhoff disease patients in individual-cases of off-label-use. Taken together, we have identified an unanticipated neuroprotective effect of acetyl-l-leucine and underlying mechanisms of action in lysosomal storage diseases, supporting its further evaluation in clinical trials in lysosomal disorders.


Introduction
Acetyl-DL-leucine (ADLL) is a racemic (1:1) mixture of enantiomers and a derivative of the branched chain amino acid leucine. It has been used as a medication (Tanganil TM ) for the treatment of acute vertigo in France and former French colonies since 1957; it is orally available and has a good safety profile according to the Periodic Updated Safety Reports from Pierre-Fabre, France. Although its mechanism of action (MOA) is not fully known, studies in a guinea pig model of acute vertigo showed that ADLL can restore membrane potential and excitability of abnormally polarized neurons of the medial vestibular nucleus (Vibert and Vidal, 2001).
Due to the phylogenetic similarities and close interactions between vestibular and deep cerebellar neurons (Highstein and Holstein, 2005), and an activation of cerebellar neurons is an essential part of central vestibular compensation, the effects of ADLL were evaluated in patients with cerebellar ataxias who have impaired function of cerebellar neurons. ADLL was shown to be beneficial in patients with cerebellar ataxia (Strupp et al., 2013;Schniepp et al., 2016). Since cerebellar ataxia is one of the leading clinical symptoms of some lysosomal storage diseases (LSDs) such as Niemann-Pick disease type C1 (NPC1) and the GM2 gangliosidoses (Tay-Sachs and Sandhoff disease), the symptomatic effects of ADLL were examined in 12 patients with NPC1 who were treated with the oral formulation of ADLL for a month (5 g/day). ADLL therapy in NPC1 patients led to significant improvements in cerebellar ataxia as well as in cognition and behaviour, suggestive of potential functional benefits beyond the cerebellum (Bremova et al., 2015). Our findings of in vitro benefit of ADLL in Tangier disease fibroblasts (Colaco et al., 2019) and in vivo slowing disease progression in Sandhoff disease mice (Kaya et al., 2020) suggest that ADLL may have the potential for therapeutic use in a broader range of diseases.
In all of these studies, ADLL was administered to patients as a racemic mixture of acetyl-D-leucine (ADL) and acetyl-L-leucine (ALL) enantiomers. Pre-clinical studies implicated ALL as the active isomer responsible for postural compensation after unilateral vestibular damage in an animal model (Gü nther et al., 2015;Tighilet Graphical Abstract et al., 2015), and three clinical trials are ongoing with ALL: NPC, GM2 gangliosidosis and Ataxia-Telangiectasia (clinicaltrials.gov NCT03759639,NCT03759665 and NCT03759678). CNS exposure of the administered compounds is equivalent within the limitations of the methods used, but L is metabolized rapidly whereas D is inert (Churchill et al., 2020). However, it remains to be determined if this proposed MOA is relevant to the effects of ADLL and individual enantiomers in NPC.
In the current study, we have therefore investigated the efficacy of ADLL and its distinct enantiomer components in a mouse model of NPC1 (Npc1 À/À ). We found that ADLL and its enantiomers significantly improved ataxia in Npc1 À/À mice, in agreement with clinical observations with ADLL (Bremova et al., 2015). It is important to note that, when administered pre-symptomatically, ADLL and ALL slowed the rate of disease progression, were neuroprotective and significantly extended lifespan. In contrast, ADL did not have these effects. Therefore, preclinical studies implicate ALL as the active enantiomer of ADLL responsible for the neuroprotective effects of the treatment.
When miglustat, the currently approved therapy for NPC1 (Patterson et al., 2007), was combined with ADLL in Npc1 À/À mice, significant synergistic benefit resulted, including further extension of life span. In a cohort of NPC1 patients, the significant neuroprotective effects of ADLL identified in Npc1 À/À mice were also observed in a 12-18-month extension phase of an ongoing observational study (Cortina-Borja et al., 2018). When the MOA of ADLL and its enantiomers were investigated in the cerebellum of treated Npc1 À/À mice, modulation of pathways involved in glucose metabolism was identified as potentially mediating the beneficial effects of the L-enantiomer. Finally, in individual-cases of off-label-use ADLL improved function in three GM2 gangliosidosis patients (Tay-Sachs and Sandhoff disease) and gait improvements were also demonstrated in a mouse model of Sandhoff disease. These findings demonstrate the distinct benefits of acetyl-leucine (AL) treatments in LSDs and show promise for clinical applications.

