Developmental delay in motor skill acquisition in Niemann-Pick C1 mice reveals abnormal cerebellar morphogenesis

Niemann-Pick type C1 (NPC1) disease is a lysosomal storage disorder caused by defective intracellular trafficking of exogenous cholesterol. Purkinje cell (PC) degeneration is the main sign of cerebellar dysfunction in both NPC1 patients and animal models. It has been recently shown that a significant decrease in Sonic hedgehog (Shh) expression reduces the proliferative potential of granule neuron precursors in the developing cerebellum of Npc1−/− mice. Pursuing the hypothesis that this developmental defect translates into functional impairments, we have assayed Npc1-deficient pups belonging to the milder mutant mouse strain Npc1nmf164 for sensorimotor development from postnatal day (PN) 3 to PN21. Npc1nmf164/ Npc1nmf164 pups displayed a 2.5-day delay in the acquisition of complex motor abilities compared to wild-type (wt) littermates, in agreement with the significant disorganization of cerebellar cortex cytoarchitecture observed between PN11 and PN15. Compared to wt, Npc1nmf164 homozygous mice exhibited a poorer morphological differentiation of Bergmann glia (BG), as indicated by thicker radial shafts and less elaborate reticular pattern of lateral processes. Also BG functional development was defective, as indicated by the significant reduction in GLAST and Glutamine synthetase expression. A reduced VGluT2 and GAD65 expression also indicated an overall derangement of the glutamatergic/GABAergic stimulation that PCs receive by climbing/parallel fibers and basket/stellate cells, respectively. Lastly, Npc1-deficiency also affected oligodendrocyte differentiation as indicated by the strong reduction of myelin basic protein. Two sequential 2-hydroxypropyl-β-cyclodextrin administrations at PN4 and PN7 counteract these defects, partially preventing functional impairment of BG and fully restoring the normal patterns of glutamatergic/GABAergic stimulation to PCs. These findings indicate that in Npc1nmf164 homozygous mice the derangement of synaptic connectivity and dysmyelination during cerebellar morphogenesis largely anticipate motor deficits that are typically observed during adulthood.


Introduction
Niemann-Pick type C (NPC) is an inherited lysosomal storage disorder, ultimately fatal and presenting with variable neurovisceral symptoms, age of onset and life span [1]. In spite of broad clinical features, impaired fine motor skills, unsteady gait and balance deficits are the earliest sign of neurological manifestation [2]. The most recent incidence estimate is 1.12 affected patients per 100,000 live births, although this value is likely underestimated because of misdiagnosis [3]. The defect is due to mutations in the genes NPC1 (95 % of cases) or NPC2, encoding for proteins that cooperatively mediate the egress from endosomes/lysosomes of exogenous cholesterol brought to the cells by the low density lipoprotein (LDL)/clathrin-coated pit pathway [4]. The role of NPC1/NPC2 proteins is particularly important in neural cells because cholesterol does not cross the blood-brain barrier once it is fully established after birth [5], making the adult brain mostly dependent on endogenouslyderived cholesterol. Accordingly, cholesterol de novo synthesis occurs in both neurons and astrocytes during early postnatal neurogenesis, thereafter becoming most prominent in astrocytes [6].
Progressive Purkinje cell (PC) degeneration [1,7] leading to ataxia, represents the most important neuropathological feature of the disease, although the reason for the selective vulnerability of this neuronal population is currently unknown. Because patients do not apparently show early developmental defects and also because most neuropathological signs appear in Npc1 −/− mice in the juvenile/young adult age, the possibility that early cerebellum development processes are impaired by NPC1-deficiency has mostly been neglected. However, the development and functional maturation of mouse cerebellar cortex is a long-lasting process encompassing the first three postnatal weeks [8], during which the need for cholesterol is likely to maximize to face the intense glial/neuronal cell proliferation and migration, neurite outgrowth, synaptogenesis and myelin formation. These findings may explain why Npc1 loss-of-function affects the cerebellum more severely compared to other brain regions such as the hippocampus and cortex, whose development is largely completed prior to birth [9,10]. It has been recently shown that, due to premature exit from the cell cycle, there are a decreased number of granule neurons (GNs) and a 20-25 % reduction in cerebellar lobule size at the end of cerebellar development [11]. This leads to a deficiency of GNs in the Inner Granular Layer (IGL), which may contribute to the later PC degeneration. In line with the robust mitogenic activity Shh exerts on GNs [12], Shh mRNA levels were found to be significantly reduced at the time of final divisions of GN precursors [11]. Besides GNs, also Bergmann glia (BG) responds to Shh [13] by differentiating in relationship with PC migration, dendritogenesis, synaptogenesis and maturation [14], suggesting that Npc1-deficiency also affects the normal pattern of BG differentiation.
Among the animal models of NPC disease, the Npc1 nmf164 mouse is of particular interest because it harbors a single nucleotide substitution (A to G at cDNA bp 3163) causing an aspartate-to-glycine substitution (D1005G) in the cysteine-rich luminal loop, conferring to the NPC1 protein a partial loss of activity as observed in most common human mutations [15]. By assessing the physical and sensorimotor development of pre-weaning Npc1 nmf164 homozygous mice, we have observed a significant delay in the acquisition of complex motor skills compared to wt littermates, which likely indicates an impairment of the cerebellar circuitry functionality. Therefore, we hypothesized that the differentiation of glial cells, including BG and oligodendrocytes, as well as the expression/localization patterns of functional markers of glutamatergic and GABAergic transmission might be altered in Npc1 nmf164 homozygous mice. The evidence we provide in this study, showing that cerebellar morphogenesis is significantly damaged in Npc1 nmf164 homozygous mice substantially confirms our hypothesis.
