The Cerebellum in Niemann-Pick C1 Disease: Mouse Versus Man

Abstract

Selective neuronal vulnerability is common to most degenerative disorders, including Niemann-Pick C (NPC), a rare genetic disease with altered intracellular trafficking of cholesterol. Purkinje cell dysfunction and loss are responsible for cerebellar ataxia, which is among the prevailing neurological signs of the NPC disease. In this review, we focus on some questions that are still unresolved. First, we frame the cerebellar vulnerability in the context of the extended postnatal time length by which the development of this structure is completed in mammals. In line with this thought, the much later development of cerebellar symptoms in humans is due to the later development and/or maturation of the cerebellum. Hence, the occurrence of developmental events under a protracted condition of defective intracellular cholesterol mobilization hits the functional maturation of the various cell types generating the ground of increased vulnerability. This is particularly consistent with the high cholesterol demand required for cell proliferation, migration, differentiation, and synapse formation/remodeling. Other major questions we address are why the progression of Purkinje cells loss is always from the anterior to the posterior lobes and why cerebellar defects persist in the mouse model even when genetic manipulations can lead to nearly normal survival.

Introduction

Niemann-Pick C1 disease (NPC, OMIM # 257,220) is an autosomal recessive, neurodegenerative lysosomal storage disorder with variable clinical phenotypes [1–4]. The most common presentation of NPC disease is a child of either sex developing coordination problems, dysarthria, and hepatosplenomegaly during early school-age years [3–5]. This is accompanied by abnormal intracellular accumulation of cholesterol and glycosphingolipids in a variety of tissues, including the liver and spleen, and progressive cerebellar degeneration [2, 3, 6, 7]. The neurological progression of the disorder is relentless and it is characterized by an increasing severity of ataxia, dysarthria, and dementia until death supervenes, usually during the second decade of life [3–5, 8].

Peter Pentchev, working in Roscoe Brady’s laboratory, discovered the NPC1’s cholesterol storage in the late 1980s [9]. This was soon found to be a lysosomal storage disease [10], although the late endosomal compartment was also implicated [11]. The French neurochemist, Marie Vanier, had also found ganglioside storage [12] and was instrumental in identifying a second gene involved, Niemann-Pick C2 [13]. The causative gene for NPC1 was soon found in man [14] and mouse [15]. Abundant expression of the gene was rapidly found in the cerebellum in vivo, and in in vitro cultures, from day 0 on (embryonic stages were not studied, [16]). These findings, and the early cerebellar dysfunction detected in patients, led to a focus on the neuropathology and altered development of the cerebellum in the disease.

The peculiar cholesterol metabolism in the brain accounts for the high vulnerability of this structure. Since circulating cholesterol does not cross the BBB, brain cholesterol is synthesized in situ [17–19]. Both neurons and glia synthesize cholesterol, although the expression of low levels of lanosterol-converting enzymes [20–22] makes neuronal cholesterol synthesis less efficient. Thus, it is the astrocyte synthesized cholesterol provided to neurons by the LDL-receptor endocytic pathway that mainly supports neuronal differentiation and functional maturation [23–25].

The biochemical defect involves a failure of endocytosed cholesterol movement between intracellular organelles [26–28]. Indeed, the Niemann-Pick C1 protein is a membrane protein with 13 transmembrane domains and a sterolsensing domain [14, 29]. A second protein, NPC2, whose lack causes an almost identical disease, is also a cholesterol binding protein [30–32]. A series of beautiful experiments, which include X-ray crystal structures, have elucidated the respective roles of NPC2 and NPC1 in handling cholesterol [31, 33–40]. Cholesterol is first bound by NPC2 [34, 38], which transfers it to an N-terminal soluble loop of NPC1 [35, 38]. This transfer involves direct contact between two opposing binding sites and a reversal of surface/interior orientations of cholesterol with the binding [36, 37]. X-ray structural analyses have revealed the presence of a “direct cholesterol transfer tunnel,” which mediates cholesterol movement from the luminal side of the lysosome to the cytoplasmic side [40, 41]. This notion is strengthened by recent computations and structural experiments predicting cholesterol transfer through a path involving a NPC1 cavity [41–45].

We have long wondered why the cerebellum is the first structure to display neurodegeneration in mice and humans. Among the various brain areas, cerebellum development goes largely beyond birth both in mice and humans [46–50]. Within the first three postnatal weeks, various cerebellar cells proliferate, migrate, and differentiate to achieve full maturation and circuit integration in mice [49, 50]; in humans, these processes take several years [49, 51]. Such an extended time length prompts us to favor the possibility that defective intracellular cholesterol mobilization alters cerebellar developmental trajectories, by reducing the “accessible metabolic cholesterol pool” that is needed to interact with and activate factors controlling signaling pathways [52, 53].

We also envision that the disruption of the autophagic flux reduces the intense cell makeover occurring within the long journey cerebellar cells, both neurons and glial cells, undertake to complete their functional maturation [54–56]. Consistent with this line of thought, brain organoids from neural stem cells of NPC patients-derived fibroblasts showed significantly reduced proliferative capacity and impaired neuronal differentiation, associated with intracellular accumulation of cholesterol [57].

We also question to what extent the much later development of cerebellar symptoms in humans (mid-first decade with the most common mutations compared to 18 days post-natal with a comparable mutation in mice) is due to the later development and/or maturation of the cerebellum. The full maturation of cerebellar cells takes several years of postnatal life in our species; Purkinje neurons (PN) markedly increase their size, determined as the area of their processes field, between birth and 2 years and between 7 and 9 years, achieving maturity by 18−20 years [49, 51, 58–60]. The complete morphological development of these cells results from a dynamic remodeling of their processes, in response to stimuli provided by internal and external milieux. Other GABAergic neurons, i.e. stellate and basket cells, are born post-natally from a neurogenic niche established in the perspective white matter [51, 61]. Granule cells (GCs) cease proliferating by 9 months [49, 58, 60, 62] and undergo a progressive clustering at the level of the granule layer, beneath the PN layer, which maximizes at 14−15 years of age [49, 51, 58]. While a thorough description of mouse and human cerebellum development is beyond the purpose of this review, a timeline of developmental events of mouse and human cerebellum is compared in Fig. 1.

