Animal models for Niemann-Pick type C: implications for drug discovery & development

ABSTRACT

Introduction: Niemann-Pick type C (NPC) is a neurovisceral, progressively detrimental lysosomal storage disease with very limited therapeutic options and no approved treatment available in the US. Despite its rarity, NPC has seen increased drug developmental efforts over the past decade, culminating in the completion of two potential registration trials in 2018.

Areas covered: This review highlights the many available animal models that have been developed in the field and briefly covers classical and new cell technologies. This review provides a high-level evaluation and prioritization of the various models with regard to efficient and clinically translatable drug development, and briefly discusses the relevant developments and opportunities pertaining to this.

Expert opinion: With a number of in vitro and in vivo models available, and with having several drugs, all with various mechanisms of action, either approved or in late stage development, the NPC field is in an exciting time. One of the challenges for researchers and developers will be the ability to make use of the lessons learnt from existing late-stage programs as well as the incorporation not only of the opportunities but also the limitations of the many models into successful drug discovery and translational development programs.

KEYWORDS:

• Animal model
• cholesterol
• drug development
• glycosphingolipids
• lysosomal storage diseases
• lysosomes
• Niemann-Pick type C
• NPC
• NPC1
• NPC2
• sphingolipids

1. Introduction

1.1. Niemann-Pick type C disease

Niemann-Pick type C (NPC) is a neurovisceral, progressively detrimental lysosomal storage disease (LSD) caused by autosomal recessive loss-of-function mutations in either the NPC1 (95% of patients) or the NPC2 (5% of patients) genes [1]. The incidence is about 1:100,000 live births, although it might be more common in adults than previously recognized [2]. NPC is clinically and genetically heterogenous with approximately 400 NPC1 and 23 NPC2 pathogenic mutations described to date, the majority being missense mutations, but insertions, deletions and duplications have also been described [3]. The exact function of NPC1 is still not completely understood, but it appears that NPC1 and NPC2 function in the efflux of various lipid species including sphingolipids and cholesterol from late endosomes and lysosomes. Patients with mutations in either NPC1 or NPC2 are clinically indistinguishable [1,4].

The clinical presentation of NPC is highly heterogeneous, but the signs and symptoms can be broadly divided into three main groups: neurological, visceral and psychiatric, which also partially allow for separation into age groups. Visceral disease is most prominent in infantile cases and usually involves hepatomegaly and splenomegaly. Moreover, cholestatic jaundice in neonates or young infants can be an early sign of NPC [3]. Hallmark clinical neurological manifestations include dysphagia, dysarthria, cerebellar ataxia, dystonia and ocular-motor abnormalities such as vertical supranuclear gaze palsy. The clinical symptoms of most importance for quality of life was recently prioritized by patients, their caregivers and clinical experts and reflected the importance of the neurological manifestations of the disease as ambulation, speech, swallowing, cognition and fine motor skills where universally agreed upon as the most important manifestations of the disease for patients [5]. Patients with NPC, particularly adolescent/adult-onset NPC patients, can present with a range of psychiatric symptoms including cognitive decline, schizophrenia-like psychosis, and mood disorders. The age of onset of neurological symptoms varies greatly amongst NPC patients, which is reflected in the heterogenous rate of disease progression and overall life expectancy. Almost all NPC patients die prematurely, and the most common reported cause of death is aspiration pneumonia following impaired swallowing [1,6].

Mutations in NPC1 constitute the majority of all cases of NPC with most of these encoding missense alleles, the isoleucine to threonine 1061 substitution (I1061T) being the most prevalent in NPC patients in western Europe and the US, accounting for approximately 20% of all mutations [1]. Studies of this and other missense alleles suggest that the primary reason for NPC1 loss of function is due to reduced conformational stability of the protein as residual NPC1 reaching the lysosome retains some functionality. Indeed, both the wild type and I1061T NPC1 proteins closely associate with several components of the cellular protein quality control system, including molecular chaperones of the Heat Shock Protein 70 (HSP70) family, and a majority of mutated NPC1 proteins appear amenable to refolding and augmentation of residual function [7–11].

1.2. NPC biomarkers

While the primary cause of NPC is the loss of NPC1 (or NPC2) protein function due to missense mutations, the downstream effects of this loss are a severely compromised late endosomal and lysosomal system, characterized by intra-lysosomal accumulation of a number of lipid species, most notably sphingolipid- and cholesterol derivatives, many of which have been proposed as potential biomarkers of the disease [3,12,13].

