The authors of this review are enthusiastic about the approach of using iPSC-based cell models for disease modeling and drug discovery, especially for the neurovisceral disease NPC1. However, we are also aware of the limitations of using iPSC-based model systems and therefore want to address at least some of these issues.
5.1. Limitation of the Applicability of iPSCs and Derived Cells
Since the ground-breaking publications by Yamanaka and colleagues in 2006, describing the reprogramming of mouse fibroblasts into induced pluripotent stem cells [
132], and of human fibroblasts in 2007 [
133,
134], a range of methods combining the use of different donor somatic cells, transcription factors, gene delivery methods and small molecules have been developed to generate disease specific iPSC lines. To date, the most popular donor somatic cells are fibroblasts, being used in more than 80% of all reprogramming experiments published [
135]. Additionally, other cell types have been used for reprogramming, such as human primary keratinocytes, peripheral blood mononuclear cells, and urine-derived renal epithelial cells [
136,
137,
138]. In general, somatic cells are discussed to have an “epigenetic fingerprint”, reflected by a pattern of chemical markers on their DNA, impacting the reprogramming and, later, the differentiation potential of the cells. Furthermore, iPSCs represent cells in an embryonic developmental stage, and, thus, one must keep in mind some drawbacks related, e.g., to the maturation of iPSC-derived neurons. This is of particular interest for neurons because the expression of ion channel subunits changes during embryonic/fetal development. For example, the subunit composition of inhibitory GABA
A receptors (GABAAR) and glycine receptors (GlyR) changes during the maturation of neurons, and thus the functional properties, such as the activation/inactivation and ion conductance of the ion channels, are altered. Moreover, GABAAR and GlyR are excitatory rather than inhibitory at early developmental stages. This is due to increased intracellular Cl
−-concentration achieved by Cl
−-transporters expressed during embryonic development [
139]. Our group observed GABA-induced depolarization in the iPSC-derived neurons used by Rabenstein and colleagues [
20], which led to the activation of voltage-gated Ca
2+ channels (unpublished data), indicating a fetal developmental stage of the neurons even after six weeks of differentiation. Therefore, whenever possible, functional changes in iPSC-derived neurons should be matched with other model systems, such as primary neuronal cultures, that preferably have a developmental stage relevant to the disease under investigation.
Regarding the developmental stage of iPSCs, several approaches have been made to overcome this issue by using stress inducers and proteins pushing forward the ageing of cells [
140]. Trans-differentiation of somatic cells, e.g., fibroblasts, into disease-affected cells, such as neurons, offer the possibility to bypass a state of pluripotency, which is linked to the reset of the age of the cells. This aspect is particularly important for diseases with progressive neurodegeneration. Trans-differentiated cells show no tumorigenicity when transplanted in vivo, but show similar functionality [
141,
142]. Still, there are limitations in the concept of trans-differentiation, as the obtained population of post mitotic cells, such as neurons, is limited, as such cells do not divide anymore, resulting in a restricted efficiency and expansion for use in high throughput drug screening [
142].
The choice of the transfection system to obtain iPSCs is without any doubt the most critical step determining the further usability of the iPSCs. While retrovirus was the method initially used for reprogramming, they only infect dividing cells, whereas lentivirus can infect both non-dividing and proliferative cells [
143]. The major risk of using retrovirus and lentivirus for reprogramming is the integration of viral vectors into host DNA, and it goes without saying that such iPSCs cannot be used for applications in humans. Non-integrative viral delivery includes adenovirus [
144] and Sendai virus [
145]. Alternatively, plasmids [
146,
147], RNA delivery [
148] and protein delivery [
149] may be used. In addition, several chemical compounds have been shown to further improve reprogramming [
150].