Fig. 1.

An overview of mouse (A) and human (B) cerebellum development. The figure outlines key events landmarking the progressive acquisition of structural and functional complexity. Drawings in the upper part of the panels display gross anatomical changes of the developing cerebellum from embryonal stages up to adulthood. Since initial specification in the embryonic rhombic lip (RL: a germinal zone that is located in the fourth ventricle), both mouse and human cerebellum undergo significant size increase. The embryonic structures where neuronal progenitors originate and the various cell layers of the postnatal/adult cerebellar cortex are as follows: embryonic rhombic lip (embryo, RL) where granule cell (GC) progenitors originate, external granule layer (EGL) where dividing GC progenitors temporarily localize and internal granule layer (IGL) where postmitotic GC localize, are in red; Purkinje cell layer (PCL) is shown in green; ganglionic eminence (embryo, GE) and perspective white matter (PWM) where various GABAergic interneurons originate in the embryonic and postnatal/adult cerebellum, respectively, are in yellow. Note that the EGL is missing in the adult cerebellar cortex, since GC progenitors complete their development during the early post-natal life by stopping dividing and migrating inward to form the IGL. A scheme of the timeline of cerebellar developmental milestones is provided below drawings of panels A and B; key events of Golgi/stellate/basket cells (GABAergic lineage), GC and PC development from embryonic stages up to the adulthood are indicated. Teardrops and dashed lines indicate the timing and the duration of major developmental processes, including cell proliferation/differentiation at neurogenic niches and subsequent morphofunctional maturation, while the color intensity of the bars indicates the progression to full maturation. Note that the maturation of PCs is a long-lasting process, encompassing several weeks in the mouse and several years in our species. Similarly, the generation and integration of GABAergic interneurons into the cerebellar circuits are a protracted process. E: embryonic stage; PN: postnatal stage; CS: Carnegie stage; pcw: postconceptional weeks; y: postnatal years

Mice

There are currently at least 8 mouse models of Niemann-Pick C1 (and at least one of NPC2). There were 3 spontaneous mutations that have been studied to various degrees. Mary Lyon described a “foam cell reticulosis” phenotype of an autosomal recessive mutation in the GM strain of mice in 1965 [63]. It was classified as NPC due to the normal level of sphingomyelinase (deficient in NPCA and B) [64] but it was not readily available. A group at NIH studied and cloned the NPC mutation [15] found and thoroughly characterized at the US National Center for Toxicological Research [65] on the BALB/cNCTR strain. This later became designated as the Npc1^nih mutation and was widely shared by the NIH group. This was a much more tedious before the gene was cloned as it required breeding of unaffected pairs to see if affecteds were born. The spm NPC mutation was found in Japan [66] and sequenced at the Jackson laboratory [67].

One mutation was found from an ethylnitrosourea mutagen screen [67], several from embryonic stem cell knockouts and/or knock-ins [68] and, most recently, a CRISPR/Cas9-created mutation [69]. Of these, there are two frequently used mouse models of Niemann-Pick C1 disease. The first, whose discovery is described above, is due to a spontaneous insertional mutation in the gene resulting in a null mutation with no detectable RNA or protein, Npc1^nih [15]. This mutation creates a model mimicking the infantile form of the human disease [70], which is most marked by its gastrointestinal rather than its neurological symptoms [71]. An ethylnitrosoura-induced-point mutation in the gene created a mouse model of the most common, juvenile form of the disease, the Npc1 ^nmf164 mouse line [67]. Analyses of lifespan, body and spleen weight, gait, and other motor activities, as well as acoustic startle responses, all revealed a more slowly developing phenotype in Npc1 ^nmf164 mutant mice in comparison to mice with the null mutations (Npc1^nih) [67, 72]. Although Npc1 mRNA levels appear relatively normal, Npc1 ^nmf164 brain and liver display dramatic reductions in Npc1 protein, as well as abnormal cholesterol metabolism and altered glycolipid expression. Furthermore, histological analyses of the liver, spleen, hippocampus, cortex, and cerebellum reveal abnormal cholesterol accumulation, glial activation, and PC loss at a slower rate than in the Npc-1^nih mouse model. Magnetic resonance imaging studies also reveal significantly less demyelination/dysmyelination than in the null alleles [67]. Thus, Npc1 ^nmf164 mice offer many advantages as a model for the late-onset, more slowly progressing forms of NPC disease that comprise the large majority of human cases. More recently, the search for a suitable model to evaluate proteostatic therapies has been fulfilled by the Npc1^I1061T mouse model [68], harboring a mutation that is highly common in our species. The mutation produces a misfolded Npc1 protein that being targeted to ER-associated degradation has a reduced half-life. Finally, the recent CRISPR mutation, a 9 base-pair in-frame deletion of 3 amino acids near the C-terminus [69], was studied with regard to prenatal and neonatal lethality. Growth restriction was found at embryonic day 16.5 and newborn lethality due to early development of lung disease [73] normally a late finding and more severe in NPC2 [74].