Animals and treatments
BALBc/NPC1 nih (Pentchev et al., 1980) and Sandhoff (Hexb À/À ) mice (Sango et al., 1995) were bred as heterozygotes to generate homozygous null (Npc1 À/À ) and mutant (Hexb À/À ) mice, along with their respective wild-type controls (Npc1 þ/þ , Hexb þ/þ ). Mice were housed under non-sterile conditions, with food and water available ad libitum. Since the phenotypic development affects both genders similarly upon Npc1 abrogation (Morris et al., 1982), all experiments were conducted on female mice to facilitate group housing from different litters using protocols approved by the UK Home Office Animal Scientific Procedures Act, 1986. All animal works was conducted under the UK Home Office licencing authority and the Project licence number is P8088558D. Mice were randomly assigned to treatment groups and the drugs coded and the staff blinded to treatment when performing behavioural analysis.

Chinese hamster ovary cells
Wild-type and NPC1-deficient Chinese hamster ovary (CHO) cells were used for in vitro experiments and have been described previously (Dahl et al., 1992;Higaki et al., 2001).

Drug treatments
AL analogues and miglustat treatment protocols are described in the Supplementary experimental procedures.

Mouse behavioural analysis
The weight and activity of each mouse were recorded weekly until reaching the humane endpoint (defined as a loss of 1 g body weight within 24 h). CatWalk [10.5 system (Noldus)], NG Rota Rod for mice (Ugo Basile) was performed as described in the Supplementary experimental procedures.

Sample preparation
Mice were saline perfused under terminal anaesthesia. Tissues for biochemical analysis were snap-frozen on icecold isopentane. For immunofluorescent staining, tissues were perfused with 4% paraformaldehyde followed by phosphate buffered saline (Gibco #14190144) and kept in 4% paraformaldehyde for 24 h then stored in phosphate buffered saline containing 20% w/v sucrose. Biochemical analyses were performed on water-homogenized tissues (50 mg/ml) and protein content determined (BCA protein assay, Thermo Fisher #23227) according to the manufacturer's instructions.

Western blot analyses
Western blot analyses were performed on mouse cerebellum as described in the Supplementary experimental procedures.
Nicotinamide adenine dinucleotide and its reduced form extraction Nicotinamide adenine dinucleotide (NAD) and its reduced form (NADH) extractions were performed on 20 mg phosphate buffered saline washed fresh tissue on wet ice homogenized with a Dounce homogenizer with 400 ll of NADH/NAD extraction buffer (Abcam NAD/NADH Assay kit #ab65348). For details, see Supplementary experimental procedures.

Sample analysis Sphingoid base measurements
Sphingoid base extraction and detection with reverse phase high performance liquid chromatography was performed as described in the Supplementary experimental procedures.

Glycosphingolipid measurements
Glycosphingolipids (GSLs) were extracted and measured by Normal Phase-high performance liquid chromatography according to published methods (Neville et al., 2004). For details, see Supplementary experimental procedures.

Cholesterol measurements
Cholesterol was measured from Folch-extracted tissues with the Amplex red Cholesterol assay kit according to the manufacturer's instructions. For details, refer to Supplementary experimental procedures.

Flow cytometry experiments of CHO cells
Relative acidic compartment volume staining of live cells was with LysoTracker TM Green DND-26 (Thermo Fisher #L7526), mitochondrial volume was determined with MitoTracker Green (Invitrogen #M7514) and mitochondrial reactive oxygen species with MitoSOX Red (Invitrogen #M36008

Western blotting
Western blotting was performed as described in Supplementary experimental procedures. The antibodies used are summarized in Supplementary Table 1.

