Reducing GBA2 Activity Ameliorates Neuropathology in Niemann-Pick Type C Mice

The enzyme glucocerebrosidase (GBA) hydrolyses glucosylceramide (GlcCer) in lysosomes. Markedly reduced GBA activity is associated with severe manifestations of Gaucher disease including neurological involvement. Mutations in the GBA gene have recently also been identified as major genetic risk factor for Parkinsonism. Disturbed metabolism of GlcCer may therefore play a role in neuropathology. Besides lysosomal GBA, cells also contain a non-lysosomal glucosylceramidase (GBA2). Given that the two β-glucosidases share substrates, we speculated that over-activity of GBA2 during severe GBA impairment might influence neuropathology. This hypothesis was studied in Niemann-Pick type C (Npc1 -/-) mice showing secondary deficiency in GBA in various tissues. Here we report that GBA2 activity is indeed increased in the brain of Npc1 -/- mice. We found that GBA2 is particularly abundant in Purkinje cells (PCs), one of the most affected neuronal populations in NPC disease. Inhibiting GBA2 in Npc1 -/- mice with a brain-permeable low nanomolar inhibitor significantly improved motor coordination and extended lifespan in the absence of correction in cholesterol and ganglioside abnormalities. This trend was recapitulated, although not to full extent, by introducing a genetic loss of GBA2 in Npc1 -/- mice. Our findings point to GBA2 activity as therapeutic target in NPC.


Introduction
Glucocerebrosidase (GBA) is a membrane-associated enzyme that hydrolyses the β-glucosyl linkage of glucosylceramide (GlcCer) in lysosomes [1]. Inherited deficiency of GBA due to Objective criterion for final euthanasia of Npc1 -/mice, along with WT controls, was weight loss greater than 30% of peak weight accompanied by hunched posture and reduced activity. We also checked that at this stage the animals were no longer able to remain upright due to trembling/shaking. Mice body weight was recorded on a Tuesday-Friday schedule. Once a week a movie of 1 min was made to record the clinical symptoms. There were no unexpected deaths of animals used in the study. When the mice presented reduced activity hydration gel packs and wet food were placed in the cage to facilitate access to fluids and food. Animals were anesthetized and sacrificed by cervical dislocation (see next section for details).
The mice were housed at the Institute Animal Core Facility in a temperature-and humidity-controlled room with a 12-h light/dark cycle and given free access to food and water ad libitum. All animal protocols were approved by the Institutional Animal Welfare Committee of the Academic Medical Centre Amsterdam in the Netherlands.