2-Hydroxypropyl-β-cyclodextrin, a drug promoting cholesterol movement from late endosomes to the metabolically active pool of cholesterol in the cytosol [16], has been shown to slow the appearance of ataxic symptoms in NPC1 disease mouse [17,18] and cat models [19], representing the major treatment currently studied in NPC1 patients. In light of this evidence we assessed whether the administration of this drug rescued the abnormal cerebellar morphogenesis of Npc1 nmf164 mice.

Animals and treatments
Npc1 nmf164/nmf164 mice with BALB/cJ background (hereafter named Npc1 nmf164 mice) obtained from heterozygous crosses were exposed to a 12 h light-dark cycle, receiving food and water ad libitum. The genotypes of pups were identified by PCR analysis of tail DNA as described [15]. Because a preliminary evaluation ruled out any gender effect on preweaning and adult behavioral performances, male and female mice were grouped together for analyses. Preweaning and adult behavioral performances were analyzed on the same cohorts of 10 Npc1 nmf164 and 10 wt littermates, obtained from 5 litters made of at least 7 pups. Treatment with 2-hydroxypropyl-β-cyclodextrin (hereafter named CD; average degree of substitution of 0.67 of hydroxypropyl groups per glucose unit, MW~1369 Da, catalog number H-107, Sigma Aldrich, Milan, Italy) was performed by two subsequent subcutaneous injection of either a 20 % solution (w/v; 4000 mg/Kg body weight) of CD in PBS, or plain PBS (sham, control) to 4-and 7-day-old mice Npc1 nmf164 and wt littermates [11,20]. The effect of CD administration on behavioral performances of preweaning pups was assessed on a cohort of 10 Npc1 nmf164 and 10 wt littermates (5 pups either PBS-or CD-injected/genotype), obtained from 5 litters made of at least 7 pups.
A scheme summarizing the time schedule of behavioral assays and expression pattern analyses is reported in Fig. 1. Experimental protocols and related procedures were approved by the Italian Ministry of Public Health. All efforts were made to minimize animal suffering, according to European Directive 2010/63/EU.

Preweaning behavior assessment
From postnatal day (PN) 3 to PN21, pups were separated from their dams daily between 9:00 a.m. and 3:00 p.m. for a maximum of 15 min, and tested for physical, postural, locomotor and complex motor behavior development in a warmed environment (30-32°C) [21][22][23]. Behavioral assessment evaluated the development of physical parameters (body weight, eye opening, fur appearance, incisor eruption), locomotion (pivoting, crawling, quadrupedal locomotion), swimming performance (direction and limb use), reflex appearance (surface righting reflex, negative geotaxis, cliff avoidance) and complex motor behaviors (ascending a ladder, crossing a narrow bridge, suspension on a wire). Besides direct behavioral observations, videos were also recorded throughout the entire test cycle. To avoid the possibility of order effect(s), the test sequence was administered to each pup in random order for each test. The attribution of the dominant behavior to a specific category in each observation period was made blindly with regard to pup's genotype. Categorization was considered reliable only when judgments were consistent (inter-rate reliability > 0.9). The test batteries used for the assessment of physical and sensorimotor development were as follows: (a) Physical development. The body weight was measured daily in the interval PN3-PN21 and eye opening, fur appearance and incisor eruption were evaluated by visual inspection. (b)Development of quadrupedal locomotion. Fluent forward movements with all limbs supporting the whole body and the pelvis elevated were analyzed from PN3 to PN15 by using Ethovision XT software (Noldus, The Netherlands). The pup was placed on a board and video-recorded for 120 s to analyze the following locomotion categories: (i) pivoting, turning movements by broad swipes with forepaws, using only one hindlimb as a pivot and having the pelvis anchored to the ground; (ii) crawling, dragging the body forward or pushing it backward by undulating movements of the trunk and often dragging the hindlimbs in an extended position with foot soles facing upward; (iii) quadrupedal locomotion, smooth and coordinated walking, in which the body is supported in sequence by different numbers of feet the acquisition of complex motor abilities requires the complete maturation of basic reflexes such as the grasping response, which normally appears by the end of the first postnatal week [22], the development of complex motor behaviors was scored from PN10 on.
(i) Ascending a ladder: the pup was placed on a steel ladder (cm 15 × 25, 20 rungs, 1 cm apart, inclination angle 25°) with top leaning against a platform holding littermates. The ability to ascend the ladder within 120 s was evaluated and the day of the first successful performance was recorded. (ii) Crossing a narrow bridge: the pup was placed on the start platform connected by a plywood bridge (40 × 1 × 3 cm) to the goal platform holding littermates. The ability to traverse the bridge within 120 s was evaluated and the day of the first successful crossing was recorded. (iii) Suspension on a wire: the pup was suspended by its forepaws on a wire (2 mm diameter and 50 cm long) extended horizontally between two poles (30 cm high). The suspension time and the first suspension with the 4 limbs (hind limb suspension) were recorded (allotted time 60 s).