Image acquisition and quantification
Imaging of brain sections and CHO cells was performed with a Leica-SP8 confocal microscope. Western blot data acquisition was conducted with LiCOR Odyssey Infrared imaging system (Model No. 9120)

Clinical studies
Demographics and statistical analysis of individual-cases of off-label-use of adult NPC1 patients treated with ADLL is described in the Supplementary experimental procedures.
Individual-cases of off-label-use in GM2 gangliosidosis patients are described in the Supplementary experimental procedures.
Blinded video-rating is described in the Supplementary experimental procedures.
All patients gave their informed consent according to the Declaration of Helsinki prior to the compassionate use study.

Statistical analysis and power calculations
To calculate the number of mice needed for the experimental groups in this study, we used G*Power software (http://www.gpower.hhu.de). The power was set as 0.8 and the significance level, a, as 0.05. The mean lifespan of the NPC1 mouse model in our facility is 87 days with a standard deviation of 3 days. NPC1 is a disease with no curative treatment and many experimental diseasemodifying drugs have reported a 10-30% increase in lifespan. Therefore, we based our power calculation on a 10-30% effect size and a sample size of minimum five and maximum eight animals per group was determined. For statistical analysis, see Supplementary experimental procedures.

Data availability
Data will be made available upon reasonable request.

Results
ADLL, ALL and ADL administered during the symptomatic phase of disease improves ataxia in Npc1 À/À mice Npc1 À/À mice have a 10-12-week life span, with onset of symptoms (gait abnormalities, tremor and weight loss) beginning at 6-7 weeks of age (Williams et al., 2014). ADLL, ALL and ADL were administered to Npc1 À/À mice in their diet (0.1 g/kg/day), with a dose identical to that used in observational clinical studies (Bremova et al., 2015). Untreated 9-week-old Npc1 À/À mice exhibit statistically significant ataxia that presents as a pronounced sigmoidal gait, relative to wild-type mice (P < 0.0001) (Supplementary Fig. 1), whereas 9-week-old Npc1 À/À mice treated with ADLL, ALL or ADL from 8 weeks of age (1 week of treatment) displayed significantly reduced ataxia as determined by measuring lateral displacement from a straight trajectory in an automated gait analysis system (Supplementary Fig. 1) ( Fig. 1A and B) (P < 0.0001, all treatments). These results indicate that the acute anti-ataxic effect of AL is stereoisomer independent in Npc1 À/À mice.
Pre-symptomatic treatment with ADLL, ALL and ADL improves ataxia in Npc1 À/À mice We investigated whether any additional benefit was conferred by AL analogues if treatment commenced before symptom onset. Npc1 À/À mice were therefore treated presymptomatically from weaning (3 weeks of age) and assessed at 9 weeks of age (6 weeks of treatment). Similar to the acute treatment response, all three AL analogues tested significantly reduced ataxia ( Fig. 1A and C) (P < 0.0001 for all treatments) with the magnitude comparable to the effects observed with 1 week of treatment ( Fig. 1A and Supplementary Fig. 1).
ADLL, ALL and ADL improve neuropathology and reduce some lipid species in Npc1 À/À mouse brain in a stereo-selective manner Since NPC1 disease is characterized by the accumulation of sphingoid bases (sphingosine and sphinganine), cholesterol, sphingomyelin, free fatty acids and GSLs (Lloyd-Evans and Platt, 2010), we measured the impact of ALs administered to Npc1 À/À mice from 3 weeks of age on lipid storage in the brain. At 59 days of age, Npc1 À/À mice exhibited increased lipid levels relative to wild-type (Williams et al., 2014) (Fig. 2A-F). In order to evaluate the differential impact of the AL analogues in the CNS, we analysed the cerebellum and the forebrain (referred to as brain in Fig. 2) separately. Total GSLs in the forebrain were not significantly altered by any of the AL treatments (Supplementary Fig. 2A), while ALL selectively reduced GM1a (20.1%; P ¼ 0.0018) and GM2 (19.6%; P ¼ 0.0222) ( Fig. 2A). Interestingly, although sphingosine levels were not significantly affected, ADLL and ADL reduced sphinganine levels by 13.5% (P ¼ 0.0456) and 18.2% (P ¼ 0.0111), respectively ( Fig. 2B and C). Analysis of the cerebellum confirmed no significant difference in total GSLs relative to untreated Npc1 À/À mice with any of the AL analogues tested ( Supplementary Fig.  2B). However, ADLL and ALL treatment significantly lowered levels of specific GSLs including GA2 (ADLL: 21.3%, P ¼ 0.0102; ALL: 23.8%, P ¼ 0.0042), whereas  (Npc1 À/À UT), ADLL (Npc1 À/À ADLL), ALL (Npc1 À/À ALL), ADL (Npc1 À/À ADL) treatments minimum five, maximum seven animals for each group. *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001. (A) Symptomatic and pre-symptomatic AL treatment y coordinates displacement of each consecutive foot from a straight-line trajectory (mean 6 SD, n ¼ 6). (B) Late stage AL treatment SEMs of y coordinate changes (mean 6 SD, n ¼ 6; one-way ANOVA. (C) Early stage AL treatment SEMs of y coordinate changes (mean 6 SD, n ¼ 6; one-way ANOVA). (D) Early stage AL treatment front and hind duty cycle measurements, mean 6 SD (two-way ANOVA). (E) Early stage AL treatments front and hind stand mean measurements, mean 6 SD (two-way ANOVA). Duty Cycle and Stand mean analyses were measured by three-recorded runs per animal and quantitative results were obtained with Noldus CatWalk 10.5 software system. (F) Early stage AL treatment motor performance measurements, mean 6 SD (one-way ANOVA). Motor function performance was measured with accelerating Rotarod (1 rpm up to 10 rpm). (G) Life expectancy percentages and median survivals (Gehan-Breslow-Wilcoxon test), n ¼ 6 a significant reduction in GA1 was only observed with ALL treatment (23.5%, P ¼ 0.0254) (Fig. 2D). Sphingosine levels were increased by 19%, (P ¼ 0.0042), but sphinganine was significantly reduced following ADL treatment (42.3%, P ¼ 0.0074) ( Fig. 2E and F). These results indicate differential biological targets of the two enantiomers. Cholesterol levels in the forebrain and in the cerebellum were not shown as total brain cholesterol levels are not changed in the NPC brain (Xie et al., 2000).
the cerebellum by western blotting and saw no changes in myelin basic protein content with AL treatments (Supplementary Fig. 2D and E).