Biomarkers are increasingly important for diagnosis and drug development, not least in rare diseases such as NPC where biomarkers, besides improved diagnosis, may provide important information about target engagement and biological impact on disease-relevant molecular events in support of clinical efficacy. The latter is of great importance in rare diseases as it may be challenging to establish clinical efficacy through standard clinical trials given the scarcity of patients and their heterogeneity in clinical presentation [3,12].

For NPC, analysis of sphingolipids and cholesterol homeostatic responses have provided a number of proposed diagnostic plasma biomarkers such as cholesterol oxidation products (oxysterols), bile acids, (glyco)sphingolipids (GSLs) and their metabolic derivatives such as Lyso-SM509 [14–20]. Other biomarkers such as Bis(monoacyl)glycerophosphate (BMP), triglycerides, Calbindin, Chitotriosidase, Cathepsin D, Fatty Acid Binding Protein 3 (FABP3), Galectin 3, Heat shock protein 70 (HSP70), high/low density lipoproteins (HDL/LDL), P-Tau and amyloid-beta have been explored in various matrices such as blood, peripheral blood mononuclear cells, various organs and tissues, urine and cerebrospinal fluid (CSF) [21–29].

Morphological changes in brain substance have also been proposed as a potential surrogate measure of disease and may be monitored by imaging techniques such as MRI and PET [6,30]. While intriguing, it is important to note that none of the above-mentioned biomarkers’ utility for disease progression monitoring and treatment effect has been validated in prospectively defined clinical trials and there is currently no single biomarker specific for NPC diagnosis [3].

1.3. Human NPC clinical trials

Miglustat (a glycosylceramide synthase inhibitor) is the only therapy approved for treatment of NPC in EU and other countries outside of the US [31], and there is, therefore, an urgent need for the development of additional therapies. A number of various experimental approaches have been, or are being, explored in small, open-label clinical studies [5,32,33](Clinicaltrials.gov ID: NCT00975689, NCT02124083, NCT03471143, NCT02912793). The argument for such exploratory studies is mainly that they might give a quick indication of whether one should pursue properly powered and controlled clinical trials. While this might seem attractive it is critical to remember that investigators should have a clear understanding of what part of the disease mechanism is targeted, how this is best measured and whether the biological rationale is scientifically sound. Furthermore, as NPC is a heterogeneous disease, any conclusions based on small, uncontrolled, open-label studies should be considered very carefully. While small exploratory trials may be warranted under certain circumstances, the above-mentioned shortcomings clearly stress why properly conducted and controlled clinical studies are needed in order to advance our understanding of NPC and its potential intervention points. In line with this, two potential registration-enabling clinical trials have recently been completed (Clinicaltrial.gov identifiers: NCT02534844, NCT02612129), studying the intrathecally administered cholesterol scavenger, VTS-270 (Hydroxypropyl-beta-cyclodextrin) and the CNS-penetrant, orally available, Heat Shock Protein amplifier, arimoclomol. The VTS-270 trial has reported that the product did not show a statistically significant effect over a sham procedure control; a result that was ascribed to a lack of disease progression in both arms of the trial [34] (Clinicaltrials.org: NCT02534844). The arimoclomol trial has reported positive results, demonstrating a clear reduction of the disease progression as measured by the primary endpoint, a 5-domain abbreviated NPC clinical severity scale focused on the disease symptoms of most importance to patients and their caregivers. Additionally, the arimoclomol clinical trial also showed a biological response to arimoclomol on key characteristics of its mechanism of action and the disease biology of NPC, including significant increases of the rescue protein Heat Shock Protein 70 and significant reduction of accumulation of unesterified cholesterol in blood cells [35] (Clinicaltrials.org: NCT02612129).

With a number of novel animal models being developed and two potential registration-enabling clinical trials having recently been completed, there is a marked interest in how these models have been and may be used for drug discovery and, in particular, translational development for NPC. This review seeks to provide an overview of the available models and a critical evaluation of their suitability as translational models for NPC.

2. Animal models of NPC

The NPC1 gene is highly conserved among eukaryotes, making it possible to generate NPC models ranging from mammals to fungi, including cat, mouse, zebrafish, fruit fly, nematode and yeast models of NPC [36] (Table 1). In addition to the aforementioned model organisms, neuronal cells differentiated from stem cells or other sources of human or murine origin have successfully been used as orthogonal model systems for drug screening and testing [37]. This review will primarily focus on animal models of the disease, with a short glimpse into the ex vivo human cell systems.

Table 1. Feline and murine NPC models.