Tissue processing and immunohistochemistry
Animals were first anesthetized with a dose of Hypnorm (0.315 mg/mL fenyl citrate and 10 mg/mL fluanisone) and Dormicum (5 mg/mL midazolam) according to their weight. The given dose was 80μL/10g body weight. Blood was collected by a heart puncture followed by cervical dislocation. Brains were dissected, rinsed with phosphate-buffered saline (PBS) and divided by median section. The left half was fixed in phosphate-buffered formalin and further used for histology. The right half of non-perfused brain was snap frozen in liquid N 2 and stored at -80°C for biochemistry. Later, homogenates from the frozen material were made in 25 mM potassium phosphate buffer, pH 6.5, supplemented with 0.1% (v/v) Triton X-100 and protease inhibitors.
In addition, after blocking endogenous peroxidase activity and epitope retrieval as indicated above, sections were subsequently incubated with 5% (v/v) normal goat serum, rabbit anti-GBA, and intestinal alkaline phosphatase-conjugated goat anti-rabbit IgG (BrightVision, ImmunoLogic). Intestinal AP activity was detected using VectorBlue substrate (SK-5300; Vector Laboratories, Burlingame, CA, USA) in presence of 0.2 mM levamisole to inhibit endogenous non-intestinal alkaline phosphatase activity. Next, epitope retrieval (10 min, 121°C) in citrate buffer (pH 5.8) was applied to remove all antibodies [18]. Subsequently, sections were incubated with normal goat serum, rabbit anti-GBA2, and HRP-conjugated goat anti-rabbit IgG. HRP activity was detected using ImmPACT AMEC Red substrate (SK-4285; Vector Laboratories). Sections were counterstained with methyl green and mounted with VectaMount (Vector Laboratories). Analysis was performed using brightfield microscopy (Leica DM5000B) with an HC PLAN APO 20x/0.70 objective. Multispectral data sets were acquired using a Nuance imaging system (Perkin Elmer, Hopkinton, MA, USA) from 420 to 720 nm at intervals of 10 nm. Spectral libraries for each chromogen were obtained from single-stained sections and were used to unmix the triple staining patterns. Nuance 3.0.2 software was used to construct composite images applying color universal design. Images were color inverted using Adobe Photoshop to enhance visibility for color-blind individuals.
In vivo labeling of GBA2 in rat brain Wistar rats (300 g) (Charles River Laboratories, Wilmington, MA, USA) were implanted with intracerebroventricular (ICV) cannulas using the coordinates: AP -0.9, L +2.0 and V -3.4. After a recovery period conduritol-β-epoxide (1 μM) was infused at a rate of 1 μL per minute for 10 min. After one hour, ABP 1 (10 nM) [19] was infused at a rate of 1 μL per minute for 10 min. After 4 h the animals were sacrificed and the brains were frozen. Thirty-μm-thick cryostat sections were prepared and extensively washed. After overnight incubation with rabbit anti-calbindin D-28K antibody, the secondary Alexa-488-conjugated donkey anti-rabbit IgG antibody (R37116, Invitrogen, Carlsbad, CA, USA) was added and incubated for 1 h at RT. The sections were mounted with Vectashield (Vector Laboratories, Burlingame, CA, USA) containing DAPI and fluorescence was imaged using confocal laser scanning microscopy (Leica TCS SP5, Leica Microsystems, Wetzlar, Germany).

Gel electrophoresis and fluorescence scanning
Electrophoresis in sodium dodecylsulfate (SDS) containing 7.5% polyacrylamide gels was performed as earlier described [19].

Western Blot
Equal amounts of protein (50 μg) were subjected to electrophoresis on 7.5% SDS-polyacrylamide gels and then transferred to nitrocellulose membranes (Whatman, Dassel, Germany) using an electroblotting apparatus (Bio-Rad Laboratories, Hercules, CA, USA). The blots were blocked in 5% (w/v) nonfat dried milk in TBST buffer (

Histochemistry and Purkinje cell (PC) quantification
Tissue sections were stained with hematoxylin and eosin (HE). Parasagittal sections of cerebellum were scanned using an Olympus BX61VS microscope with UPlanSApo 20x/0.75 objective and TIFF images were acquired using the Olympus dotSlide system (Olympus, Tokyo, Japan). All those PCs were counted in which a nucleus with a nucleolus was observed. The PC layer was traced and length of the trace was measured using ImageProPlus 7.0 software (Media Cybernetics, Rockville, MD, USA). The number of PCs per unit of PC layer was quantified for the anterior, central, posterior, and flocculonodular transversal zones of the cerebellum [21].

Erasmus ladder
We tested motor coordination of untreated and iminosugar treated (5 mg/kg.bw/day of MZ-21) Npc1 +/+ , Npc1 +/and Npc1 -/mice using the Erasmus Ladder. Details on the device and its software have been published before [22][23][24]. Both the recording and stimulating parts of the setup are controlled by software written in LabView (National Instruments, Austin, TX, USA) operating at a fixed cycle of 2 ms. All data collected are stored in a relational database (MySQL, Oracle, Redwood Shores, CA, USA) and subsequently processed and analyzed off-line using custom-written software in LabView and Python (Python Software Foundation, Beaverton, OR, USA) as well as SPSS (IBM Corporation, Armonk, NY, USA). For the current study we followed a modified version of a previously developed protocol [24]. In principle, if health status permitted, each mouse had to perform 1 daily session during 5 days at the age of 6 weeks and one daily session during 2 days at the age of 9 and 10 weeks. The sessions for the 6 weeks-old stage included in consecutive order 3 non-perturbed sessions so as to adjust to the apparatus, 1 fix-obstacle session during which rung number 19 remained always elevated by 18 mm (i.e. 12 mm above the walking path), and 1 paired session in which the unconditioned stimulus (i.e. elevated rung) was provided at a random location on the right side 200 ms following the onset of the conditioned stimulus (i.e. a tone), which in turn depended on the predicted trajectory of the mouse (for details see [22]). The sessions for the 9 and 10 weeks-old stages included 1 nonperturbed session followed by 1 fix-obstacle session. During all sessions mice had to walk back and forth between the 2 shelter boxes, and during all sessions, which always included 45 trials, we took step time as the main outcome parameter, which was defined as the time (in ms) that elapses between the onsets of two consecutive touches on the rungs.