Adult behavior assessment
PN30, PN60 and PN90 Npc1 nmf164 and wt littermates were subjected to two daily sessions (morning and afternoon) of the following consecutively administered tests assessing motor behavior [24]: (i) Vertical screen: the mouse was placed on a horizontal wire screen (cm 15x25, wire diameter 2 mm, spaced at 1 cm). The screen was rapidly turned to vertical position with the mouse facing the floor at the lower edge. The latency to turn upward and to climb to the upper edge was measured during 60 s. This test was performed as the first one of the morning session and was not repeated in the afternoon of that day. (ii) Balance beam: the mouse was placed perpendicularly at the center of a horizontal round beam (covered with paper tape, outer diameter 2 cm, length 1 m, divided into 10 sections and placed 50 cm above a padded surface). The retention time and the number of beam sections crossed during 180 s were recorded and the results of morning and afternoon trials were averaged. (iii) Coat hanger: the mouse was suspended in the middle of the horizontal bar of a coat hanger (diameter 3 mm, length 35 cm, placed 30 cm above a padded surface) with its forepaws. The body position of the animal was observed for 60 s and scored as follows: 0, a fall within 10 s; 1, grasping the hanger with one limb; 2, grasping the hanger with two limbs; 3, grasping the hanger with three limbs; 4, grasping the hanger with four limbs; 5, actively escaping to the end of the bar. The values of morning and afternoon trials were averaged. These tests were selected because they were similar to those we had exploited in behavioral analyses of preweaning pups in terms of functions evaluated and experimental setting. and then processed for Western blot analyses. When proteins of interest had very different electrophoretic migrations, such as in the case of glutamine synthetase and MBP, membranes were cut into strips to be probed with different antibodies. The primary and secondary antibodies used are reported in Table 2. To evaluate the effect of CD administration on protein levels, similar assays were also performed on PN15 wt and Npc1 nmf164 , either sham-or CD-treated (4 mice/genotype/treatment) mice. To this end, fixed brains were cryoprotected with sucrose (30 %, w/v, in PBS), embedded in FSC22 Clear R Frozen Section Compound (Leica Biosystems) and serially sectioned (slice thickness 8 μm) using a Leica CM 1900 cryostat. For GLAST detection, cryosections were subjected to 20 min fixation in acetone (−20°C), which significantly improved antigen detection [25]. Paraffin sections and cryosections were then processed for epitope unmasking and endogenous peroxidases inactivation. For antigen unmasking, sections were incubated (5 min × 2) in 10 mM sodium citrate, pH 6.0 in a microwave oven and then in 0.3 % H 2 O 2 for 15 min at RT to inactivate endogenous peroxidases. A 2 h incubation in a blocking solution made of 0.5 % BSA in PBS preceded the incubation of sections with anti-GLAST and anti-MBP antibodies. For the detection of VGlut2, GFAP, Glutamine synthetase and GAD65 the blocking solution was supplemented with 0.1 % Triton X-100. The incubation of sections with primary antibodies lasted approximately 18 h at 4°C and was followed by several washes in PBS before exposure to the appropriate secondary antibody (see Table 2 for details). Antibody-antigen complexes were revealed with Vectastain Elite ABC Kit (Vector Laboratories Inc., Burlingame, CA, USA) followed by DAB Peroxidase Substrate Kit (Vector Laboratories Inc.), according to manufacturer's instructions. Immunodetection specificity was assessed by omitting the primary antibody. Images were obtained using a Zeiss Axioplan microscope equipped with a Sony nex-3 N mirror-less camera (Sony Europe Limited, Milano, Italy) and processed using ImageJ NIH software (National Institutes of Health, Bethesda, MD). VGluT2-and GAD65-positive puncta were quantitated in 3-4 sagittal sections of 4 mice/genotype as previously  [26,27], with slight modifications. Images were acquired using a Zeiss Axioplan microscope at 100X magnification (Neofluar, 0.7-1.30) and a Sony nex-3 N mirror-less camera. For each antibody, at least 8 image fields of lobule II and lobule X were acquired along the molecular layer starting from the pial surface. The abundance of VGluT2-and GAD65-positive puncta was determined in regions of interest (ROI) of 6500 μm 2 and 3200 μm 2 , respectively, randomly selected in outer and inner molecular layers by the "cell counter" function of ImageJ NIH software. The number of GAD65-positive puncta around the PC's soma was also determined. Only VGluT2-and GAD65-positive puncta having a high-tomoderate staining and a diameter of 0.3-1.3 μm were counted. All determinations were performed blindly and independently by two investigators. Because no significant difference was observed between counts of lobule II and lobule X microscopic fields of wt or Npc1 nmf164 mice, data were pooled.

Statistical analyses
Statistical analyses were performed with STATISTICA 8 (StatSoft) software. Data were first tested for normality (Wilk-Shapiro's test) and homoscedasticity (Levene's test), and then analyzed by unpaired two-tailed Student's t test or two-way ANOVAs for independent (genotype, treatment) and repeated (age) measures, followed by Bonferroni's post-hoc test. When data did not fully meet parametric assumptions or were ordinal (locomotion and swimming measures), comparisons between groups were performed by Mann-Whitney's U test. To control for alpha inflation, i.e. the proportion of type I errors among all rejected null hypotheses, the False Discovery Rate (FDR) was set to 0.05 and estimated through a bootstrap procedure [28]. Differences were considered to be significant at the p ≤ 0.01 level.