Species	Name	Affected Gene	Mutation (cDNA position)	Mutation outcome	Strain of origin	References
Feline	NPC1	Npc1	c.2864G>C	p.Cys955Ser (C955S)	N.A.	[38,39]
Feline	NPC2	Npc2	c.82 + 5G>A	p.Gly28_Ser29ins35 (G28_S29ins35) (retention of 105 bp in the mature mRNA leading to an in-frame insertion of 35 amino acids between residues 28 and 29 of NPC2 protein	N.A.	[45]
Murine	Npc1^nih (Npc1^m1n)	Npc1	824 bp of MaLR retrotransposon-like DNA replacing 703 bp of wild-type genomic sequence.	Frameshift + premature stop codon (null allele)	BALB/c	[46–48]
Murine	Npc1^spm	Npc1	c.2911 + 3A>C	p.Val971GlyfsTer31 (V971GfsTer31) (cDNA retains 43 base pairs of intron 19 shifting the reading frame, replacing p.Val 971 for a Gly, adding 30 novel residues followed by a stop codon)	C57BL/Ks	[49,72]
Murine	Npc1^pf/pf	Npc1	c.604C>G; c.607-608TT>GC	p.Pro202Ala;Phe203Ala (P202A/F203A)	C57BL/6	[76]
Murine	Npc1^nmf164	Npc1	c.3163A>G	p.Asp1005Gly (D1005G)	C57BL/6	[49]
Murine	Npc1^tm(I1064T)Dso (Npc1^I1061T)	Npc1	c.3179T>C+ loxP sites flanking exons 14–20	p.Ile1064Thr (I1064T)	C57BL/6	[70]
Murine	Npc1^imagine	Npc1	c.1554-1009G>A	Inclusion of 194-bp pseudoexon in intron 9 resulting in a premature stop codon	C57BL/6	[83]
Murine	Npc1^pioneer	Npc1	c.1920delG	p.His641ThrfsTer2 (H641TfsTer2) (frameshift replacing p.His641 with Thr and resulting in a premature stop codon two residues downstream)	C57BL/6	[83]
Murine	Npc2^tm1Plob	Npc2	Insertion into intron 2 resulting in splice junction mutation	Hypomorhic allele (decrease in correctly spliced Npc2 mRNA)	BALB/c (or mixed)	[73]
Murine	Npc1 ASO	Npc1	N/A	Antisense-mediated knockdown of hepatic Npc1	BALB/c	[79]

2.1. Feline NPC models

2.1.1. NPC1 cat

A feline NPC1 model was established from a litter of domestic kittens where some offspring showed progressive neurological disease. The initial diagnosis of NPC was based on biochemical findings of increased levels of hepatic cholesterol, glucosylceramide (GlcCer), galactosylceramide (GalCer) and sphingomyelin [38]. Thirteen years later, the underlying NPC1 mutation in the cat colony was identified as a point mutation in the feline NPC1 gene giving rise to a NPC1 protein with a C955S conversion [39] (Table 1).

The disease in NPC1 cats resembles the juvenile form of human NPC by replicating several clinical features including hepatomegaly, pulmonary complications and central nervous system (CNS) disease including ataxia [38,40,41]. The neurologic dysfunction renders the NPC1 cats unable to walk or stand without assistance by around 19 weeks of age [41].

NPC1 cats have profound visceral disease involving liver, lung, and spleen as well as increased serum levels of liver enzymes (e.g. ALT and AST) and chitotriosidase [41,42]. Peripheral nerve involvement has also been demonstrated [43]. The feline CNS disease is similar to human NPC and is characterized by lysosomal storage, gliosis, meganeurite formation, ectopic dendritogenesis, neuroaxonal dystrophy and loss of cerebellar Purkinje cells. In contrast to human disease, neurofibrillary tangles have not been observed in NPC1 cats (reviewed in [44]). The NPC1 feline model replicates the biochemical hallmarks of NPC such as accumulation of free cholesterol, sphingosine, GlcCer, Lactosylceramide (LacCer), and the gangliosides GM2 and GM3 in various target organs [42] and has been used in the identification of potential biomarkers including plasma cholestane-3β,5α,6β-triol (C-triol), 7-Ketocholesterol (7-KC) [20], GlcCer [18], and CSF calbindin [23].

2.1.2. NPC2 cat

A spontaneous feline model of NPC2 deficiency has also been described. The cats were homozygous for an intronic mutation giving rise to an in-frame insertion of 35 amino acids between residues 28 and 29 in the NPC2 protein (Table 1). The phenotype of the NPC2 cats was comparable to the juvenile form of the NPC disease in humans including loss of cerebellar Purkinje cells, gliosis, demyelination, swollen hepatocytes and foamy macrophages in the lungs. Moreover, ex vivo fibroblasts were positive for filipin staining of free cholesterol [45].