Cerebellar Changes and Clinical Signs in Mice

Many early studies were performed only with the Npc1^nih mouse, which displayed a conspicuous atrophy of the cerebellum at 40 days of age [75, 76]. A first, histologically appreciable, sign of PC liability at this age was the presence of small concentric membranous bodies in their soma and dendrites, causing focal swelling of dendrites (megadendrites) in the molecular layer and of axons in the granular layer [75–77]. A fairly rapid PC demise was then witnessed by the marked loss of these cells throughout the cerebellum and by the worsening of dendrite and axon anomalies by the age of 60−80 days [75, 76], with particular reference to the presence of ectopic dendritogenesis, atrophic dendrites, and axonal spheroids in spared PC [75, 76, 78]. Inclusion bodies were detected in microglia and astrocytes [77] and pronounced myelin degeneration preceded axon atrophy [79]. The cerebellum of mice carrying the spm NPC mutation [66] displayed similar morphological changes, although PC loss was faster [76, 80]. Cytoplasmic concentric membranous lamellae were detected in PC at an early stage and thereafter in granule neurons, GABAergic interneurons, and various glial cells [80]. With a few species-specific differences, the appearance of neuronal anomalies follows a similar path in murine, feline, and cat models (reviewed in [81]).

The earliest grossly visible cerebellar motor sign in the Npc1^nih mouse model is continuous extensor tremor, which is most noticeable when holding the mouse by its tail. When continuously present for 6−7 s, it is a very highly accurate determination of “onset of symptoms” [82, 83]. It allowed earlier classification, read blind, for fine mapping [82] of the of the Npc1^nih locus to an interval bridged by a single yeast artificial chromosome [83]. This parameter is so highly reproducible that, again when read in a blinded manner, an 11 day shift, elicited by what were at that time suboptimal doses of hydroxypropyl-beta-cyclodextrin (HPBCD), was statistically significant to 1 part in a thousand probability of an error [84, 85]. However, much more subtle motor defects could be detected earlier, even in the juvenile model of the disease [67, 85, 86].

In the mouse, complex motor abilities requiring fine limb coordination, balance, and muscle strength are normally acquired by the end of the second postnatal (PN) week [85]. Three tasks (ascending a ladder, crossing a narrow bridge and suspension on a wire) allowed the differentiation of the contribution of motor coordination and balance from that of grip and muscle strength [86, 87]. Npc1^nmf164 pups acquired these abilities with a significant delay compared to wild-type littermates [85]. Indeed, whereas wild-type pups crossed the narrow bridge in its entire length and hanged onto the wire with four limbs at PN14, Npc1^nmf164 mice crossed the bridge only at PN17 and developed the four-limb hanging ability at PN18. In contrast, grip ability and muscle strength developed similarly in Npc1^nmf164 and wild-type littermates, as shown by their similar ability to ascend the ladder after PN15 and to hang on the wire for increased time with increasing age.

Using a 7 Tesla magnet (made possible by a much smaller bore), MRIs of mice can be obtained which, because of the high magnetic intensity, are nearly as highly resolved as are those of humans with the usual 2.5 Tesla machine (lower because of the much larger bore). Such MRIs allow the detection of anatomical changes in living, pre-weaning mice [88–90] T2-weighted images from the 23 day old Npc1^nih mutant mice did not appear to differ except in the corpus callosum (CC) region, which was slightly hyperintense with respect to the surrounding gray matter for the Npc1^nih mice [88]. A valuable mode of analysis of the MRIs for white matter changes is diffusion tensor imaging (DTI) which was carried out in the brains of these very young Npc1^nih mice [88]. The diffusion of water in white matter tracts of Npc1^nih mice at this young age was already abnormal, exhibiting decreased anisotropy, as quantified by fractional anisotropy (FA), compared with their wild-type littermates, the controls [88]. The reduction in FA was general in the white matter tracts, statistically significant in the corpus callosum and highly statistically significant in the internal capsule and the fimbria of the hippocampus [88]. The postmortem histological staining revealed myelin deficiencies in Npc1^nih mice, consistent with the reduction in FA measured in vivo [88]. Beneficial effects of treatment with HPBCD were also detectable at this age by FA, which correlated with increased myelination as seen by histology [88]. This was, at the time, the earliest detection of a therapeutic effect in Npc1 homozygous mice. Magnetic resonance spectroscopy in the Npc1^nih model only showed increased myo-insoitol and decreased taurine in posterior regions of the brain in late stages of the disease [89]. Thus, it was the white matter that was different by DTI and the cerebellum was not particularly remarkable. Similar analyses in the Npc1^nmf164 mice were also performed. T2-weighted MRI disclosed little decrease in myelination in mutant Npc1^nmf164 mice relative to the wild-type mice [67]. This was carefully quantified in four regions of interest (ROIs). The Npc1^nmf164 mutants were not statistically different from the wild-type controls for the T2 signal (which represents water replacing fat), whereas Npc1^nih mutant mice had shown a very significant increase in this signal in the corpus callosum (CC) and external capsule (EC) and internal capsule (IC) [67]. When careful volumetric studies were performed, the whole brain of the Npc1^nih mice decreased in volume from 3 to 9 weeks of age (− 6%) while the normal brain increased in size over this interval (+ 4%) [90]. Over this same time period, the cerebellum shrunk by 9% and the ventricles increased by 17% (compared to growth of the normal cerebellum over this time period by 9% and a growth of the ventricles in the normal mice of 22%).

Altered Cerebellar Development in Niemann-Pick C1 Model Mice

There is a huge cholesterol demand to build-up the enormous membrane surface of PCs and surrounding astrocytes [91]. Long axons and complex dendritic trees thereby have a decreased cholesterol supply from the very distant PC soma, whereas widespred astrocyte end-feet appear to secure a more efficient delivery of ApoE-associated cholesterol [92, 93]. Synaptic formation and remodeling require efficient cholesterol mobilization, due to the high cholesterol content of synaptic vesicles [93]. Climbing fibers and PC’s innervation provides one of the most striking event of synapse remodeling occurring in the postnatal cerebellum [94, 95]. At birth, the majority of PCs receive input from several climbing fibers (CFs), typically 5, whereas by PN7, there is only one CF innervating a PC soma. Thereafter, this single CF moves upward to form synapses with PC dendrites, while PC soma receives inputs from inhibitory basket cells [96–98]. Several studies have recently shown that the integration of neuronal cell types and synapse refinement are altered in the early postnatal developing cerebellum of Npc1 mice [85, 99, 100] and that PC degeneration is preceded by a sharp decrease of calcium-binding proteins [101], glutamate transporters [85, 102], perturbations of GABAergic inputs [85, 103]. Glial cells have been also found to display a number of morphofunctional alterations, including thicker and less ramified shafts of Bergmann glia [85], reduced differentiation of oligodendrocytes and myelin formation [104] and astrocyte and microglia activation [85, 100, 105–108]. The latter was found to precede neuronal degeneration [105].

