Induced Pluripotent Stem Cells: A Tool for Modelling Parkinson’s Disease

Abstract

Parkinson’s disease (PD) is the second most prevalent neurodegenerative disorder. Among its pathologies, progressive loss of dopaminergic neurons in the substantia nigra is characteristic and contributes to many of the most severe symptoms of PD. Recent advances in induced pluripotent stem cell (iPSC) technology have made it possible to generate patient-derived dopaminergic neuronal cell culture and organoid models of PD. These models have contributed to understanding disease mechanisms and the identification of novel targets and therapeutic candidates. Still needed are better ways to model the age-related aspects of PD, as well as a deeper understanding of the interactions among disease-modifying genes and between genetic and environmental contributions to PD’s etiology and progression.

Parkinson’s disease and its cellular models

Parkinson’s disease (PD) is the second most common age-related neurodegenerative disorder, characterized primarily by progressive loss of motor and non-motor functions as a result of the loss of dopaminergic (DA) neurons from the substantia nigra. About 10% of PD cases are familial, and both autosomal recessive and autosomal dominant causal PD genes are known. The remainder of cases are sporadic and idiopathic, but mutations in a number of genes, including some with causal alleles, can confer increased risk for the sporadic form of PD. Aggregation of the product of the α-synuclein gene is observed in most, but not all, familial PD cases and in nearly all sporadic cases, and is widely assumed to be a critical cytopathological event.

Animal models are an important tool for studying human disease, but have various critical limitations, particularly when modeling age-related diseases of the human brain. With the advent of iPSC technology, where skin cells (for instance) can be dedifferentiated into pluripotent stem cells, and then differentiated into specific cell types (for example dopaminergic neurons), it has become possible to derive cellular models of neurons from live PD patients. Figure 1 provides a schematic representation of using iPSCs for modelling PD. One complication is that, the act of turning adult cells into pluripotent stem cells resets the chronological age of these cells back to an almost fetal state, making their relevance to idiopathic adult-onset disease uncertain. To some extent, this has been compensated for by selecting genetic forms of PD, on the reasonable grounds that such patients have been living lifelong with the effects of the mutation, and therefore if nothing else, cell culture models with such mutations allow one to study the earliest disease manifestations.

Figure 1. Schematic representation of the workflow for using iPSC’s for modelling Parkinson’s disease.

Cells from PD patients are isolated by biopsy, which are then reprogrammed to iPSC. These can be differentiated into dopaminergic neurons which can be used for phenotyping assays, drug screening, and in vivo transplantation in animal models to understand the mechanism underlying PD

The body of literature on using iPSCs for modelling PD has been growing rapidly in recent years. DA neurons have been generated from patients carrying mutations in the PD genes leucine-rich repeat kinase 2 (LRRK2), α-synuclein (SNCA) (both point mutations and wild-type gene multiplication), phosphatase and tensin homolog induced putative kinase 1 (PINK1), β-glucocerebrosidase (GBA), vacuolar protein sorting ortholog 35 (VPS35), and from patients with idiopathic PD. In this review we discuss the use of iPSCs for modelling PD, and the potential of iPSCs for developing better in vitro models for understanding the mechanisms of disease and finding new therapeutics. We apologize to those whose excellent work we have not been able to include, as our attempt here is to cover representative examples, and to offer a concise synthesis highlighting some of the major issues currently faced in this area.

iPSC models of SNCA

Intracellular aggregates of α-synuclein (encoded by SNCA), a protein predominantly found in the CNS and red blood cells, are a histopathological hallmark of most cases of PD. Various SNCA iPSC models have been used for modeling PD. Both point mutations in the SNCA gene and multiplication of the wild-type gene are causal for familial PD [1]. Triplication of the wild-type SNCA locus causes an early-onset, fully penetrant and aggressive form of PD with rapid cognitive decline [2]. DA neurons derived from 3X SNCA patients contain twice as much α-synuclein protein and mRNA compared to neurons derived from controls [3]. α-synuclein protein was increased in DA neurons derived from 3XSNCA patients, but no significant difference in SNCA gene expression was observed between these DA neurons and those derived from controls [4]. SNCA triplicate neurons have been shown to be more prone to oxidative stress, and the expression of oxidative stress-related genes such MAO-A was upregulated [3,4]. DA neurons generated from a PD patient harboring the SNCA triplication recapitulated an endoplasmic reticulum (ER) stress phenotype, specifically the induction of IRE1α/XBP1 of the unfolded protein response pathway (UPR) [5]. UPR activation has also been found in postmortem tissue [9], highlighting the utility of the iPSC model for studying disease mechanisms related to stress pathways.