2.2. Murine NPC models

2.2.1. NPC mouse models with null alleles

2.2.1.1. Npc1^nih (Npc1^m1n)

The standard model in the field is the Npc1^nih (also named Npc1^m1n) mouse model which arose as a result of insertional mutagenesis in the BALB/c strain [46] resulting in a Npc1 null allele (Table 1) [47]. The loss of Npc1 results in an early-onset, rapid progression of NPC disease resembling late infantile disease in humans [44]. The disease in these mice is characterized by progressive weight loss, loss of motor coordination, visible tremor, ataxia, loss of ambulation, hindlimb paralysis, poor feeding and terminal disease before 90 days of age [48–50]. Npc1^nih mice also display an array of motor impairment phenotypes from 4 to 6 weeks of age that can be studied with behavioral assays such as gait analyses, SHIRPA, rotarod, balance beam, coat hanger, and open field locomotor assays. These mice also show cognitive impairments, as measured by learning and recall assays [25,51–53].

The Npc1^nih mouse model shows progressive neuropathology including lysosomal storage, axonal swelling, demyelination, and gliosis, whereas meganurite formation and ectopic dendritogenesis are limited [44,53]. Moreover, profound neurodegeneration results in an orderly patterned loss of cerebellar Purkinje cells, CNS demyelination, and a more general global brain atrophy that can be quantified using high-resolution volumetric MRI [54–56]. Cardinal CNS-related pathological features of NPC include loss of white matter and cerebellar atrophy [30,57–60]. The Npc1^nih model has been used to study the demyelination processes and establish a clear function for NPC1 in neurons and glia in the maintenance and formation of CNS myelin, the loss of which precedes cerebellar atrophy and the loss of Purkinje neurons [61–63]. Histopathologically, the model shows some differences compared to the human NPC disease as neurofibrillary tangles (tau aggregates) and β-amyloid plaques are not readily observed in murine Npc1-deficient neurons, whereas this is the case in human mutant NPC1 neurons [44].

Biochemical NPC hallmarks such as accumulation of free cholesterol, GlcCer, LacCer, sphingomyelin, sphingosine and gangliosides (GM2 and GM3) in various target organs are also recapitulated in the Npc1 null mice (reviewed in [64]). Fan and colleagues used sphingolipid profiling of Npc1^nih brains and plasma from human NPC patients to identify sphingolipid plasma markers that show similarly abnormal levels between mice and NPC patients [18]. Likewise, several oxysterols including C-triol were elevated in plasma, liver, and brain of Npc1^nih mice similar to what was seen in plasma from NPC patients [20].

Concomitant with lipid accumulation, murine NPC1 models also show inflammatory changes in liver and brain [65,66]. These include changes in the expression of innate immunity genes such as Lysozyme 1, and the activity of lysozyme has been suggested as a plasma biomarker of NPC disease [67].

It is important to pay close attention to the genetic background of the experimental animals as the lifespan of the Npc1^nih mouse is influenced by housing conditions, genetic drift of the mouse colony and strain background [68,69]. A more severe lipid storage in the livers of Npc1^nih mice on the C57BL/6 background compared to the BALB/c Npc1^nih has also been reported [70]. While the Npc1^nih model is the standard model for NPC animal studies, a recent report emphasizes the need to carefully evaluate data generated in light of the background strain used for the breeding of the mice as the same drug can apparently have opposite effects in different genetic backgrounds of the Npc1^nih mouse [71].

2.2.1.2. Npc1^spm

The Npc1^spm mouse model arose spontaneously on the C57BL/Ks background, and was early on recognized as having a ‘sphingomyelinosis’, hence the name spm [72]. Several years later, the underlying genetic cause was identified as a point mutation in Npc1 leading to a null allele (Table 1) [49]. The disease in the Npc1^spm and Npc1^nih models are highly similar [56,72].

2.2.1.3. Npc2^tm1Plob

The Npc2^tm1Plob mouse model contains a splice junction mutation in Npc2 resulting in a ≈ 96% reduction in NPC2 protein levels. The phenotype of the NPC2 hypomorph mouse model is highly similar to the Npc1^nih mouse and includes a similar pattern of cerebellar Purkinje cell loss, as well as storage of cholesterol and glycolipids such as GM2 and GM3 [73]. Despite these similarities, the NPC2 hypomorph mice have slightly delayed onset and slower disease progression than the Npc1^nih mouse model [74].