Results
Npc1 nmf164 mice show a delay in the acquisition of complex motor skills requiring fine motor coordination and balance Because sensorimotor reflexes and motor skills normally appear with a definite timing during the first 3 weeks after birth, they represent a useful tool to assess early postnatal neural development [29]. We therefore evaluated the acquisition of several developmental milestones in the physical and sensorimotor development of Npc1 nmf164 mice from PN3 until weaning (PN21). Body weight, fur appearance, incisor eruption and eye opening were recorded as indexes of physical growth and development, observing no difference between Npc1 nmf164 and wt littermates (Fig. 2a).
All pups similarly increased their body weight in the interval PN3-PN21 (main effect of genotype: F 1,18 = 0.80, p = 0.78; main effect of age: F 18,324 = 364.14, p < 0.00001; To analyze the locomotor development we determined the appearance of the dominant locomotory categories pivoting (turning with circular motions), crawling (moving forward/pushing backward the body) and quadrupedal locomotion (showing fluent and swift forward movements), observing no difference between Npc1 nmf164 and wt littermates (Fig. 2b, Table 3). Namely, pups showed pivoting from PN3 to PN9, crawling at PN10-11 and quadrupedal locomotion since PN12. We also determined the development of swimming abilities and observed no effect of genotype: all pups floated with asynchronous limb movements at PN4, swam in circles at PN5, swam in a straight line at PN12 and displayed the adult swimming pattern (paddling only the hindlimbs) after PN14 ( Fig. 2c-d, Table 3). We then recorded the appearance of reflexes as surface righting reflex, negative geotaxis and cliff avoidance, which involve vestibular, tactile and proprioceptive systems [30]. Negative geotaxis and cliff avoidance are more representative of sensory ability, whereas the surface righting reflex is more representative of motor ability [22]. Npc1 nmf164 mice displayed a timing of reflex appearance that matched that of wt littermates (Fig. 2e), exhibiting similar appearance of surface righting reflex (main effect of genotype: Z = −1.70, p = 0.10) and negative geotaxis since PN4 (main effect of genotype: , as well as cliff avoidance since PN7 (main effect of genotype: Z = 0.20, p = 0.85).
In the mouse, complex motor abilities requiring fine limb coordination, balance and muscle strength are normally acquired by the end of the second postnatal week. Three tasks (ascending a ladder, crossing a narrow bridge and suspension on a wire) allowed us to differentiate the contribution of motor coordination and balance from that of grip and muscle strength. Npc1 nmf164 pups acquired these abilities with a significant delay compared to wt littermates (Fig. 2f). Indeed, whereas wt pups crossed the narrow bridge in its entire length and hanged on the wire with four limbs after PN14, Npc1 nmf164 mice crossed the bridge only at PN17 (main effect of genotype: Z = −2.54, p = 0.01) and developed the four-limb hanging ability at PN18 (main effect of genotype: Z = −2.98, p = 0.004). In contrast, grip ability and muscle strength developed similarly in Npc1 nmf164 and wt littermates, as shown by their similar ability to ascend the ladder after PN15 (main effect of genotype: Z = 0.27, p = 0.80) and to hang on the wire for a longer time with increasing age (main effect of genotype: F 1,18 = 1.09, p = 0.31; main effect of age: F 10,180 = 3.23, p = 0.0008; interaction between genotype and age: F 10,180 = 0.20, p = 0.99).
The possibility of evaluating the efficacy of CD to rescue the developmental delay in motor skills acquisition of Npc1 nmf164 and wt littermates was hampered by the hyperactivity of mouse pups elicited by the injection per se. Both CD-treated and sham group pups, regardless of genotype resisted our attempts to perform motor behavior assessments.
Motor deficits of Npc1 nmf164 mice become more severe in adulthood To fully characterize motor phenotype in adults, PN30, PN60 and PN90 Npc1 nmf164 and wt littermates were subjected to a battery of tests including Vertical screen, Balance beam, and Coat hanger.
The Vertical screen test (similar to the ascending on a ladder) investigates the climbing response that requires good grip and muscle strength (Fig. 3a). In this test Npc1 nmf164 mice reached the upper edge of the screen more slowly than wt littermates, even if both genotypes turned upwards with similar time (turning upward: main effect of genotype: F 1,18 = 0.12, p = 0.73; main effect of age: F 2,36 = 1.91, p = 0.16; interaction between genotype and age: F 2,36 = 1.52, P = 0.23); (climbing to the upper edge: main effect of genotype: F 1,18 = 11.31, p = 0.004; main effect of age: F 2,36 = 0.59, p = 0.57; interaction between genotype and age: F 2,36 = 2.63, p = 0.09).
The Balance beam test (similar to crossing a narrow bridge) measures fine motor coordination and balance (Fig. 3b). When placed on an elevated round beam, Npc1 nmf164 mice crossed significantly fewer beam sections than wt mice did and significantly fewer sections as days   The Coat hanger test (similar to suspending on a wire) further characterizes motor coordination by providing an "agility score" (Fig. 3c). Npc1 nmf164 mice obtained scores lower than those of wt mice when suspended on the coat hanger. In fact, while wt mice rapidly escaped to the bar end, Npc1 nmf164 mice did not progress to the end of the bar although they were able to grasp the bar with four limbs (main effect of genotype: F 1,18 = 18.81, p = 0.0004; main effect of age: F 2,36 = 3.80, p = 0.03; interaction between genotype and age: F 2,36 = 2.30, p = 0.11).