Cholesterol measurement
Total cholesterol levels in brain homogenates were determined after lipid extraction according to Folch [25] using a colorimetric enzymatic kit (Biolabo, Maizy, France).

Ganglioside quantification
Ganglioside composition was determined as previously described [26]. Asialo-ganglioside GM1 from bovine brain (Sigma, St Louis, MO, USA) was used as an internal standard.

(Glyco)sphingolipid quantification
Levels of ceramide and of the isomers glucosylceramide and galactosylceramide were determined in homogenates of total brain after lipid extraction by modified Bligh and Dyer method [27]. Neutral glycosphingolipids were separated from neutral lipids (ceramides and phospholipids) by Solid Phase Extraction (SPE) using a Backergrond Silica Gel (SiOH) SPE column (1 mL, 100 mg per column). After deacylation with sphingolipid ceramide N-deacylase (SCDase, Takara Bio Inc., Japan), glucosylceramide-borne glucosylsphingosine was digested with recombinant GBA (imiglucerase, Genzyme) yielding freed sphingosine. Sphingoid bases and lyso-glycosphingolipids were derivatized with O-phthaldialdehyde and analyzed by HPLC as described before [28].

Glucosylsphingosine quantification
Glucosylsphingosine analysis was performed according to a previously described protocol [29].

Statistical analysis
Values presented in figures concerning the Npc1 -/mice represent mean ± s.e.m.. Statistics were performed using the nonparametric Kruskal-Wallis test followed by a post hoc Dunn's multiple comparison test using GraphPad Prism 5 (Graph Pad Software, Inc., San Diego, USA): Ã P < 0.05; ÃÃ P < 0.01; ÃÃÃ P < 0.001.