Potential targets of leucine analogues in the cerebellum
We investigated the potential targets that ALs could affect in the cerebellum since this brain region is particularly important in NPC1-related pathology. As L-leucine is a potent activator of the mammalian target of rapamycin (mTOR) (Kimball et al., 1999), which negatively regulates autophagy (Klionsky and Emr, 2000), we investigated mTOR and its phosphorylation status in the cerebellum of AL-treated Npc1 À/À mice. Western blotting indicated that levels of mTOR and its phosphorylation on serine 2448 (Yanagisawa et al., 2017) [phosphorylated/total protein expression ratios (p/t)] were not changed by ADLL, ALL or ADL (Fig. 5A and  D). Autophagic vacuoles accumulate in NPC1 disease, consistent with impaired lysosomal flux (Boland et al., 2010). None of the compounds significantly changed the ratio of LC3-I and LC3-II ( Fig. 5B and D). Furthermore, none of the ALs changed autophagic function or flux ( Fig. 5B and D). In addition, we were unable to detect a significant change in levels of its substrate p62/ SQSTM1 (Pankiv et al., 2007) (Fig. 5C and D).

Effects on energy metabolism and antioxidant pathways
In order to investigate potential disruption of branched chain amino acid metabolism, we examined mitochondrial branched chain keto acid dehydrogenase enzyme A subunit (BCKDHA), which functions in the final step of the pathway that yields acetyl-CoA (Harris et al., 2005). Phosphorylation of BCKDHA (that inactivates the enzyme) was significantly reduced in untreated Npc1 À/À cerebellum relative to wild-type (37.6%, P ¼ 0.0254) and the ratio of phosphorylated enzyme (p) and total enzyme (t), p/t, was significantly increased (36.6%, P ¼ 0.0269) (Fig. 5E and F) in Npc1 À/À cerebellum. None of the ALs tested had a significant impact on either total BCKDHA levels or the extent of its phosphorylation ( Fig. 5E and F). Branched chain amino acids, especially leucine, are important for glutamine/glutamate balance and for energy metabolism when glucose metabolism is deficient or impaired (Yudkoff, 1997;Kennedy et al., 2016). However, there was no difference in mitochondrial enzyme glutamate dehydrogenase protein levels between untreated Npc1 À/À and Npc1 þ/þ mice or following AL treatments ( Fig. 5F and G).
The ratio of NAD/NADH regulates various metabolic pathway enzymes, such as those involved in glycolysis and the tricarboxylic acid (TCA) cycle [including pyruvate dehydrogenase (PDH) enzyme] (Srivastava, 2016). NAD/ NADH production (mostly derived from the TCA cycle) fuels the mitochondrial respiratory chain and therefore is of high importance in energy metabolism (Da Veiga Moreira et al., 2015). We therefore measured the changes in NAD/ NADH content in the cerebellum and their ratio upon AL treatments. NAD and NADH were decreased in Npc1 À/À relative to Npc1 þ/þ mice achieving statistical significance for NADH (55.1%, P ¼ 0.0314) and the sum of NAD and NADH (49.2%, P ¼ 0.0001) (Fig. 6A). Although there was no change in the levels of the coenzymes following AL   treatments, in comparison with untreated NPC1 and WT (Fig. 6A), the NAD/NADH ratio was significantly decreased with ADLL treatment (30.7%, P ¼ 0.0240) (Fig. 6B), which might be an indication of active TCA cycle machinery and glycolysis. The status of energy metabolism also can be assessed by determining ADP and adenosine triphosphate ratio (ADP/ATP). ADP/ATP ratio was also increased with ADLL (Fig. 6C) (P ¼ 0.0101), which might indicate increased energy expenditure and a glycolytic state.

Neuroprotective effects of ADLL in individual-cases of off-label-use in NPC1 patients
Having observed unanticipated beneficial effects of ALs in Npc1 À/À mice we investigated whether neuroprotective effects also occurred in NPC1 patients treated with ADLL enrolled in an observational clinical study (Cortina-Borja et al., 2018) (for demographics of patients see Supplementary Table 2). Total clinical severity scores [with higher values equating to increasing levels of disability (Yanjanin et al., 2010)] were plotted prior to initiation of treatment with Tanganil TM (ADLL), incorporating available retrospective data (Fig. 7A). All 13 patients had positive slopes of disease progression before ADLL treatment (1.8 severity units/year was the average slope). Following initiation of ADLL treatment, the average slope was À1.78 (three patients had no slope, 10 had negative slopes) equating to a significant reduction in disease progression and improvement in the majority of patients (one-sided sign test P ¼ 0.0002) (Fig. 7A). The data were also computed as annual severity increment scores (Cortina-Borja et al., 2018) that measure the rate of disease progression. A mean of À9.1%/year (P < 0.001) in annual severity increment scores was observed (Fig. 7B). When the slope each patient's pre-treatment total severity score per year was plotted, all patients had positive slopes, whereas posttreatment slopes were either zero or negative, consistent with stabilization or improvement in disease progression (Fig. 7C). When the treated patient data were analysed by neurological subdomains ( Table 1) the majority of patients improved or stabilized on treatment in functional and cognitive subdomains (Table 1).