2.2.2. NPC mouse models with point mutation alleles

Even though the Npc1-null mice models mimic the progression of human disease, very few NPC patients are homozygous for truncating mutations, and most of the pathogenic alleles are point mutations in NPC1 (or NPC2) leading to misfolded NPC proteins [9,10,75].

2.2.2.1. Npc1^pf/pf

The Npc1^pf/pf mouse model is the first described NPC mouse model with Npc1 missense mutations. This mouse model has a double mutation (P202A/F203A) in the cholesterol binding pocket of NPC1 (Table 1), thus creating a loss-of-function mutant. Npc1^pf/pf mice develop NPC disease similar to Npc1^nih mice, despite a high expression level of a correctly localized mutant NPC1 protein [76].

2.2.2.2. Npc1^I1061t (Npc1^{tm(i1064t)Dso})

The most common human NPC1 mutation gives rise to the I1061T missense substitution with an estimated frequency of approximately 20% of all pathogenic NPC alleles [75]. The Npc1^I1061T mouse model (also named Npc1^{tm(I1064T)Dso}) was engineered to harbor the I1061T mutation, as well as conditional deletion of exons 14–20 for optional generation of a Npc1 null allele [70] (Table 1). In humans, the I1061T missense mutation leads to a misfolded NPC1 protein, which is subject to premature degradation via endoplasmic reticulum-associated degradation (ERAD) in fibroblasts and thus is rescuable by approaches that increase the levels of ER and cytosolic chaperones such as calreticulin and HSP70 [7,8]. This is recapitulated in the Npc1^I1061T mouse, which displays a misfolded NPC1 protein that is prematurely degraded [70].

Phenotypically, the I1061T homozygous animals develop a NPC1-characteristic disease with decreased lifespan (mean survival is 125 days), age-dependent weight loss, visible tremor and progressive decline in motor coordination, albeit the disease is less severe compared to the BALB/c Npc1^nih mouse strain [70]. Intriguingly, C57BL/6 Npc1^I1061T mice almost completely lost their phenotype when backcrossed into the BALB/c strain emphasizing that genetic background is an important parameter to include in the evaluation of the model [70].

Other hallmarks of human NPC1 disease recapitulated in the Npc1^I1061T mouse are the progressive loss of cerebellar Purkinje cells, cerebellar storage of cholesterol (only in specific neurons, not in bulk tissue) and glial activation. Sphingolipid accumulation in bulk brain tissue was detected for LacCer, sulfatide, sphingosine, and GM3. These sphingolipids, amongst others, were also found to be elevated in liver tissue, together with free cholesterol and C-triol accumulation. Liver disease was also evident by the presence of lipid-laden foamy macrophages, astrogliosis and increased expression of inflammatory genes [70].

2.2.2.3. Npc1^nmf164

The Npc1^nmf164 mouse model was generated by ethyl-nitrosourea mutagenesis by the Jackson Laboratory [49]. The Npc1^nmf164 mouse has a point mutation generating a D1005G mutation in the NPC1 protein (Table 1). This mutation has not been described in human NPC patients, but the location in the large cysteine-rich luminal loop of the NPC1 protein makes it relevant to human disease as this is one of the mutational hotspots also accommodating the two most frequent pathological NPC1 mutations encoding the P1007A and I1061T missense mutations [1,75]. The D1005G mutation results in a misfolded and highly unstable mutant NPC1 protein that is prematurely degraded via ERAD [49,77].

Npc1^nmf164 presents with a similar phenotype but an overall less severe form of NPC than what is observed in Npc1 null mice, correlating with the partial loss of functional NPC1 protein [77]. The average lifespan is between 112 and 125 days in the C57BL/6 and BALB/c strains, respectively [49,78]. In addition to the progressive weight loss, Npc1^nmf164 mice display the characteristic lipid accumulation, hepatomegaly with accompanying decreased liver function and signs of injury, splenomegaly, brain involvement with cerebellar Purkinje cell loss, axonal abnormalities and increased astrocyte and microglial activation [49,77]. However, Npc1^nmf164 mice lack a profound demyelination phenotype (non-detectable by MRI), thus making this model unfit for analysis of this characteristic disease-associated parameter [49]. Alam and colleagues have developed a disease severity scale equivalent to major human NPC1 disease domains that can be used to assess the effect of therapy in this mouse model [78].