GBA and GBA2 in wild-type and Npc1 -/mouse brain
The enzymes GBA and GBA2 in brain of end-stage Npc1 -/mice were examined employing a recently developed activity-based probe to visualize these two β-glucosidases [19]. The activitybased probe (ABP 1), a fluorescent aziridine phellitol-type suicide inhibitor (Fig 1A), labels with high specificity and sensitivity through covalent binding to the catalytic nucleophile residue in active GBA (half maximal inhibitory concentration (IC 50 ) = 1.15 nM) and GBA2 (IC 50 = 0.15 nM) enzyme molecules [19]. The two fluorescently-labeled enzymes can each be quantitatively visualized following separation by SDS-PAGE electrophoresis.
While relative content of active GBA as assessed by ABP 1 labelling was not markedly reduced ( Fig 1B), active GBA2 was nevertheless found to be increased in lysates of brain of 85-day-old Npc1 -/mice ( . Increased GBA2 in brain of Npc1 -/mice was also observed at the protein level by Western blot analysis with anti-GBA2 antibody ( Fig 1B and S1 Fig). The cerebellum is one of the brain areas most affected by neuropathology in NPC patients and Npc1 -/mice. In dissected cerebella of 75-day-old Npc1 -/mice changes in relative amount of active GBA and GBA2 as detected by ABP 1 labelling were less clear than in total brain ( Fig 1D and S1 Fig) and the ratio of GBA2/ GBA activity assayed with artificial substrate was slightly (but not significantly) increased ( Fig  1E and S1 Fig) compared to those in control wild-type (wt) mice. A similar increase in the ratio was observed for the other brain areas analyzed (S2 Fig). Next, we performed intracerebroventricular (ICV) infusion of ABP 1 into living rats. ABP 1 localized mainly to calbindin D-28K positive Purkinje cells (PCs) in the cerebellum (Fig 1F). GBA labeling was blocked by pre-infusion with conduritol-β-epoxide. These findings strongly suggest that PCs express active GBA2 in vivo (Herrera Moro et al. to be published in detail). Using immunostaining on mouse cerebellum we found that GBA2 is particularly abundant in these sole output neurons of the cerebellar cortex that are critical for motor coordination ( Fig  1G) [30]. Double immunostaining revealed GBA2 to be present in the PC cell body as well as with ABP 1 and Western blotting with anti-GBA2, anti-GBA and anti-tubulin antibodies in brain homogenates of 85-day-old Npc1 +/+ , Npc1 +/and Npc1 -/mice. (C) Ratio of GBA2 and GBA enzymatic activities (assayed with 4MU-β-D-Glc substrate) in brain homogenates of 85-day-old Npc1 +/+ , Npc1 +/and Npc1 -/mice. (D) Fluorescent labelling of GBA and GBA2 with ABP 1 and Western blotting with anti-GBA2, anti-GBA and anti-tubulin antibodies in the dendrites, whereas GBA is largely restricted to the PC cell body in a punctate distribution pattern in agreement with lysosomal localization (Fig 1G).
We then studied the natural history of neuropathology in Npc1 -/mice with special attention to GBA and GBA2. As reported earlier by others [31], progressive loss of PCs started in the anterior cerebellum, as reflected by decreased staining for calbindin D-28K. Concomitantly, astrogliosis (α-GFAP) increased in aging Npc1 -/mice. Surviving Npc1 -/-PCs in the central and more posterior cerebellum continued to express GBA and GBA2 (S3 Fig). Enlarged vesicular structures in Npc1 -/-PCs at day 80 did not stain for either GBA or GBA2, whereas the remainder of the PC cell body as well as the disorganized dendrites often co-stained for both GBA and GBA2 ( Fig 1G). Both GBA and GBA2 accumulated in axonal swellings that were observed at days 60 and 80 in Npc1 Impact of genetic deficiency of Gba2 on Npc1 -/mice Since GBA2 activity was found to be relatively increased in Npc1 -/mice (Fig 1B and 1C), we tested the consequence of introducing a genetic Gba2 loss in these animals. For this we crossed Gba2-deficient mice (C57BL/6-129S6/SvEv) without an overt phenotype [17] with Npc1 -/mice (BALB/c) to generate combined Npc1 -/-/Gba2 -/-. The genetic background of Npc1 -/mice is reported to influence disease severity, being least severe in BALB/c animals [32]. This finding was recapitulated: the average lifespan of Npc1 -/mice with mixed background being 73 days (Fig 2A and 2B) compared to 83 days for animals with BALB/c background (see also Fig 3A  and 3B). In brains of Npc1 -/-/Gba2 +/+ mice with mixed background total GBA2 activity and protein were elevated at end-stage (Fig 2C and 2D, S4 Fig), similar to that seen for animals with the BALB/c background (Fig 1B and 1C).