Bergmann glia morphogenesis and functions are defective in Npc1 nmf164 mice Our analysis of the gross morphology of PN15 Npc1 nmf164 mouse cerebellum showed that the number of GNs forming the external granule layer was significantly reduced compared to age-matched wt mice (Additional file 1 and Additional file 2: Figure S1A-B), suggesting a defective proliferation of GN precursors similar to that previously observed in Npc1 −/− mice [11]. The quantification of cells incorporating BrdU (Additional file 2: Figure S1C-D) confirmed this possibility and prompted us to further analyze the cerebellar morphogenesis of these mice.
During the first week of postnatal development, BG radial shafts span the entire molecular layer, providing the scaffold for GN migration [31] and directing the distal growth of the PC dendritic tree [32]. Further BG development favors PC dendritic arborization and synapse formation, leading to the complex reticular meshwork of the adult cerebellar cortex [14]. To determine whether Npc1-deficiency affected BG morphology and/or functional differentiation, we assessed the expression and localization pattern of glial fibrillary acidic protein (GFAP), glutamate transporter (GLAST) and Glutamine synthetase by immunohistochemistry and Western blot analysis. BG morphology was thus assessed by immunostaining histological sections of PN11 and PN15 Npc1 nmf164 and wt cerebella with antibodies directed to GFAP. While no significant difference was found between Npc1 nmf164 and wt mice at PN11 (Additional file 3: Figure S2), BG of PN15 Npc1 nmf164 mice had radial shafts, which were enlarged and irregular in caliber and displayed hypertrophic astrocytes in the internal granule layer (IGL) (Fig. 4a). The overall increase in size of BG and astrocytes of Npc1 nmf164 mice was accompanied by an abnormal increase in GFAP expression, as quantified by Western blot analysis (Fig. 4b). It is worth noting the presence of two GFAP protein bands having an apparent MW of 50 and 48 kDa, respectively, both more abundant in Npc1 nmf164 mice compared to wt littermates (main effect of genotype: 48 kDa, t 6 = 4.34, p = 0.005; 50 kDa, t 6 = 3.44, p = 0.01). The 48 kDa protein band is generated by calpain I proteolitic cleavage [33] and increases during neurodegenerative processes [34].
BG is normally provided with a large amount of GLAST, which is particularly abundant in the cell body rating score values in the Coat hanger test (c); body weight values (d) of experimental group mice of increasing age. All data are expressed as mean ± SEM. * p < 0.01, ** p < 0.001, *** p < 0.0001 and perisynaptic membranes, here preventing glutamate spillover between adjacent PCs [35]. We determined GLAST expression by immunostaining and Western blot analyses, observing a significant GLAST reduction in Npc1 nmf164 compared to wt littermates (main effect of genotype: t 6 = 4.27, p = 0.005) (Fig. 5a, c). Such GLAST reduction was particularly evident around PC soma, which are normally enwrapped by lamellar processes arising from BG cell bodies [8,14,36] and in the distal BG radial shaft close to the pial surface. Npc1 nmf164 cerebella also displayed a significant decrease in Glutamine synthetase expression, as evaluated by both immunohistochemistry and Western blot analyses (main effect of genotype: t 6 = 4.79, p = 0.003) (Fig. 5b, d). The decrease in Glutamine synthetase was stronger at the level of BG soma and milder along BG radial shafts. In spite of the abnormal morphological/ functional development of BG processes, the number and localization of BG soma around PC cell bodies were apparently normal (Additional file 4: Figure S3).

Purkinje cells of Npc1 nmf164 mice display a reduced number of glutamatergic and GABAergic inputs
PCs display distinct anatomical and physiological compartments, which receive at least two excitatory and two inhibitory inputs on different proximal and distal subcompartments of cerebellar cortex [36] respectively, dividing the molecular layer into outer and inner parts. In fact, thin distal PC dendrite branchlets receive glutamatergic inputs from parallel fibers and GABAergic inputs from stellate interneurons [37], whereas the thick proximal PC dendritic shafts receive synapses mostly from GABAergic basket interneurons and glutamatergic climbing fibers [38]. We studied PC glutamatergic and GABAergic inputs to PCs by immunostaining histological sections of PN15 Npc1 nmf164 and wt cerebella with antibodies directed to vesicular glutamate transporter subtype 2 (VGluT2, labeling glutamatergic terminals) and glutamic acid decarboxylase 65 (GAD65, labeling GABAergic terminals). Compared to wt littermates, the molecular layer of Npc1 nmf164 mouse cerebella displayed a reduced number of VGluT2-positive puncta, which was particularly pronounced at the level of outer part of molecular layer (main effect of genotype: t 6 = 3.87, p = 0.008), whereas differences at the level of inner molecular layer didn't reach statistical significance (main effect of genotype: t 6 = 2.55, p = 0.04) (Fig. 6a). As expected, VGluT2 immunostaining was also detected at the level of glomeruli, where glutamatergic afferent mossy fibers synapse with granule neuron dendrites, with similar expression patterns in wt and Npc1 nmf164 mice. Finally, Western blot analysis revealed a significant reduction of VGluT2 protein levels in the cerebellum of PN15 Npc1 nmf164 mice (main effect of genotype: t 6 = 4.75, p = 0.003) (Fig. 6b).