The main advantage of the Npc1^nmf164 model over the null mouse models is that the disease resembles human juvenile NPC that comprises the majority of human cases. The presence of residual NPC1 protein also allows for testing therapies that enhance the stability of the mutant NPC1 protein. Moreover, the slower phenotypic progression facilitates longer time for intervention and allows for homozygous breeding pairs. However, the low percentage of homozygous mutant offspring obtained from heterozygous breeding pairs is a clear limitation of this mouse model [49].

2.2.3. Other murine NPC models

Antisense oligonucleotide knockdown approaches have been used in vivo to knock down hepatic expression of Npc1 in wild type mice creating a hepatic phenotype similar to the Npc1^nih model including hepatomegaly, accumulation of lipids and foamy macrophages, cell death and elevated serum ALT activity [79,80]. This approach allows for a fast generation of model animals and has been used in the study of various therapeutic concepts aimed at the liver pathology of NPC [80–82]

Recently, two novel NPC mouse models with pseudoexon-generating Npc1 mutations, Npc1^{imagine/imagine} and Npc1^{imagine/pioneer}, have been described (Table 1) [83]. These mice do not produce functional NPC1 protein and they recapitulate the main hallmarks of NPC.

2.3. Other NPC animal models with drug discovery relevance

NPC models of the common laboratory model organisms such as the zebrafish Danio rerio, the fruit fly Drosophila melanogaster, the nematode Caenorhabditis elegans, and the yeast Saccharomyces cerevisiae have also been developed and used to gain insight into NPC cellular pathways. One advantage of these models in respect to drug discovery is that genetic manipulation is relatively easier, thus allowing for modifier screens. Furthermore, such models might prove suitable for developing higher throughput of early drug discovery and development platforms.

2.3.1. Zebrafish (Danio rerio)

A zebrafish model with npc1 deficiency obtained via morpholino oligonucleotides has shown that npc1 depletion results in early developmental defects and accumulation of free cholesterol in zebrafish [84,85]. Recently, two genetic npc1-null zebrafish models have been developed [86]. The npc1^hg37 zebrafish have an intracellular accumulation of free cholesterol in the developing embryo and in the liver of the larvae. In addition, disorganization of cerebellar Purkinje cells was found in older animals, which also showed late-onset neurological symptoms and decreased growth and lifespan. One limitation of the npc1-null zebrafish model is that the homozygous animals are infertile, but this may be solved by using LysoTracker red staining of the npc1-null larvae as selection criteria. Moreover, zebrafish larvae contain neuromasts, a sensory organ, that are in direct contact with the surrounding media. The npc1-deficient neuromasts are brightly stained with filipin and LysoTracker red, which offers the possibility of testing compounds in vivo that do not cross the blood-brain barrier. However, analogous to the npc1-null mice models, these mutant zebrafish models are completely devoid of npc1 protein and there is still a great need for zebrafish models with npc1 missense mutations.

2.3.2. Fruit fly (Drosophila melanogaster)

The fruit fly genome contains two npc1 genes, npc1a and npc1b, as well as eight orthologs of the mammalian npc2 family. Insects and other arthropods require the dietary intake of cholesterol that is converted to steroid hormones that are essential for development (reviewed in [87]). A series of genetic mutations in the fruit fly has revealed that npc1a and several npc2 members are essential for steroid hormone biosynthesis via regulation of cholesterol uptake and/or trafficking [88–90].

2.3.3. Roundworm (Caenorhabditis elegans)

C. elegans has two homologs of the human NPC1 gene [91]. Double mutants display a phenotype with constitutive dauer formation, which is an alternative developmental stage whereby the larva enters a type of stasis and can survive harsh conditions. This phenotype can be functionally rescued by human NPC1 protein [92], but the model has to our knowledge not yet been explored for drug discovery or development.

2.4. Human cellular systems

In addition to the whole-organism models of NPC, patient-derived cell lines have proven to be an invaluable tool for studies of molecular disease mechanisms and effects of potential therapies at a (sub)cellular level. The most widely used cell lines are fibroblasts established from skin biopsies which naturally facilitate the study of disease-associated NPC1 or NPC2 mutations and the effects of targeting various cellular disease mechanisms such as NPC1 protein refolding as well as organellar systems involved in the disease including mitochondria and the lysosomal compartment and their interconnecting systems [93,94].