The loss of PCs is in an anterior to posterior order. In general, the loss is preferentially in aldolase negative stripes, which occur in a parasagittal pattern [76, 109–112]. Because the striping is “zebra-like,” when first discovered, the monoclonal antibody detecting these stripes was called “antizebrin” and only later found to detect aldolase. Since aldolase is involved in both glycolysis and gluconeogenesis, these might be more energetically active cells. The concentration of aldolase-positive PCs progressively increases toward the posterior lobules, which are also frequently cells that express heat shock protein (Hsp) 25, which might protect cells from stress [113, 114]. Before death, the PCs have focal axonal swellings with cytoplasmic inclusions and frequent megadendrites. This pattern of loss of PCs, progressively from anterior to posterior, is one seen in numerous neurological disorders with cerebellar degeneration (e.g., spinocerebellar ataxia 6 and multiple system atrophy; reviewed in [76]). With mouse models of NPC1, the rate of progression does not correlate with survival time. The sphinomyelinosis (spm [66]), an intronic 45 bp insertion resulting in a frameshift and leading to a nonsense codon terminating the protein in loop 1, and the Npc1^nmf164/164 [67] homozygous Npc1 mutations have a much more rapid progression of loss of PCs yet live substantially longer than does the Npc1^nih homozygote [76]. Silver staining of degenerating neurons suggested that the PC death occurred in a retrograde manner starting from the dendrites [115]. A certain extent of regional differences of cerebellar defects could be secondary to other brain region damages. In particular, impairments in the thalamus, which is one of the first region displaying pathologic defects in the Npc1^−/− mice [116–118] or in the inferior olivary nuclei, whose climbing fibers autophagic impairments correlate with Purkinje neuronal death in a cat model of the NPC pathology [119].

The cerebellum may differ greatly from other parts of the brain in regards to the question of whether the defect is cell-autonomous. This might reflect differences between brain regions in the interchange of cholesterol-bearing lipoproteins between cells. Ko et al. used chimeras of wild-type and Npc1^nih homozygous embryos and found dying Bergmann glial cells adjacent to healthy PCs and dying PCs adjacent to healthy Bergmann glial cells [120]. Thus, they concluded that the intracellular cholesterol transport defect in NPC1 disease is cell-autonomous. The cell death was contributed to autophagy although mice with a deletion of the autophagy-controlling protein, Bcl2, do not have a change in survival [121]. Cell autonomy is clearly not true for most of the brain in NPC1. Homozygous deficient Npc-1^nih mice, which have the normal gene expressed only in glia using a Glial Fibrillary Acid Protein (GFAP)-promoted cDNA transgene, have greatly extended survival [122, 123]. The addition of neuron-specific expression of the normal cDNA using the neuron specific enolase (NSE) promoter further extended survival [124] but in both cases, cerebellar signs developed and persisted. The difference between this apparent cell autonomy in the cerebellum and the rest of the brain is not due to the lack of local synthesis of carrier lipoproteins or their receptors—PCs synthesize cholesterol and progesterone [125] and, also, are especially abundant in the ApoE2R lipoprotein receptor [126]. However, strong evidence for a mutualism exists. In cultures of PCs, lack of NPC1 in either PCs or glial cells did not affect PC excitability but the absence in both PCs and glia impaired presynaptic input to PCs “suggesting a cooperative effect of neuronal and glial NPC1 on synapses” [127]. Pfrieger’s laboratory’s classic work on neuronal cell dependence on co-cultures with glia had shown that the active factor was apolipoprotein-carried cholesterol [93] but that possibility has not been shown in these PC-glia cultures.

Indirect proof of subtle anomalies of developing trajectories in NPC1 disease is provided by the comparison of microglia transcriptome profile of pre-symptomatic (i.e. 3 weeks old) and symptomatic (i.e. 7 weeks old) Npc ^nih mice [128]. Although only a few transcriptomic changes were found in 3 weeks old Npc ^nih compared to wild-type mice, multiple changes at this stage were observed for microglia transcripts. In particular, Tmem119 expression was reduced, while Cd68, Igf1, and Gapdh markers were upregulated, depicting a pattern of microglia activation that is consistent with the increased microglia-Purkinje cell neuronal dendrites interaction in the cerebellar molecular layer during the first post-natal week [99]. Interestingly, changes of microglia transcriptome preceded changes of neuronal transcriptome, outlining the neuroprotective surveillance activity of microglia in the post-natal cerebellum, which is completely independent from the neurotoxic activity that it displays at overt stages of the disease [105, 107, 108, 128, 129].

An overview and a summary of developmental anomalies of Npc1 mice are provided in Fig. 2 and Table 1.

Fig. 2.