ADLL shows benefit in other LSDs: effects in Sandhoff mice and GM2 gangliosidosis patients
The Sandhoff (Hexb À/À ) mouse model is pre-symptomatic up to 6-8 weeks of age. Subsequently, they develop tremor and their motor function begins to decline (Jeyakumar et al., 1999). By the later stages of the disease (12-15 weeks), they are inactive and are unable to complete motor function tests such as bar crossing (Jeyakumar et al., 1999). Hexb À/À animals were treated from weaning with ADLL (0.1 mg/kg/day, the same dose used for treatment of Npc1 À/À mice). ADLL-treated mice had improved gait parameters; including hind stand mean (P ¼ 0.0323) (Fig. 8A), front (P ¼ 0.0039) and hind step cycle (P ¼ 0.0062) (Fig. 8B). To date, ALL and ADL have not been evaluated in the Sandhoff mouse model.
Our findings in Sandhoff mice were extended to individual-cases of off-label-use in three patients with a confirmed GM2 gangliosidosis diagnosis (two Tay-Sachs and one Sandhoff disease) [the latter case in press (Bremova-Ertl et al., 2020)] treated with ADLL. We found a mean improvement of the Scale for Assessment and Rating of Ataxia by 20.3%, the Montreal Cognitive Assessment by 17.8% and the 8-meter-walking-test by 42%, as shown in Fig. 8C. All patients and caregivers also reported a subjective improvement and have continued treatment at the same dosage. Videos of the effect of treatment on gait and postural instability are shown in Videos 1 and 2. In addition, three experienced movement disorder experts performed blinded analysis of the videos and rated the videos based on the Clinical Impression of Change in Severity (1 ¼ normal, not at all ill; 2 ¼ borderline ill; 3 ¼ mildly ill; 4 ¼ moderately ill; 5 ¼ markedly ill; 6 ¼ severely ill; 7 ¼ among the most extremely ill). The unbiased observation before and after ADLL treatment showed statistically significant improvements on overall scoring (t-test P ¼ 0.0039) ( Fig. 8D and Table 2). ALL and ADL have not been evaluated in clinical settings, but trials with ALL are being conducted [NPC (NCT03759639) and GM2 gangliosidosis (NCT03759665) and Ataxia-Telangiectasia (NCT03759678)].