Whereas skin fibroblasts are a suitable model system for investigating the fundamental aspects of human NPC disease, they may not be representative for the specific cellular changes observed in NPC1-mutant neurons, a key target cell population [37]. To this end, recent advances in human stem cell technology have paved the way for the generation of human neuronal models of NPC by differentiation of stem cells from different souces of origin. Efforts of NPC1 knockdown in human embryonic stem cells and transcription factor-mediated reprogramming of patient-derived fibroblasts to induced pluripotent stem cells (iPSCs) followed by neuronal differentiation have created human neuronal cell cultures with typical morphology, lineage-specific marker expression and electrophysiological activity [95–97]. Direct reprogramming of skin-derived fibroblasts from NPC patients to induced neural stem cells (iNSCs) has also been reported [98]. The iNSCs display increased cholesterol accumulation and thus hold potential as an alternative platform for drug screening. Multipotent adult stem cells (MASCs) derived from early passage fibroblast cultures or directly from skin biopsies of NPC patients can also be neuronally differentiated without prior reprogramming, providing yet another option for the generation of neuronal-like cellular model systems [99].

Another advantage of using NPC patient-derived iPSCs, iNSCs or MASCs is that these cells can be differentiated into other disease-affected cell types such as astrocytes, oligodendrocytes and hepatic-like cells [96,98]. Moreover, as one of the pathological manifestations of NPC is cerebellar Purkinje cell neurodegeneration, future advantages in protocols for generating this particular neuronal subtype from human iPSCs could be of great value for increasing the understanding of this particular aspect of the disease biology, and potentially a useful tool for drug development [37,100].

While patient-derived fibroblasts and iPSCs are excellent models for disease-associated NPC1 mutations, most of these cell lines do not permit studies of individual NPC1 mutations due to compound heterozygosity. This limitation can be bypassed by current genome editing techniques that offer the opportunity of generating iPSC cell lines with specific disease-associated NPC1 (or NPC2) mutations and isogenic control lines which can be used for mutation-targeted drug discovery in human neurons [37]. A mutation-specific drug screening platform has also been obtained by the engineering of the human U2OS osteosarcoma cell line to supress endogenous NPC1 expression and subsequent transient transfection of constructs encoding mutant NPC1 protein in order to test the refolding capacity of specific mutations [9].

3. Conclusion

NPC is an extraordinarily complex disease and no single animal model can appropriately cover all aspects of the disease pathology as it presents clinically. It is therefore important to be familiar with and critically evaluate the available models in order to select the most appropriate for the individual study’s objectives. For drug discovery, development and translation, it is important to focus on the disease mechanism one seeks to address, the potential to establish pharmacodynamic response biomarkers as well as which clinical symptoms are most important to patients and if they or the underlying disease mechanism can be evaluated in a given model.

With regard to the latter, the five most important clinical symptoms affecting patients’ quality of life are ambulation, speech, swallowing, cognition, and fine motor skills as prioritized by patients, their caregivers and NPC clinical experts [5]. While in vitro models and simpler organisms hold merit for drug discovery and some types of proof-of-concept studies, the capacity of a given model to accurately replicate these clinical symptoms in a quantifiable fashion should be considered of particular importance when deciding upon what model(s) to use for translationally relevant studies.

4. Expert opinion

Miglustat is currently the only approved treatment for NPC, although not approved in the US [6]. There is therefore still an urgent need for the development of additional therapies for this devastating disease and a marked interest in understanding which and how animal models have been and may be used for drug discovery and development for NPC.

While a number of animal models now exist for NPC (reviewed herein), only a few have been used as part of translational drug development programs with the Npc1^nih mouse and the NPC1 cat being the models used in preclinical studies of miglustat and the recently completed potential registration-enabling clinical trials of VTS-270 (cyclodextrin) and arimoclomol (Clinicaltrial.gov identifiers: NCT02612129, NCT02534844) [25,101–103].

Many of the model systems reviewed here, ranging from cell models to zebrafish, could be relevant for drug discovery purposes pending the given situation and the mechanism of action of the drug class under consideration. Also, some of the animal models have particular merits such as testing of chaperoning strategies, which will be discussed later. However, the best characterized and most readily available animal model for translational studies remains the Npc1^nih mouse, both of which are important practical considerations when designing a translational program.

The Npc1^nih model displays many of the clinical, pathological and biochemical hallmarks of NPC and, in particular, replicates the clinically important symptoms of ambulation challenges and loss of motor coordination while also presenting with demyelination and a biomarker profile reflecting many of the biological changes also observed in human patients [25,30,104–106].