After the crucial first step of generating iPSC lines carrying the molecular defects of interest, efficient cell-specific differentiation protocols, in the case of NPC1 with neurovisceral clinical presentation, are required to obtain neuronal and hepatic cells. The use of adequate protocols is challenging, as an almost unmanageable number of differentiation protocols have been published. In terms of NPC1 disease modeling, most neuronal cell systems have been created without specification for differentiation into a particular type of neuron. None of these publications (
Table 1) have yet successfully recapitulated, e.g., the specific cerebellar neuronal dysfunction and degeneration. Cerebellar Purkinje cells are particularly susceptible to mutations of the NPC1 protein, and their progressive loss is one of the hallmarks of neurodegeneration observed in NPC1. Due to their complex morphology and circuitry, unique firing properties, and long maturation time, it is certainly challenging to develop appropriate protocols for in vitro differentiation. However, a significant breakthrough has been achieved recently. Muguruma and colleagues [
151] described the generation of Purkinje cell precursors from mouse embryonic stem cells, which could be differentiated into neurons showing Purkinje cell-specific characteristics. Silva and colleagues [
152] described the maturation of human iPSC-derived cerebellar neurons. The authors developed protocols to generate cerebellar cell types such as Purkinje cells or granular cells, which have been shown to be electrically active cells. It is expected that such model systems will help to decipher the processes underlying the high vulnerability of human Purkinje cerebellar cells, and it is of great interest to what extent human iPSC-derived cerebellar neurons can reflect the in vivo situation. Although model systems based on organoids or 3D-culture systems have great potential to reproduce the in vivo situation in a more complex way than 2D-culture systems, these models also have disadvantages. They can only represent the interaction between different cell types to a certain extent, and the interactions between different organs, affected by a disease, cannot be studied. This is of particular importance, for example, for pharmacokinetics and the distribution of a drug in the organism, especially in neurodegenerative diseases, where potential drugs must be able to cross the blood-brain barrier.
Regarding the differentiation of iPSCs into HLCs, the majority of available protocols are based on three steps: (1) induction of definitive endoderm, (2) specification into hepatic progenitor cells and (3) maturation into HLCs [
153,
154]. Publications using iPSC-based model systems for NPC1 mostly use one of these strategies, but there are significant differences between the protocols. This has to be considered critically, as it may lead to differences in the appearance of pathophysiological features and their attenuation by potential substances for the treatment of NPC1.
A further important aspect of using iPSCs is the availability of appropriate control cells. Differences in genetic background between iPSC lines can hamper the interpretation of experimental results. For genetic disorders, genetic editing is a powerful tool to generate gene-corrected iPSCs for disease modeling because it produces cells with identical epigenetic backgrounds. Technologies such as CRISPR/Cas9 can correct or induce a specific mutation, thus leading to an isogenic control that only differs in the underlying mutation compared to its control cell line. In NPC1, it is even more difficult to select an appropriate control, because the pathophysiology of NPC1 is extremely diverse. Selecting mutations to study is particularly difficult in NPC1 because there is no really strong correlation between mutations and phenotypic expression in this disease. Patients with the same mutation, even identical twins affected by NPC1 [
155], show strong differences in clinical expression. This example shows that epigenetic aspects can only be poorly represented in cell culture models, and it is questionable whether they play a role at all in iPSC-based models, since the cells undergo an epigenetic reset during reprogramming.
The limitations of iPSCs and their cellular derivatives mentioned above are general and apply to disease modeling and drug discovery for all types of diseases. The wide variety of protocols to generate iPSCs and differentiation into desired cell types make it difficult to compare studies that have used different protocols. On the other hand, iPSC-derived cells used for disease modeling of NPC1 show comparable pathophysiological manifestations or pathogenic mechanisms of NPC1, regardless of the protocols used. Thus, it can be concluded that the observed disease features are due to the NPC1 mutations and not to the protocol used for reprogramming or differentiation, demonstrating the applicability of iPSC-derived cells for disease modeling.
5.2. Perspectives of the Applicability of iPSC-Derived Cells
As summarized above, several laboratories showed that iPSC models are a valuable tool for NPC1 disease modeling, revealing pathophysiological features and, moreover, elucidating the pathogenic mechanisms of NPC1.