Discussion
In this study, we have investigated the effects of AL in mouse models of NPC1 and Sandhoff disease and in patients to better understand its therapeutic potential and to gain insights into its MOA.
We found that ADLL, ALL and ADL significantly improved ataxia when symptomatic Npc1 À/À mice were treated acutely for 7 days; this is in agreement with observational studies in NPC1 patients (Bremova et al., 2015) treated with ADLL, using the same dosage per kg and day. The individual enantiomers provided similar benefit to the racemic mixture for the symptomatic treatment of ataxia. The MOA explaining how ADLL and the enantiomers improve symptoms remains unknown. All AL analogues tested reduced lipid storage in neuronal and non-neuronal tissues in Npc1 À/À mice and in CHO cells null for NPC1 suggesting a MOA that is not confined to neuronal cells. The AL analogues were also observed to differentially reduce the levels of stored lipids in the liver and, to a lesser extent, in the brain of treated Npc1 À/À mice. The mechanism that underpins this 'substrate reduction' action of ALs currently remains unclear, but in view of the high degree of synergy when combined with the substrate reduction therapy drug miglustat, it may not be a major contributor to AL's therapeutic effect in NPC1 disease.
Another major finding of the current study was that ADLL and ALL (but not ADL) slowed disease progression when treatment was initiated before symptom onset, consistent with a neuroprotective mechanism. Since the neuroprotective effects were only observed with ADLL and ALL, this implicated ALL as the active enantiomer and demonstrates that the symptomatic improvement in ataxia and neuroprotection are achieved through different MOA. ALL significantly reduced neuroinflammation, which is important as this actively contributes to disease progression and reducing inflammation using nonsteroidal anti-inflammatory drugs has previously been shown to be beneficial in Npc1 À/À mice (Williams et al., 2014).
The neuroprotective effect of ADLL and ALL prompted us to determine whether similar effects occur in NPC1 patients treated with ADLL. We took advantage of an ongoing observational study in which 13 NPC1 patients have been treated with Tanganil TM (ADLL) continuously for $1 year and found that all patients showed stabilization or improvement in clinical scores, which were across all neurological domains, not just those relating to ataxia, supporting a more global neuroprotective effect in patients, analogous to those observed in the Npc1 À/À mouse.
One central question arising from these studies is the nature of the underlying MOA and in the case of ADLL and ALL, neuroprotective benefit in NPC1. Therefore, we investigated aspects of cell biology and metabolism known to be sensitive to leucine. Leucine has been shown to activate mTOR and reduce LC3-II and p62 in NPC1 cells (Yanagisawa et al., 2017). However, in this study, ADLL, ALL and ADL did not significantly affect either mTOR or autophagy in Npc1 À/À mouse cerebellum. This might be due to the presence of the acetyl group, as opposed to a primary amine in leucine, blocking the interaction with mTOR, shown in HeLa cells (Nagamori et al., 2016) or that AL analogues distribute at the cellular level in a manner that prevents their interaction with mTOR. Catabolism of leucine serves as a source of acetyl-CoA (Harris et al., 1985), which activates mTOR and nutrient sensing pathways (Son et al., 2019). However, the NPC1 cerebellum has significantly decreased levels of acetyl-CoA (Kennedy et al., 2016). This low acetyl-CoA content in NPC1 might therefore prevent nutrient sensing pathway activation. The complexity of mTOR pathways and altered metabolism more generally in NPC1 makes it likely that the effects of AL will likely be context specific.