The analysis of GAD65 expression patterns also showed that GABAergic inputs were significantly reduced in Npc1 nmf164 cerebella. To investigate this issue, we arbitrarily divided the molecular layer into outer, inner and PC layers and determined the density of GAD65-positive puncta in each layer, observing a significant reduction of puncta in molecular and PC layers of Npc1 nmf164 vs wt mice (main effect of genotype: outer molecular layer: t 6 = 3.64, p = 0.01; inner molecular layer:  (Fig. 7a). Reduced GAD65 expression was also confirmed by Western blot analysis (main effect of genotype: t 6 = 3.71, p = 0.01) (Fig. 7b). In spite of the reduced abundance of GAD65-positive puncta, the number and localization of GABAergic interneurons along the molecular layer appeared similar in Npc1 nmf164 and wt mice, as determined by hematoxylin/eosin Y staining and parvalbumin immunostaining (Additional file 4: Figure S3).

Npc1 nmf164 mice display defective myelin maturation
It was recently shown that the selective ablation of Npc1 expression in oligodendrocytes results in defective myelin formation in the forebrain and corpus callosum of PN16 mice [39], indicating that these cells need exogenous cholesterol uptake at least during early postnatal life. This finding is in agreement with previous observations showing the expression of low-/very lowdensity lipoprotein receptors by oligodendrocytes [40] and the dependence on glia-derived cholesterol of Npc1deficient brains [41,42]. Moreover, dysmyelination and myelin loss were previously reported in prefrontal cortex, corpus callosum and hippocampus of Npc1 −/− mice [43] and found to be associated with defective genetic control of oligodendrocyte differentiation [44]. To investigate whether/how Npc1-deficiency also affected myelin Fig. 5 Bergmann glia function appears to be defective in Npc1 nmf164 mice. a Immunostaining with antibodies directed to GLAST (brown) shows that PN15 Npc1 nmf164 mice display a reduced expression of GLAST at the level of BG processes in the outer part of molecular layer (arrowheads) and around Purkinje cell soma (arrows) compared to wt littermates. Representative fields of parasagittal sections of wt and Npc1 nmf164 mouse cerebella are shown in the Fig.; scale bars: 10 μm. Higher magnification fields are shown on the right; scale bars: 5 μm. b Immunostaining with antibodies directed to Glutamine synthetase (brown) shows that PN15 Npc1 nmf164 mice display a reduced expression of Glutamine synthetase at the level of BG soma (arrowheads) and processes compared to wt littermates. Representative fields of parasagittal sections of wt and Npc1 nmf164 mouse cerebella are shown in the Fig.; scale bars: 20 μm. Higher magnification fields are shown on the right; scale bars: 5 μm. ML: Molecular Layer; PCL: Purkinje Cell Layer. c-d Western blot analyses of GLAST (c) and Glutamine synthetase (d) protein expression in cerebella of PN15 wt and Npc1 nmf164 mice. Histograms indicate GLAST (c) and Glutamine synthetase (d) abundance (mean ± SEM) determined by densitometry of protein bands obtained in at least 3 independent experiments taking the β-actin as internal reference. * p < 0.01 formation during cerebellum development, we determined the expression of myelin basic protein (MBP), a well-established marker of mature myelin [45] in PN11 and PN15 cerebella of Npc1 nmf164 and wt mice. As shown by Western blot analysis (Fig. 8a), the level of MBP isoforms was significantly reduced with respect to wt at either PN11 (main effect of genotype: 17. Western blot analysis of GAD65 protein expression in cerebella of PN15 wt and Npc1 nmf164 mice. Histograms indicate GAD65 abundance (mean ± SEM) determined by densitometry of protein bands obtained in at least 3 independent experiments taking the β-actin as internal reference. * p = 0.01 0.01; 21.5 kDa: t 6 = 21.86, p = 0.000001). Various MBP isoforms are generated by alternative splicing and exert specific functions in different intracellular compartments. Namely, the 17.2 and 21.5 kDa isoforms are highly expressed in the cell body and nucleus of developing oligodendrocytes, playing a regulatory role in the genetic program of oligodendrocyte differentiation [46]. In contrast, the 18.5 kDa isoform localizes at the plasma membrane and actively participates in membrane compaction typical of mature myelin [47].
Immunohistochemistry of PN15 cerebellar sections fully confirmed the impairment of myelin formation in Npc1 nmf164 mice, showing a significant reduction of MBP immunostaining at the level of PC axons and white matter (Fig. 8b). To further characterize this defect, similar analyses were also performed on cerebral cortex and corpus callosum, i.e. brain areas in which individual oligodendrocytes are more easily detected (Fig. 8c). Based on these analyses, the dysmyelination of Npc1 nmf164 mice appeared to be associated with poor oligodendrocyte differentiation, as indicated by the reduced length of the processes that typically radiate from oligodendrocyte soma. Accordingly, the MBP immunostaining of cerebral cortex and corpus callosum was strongly reduced, in agreement with previous observations [43].