A scheme of cerebellar cortex organization in wild-type and PN14 Npc1 mice. During the first two weeks of postnatal life, the cerebellar cortex is organized into four layers: external granule layer (EGL), molecular layer (ML), Purkinje cell layer (PCL), and internal granule layer (IGL). Granule cells (GC, red) differentiate and migrate along Bergmann glia shafts (BG, gray), passing through the PCL layer (PC, green) to reach their final destination within the IGL. GCs project their axons (also named parallel fibers, thin red lines) into the ML and form excitatory synapses with PC dendrites and with GABAergic interneurons, i.e. stellate (SI, yellow) and basket (BI, yellow) cells. BI/SI establish inhibitory synapses with PC soma and dendrites. PCs receive excitatory inputs from climbing fibers (CF, black), whereas glomeruli (made of a complex array of intertwined nerve terminals and synaptic contacts, deep gray oval) receive excitatory inputs from mossy fibers (MF, black). Excitatory and inhibitory synapses are indicated by red triangles and yellow square, respectively. Astrocytes (AC, purple) and microglia (MG, blue) are also displayed. PC myelinated axons (ms, brown) carry the sole inhibitory output from the cerebellar cortex to deep cerebellar nuclei; from here cerebellar input reaches the motor cortex via the thalamus. Dashed lines highlight the presence of morphofunctional defects at cell-to-cell communication and overall circuitry. A summary of developmental anomalies displayed by Npc1 mice compared to wild-type mice is reported below drawings. Defects include the following: (1) decreased GC proliferation, leading to reduced size of cerebellar lobules/overall cerebellum; (2) poor morphological differentiation of BG; (3) a general imbalance of glutamatergic/GABAergic stimulation that PCs receive from CF/parallel fibers and BI/SI, respectively; (4) astrocyte and microglia activation; (5) reduction of glomerulus size and complexity; (6) defective oligodendrocyte differentiation, causing a reduction of myelin sheats

Table 1. Main Niemann-Pick type C disease-associated cerebellar defects in mouse models and human patients.

	Model	In text ref.
Mouse
Defect: clinical symptoms
Continuous extensor tremor	Npc1nih	[84, 85]
Delay of motor coordination, balance, grip, and muscle strength	Npclnmf164	[85]
Decreased anisotropy in white matter, myelin deficiencies	Npc1nih	[88]
Decreased myelination	Npclnmf164	[67]
6% volume decrease from 3 to 9 weeks of age	Npc1nih	[89]
Defect: cerebellum development
Synaptic defects	Npc1nih	[85, 99, 100]
Retrograde PC degeneration	NCTR-BALB/c	[115]
Alteration of microglial transcriptome	Npc1nih	[128]
Increased microglia-Purkinje cell dendrites interaction	Npc1nih	[99]
Decreased number of GCs and 20-25% reduction of cerebellar lobule size	Npc1nih	[152]
Reduction of Shh mRNA expression	Npc1nih	[152]
Decrease of GC proliferation	*Nsdhltml.1Hrm	[154]
Thicker and less ramified Bergmann glia shafts	Npc1nih	[85]
Reduction of GLAST, glutamine synthetase, VGLUT2, and GAD65	Npc1nih	[85]
Cerebellar dysmyelination and oligodendrocyte differentiation impairments	Npc1nih	[85]
Decreased fraction of ciliated cells, shortening of the primary cilium	Npc1nmf164	[160]
Reduced expression of Shh, Ptch, and Smo	Npc1nmf164	[160]
Other defects indirectly caused by cerebellum disfunctions
Thalamus impairments	Npc1nih	[116–118]
Climbing fibers autophagic impairments in olivary nuclei	Npd-/- cat	[119]
Presynaptic input to PC impairments	Npc1nih derived cells	[127]
TDP-43 mislocalization and hyperphosphorylation	Npc1nih	[176]
Humans
Defect (clinical symptoms)
Neonatal jaundice and enlargement of the liver and/or spleen	NPC patients	[4–5, 7]
Enlargement of the ventricles; atrophy of the cerebrum and cerebellum and reduction of the corpus callosum	NPC patients	[186]
Gray matter reductions in hippocampus, thalamus, superior cerebellum, insula, and cortex	NPC patients	[187]
Neuronal morphological changes	NPC patients	[78]
Defect (cerebellum development)
Decreased fraction of ciliated cells	NPC patients fibroblasts	[160, 161]
Reduced expression of Shh, Ptch, and Smo	NPC patients fibroblasts	[160]
Reduction of ciliated cells and a shortening of the primary cilium	NPC patients fibroblasts	[160]
Reduction of PTCH cytoplasmic levels	NPC patients fibroblasts	[161]
Other defects
Upregulation of miRNA 196 and 296 and the downregulation of miRNA 143	NPC patients fibroblasts	[174, 175]
TDP-43 mislocalization and hyperphosphorylation	NPC patients derived neuronal cells and brain sections	[176]
Severe pulmonary involvement	NPC patients	[12]

Abbreviations: NPC, Niemann-Pick type C; Npc1, NPC intracellular cholesterol transporter 1; NCTR, National Center for Toxicological Research; Nsdhl, Sterol-4-alpha-carboxylate 3-dehydrogenase, decarboxylating; GC, granule cell; Shh, Sonic hedgehog; BG, Bergmann glia; GLAST, glutamate/aspartate transporter; VGLUT2, vesicular glutamate transporter 2; GAD65, glutamate decarboxylase 65; Ptch/PTCH, patched; Smo, smoothened; TDP-43, TAR DNA-binding protein 43

Defective Shh Signaling Underscores Altered Cerebellum Development in Npc1 Mice

Clearly, there is not a global lack of cholesterol (Chol) for Sonic hedgehog (Shh) signaling which is important in the developing cerebellum—the phenotype of NPC1 diseased patients is quite different from that of Smith-Lemli-Opitz syndrome due to 7-dehydrocholesterol reductase deficiency where facial (anteverted nares), CNS (microcephaly and brain structural abnormalities, including hypoplasia of the cerebellum), limb (toe 2/3 syndactyly), and other birth defects indicate a more global lack of Shh signaling [130–132]. However, at least two lines of thinking indicate that Shh signaling can be responsible of altered cerebellar morphogenesis as is suggested by the cerebellar defects in Smith-Lemli-Opitz [133, 134].