Altered glucose/energy metabolism has previously been documented in the pre-symptomatic Npc1 À/À mouse cerebellum (Kennedy et al., 2013). PDHE1 alpha was found to be inactivated and there was a shift from pyruvate/ PDH dependency (upon which neurons rely for aerobic respiration) towards lactate/LDH (i.e. anaerobic respiration). The consequence of this would be progressive impairment of energy generation (Kennedy et al., 2013). In this same study, alterations in cerebellar amino acids (including leucine) and low acetyl-CoA were also reported (Kennedy et al., 2013). Independent metabolomics studies on Npc1 À/À mouse liver reported imbalances in amino acid levels (Ruiz-Rodado et al., 2016).
In view of data implicating altered energy metabolism in Npc1 À/À mice, we measured ADP/ATP and NAD/ NADH ratios, as a sensitive measure of energy status (Stein and Imai, 2012;Murphy and Hartley, 2018) and found that ADLL treatment might have shifted the system towards a more TCA-dependent glycolytic state, in the absence of significant changes to NAD-NADH coenzyme levels (which are lower in Npc1 À/À cerebellum), in agreement with previous findings in Npc1 À/À liver (Ruiz-Rodado et al., 2016). We then studied whether the shift in glycolysis is towards pyruvate utilization to produce lactate (anaerobic pathway) or to produce acetyl-CoA (aerobic pathway). We found that ADLL treatment significantly increased PDHE1 alpha levels, and decreased PDH enzyme phosphorylation (causes enzyme inactivation) in the cerebellum suggesting activation of the aerobic pathway. Interestingly, PDH-deficient mice also display Purkinje neuron degeneration and relapsing ataxia (Pliss et al., 2013), suggesting the ability of ADLL to protect Purkinje cells against degeneration may be via activation of PDH. However, our findings are based on whole brain lysate and more detailed studies on discreet cell populations would be informative. This effect on PDH did not reflect a change in PGC1-alpha protein levels (indicator of oxidative phosphorylation, OXPHOS), which might be due to the fact that up-regulating OXPHOS requires coenzyme NADH and ADLL treatment does not increase NADH levels.
Individual enantiomers had distinct effects in NPC1 cerebellum; while ALL normalizes altered levels of PDH and LDH and mildly reduces SOD1 levels, ADL enhanced levels of the mitochondrial reactive oxygen species scavenger SOD2 (manganese dependent SOD2). The  patients. Sandhoff untreated (Hexb 2/2 UT), ADLL (Hexb 2/2 ADLL). Six to eight animals per group. (A) Front-hind stand mean measurements of Hexb À/À UT and Hexb À/À ADLL, mean 6 SD, *P ¼ 0.0323, two-way ANOVA. (B) Front-hind step cycle measurements of Hexb À/À UT and Hexb À/À ADLL, mean 6 SD, **P < 0.0063, two-way ANOVA. (C) Percent improvement in clinical scores in three patients with GM2 gangliosidosis on baseline and after 1 month on medication with ADLL. Scale for Assessment and Rating of Ataxia Montreal Cognitive Assessment and 8-meter-walking-test were assessed. Third patient was excluded from Montreal Cognitive Assessment because the test is not approved for children. See also Videos 1 and 2. (D) Unbiased Clinical Impression of Change in Severity scores of GM2 patient videos. Clinical Impression of Change in Severity Skala; 1 ¼ normal, not at all ill; 2 ¼ borderline ill; 3 ¼ mildly ill; 4 ¼ moderately ill; 5 ¼ markedly ill; 6 ¼ severely ill; 7 ¼ among the most extremely ill patients. Paired t-test, P ¼ 0.0039. For individual scoring data see Table 2. biological relevance of subtle change in SOD1 level with ALL is unknown and functional readouts are needed to validate this result. In vivo, ADLL and ALL treatments were associated with a significantly greater benefit in motor function and lifespan relative to ADL (Fig. 1). This study suggests that targeting altered cellular metabolism in neurodegenerative diseases may achieve greater efficacy than using antioxidant approaches. The metabolic changes in Npc1 À/À mice in response to AL treatments are summarized in Fig. 9. Finally, we explored whether the benefits of AL extend to other LSDs. We observed efficacy of ADLL in a mouse model of Sandhoff disease, one of the GM2 gangliosidoses (Platt et al., 2012), a subgroup of GSL lysosomal storage diseases (LSDs) that includes Tay-Sachs disease: we demonstrated that ADLL-treated Sandhoff mice had a significant increase in lifespan and improvements in motor function (Kaya et al., 2020); however, the individual isomers ALL and ADL were not assessed.