CD treatment partially rescues the abnormal development of glial and neuronal cells in Npc1 nmf164 mice A single CD administration to PN7 Npc1 −/− mouse pups was shown to rescue cholesterol defects, extend life span [17] and restore normal patterns of cerebellar granule proliferation [11]. Therefore, to determine whether early postnatal CD treatment re-established normal patterns of glial and neuronal morphological/functional markers, we performed Western blot analyses of protein extracts obtained from cerebella of PN15 wt and Npc1 nmf164 mice, either sham-or CD-treated as previously described [20]. Results of this survey (Fig. 9) can be summarized as follows: first, wt and Npc1 nmf164 sham-treated mice displayed differences in protein levels similar to those observed between wt and Npc1 nmf164 naive mice (compare data of Figs. 4, 5, 6, 7, and 8 to those of Fig. 9), ruling out the possibility that the injection per se altered protein expression; second, CD administration somehow influenced the expression of GFAP, GLAST and MBP of wt mice, while having no apparent effect on Glutamine synthetase, VGluT2 and GAD65 (see Table 4 for two-way ANOVA analyses); third, CD administration to Npc1 nmf164 mice fully rescued the decrease of Glutamine synthetase, VGluT2, GAD65 and MBP, restoring the protein levels to those of either sham or CD-treated wt mouse cerebella (Table 4). By contrast, CD administration did not rescue GFAP and GLAST expression levels in Npc1 nmf164 mice. Indeed, GFAP and GLAST protein levels of either sham-or CD-treated Npc1 nmf164 mice were significantly higher (GFAP) and lower (GLAST), respectively, of those of wt mice. Immunohistochemical assays confirmed results obtained by Western blot analyses, showing that CD treatment re-wired the expression of VGluT2 (paradigmatic of neuronal functional marker), but not GFAP (paradigmatic of a glial morpho- Fig. 9 CD administration partly rescues morpho/functional markers of glial and neuronal cells. Representative western blot analyses of total protein preparations obtained from PN15 wt and Npc1 nmf164 mice, either sham-or CD-treated, and probed with specific antibodies. Histograms indicate the abundance (mean ± SEM) of each protein determined by densitometry of protein bands of at least 3 independent experiments taking β-actin as internal reference. * p ≤ 0.01, ** p < 0.001, *** p < 0.0001 functional marker) (Additional file 5: Figure S4, Additional file 6: Figure S5).

Discussion
We have shown that Npc1 nmf164 mice acquire fine motor coordination and balance as well as complex abilities depending on cerebellar maturation [48,49] with a significant delay compared to wt littermates, in spite of their normal physical and postural development. An overall disturbance of cerebellar morphogenesis underscores this phenotype, emphasizing the relevance of exogenous cholesterol uptake and Npc1-mediated intracellular trafficking for proper cerebellar development. The presence of a developmental delay instead of a more severe deficit of these abilities is likely explained by the greater plasticity of developing cerebellum and/or the availability of cholesterol of neuronal origin before the full shift to astrocyte-derived cholesterol, which compensates the deficit of Npc1 function.
The normal appearance of sensorimotor reflexes and locomotion development of Npc1 nmf164 pups within the first two postnatal weeks indicate that vestibular, tactile, and proprioceptive systems; descending motor pathways; and brain stem-spinal networks [30,49] are not apparently affected by Npc1-deficiency. Conversely, the domain of complex motor abilities is damaged by Npc1deficiency because they are also prematurely lost in the adulthood. In fact, motor coordination and balance are more severely impaired than grip capacity and muscle strength as early as at PN30 and these motor defects thereafter translate to severe ataxia, because of the massive PC degeneration [15,24].
The acquisition of complex motor abilities depends on proper sequencing and coordination of motor outputs. These are prominent properties of cerebellar circuitry [48], consisting of several functional modules that allow the real-time control of movements and the long-term changes underling motor learning, by finely regulated signal generation and flow that ultimately converge on PCs.
The inhibitory activity of PCs is dynamically orchestrated at the level of both dendritic shafts and cell body by a number of excitatory and inhibitory neurons, while PCs in turn modulate the excitability of deep cerebellar nuclei. Therefore, an altered pattern of synaptic inputs to PCs may affect the timing of their firing and finally result in behavior abnormalities. For example, it was recently demonstrated that an altered GN development results in impaired motor coordination [50,51]. Similar features were also reported as a consequence of a defective development of BG processes [52][53][54], likely because correct BG development is crucial for cerebellar cytoarchitecture and function [52]. Conversely, a precocious BG and PC maturation is associated with an earlier acquisition of motor abilities in young and improved motor learning and coordination in adult mice [55]. In light of these findings, it is possible that the developmental delay in the acquisition of complex motor skills we have observed in Npc1 nmf164 pups results from a derangement of synaptic inputs to PCs.
Indeed, we found several developmental anomalies that impinge on the functionality of PCs, suggesting the possibility that the selective vulnerability of these cells represents the final outcome of a number of developmental defects in glial and neuronal cells forming the ordered pattern of cell-to-cell interaction and synaptic connectivity of cerebellar cortex [56]. For instance, the abnormal BG differentiation (thicker radial shafts and a less elaborate reticular pattern of lateral processes) we observe in PN15 Npc1 nmf164 mice may be particularly relevant. In fact, BG processes organize the compartmentalization pattern of synaptic inputs that reach PCs [36], playing a prominent role in the differential guidance and targeting of basket and stellate cell axons [37]. In addition, BG processes finely regulate cerebellar synaptic activity [57] by almost completely enwrapping the synapses that parallel and climbing fibers establish with PCs [14,36].