Body weight, physical activity and motor coordination of mice with various combinations of deficiencies in Npc1 and Gba2 were monitored. Mice were euthanized when weight loss exceeded 30% of peak weight and the animals presented hunched posture, reduced activity and were no longer able to remain upright due to trembling/shaking. As shown in the Kaplan-Meier survival curves (Fig 2A), combined loss of Gba2 prolonged lifespan significantly in Npc1 -/mice (P logrank = 0.0180 from Npc1 -/-/Gba2 +/+ ) and increased maximal bodyweight (Fig 2B) compared to animals wild-type or heterozygous for this enzyme. Neurological symptoms-defective coordination and shaking (ataxia)-were delayed by 15 days in double-knockout mice (Fig 2B). Of note, Gba2 deficiency did significantly delay, but not fully prevent complications. PC number appeared higher in the transversal cerebellar zones of 83-day-old Npc1 -/-/Gba2 -/mice compared to 73-day-old Npc1 -/-/Gba2 +/+ animals (Fig 2E and 2F), nevertheless this difference did not reach statistical significance (P = 0.1897, Fig 2F). Cerebellar calbindin D-28K expression was more widespread and neurofilament containing axonal swelling was less abundant in double-knockout mice compared to Npc1 -/-/Gba2 +/+ controls (Fig 2E). In contrast, astrogliosis was not diminished by the genetic deficiency of Gba2, as evidenced by abundant GFAP staining (Fig 2E).  Having established that genetic loss of Gba2 ameliorates neuropathology in Npc1 -/mice, we investigated whether such beneficial effect can also be obtained by pharmacological inhibition of GBA2 activity. We have previously designed very potent inhibitors of GBA2, the hydrophobic aza-sugars AMP-DNM (MZ-21) and ido-AMP-DNM (MZ-31) (Fig 1A) [33]. Both azasugars enter the brain well and inhibit GBA2 at low nanomolar concentration [34].
Npc1 -/mice (BALB/c background) were fed a diet calculated to deliver 1 or 5 mg/kg.bw/day of MZ-21 or MZ-31. Again, iminosugar-mediated reduction of GBA2 activity increased lifespan by 20-30 days (Fig 3A and 3B) and significantly delayed onset of weight loss and clinical manifestations of neurodegeneration ( Fig 3B). We found that in iminosugar-treated animals GBA2 enzyme activity (Fig 3C and 3D), but not GBA2 protein (Fig 3C and S5 Fig), was completely abrogated with either iminosugar at both doses. Additionally, the number of PCs in the cerebellum of 85-day-old iminosugar-treated Npc1 -/mice was significantly higher (P < 0.05, Fig 3F) than in age-matched untreated animals as assessed by PC counts in HE stainings ( Fig 3F) and calbindin D-28K immunostaining (Fig 3E). Further consistent with improved clinical status, axonal swelling in Npc1 -/mice as indicated by neurofilament immunostaining was less extensive following iminosugar treatment (Fig 3E). Extent of GFAP-positive astrogliosis however was not affected (Fig 3E). Also staining for activated microglia and macrophages, using Iba-1 and F4/80, were largely similar in Npc1 -/mice with and without iminosugar treatment (S6 Fig). To evaluate the effect of pharmacological inhibition of GBA2 on motor coordination we tested iminosugar-treated and untreated Npc1 +/+ , Npc1 +/and Npc1 -/mice on the Erasmus Ladder [24]. Untreated Npc1 -/mice needed significantly longer time to make a single step than Npc1 +/+ controls at the ages of 6 (paired session, P = 0.0022) and 9 weeks (non-perturbed session, P = 0.001; fix-obstacle session, P = 0.002) (Fig 3G). Remarkably, these differences were not observed for the iminosugar-treated animals during the fix-obstacle session (P = 0.060) (Fig 3G). Only at 10 weeks of age did iminosugar-treated Npc1 -/mice need significantly longer time than controls to make a single step during the non-perturbed (P < 0.001) and the fixobstacle (P < 0.001) sessions ( Fig 3G). Thus, as time progressed iminosugar-treated Npc1 -/mice also developed impaired motor coordination, but the onset of the impairments was delayed by about four weeks.