Gait parameters
Gait variation/Acute  the proposed mechanism of AL treatments in the cerebellum. Red arrows indicate the pathologies in NPC1 relative to its wild-type counterparts. Green arrows indicate the significant changes in response to AL treatment.
Video 1 A 28-year old male patient with juvenile-onset Tay-Sachs disease, 8-meter-Walk-Test is shown. Note the gait instability with increased fall risk at baseline. The patient is not able to walk independently, constant support of either assistant or the wall is required. After 1 month on medication with ADLL, the patient has more postural stability, the gait pattern changes with increased pace rate and reduced variability and postural sway.
Video 2 8-meter-Walk-Test of a 32-year-old female patient with adult-onset Tay-Sachs disease is shown prior to and after 1 month on medication with ADLL. Note the more stable nature of the gait pattern with decreased fall risk on medication.
We also found improvement in gait parameters in Hexb À/À mice upon ADLL therapy consistent with the observational clinical studies we conducted. In three patients with GM2 gangliosidosis with ataxia that were treated with ADLL in individual-cases of off-label-use gait significantly improved, supporting the concept that ADLL and ALL may also be efficacious in multiple LSDs and other neurodegenerative diseases.
In conclusion, we have found that a well-tolerated drug, ADLL, and its enantiomer ALL slow disease progression and improve motor function and lipid accumulation in murine and cell models of NPC1. Furthermore, in observational studies in NPC1 patients, treatment with ADLL was associated with improvement in multiple neurological domains and a significant reduction in the rate of disease progression, potentially via restoration of aerobic metabolism based on the mechanistic studies conducted in the Npc1 À/À mouse model. The striking synergy of ADLL with miglustat in Npc1 À/À mice suggests that combining these two drugs will lead to improved short-and long-term clinical outcomes and merits clinical trials. It is interesting to note that 12 of the 13 NPC1 patients in the observational clinical study were on the standard of care miglustat, so part of their clinical improvement may be the result of the synergy between the two disease-modifying drugs. The individual enantiomers ALL and ADL have not yet been tested in combination with miglustat. The finding that both a mouse model of Sandhoff disease and GM2 gangliosides patients, regardless of age of onset, also showed clinical improvement when treated with ADLL suggest that this, and related analogues, may have broader utility in LSDs and more common neurodegenerative diseases. Based on ALL being the neuroprotective enantiomer, clinical trials with ALL are currently being conducted [NPC (NCT03759639), GM2 gangliosidosis (NCT03759665) and Ataxia-Telangiectasia (NCT03759678)].

Supplementary material
Supplementary material is available at Brain Communications online.