In a study using iPSCs from a PD patient carrying an SNCA multiplication, α-synuclein oligomers caused impairment of anterograde axonal transport of mitochondria, which was restored by inhibition of oligomer formation [6]. α-synuclein oligomers also caused subcellular relocation of the transport-regulating proteins KLC1, Miro1 and tau. α-synuclein oligomers also reduced ATP levels, leading to axonal transport disruption and energy deficits causing synapse loss in DA neurons [6]. In a different study, iPSC-derived DA from SNCA triplication patients generated α-synuclein aggregates that caused mitochondrial-linked neuronal death, suggesting that the transition from monomeric α-synuclein to aggregates has functional implications in PD [7].

The phenotypes of neurons from patients with the causal SNCA A53T mutation have also been studied for modeling PD [8] [9]. Intraneuronal nitric oxide (NO) and protein nitration were increased in DA neurons derived from patients with the A53T-mutation compared to isogenic controls. DA neurons from patients harboring the A53T mutation were also shown to have higher basal and toxin-induced NO [8]. Free radicals inactivated the transcription factor MEFC2 by S-nitrosylation of Cys39 in α-synuclein mutant DA neurons [8]. Transcription of genes such as PGC-1α, downstream of MEFC2, was decreased in the DA neurons derived from patients harboring the A53T mutation, and was partially rescued by an inhibitor of Nitric Oxide Synthase, L-NAME [8]. The effects of mitochondrial toxins such as rotenone were completely rescued by activating MEFC2 by using isoxazole [8]. In other work, DA neurons derived from patients harboring A53T under basal conditions and stress also demonstrated defective neurite outgrowth and fragmented axons with swollen structures containing α-synuclein and tau [10]. These neurons also showed disrupted synaptic connectivity and alterations in transcriptional genes, aberrant phenotypes that could be rescued by small molecules [10].

In a study examining DA neurons derived from PD patients with the SNCA-A53T mutation and isogenic controls, axonal pathology in DA neurons with the A53T mutation was shown to be caused by loss of Nrf2 transcriptional activity, affecting expression of microtubule binding genes [11]. In a separate study, DA neurons from PD patients with the A53T mutation and SNCA triplication displayed altered mitochondrial morphology and decreased mitochondrial membrane potential compared to control neurons.[12] Another study further showed that DA neurons from both A53T and E46K SNCA-mutant iPSCs displayed fragmented mitochondria and accumulation of α-synuclein deposits clustered to mitochondrial membranes when exposed to cardiolipin [13].

α-synuclein accumulation has been shown to interfere with lysosomal function by disrupting lysosomal hydrolase trafficking. This effect could be partially rescued in a variety of iPSC-derived midbrain DA neurons by overexpression of key mediators of endolysosomal trafficking such as RAB1 or by pharmacological activation of GBA activity (see below) [14]. In other work, DA neurons derived from patients with young-onset PD but with no known mutations for PD displayed increased accumulation of soluble α-synuclein and phosphorylated protein kinase Cα, and a reduction in LAMP1 [15]. PEP005, an activator of lysosomal function, reduced α-synuclein and phosphorylated protein kinase Cα levels and increased LAMP1.

DA neurons derived from patients with SNCA-A53T and SNCA triplication have been differentiated to macrophages, recapitulating many features of microglia [16]. SNCA triplication but not A53T macrophages demonstrated increased intracellular levels of α-synuclein versus controls, released significantly more α-synuclein to the medium, and showed reduced phagocytic capability. The macrophages degraded α-synuclein, which could be inhibited by blocking proteasomal and lysosomal pathways [16].

As for causal α-synuclein variants and α-synuclein protein pathology, E46K DA neurons exhibited increased levels of α-synuclein, which were reduced by Stearoyl CoA desaturase [17]. DA neurons derived from iPSCs of PD patients with the A53T mutation displayed increased intracellular α-synuclein accumulation [18], and DA neurons from patients carrying the SNCA triplication showed oligomeric α-synuclein pathology and increased levels of α-synuclein extracellular release. Transcriptomic analysis revealed perturbations in expression of genes linked to mitochondrial function, aberrant mitochondrial morphology and impairment in membrane potential, as well as decreased levels of phosphorylated DRP1Ser616[12]. Increased ER stress, impaired cholesterol, and lipid homeostasis were observed in these DA neurons, suggesting a correlation between α-synuclein cellular pathology and defects in metabolic and cellular bioenergetics in PD [12]. In DA neurons derived from iPSCs of PD patients, aggregated α-synuclein disrupted the ability of neurons to respond to misfolded proteins in the ER. [19]. A combination of synergistic increase of many proteostasis pathways may also open new therapeutic avenues for PD [19].