First, active Shh is Chol-modified; the Shh precursor is processed by an enzymatic reaction performed by its C-terminus that generates a Chol-modified N-terminal fragment responsible for the signaling activity [135]. The palmitoylation of the N-terminus of this fragment [136, 137] generates a double lipidated protein [135] that is recognized by the sterol-sensing domain of Dispached (Disp), allowing Shh secretion [137]. Shh mutant proteins not suitable for cholesteroylation are rapidly degraded in the ER, leading to no active ligand secretion and determining congenital brain malformations [138, 139]. Recently, it has been found that the extracellular cysteine-rich-domain (CRD) of Smo binds Chol and that targeted mutations of key binding residues prevent Shh signaling [140]. In a search for Chol-modified proteins, only Smo was found [141, 142]. Chol modification of Smo is required for its translocation to the primary cilium plasma membrane [141, 142]. In particular, the binding of Shh to Ptch1 determines the relief of Ptch1 inhibitory activity on Smo, allowing the binding cleft of the CRD and TMD Smo domains to be accessed by Chol [142]. However, how Ptch1 inhibits Smo is still debated. Based on Ptch1 ability to bind and mediate the efflux of Chol from cells [143], which is in agreement with its structural homology to NPC1 [144], it has been proposed that Shh-mediated Ptch1 inactivation leads to an increase of the Chol pool required for Smo modification at the primary cilium [145]. In line with this, Chol loading synergizes with Shh to activate Smo [140, 146] and leads to Smo activation even in the absence of Shh ligand [146].

Both lipid modifications of Shh play a role in inhibiting Ptch [147]. X-ray crystallography studies of Ptch show that a dimer of Ptch interacts with Shh with a tunnel in the Ptch protein appropriate for cholesterol passage and into which the palmitoyl-modified N-terminus of the N-terminal fragment of Shh fits to block transport [148, 149]. Additional studies by cryo-electronmicroscopy provide evidence for a tetrameric structure of the two dimers and that the C-terminal cholesterol of Shh fits into an extracellular sterol binding pocket on Ptch [147].

Second, Shh activity goes far beyond the classic morphogen activity, regulating stem and progenitor cell differentiation and integration within circuits, synapse formation and communication between neurons and glial cells [150]. In the post-natal cerebellum, PCs maintain a bi-directional— upward and downward—signaling that is fundamental for the generation of a balanced number of excitatory and inhibitory interneurons. The downward Shh signaling acting on the transient germinal compartment of the perspective white matter (PWM) controls the generation of post-natally born GABAergic interneurons and astrocytes [151].

This line of thinking suggests that the dysregulation of Chol handling in Npc1-deficient neurons decreases the metabolic active pool of Chol needed for covalent modification of Shh and its downstream secondary target, Smo, thus affecting a number of developmental trajectories within the developing cerebellum [135, 140, 141]. Due to premature exit from the cell cycle of proliferating GC precursors, there is a decreased number of GCs and a 20−25% reduction in cerebellar lobule size at the end of cerebellar development in Npc1^nih mice [152]. This causes a decrease of GCs in the granular layer beneath the PN cell layer that impinges on cerebellar glomerulus size and synaptic connectivity. In line with the robust mitogenic activity Shh exerts on GCs [153], the content of Shh transcripts was found significantly reduced since PN8 up to the time time of final GC precursors divisions [152]. A similar finding was reported in a mutant mouse line harboring a mutation in the cholesterol synthesis pathway. Without Chol, there was a significant decrease in GC proliferation [154], as it is expected in NPC1 deficiency where cholesterol availability for C-terminal attachment to Shh is decreased. Chol modification of Shh is crucial for generating temporal and spatial gradients of this factor [132, 155, 156]. Starting from mouse embryonal day 17.5 (E 17.5) Shh is continuously secreted by PCs and diffuses upward and downward to the so-called prospective white matter (PWM, a secondary germinal zone of the postnatal cerebellum). Such bi-directional Shh signaling drives the generation and differentiation of both excitatory (GCs) and inhibitory interneurons (basket and stellate) respectively [151]. In response to Shh, a niche of progenitors residing in the PWM also generates astrocytes [157]. Therefore, virtually any cerebellar cell type relies on Shh signaling. Shh shapes BG differentiation in relationship to migration, dendritogenesis, synaptogenesis, and maturation of PCs [158, 159], suggesting that Npc1-deficiency can also affect normal pattern of BG differentiation. Accordingly, BG processes of PN15 Npc1-deficient mice were found to display thicker radial shafts and a less elaborate reticular pattern compared to wild-type age-matched littermates. This feature rebounds on the glutamate clearance efficiency, which appears affected by the significant reduction of glutamate transporter (GLAST) and Glutamine synthetase expression levels [85]. Meanwhile, VGLUT2 and GAD65 expression level reductions pinpoint an overall deranged glutamatergic and GABAergic stimulation that PCs receive by climbing/parallel fibers and basket/stellate, respectively. Lastly, oligodendrocyte differentiation is not spared by Npc1-deficiency as indicated by the severe cerebellar dysmyelination, which was detected as early as at PN11, appearing very pronounced at the level of PC axons [85].

One study, by exploiting the Npc1^nmf164 mouse line and NPC1 patient fibroblasts, showed that the expression level of Shh, Ptch, and Smo was significantly reduced in the second postnatal week [160]. Both Ptch and Smo appeared targeted to degradation since by electron microscopy, they were found to localize in multivesicular bodies [160]. The fraction of ciliated cells was found reduced either in the mouse model or in patient fibroblasts. In addition, primary cilia were shorter and stubby compared to controls [160]. The second study, by exploiting NPC1 patient fibroblasts, showed that cytoplasmic levels of PTCH were significantly reduced and consistently with the above mentioned study reported a reduction of ciliated cells and a shortening of the primary cilium of NPC1 fibroblasts compared to controls [161].