Brain lipids and lysosomes in iminosugar treated Npc1 -/mice
Given the primary role of cholesterol and ganglioside accumulation in lysosomes in NPC, we next analyzed the impact of iminosugar treatment of Npc1 -/mice on brain lipid composition and lysosomes. No significant differences between total brain cholesterol and ceramide levels of age-matched (85-day-old) Npc1 +/+ , treated or untreated Npc1 -/mice were detected (Table 1). Although not statistically significant, ceramide levels showed the trend to be elevated in total brain of Npc1 -/mice compared to Npc1 +/+ controls and tended to normalize with most iminosugar treatment conditions (Table 1). In contrast, GlcCer and the gangliosides GM3 and GM2 were increased in the brain of Npc1 -/untreated mice and these levels were not normalized by MZ-21 or MZ-31 treatment (Table 1). Rather, GlcCer was more than 10-fold increased by iminosugar treatment ( Table 1). The de-acylated form of GlcCer, glucosylsphingosine (GlcSph), was also elevated in untreated animals and further increased by the treatment (Table 1). Galactosylceramide (GalCer), the main component neuronal myelin sheaths, levels were reduced by half in the brain of Npc1 -/mice, confirming that the profound CNS demyelination in these animals was not corrected by the treatment.   Levels of GalCer, GlcCer, Cer, GlcSph and gangliosides (GM3 and GM2), and activity of lysosomal enzymes β-hexosaminidase in whole brain homogenates of mice resulting from the crossing of Gba2 -/mice with Npc1 -/mice. Data (n = 5-6 per group) were analyzed using the Kruskal-Wallis test with post hoc Dunn's multiple comparison test. * P < 0.05, ** P < 0.01 and *** P < 0.001 from Npc1 +/+ /Gba2 +/+ doi:10.1371/journal.pone.0135889.t002 Lipid analysis of dissected cerebella gave very similar results to total brain (S7 Fig). Ceramide (P < 0.05), GlcCer (P < 0.05) and GlcSph (P < 0.001) levels were increased in the cerebellum of 75-day-old untreated Npc1 -/mice compared to age-matched wt animals. In endstage (115-day-old) Npc1 -/mice, treatment did not affect ceramide levels and aggravated elevations in GlcCer and GlcSph. GalCer was reduced in cerebellum of Npc1 -/mice.
As previously reported [12], lysosome markers (β-hexosaminidase and β-glucuronidase) were elevated in brain of Npc1 -/mice as a consequence of lysosomal hypertrophy. Iminosugartreatment did not normalize these markers (Table 1). Noteworthy, the levels of cholesterol, GSLs, GlcSph, and lysosome markers were similar in 85-day-old brains of Npc1 -/-/Gba2 -/mice compared to those in brains of age-matched iminosugar-treated Npc1 -/animals (Table 2). Hence, pharmacological or genetic abrogation of GBA2 enzymatic activity while ameliorating neurological manifestation does not concomitantly correct the lysosomal hypertrophy or the accumulation of GSLs in the central nervous system of Npc1 -/mice.