The glutamate transporter GLAST finely regulates PC firing at BG perisynaptic processes by preventing glutamate spillover between adjacent PCs [35] and wt and Npc1 nmf164 littermates received subcutaneous injections of plain PBS or CD, according to the schedule of Fig. 1 maintaining the one-to-one functional relationship between climbing fibers and PCs that is crucial for cerebellar control of motor function [54,58,59]. Because the glutamate recovered by GLAST is metabolized to glutamine by Glutamine synthetase [60], the reduced expression of GLAST and Glutamine synthetase in Npc1 nmf164 mice is in line with the proposal that GLAST is a limiting factor in glutamate synthesis [61]. Also, GFAP plays a key role in astrocyte-neuron interactions, by modulating the trafficking and function of astrocytic and neuronal glutamate transporters, as well as glutamine production [62]. All together these findings suggest that the abnormal morphological differentiation of BG affects the functional specialization of their processes, as also indicated by the reduced expression of GLAST and Glutamine synthetase. In addition to BG, a decrease of GLAST and Glutamine synthetase expression is likely to occur in astrocytes, the functional impairment of which is indicated by astrocytosis typically displayed by Npc1 nmf164 cerebella. In this regard it is worth noting that, in Npc1-deficient mice, astrocytosis is consistently accompanied by microglia activation [15,63,64] and the down regulation of GLAST has been found to correlate with the release of inflammatory cytokines by activated microglia [65]. Npc1 is also abundant in the recycling endosomes of presynaptic terminals. In fact, Npc1-deficiency results in morphological, biochemical and functional modification of both excitatory and inhibitory presynaptic terminals and synaptic vesicle turnover [66]. This may explain the reduction of both excitatory and inhibitory inputs received by PCs, as indicated by the significant reduction of VGluT2 and GAD65 puncta. This imbalance of synaptic inputs associated with Npc1-deficiency is in agreement with previous findings showing a decrease of synaptic inputs to PCs in co-cultures of Npc1-deficient neurons and glial cells [67]. The lower number of GNs that are generated in the cerebellum of Npc1 nmf164 mice likely contributes to the reduction in glutamatergic inputs, as indicated by our finding that VGluT2 puncta reduction is prominent in the outer molecular layer, which is mostly made of GN axons. On the other hand, the GN reduction may also impinge on the full differentiation of basket/stellate interneurons, which, among other intrinsic genetic programs and extracellular cues, depends on connectivity with GN axons [68,69]. Along the same line, the lower number of GNs may also be responsible for the abnormal differentiation of BG processes in Npc1 nmf164 mice, because the glutamate released by parallel fibers modulates the degree of BG perisynaptic envelopment acting through calciumpermeable AMPA receptors [57].
Defective oligodendrocyte maturation likely underlines the overall reduction in MBP we have observed in Npc1 nmf164 cerebellum, in line with previous studies showing dysmyelination in both NPC patients and Npc1 −/− mice [43,70]. Moreover, the decreased MBP expression not only affected the 18.5 kDa (specific to mature oligodendrocytes), but also the 17.5 kDa and 21.5 kDa isoforms (specific to developing oligodendrocytes), indicating that oligodendrocyte differentiation per se is also impaired in Npc1-deficient mice. Although Npc1 deletion in neurons triggers the block of oligodendrocyte maturation and thus leads to a subsequent failure of myelin formation [39], the exogenous cholesterol uptake by oligodendrocytes coupled to Npc1-mediated intracellular trafficking is also relevant for the formation of myelin sheaths [40,44]. Accordingly, oligodendrocyte ablation during the first postnatal weeks gives rise to ataxia and motor deficits in the mouse [39,71]. By showing that a significant myelin reduction is prominent at the level of PC axons, our results further corroborate the convergence of various Npc1 deficiency-dependent abnormalities on PC functionality.
Present findings also demonstrate that early postnatal CD treatment effectively re-wires developmental trajectories, by partly rescuing the defective cerebellar morphogenesis and thus explaining the well-established beneficial effect a single CD administration to PN7 Npc1 −/− mice has in rescuing lysosomal cholesterol accumulation and slowing down the appearance of ataxic symptoms [17-19, 72, 73]. In fact, this treatment is particularly timely because cerebellar morphogenesis maximizes the need for cholesterol, making the exogenous LDL-uptaken cholesterol a rate-limiting factor for neurons [74,75]. Noteworthy, CD administration didn't rescue either GFAP hyper-or GLAST hypo-expression of Npc1 nmf164 mice, although it fully rescued Glutamine synthetase levels in these mice. Because Glutamine synthetase is mainly expressed by astrocytes, this observation rules out the possibility that astrocytes are not influenced by CD, making this issue worthy of further investigation, also in light of the ability of CD administration to influence GFAP and GLAST expression of wt mice.

Conclusions
In conclusion, we correlate the delay of complex motor skills acquisition by Npc1 nmf164 mice to a number of glial cell differentiation anomalies and derangement of synaptic input to PCs. We believe that these findings are relevant because: i) delineate a novel perspective to explain the selective Purkinje cell vulnerability in NPC1 mouse models and patients; and, ii) emphasize the need of early diagnosis to secure the best treatment efficacy in patients.

Additional files
Additional file 1: Supplementary materials and methods. (DOCX 127 kb) Additional file 2: Figure S1. Npc1 nmf164 mice display a reduced density of GNs in the external granule layer (EGL), which is due to reduced proliferation of GN precursors. A Representative sections are shown in