Novel Pathways Under Scrutiny: NPC1 Couples LDL-Derived Cholesterol to mTORC1 Activation

Recently, the finding that the lysosomal transmembrane SLC38A9 protein activates mTORC1 upon cholesterol interaction [162] has outlined how proper trafficking of LDL-derived cholesterol within late endosomes/lysosomes is crucial for signaling pathways controlling cell-metabolism. SLC38A9 belongs to the family of solute carrier group protein and was first identified by its ability to activate the mTORC1 pathway depending on lysosomal aminoacid content, particularly arginine and leucine [163, 164]. The presence of a CARC/CRAC consensus motif [165], which is typically harbored by cholesterol-regulated proteins, suggested that SLC38A9 activity might be influenced by lysosomal cholesterol content. Actually, this issue was elegantly investigated by Zoncu’s group, which provided evidence that NPC1 interacts with SLC38A9 and inhibits cholesterol-mediated mTORC1 activation by cholesterol export outside the lysosomal lumen [162]. Thus, SLC38A9 translates the increase of LDL-derived cholesterol to mTORC1 pathway activation, whereas NPC1 acts in the opposite direction by decreasing the lysosomal cholesterol content.

By regulating the lysosomal cholesterol pool, NPC1 finely tunes metabolic activities controlled by mTORC1, which in the postnatal developing cerebellum include cell proliferation, migration, differentiation, and synaptic remodeling [166]. Hence, the disruption of NPC1 function is expected to upregulate mTORC1 pathway and to cause cerebellar alterations [162, 167, 168]. Solid evidence of this is provided by the recent finding that the selective activation of mTORC1 in mouse PCs leads to defective differentiation of these cells and also alters the cerebellar synapse connections [169].

Humans

Up to now, 486 mutations of NPC1 gene have been identified, out of which 256 consist of nonsense or frameshift mutations, determining the premature termination of NPC1 translation [2, 3, 7, 170]. Because a quality control mechanism, known as non-sense mediated decay (NMD), determines the degradation of the mutated transcripts, truncated NPC1 proteins are usually not produced [171, 172]. In particular, NMD-dependent degradation of NPC1 mRNA has been observed for nine frameshift/non sense mutations and one deletion [171, 173], as both mRNA and protein products of NPC1 were undetectable in patient samples. Studies of post-transcriptional regulatory mechanisms have also provided interesting hints. MicroRNA profiling of NPC patient fibroblasts has uncovered the regulation of a large number of miRNA, including the upregulation of miRNA 196 and 296 and the downregulation of miRNA 143, which are strictly involved in lipid homeostasis [174, 175]. Evidence of mechanisms regulating RNA metabolism contribute to the disease is provided by the mislocalization and hyperphosphorylation of TDP-43 in Npc1^nih PCs and brain sections of NPC patients [176] provide additional evidence that RNA metabolism dysregulation could contribute to cerebellum development alteration.

The clinical disorder, briefly described in the introduction, can be caused by mutations in either of the 2 genes encoding the 2 proteins involved in the egress of cholesterol from the lysosome [33, 37]. The patients with NPC2 mutations cannot generally be distinguished by clinical criteria, although severe pulmonary involvement is a typical complication in these patients, which only represent 5% of the total [13]. This much lower frequency of this mutation may reflect the major role of this protein in seminal fluid [177]. As mentioned, mutations in the NPC2 gene have been identified as being in the previously recognized HE1 gene, encoding an abundant seminal plasma protein [30]. An early adult form of NPC1 is abundant in Nova Scotia due to a founder effect and was distinctive enough to be labeled as “Niemann-Pick D” disease [178]. In recent years, many cases of adult onset, frequently presenting to psychiatrists, have been found [179].

Clinical Symptoms and Signs in Humans

The earliest manifestations of NPC1 disease in children are usually of cerebellar origin although neonatal jaundice and enlargement of the liver and/or spleen were common in the newborn period in retrospect [4, 5, 7, 180]. Clumsiness and learning problems were frequent early complaints with frank ataxia and paralysis of upward gaze leading to the diagnosis with an average age of 10.4 years in the group of 87 patients surveyed, although half were diagnosed before the age of 6.9 years [7]. Clinical scales, heavily dependent on cerebellar signs, have been developed to monitor the progress of the disease and the potential of therapies to slow it [4, 181, 182]. Eight of the 25 criteria used to train students to assess NPC disease patient charts were cerebellar [183]. Of a scoring system used extensively, for instance with the intrathecal delivery of HPBCD [181], 5 of the 9 parameters (ambulation, fine motor skills, swallowing, eye movement, and speech) are primarily determined by cerebellar function [184, 185].

Magnetic resonance imaging (MRI) results from a 15-year-old patient were compared to those of 3 relatively well age-matched controls [186]. Compared with the control subjects, the NPC patient manifested enlargement of the ventricles [186]; atrophy of the cerebrum and cerebellum, especially the vermis [186]; and reduction in the size of the corpus callosum [186]. Diffusion tensor imaging (see above) was also performed showing major changes in white tracts but, as in mice, this technique was of little value for the shrunken cerebellum in NPC1 patients [186].

T1-wighted MRI studies of 6 adult patients compared to 18 age-matched controls were more revealing of cerebellar, and other changes [187]. The patients demonstrated bilateral gray matter reductions in large clusters in bilateral hippocampus, thalamus, superior cerebellum, and insula and in smaller regions of inferoposterior cortex. As found previously (above), patients demonstrated widespread reductions in fractional anisotropy in major white matter tracts [187]. Analyses of measures of axial and radial diffusivity suggest that these changes were contributed to by both impaired myelination and altered axonal structure [187].