Discussion
In the present study we examined the potential role of the non-lysosomal glucosylceramidase GBA2 in neurological manifestations in Npc1 -/mice. Firstly, we established that GBA2 activity is indeed abnormally high in the brain of Npc1 -/mice. This was indicated by analysis of brain using ABP-labelling of GBA2, as well as by detection of GBA2 in brain lysates by Western blot and the measurement of enzymatic activity. The enzyme was also found to be particularly abundant in PCs. Even though we did not observe a striking reduction in GBA activity in the Npc1 -/-CNS, as assayed in vitro with ABP 1 and artificial substrate, the accumulation of GBA-substrates (GlcCer and GlcSph) points to an impairment of the enzyme in vivo. Secondly, we examined the outcome of introducing a genetic loss of GBA2 in Npc1 -/mice. This ablation led to a significant increase in lifespan and amelioration of motor coordination, reflected by prolonged survival of PCs. Next, we investigated whether the loss of the GBA2 protein as such or the absence of its activity exerted the beneficial effect. Employing extremely potent GBA2 inhibitors (IC 50 values < 5 nM), we observed that administration of as little as 25 μg per mouse daily led to similar beneficial responses as observed with genetic loss of GBA2, such as increased maximal body weight, delayed PC loss and delayed onset of motor deficits. The genetic background of mice is known to influence severity of disease manifestation in Npc1 -/animals [32]. Pharmacological inhibition of GBA2 in BALB/c Npc1 -/mice led to a relatively larger increase in life span (27-38%, Fig 3B) than genetic loss of GBA2 in Npc1 -/mice in mixed background (15% lifespan increase, Fig 2B). Whether this difference can be accounted to the different genetic backgrounds or points to an additional target for the iminosugars next to GBA2 is presently unclear. It has to be stressed that low nanomolar concentrations of iminosugars were applied at which no other enzymes are known to be inhibited.
Our combined findings point out that GBA2 plays a role in the early neuropathology of NPC disease in mice. Our study also reveals that pharmacological inhibition of GBA2 offers an avenue to ameliorate neurological manifestation of NPC, in particular to delay the loss of motor coordination.
Our study led to the unexpected finding that while genetic or pharmacological reduction of GBA2 activity exerts beneficial effects in Npc1 -/mice this is not accompanied by reduction of the ganglioside accumulation and lysosomal hypertrophy in the brains of these animals. Clearly, reduction of GBA2 does not entirely prevent neuropathology, but rather delays this process. It seems likely that for complete prevention of disease, the primary underlying defect at the level of lysosomes should be resolved and further prevented. A combination therapy with a low dose of cyclodextrins might offer good prospects in this regard [35]. Nevertheless, a significant delay in loss of motor coordination in NPC patients through GBA2 activity, as observed in the mouse model, would be of great clinical value. In addition, our findings suggest that extra-lysosomal processes during lysosomal deficiency contribute to disease manifestation, a notion prior formulated by other researchers [8,36].
The molecular mechanism underlying the detrimental effect of cytosolic GBA2 activity on NPC pathology is presently not known. It might be envisioned that over-production by GBA2 of cytosolic ceramide and/or sphingosine from GlcCer and GlcSph is harmful, for example through induction of apoptosis or other cell death pathways in neuronal cells, particularly those with high GBA2 activity such as PCs. Of note, excessive formation of ceramide by GBA2 has recently been reported to promote apoptosis in human melanoma cells [37]. Detection of excessive cytosolic ceramide in NPC mice cerebella would require high-resolution in situ massspectrometry analysis, which at present is technically challenging. Our study renders no indication that the slightly elevated GlcSph in brains of Npc1 -/mice causes PC loss. The beneficial pharmacological inhibition or genetic ablation of GBA2 both led to a small further increase in GlcSph. At present it cannot be excluded that GBA2 has another unknown activity leading to formation of toxic compounds, for example through trans-glycosylation of other lipid species.
A currently approved drug for NPC disease is N-butyl-1-deoxynojirimycin (Zavesca, Actelion). N-alkylated iminosugars such as Zavesca inhibit GlcCer synthase (GCS), the enzyme catalyzing the first step in the biosynthesis of GSLs [38]. Zavesca has been found to ameliorate GD in mildly affected type 1 patients and was therefore approved for treatment of this condition [39]. Since excessive gangliosides have been postulated to contribute to neuropathology in NPC disease [40], the effect of Zavesca on NPC was studied in mouse models. At high concentrations Zavesca was found to reduce inflammation and ganglioside accumulation in the CNS of NPC mice, concomitantly delaying the onset of neuromotor symptoms [40]. In humans, treatment with this iminosugar was reported to stabilize and in some cases improve neurological aspects in NPC patients [41], leading to its registration as "orphan drug" for NPC disease in Europe. It is relevant to point out in this context, that iminosugars like Zavesca are not specific GCS inhibitors. They also inhibit the hydrolases involved in the degradation of GlcCer to ceramide: GBA and GBA2 [4,6]. Zavesca inhibits GBA2 with much higher affinity than GCS (IC 50 = 230 nM versus IC 50 = 50,000 nM, respectively) [42]. Inhibition of GBA2 activity may therefore partially underlie the observed beneficial effect of Zavesca in NPC patients. The inhibitors used in our study, AMP-DNM and L-ido-AMP-DNM, are far more specific GBA2 inhibitors than Zavesca. Oral administration of AMP-DNM was reported to increase survival of NPC mice and Sandhoff disease mice [34,43]. The beneficial effects of AMP-DNM in the present study occurred at drug concentrations at which effective inhibition of GCS is highly unlikely, whereas complete inhibition of GBA2 activity occurs. For further optimization of iminosugar-type drugs for NPC and comparable disease conditions it will be key to establish firmly whether inhibition of GBA2 or that of GCS, or even both enzymes, is required to exert an optimal clinical response.
In conclusion, GBA2 inhibition suffices to significantly increase lifespan of Npc1 -/mice despite persistent lysosomal abnormalities. Evidently, reduction of this enzymatic activity in Npc1 -/mice, or even genetic absence of the enzyme, does not completely prevent neurodegeneration. Nevertheless, our results suggest that pharmacological inhibition of GBA2 might result in a clinically significant delay of neurological complications. Given their advantageous inhibitory profile GBA2 inhibitors AMP-DNM (MZ-21) and l-ido-AMP-DNM (MZ-31) may serve as lead components for drug development for NPC.