Overall, in the studies reviewed here, all major hallmarks of NPC1 have been observed. With one exception, iPSC-based NPC1 disease models showed an increased cholesterol level—the main characteristic feature of NPC1. The same holds true for increased levels of LC3BII and p62, indicating disturbances of autophagy in NPC1. Autophagy is closely related to mitochondrial functions, as the degradation of damaged mitochondria by mitophagy is a major process of the autophagic system in the cell. A proper clearance of damaged mitochondria is essential to maintain cellular homeostasis, especially as different pathophysiological phenotypes, linked to mitochondria, have been observed in NPC1. To cut a long story short, the proof of major pathophysiological features of iPSC-based model systems for NPC1 demonstrates the applicability of such systems. However, the number of studies involving neurons or glial cells is currently low, although these cell types are, besides hepatocytes, the most affected cells in NPC1.
NPC1, like other neurodegenerative diseases, is characterized by the selective loss of neurons. In the case of NPC1, cerebellar Purkinje cells are primarily affected. Of course, there are technical hurdles to obtain such cells from iPSCs in vitro, but it is surprising that hardly any research directly addresses the loss of neurons, at least for NPC1. For example, it is currently poorly understood which pathogenic mechanisms underlie the degeneration of Purkinje cells. In addition to apoptosis as the main cause of cell death, necroptosis might have a major influence on the loss of neuronal cells in NPC1. Cougnoux and colleagues [
38] evaluated this possibility in NPC1-deficient fibroblasts, iPSC-derived neuronal precursor cells, and in the cerebellum of NPC1-deficient mice. They found a decreased cell viability and an increase in the receptor interacting protein kinase 1 and 3 (RIP1, RIP3) in all cell types analyzed. Both kinases are components of a protein complex, the so called necrosome, which is known to initiate necroptosis. The inhibition of RIP1 and 3 with Necrostatin-1 led to an increase in the cell viability of fibroblasts and prevented the loss of Purkinje cells in NPC1-deficient mice, prolonging their lifespan. These results suggest necroptosis as a potential target for pharmacological intervention. Sung and colleagues [
49] did not analyze cell death directly, but described the defective neuronal differentiation of cells generated by the direct conversion of fibroblasts. The extent to which this may contribute to the neuropathology of NPC1 needs further clarification. However, one can be optimistic that further work will make use of iPSC models and will further contribute to the elucidation of the pathophysiological aspects of NPC1.
Studies of the pathogenic mechanisms of neurodegenerative diseases using iPSC-based models have revealed similarities in the pathology of various neurodegenerative disorders. In terms of NPC1, with an incidence of 1/120,000 live births, such similarities are described with Alzheimer’s disease, with a prevalence of 30 million people worldwide. In this regard, NPC1 and AD certainly represent an unequal pair for comparative studies, but in terms of pathogenic mechanisms there are some remarkable similarities, such as abnormal cholesterol metabolism, amyloid-β (Aβ) and tau pathology [
156]. Interestingly, neurofibrillary tangles, the predominant pathophysiological feature of AD, can be observed in NPC1, and patients also show an accumulation of Aβ peptides. However, no amyloid plaques were observed, which could be explained by the fact that the vast majority of NPC1 patients die before the manifestation of amyloid plaques. It is discussed that cholesterol influences the processing of amyloid precursor protein (APP), leading to amyloid plaques by altering the endocytotic trade of APP due to a change in lipid composition in the membrane. Conversely, Alzheimer’s disease shows features of disturbed cholesterol metabolism with lipid deposits, which are the hallmark of NPC1. Interestingly, a variant of ApoE, which is the main carrier of cholesterol in the brain (apolipoprotein type 4 allele, ApoE-ε4), is one of the main risk factors for late-onset AD.
The examples of NPC1 and AD impressively demonstrate the potential of iPSC-based disease models. It is likely that without iPSC technology it would be much more difficult to compare and elucidate pathogenic mechanisms in disease affected cell types. Certainly, comparative studies with post-mortem preparations are possible in this regard, but no insights into early phases, without clear signs of neurodegenerative processes, can be obtained from these, and, of course, longitudinal studies are not possible. It remains extremely exciting to see how future studies using iPSC-based model systems will contribute to the understanding of neurodegenerative diseases, and one may be optimistic that these findings will advance the development of new therapeutic approaches for NPC1 and neurodegenerative diseases in general.