While the Npc1^nih mouse model will continue to be of central importance for NPC drug discovery and development, this model is essentially a null model. Thus, it is useful for testing various therapeutic concepts such as replacement strategies (e.g. gene therapy) [107–109], metabolite storage reduction (scavenging, GSL synthesis inhibition (e.g. cyclodextrins and miglustat)) [101,102], lysosomal augmentation (HSP-based therapies, such as recombinant HSP70 and arimoclomol) [25,110], anti-inflammatory [52] and combination therapies [111]. However, it does not replicate one of the fundamental aspects of NPC biology, the misfolding and loss of function of NPC1 protein as a consequence of missense mutations, which constitute the majority of mutations in NPC.

To this end, other models such as the Npc1^nmf164 and Npc1^I1061T mouse models have been developed that could be feasible to test various chaperoning and proteostasis targeted strategies including proteasome inhibition and modulation of cellular molecular chaperones (e.g. HSP70) [8,10,11,25,49,70,112]. In this regard, in vitro models including patient cells and their derivatives might present superior systems to properly address chaperoning strategies as the mutations addressed here are of immediate clinical relevance and the systems allow for a much higher rate of analysis of various genotypes. The potential of this approach was recently demonstrated by the approval by EMA and FDA of migalastat for Fabry disease patients >16 years who have an amenable mutation. The amenability of a given mutation is assessed by an in vitro assay measuring alpha-Galactosidase A activity which forms the basis for classifying the mutation as amenable or not, and hence whether the therapy is indicated [113–116].

As such, in vitro studies can provide valuable information, not only during the early stages of drug discovery and screening but also late stage processes of a drug’s life cycle. The example of migalastat illustrates how important it is to consider what studies to perform to inform the development of a drug candidate, from its earliest screening to the design and analysis of clinical studies [117].

With the increased development efforts in rare diseases, not least NPC, it becomes increasingly important to critically address what parts of the mechanism of disease one seeks to address, as well as the mechanism of action of the drug in development, and how this is expected to ultimately provide a clinically meaningful outcome. These considerations should be carefully addressed in the drug discovery and development programs and, to this end, some learnings can be derived from the currently approved or late-stage clinical programs in NPC. As described above, the Npc1^nih model has seen use across all of these programs and while the use of the model for miglustat studies was limited prior to the approval of miglustat, the positive data from the trial of the HSP-amplifier arimoclomol suggest that the Npc1^nih model might soon be validated on some of its clinical and biochemical endpoints from a translational perspective.

As part of the arimoclomol development program, the Npc1^nih mouse was used to test augmentation of HSP70 levels and its impact on lysosomal function, which was supported by extensive prior mechanistic insight into this mechanism of action [25,110]. The studies focused on this mechanism of action and its impact on well-studied and disease mechanism-relevant pathological hallmarks of the disease, such as cerebellar demyelination as well as more readily obtainable suggested biomarkers of the disease. Phenotypically, the emphasis was placed on objectively quantifiable, clinically translatable endpoints closely or directly related to an addressable disease mechanism and known to be important to patients, such as ambulation and motor coordination [5,25].

This approach has provided a remarkably rapid and successful development of a potential drug candidate for NPC from a basic biological discovery to a successful, completed potential registration trial in just 8 years [25,110] (www.orphazyme.com). However, no single translational program will prove the solution for future drug development for NPC. Fortunately, a number of animals and cellular models have been developed for NPC, and while none will be ideal across various development programs and their individual stages, the NPC field is in the positive situation that a number of models exist and that some of these models’ translational value might soon be validated. This will make future drug discovery and development efforts, if not more simple, at least with one path already trodden – a situation not all rare diseases are fortunate enough to be in.

Hopefully, this situation will not only lead to new drug discovery and translational strategies, but solicitous use of these models and design of studies could also provide a rationale for selection of patient populations for more effective development programs and clinical trials to ultimately yield more efficacious treatment regimes.

Article highlights

• With no drugs approved for NPC in the USA and only one drug approved outside the USA there is a very large unmet need for disease-modifying therapies for Niemann-Pick type C.

• A number of in vitro and in vivo models have been developed over the past years, providing a wide arena to discover and develop new potential therapies.

• The historical ‘standard’ models will still be of great importance.

• Two potential registration trials have recently been completed for Niemann-Pick type C.

• The NPC field is in an advantageous situation: New drug development programs can utilize the many various models, and the knowledge gained from the recently completed Phase II/III clinical trials and their preceding drug development programs to inform new drug development programs and clinical trials which will hopefully lead to even more treatments being developed for NPC.

This box summarizes key points contained in the article.

Declaration of interest

Both authors are employees of Orphazyme A/S. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewer Disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

The authors are supported by Orphazyme A/S.