There have been multiple PET (positon emission tomography) and SPECT (single-photon emission computed tomography) scans searching for specific molecular defects in NPC1 patients. PET scans for neurofibrillary tangles (NFTs) and amyloid in 8 patients [188], which outlines some convergent pathological hallmarks of NPC and Alzheimer’s diseases [189]. Cerebral, but not cerebellar NFTs were found in 4 of the 8 patients; the amount was correlated with the duration of symptoms but not the degree of cognitive decline. They did not detect significant amyloid plaques. Of note, NFTs had been only detected in humans but not in NPC1 models in other species [77]. Their formation is not related to aging, as indicated by the fact that they were observed in a 4-year-old NPC brain [190]. Dopamine transporter imaging by SPECT was performed in singleton adult cases showing pronounced bilateral nigrostriatal degeneration, but again no cerebellar abnormalities [191, 192].

While data on cellular pathology occurring at the onset and in the course of the disease are very limited in our species, neuronal morphological changes, including ectopic dendrites with supernumerary spines and axon hillock enlargement and axonal spheroids formation [78], indicate that normal developmental trajectories are deranged. Although species differences had been reported in the extent of ectopic spines and meganeurites [193, 194], ectopic dendritogenesis and axon swellings are typical of large glutamatergic pyramidal neurons and GABAergic interneurons, respectively. Anomalies at the axon hillock are expected to perturb normal synaptic integration and the subsequent firing properties of GABAergic interneurons, with serious consequences on the modulatory activity, by means of inhibition, exerted by these neurons on various neuronal populations. As this inhibition is fundamental for sculpting neuronal network activity, it is expected to have profound effects on brain functions. Ectopic dendritic outgrowth and subsequent increase of spines are more pronounced in humans than in rodents and have been linked to ganglioside, particularly GM2, overexpression [195].

The Interplay Between Altered Cerebellar Development and Clinical Manifestation of NPC Disease

The timeline of cerebellum development identifies critical periods that, in addition to cerebellum vulnerability, may cause the derangement of developmental processes and skill acquisition involving other brain areas [196–200].

The gross anatomy of the cerebellum underscores the existence of a closed-loop circuit between the cerebellar cortex and the cerebral cortex via deep nuclei [197, 200]. Four different deep nuclei embedded in the white matter receive projections from anterior (I−V), posterior (VI−IX), and flocculonodular (X) lobules of the cerebellar cortex [196, 197]. Projections from deep nuclei reach the cerebral cortex via thalamus and spinal cord via brainstem nuclei [196, 197]. This array of anatomical connections posits the basis of a functional topography whereby the various cerebellum lobules process information routing from different regions of the cerebral cortex and spinal cord [198]. As a consequence, sensorimotor regions of the cerebellum—mostly anterior lobules—mature earlier compared to posterior ones consistently with the timing of developmental milestones of the cerebellar cortex. In this perspective, the graded anterior-to-posterior degeneration of PCs in NPC disease [76] appears somehow related to the timing of their functional maturation.

In addition, the functional interconnection between the cerebellum and the cerebral cortex via the thalamus is consistent with anomalies in thalamic, cortical, hippocampal, and basal ganglia regions observed in NPC patients [187, 199, 201], as well as with the significant neuronal/glial pathology of thalamocortical path at early stages of the disease that had been reported in Npc1^nih mice [116, 117].

Conclusions

In answer to the questions we raised, our best estimates are as follows: it does seem that the cerebellum is more dependent on the availability of intracellular cholesterol than other parts of the brain such that NPC1/2 deficiency affects signaling pathways fundamental for neural progenitor proliferation, differentiation, and functional maturation. While roughly completed at the birth in many brain areas, the accomplishment of these developmental processes takes up the first three post-natal weeks in the mouse cerebellum. The evidence that NPC1 deficiency affects signaling pathways is fairly new and underscores a role of cholesterol as signaling molecule that is fairly new, as well. To date, it is the accessible metabolic cholesterol required for the generation of cholesterol-modified Shh and Smo active forms or controlling cell-metabolism through the TORC1 pathway that appears particularly significant for cerebellum development. To what extent this is also true in humans needs further study.

Secondly, the general progression of anterior to posterior PC loss in the cerebellum may reflect higher energy metabolism and increased levels of HSP25 protecting the more posterior cells. In addition, it may be associated with the functional topography of the various cerebellar lobules, which communicate with different regions of the cerebral cortex via the thalamus.

Finally, the much slower progression of maturation of the human cerebellum, which does not reach final maturation until the middle of the second decade, may explain the delayed arrival of cerebellar symptoms compared to the very early onset in mice.

Abbreviations

ApoE

Apolipoprotein E

BBB

Brain blood barrier

CARC/CRAC

Cholesterol recognition aminoacid sequence

CC

Corpus callosum

CF

Climbing fiber

Chol

Cholesterol

CNS

Central nervous system

CRD

Cysteine-rich-domain

Disp

Dispached

DTI

Diffusion tensor imaging

EC

External capsule

ER

Endoplasmic reticulum

FA

Fractional anisotropy

GAD

Glutamic acid decarboxylase

GC

Granule cell

GLAST

Glutamate transporter

GM

Ganglioside

HPBCD

Hydroxypropyl-beta-cyclodextrin

HSP

Heat shock protein

IC

Internal capsule

LDL

Low density lipoprotein

MRI

Magnetic resonance imaging

mTORC

Mechanistic target of rapamycin

NFT

Neurofibrillary tangles

NMD

Non-sense mediated decay

NPC

Niemann-Pick C

NSE

Neuron specific enolase

PC

Purkinje cell

PET

Positron emission tomography

PN

Postnatal

PTCH

Patched

PWM

Perspective white matter

ROI

Region of interest

Shh

Sonic hedgehog

Smo

Smootened

SPECT

Single-photon emission computed tomography

TDP-43

TAR DNA binding protein 43

TMD

Transmembrane domain

VGLUT2

Vesicular glutamate transporter