Some models have been used to test the hypothesis that reducing pathological α-synuclein should be of therapeutic benefit. A-443654 – an AKT modulator - normalized levels of α-synuclein in iPSC-derived DA neurons from a patient carrying triplication of the SNCA gene [20]. In DA neurons harboring the E46K α-synuclein mutation, the lysophospholipase 1 inhibitor ML348 reduced total α-synuclein levels and phosphorylated α-synuclein at Ser129, demonstrating a link between palmitoylation and α-synuclein pathology [21]. DA neurons derived from iPSCs of monogenic forms of synucleinopathy caused by mutations in ATP13A2/PARK9 indicate that lysosomal exocytosis regulates intracellular levels of α-synuclein [22]. Loss of PARK9 function reduced lysosomal calcium storage, disrupted lysosomal calcium homeostasis, led to increased cytosolic calcium, and impairment of lysosomal exocytosis. Dysfunctional lysosomal exocytosis caused defective α-synuclein secretion from axons and soma, leading to α-synuclein accumulation and activation of the lysosomal calcium ion channel transient receptor potential. Mucolipin 1 (TRPML1) rescued these abnormalities and may represent a potential therapeutic strategy for PD and other synucleinopathies [22]. In DA neurons derived from iPSCs of α-synuclein triplicate patients, overexpression of α-synuclein reduced the functional availability of D2 receptors, caused a significant dysregulation in firing activity, dopamine release, and neuronal morphology [23]. Quinpirole, a selective D2 and D3 receptor agonist, restored the defective firing activity of these dopaminergic neurons [23]. Lysosomal cathepsins caused selective reduction of long-chain glycosphingolipids, reducing α-synuclein pathology [24]. SNCA DA neurons treated with the cathepsin inhibitor cysteamine exhibited significantly more intact/healthy neurites, suggesting that cysteamine can act as a disease modifying compound [25].

These findings illustrate the notion that DA neurons derived from SNCA iPCSs closely mimic many pathological features of PD and should be a promising tool not only for studying disease mechanisms, but also for preclinical work aimed at finding new therapeutics for PD.

iPSC models of LRRK2

A common form of familial PD is caused by mutations in the leucine rich repeat kinase 2 (LRRK2) gene [26], most frequently the highly pathogenic G2019S mutation (~5% of familial PD, also found in about 1% of sporadic PD), which biochemically increases the kinase activity of LRRK2. How mutations in LRRK2 cause neuronal degeneration in PD has been studied in patient-derived DA neurons [27-30]. In various studies, mutations in LRRK2 have been found to lead to accumulation of α-synuclein [28,30-32]. Some studies demonstrated an increase in SNCA transcription in LRRK2-G2019S mutant neurons [28], but others showed no upregulation of SNCA [30,31]. DA neurons derived from iPSCs of LRRK2-G2019S patients responded to defects in chaperone-mediated autophagy (CMA) by increasing levels of CMA lysosomal receptors [32]. Disrupting CMA led to increase in α-synuclein levels [32].

Oxidative stress and mitochondrial dysfunction have been both implicated in PD. In LRRK2-G2019S and LRRK2-R1441C-mutant DA neurons, basal and maximal mitochondrial respiration rates were low compared to control neurons [27]. Gene correction demonstrated that the presence of the LRRK2-G2019S mutation increased damage to mitochondrial DNA [33]. Impaired capacity of the neurons to meet local energy requirements in turn can contribute to oxidative stress [34], suggesting that improving mitochondrial function might be a promising therapeutic strategy for PD. LRRK2-G2019S mutant DA neurons also showed reduction in mitophagy due to interference with the normal functional participation of LRRK2 in removal of the mitochondrial membrane protein Miro [35]. Conversely, reducing Miro gene dosage could rescue disease-related phenotypes in LRRK2 mutant DA neurons [35]. LRRK2-G2019S mutant DA neurons displayed abnormal distribution of mitochondria and trafficking, which corresponded to decreased sirtuin deacetylase activity and NAD+ levels [36].

DA neurons derived from LRRK2-G2019S iPSC’s were also shown to be more susceptible to oxidative stress than control neurons [28]. Mutations in LRRK2 exacerbated oxidative stress, and increased expression levels of oxidative stress genes, such as Monoamine Oxidase-B (MAO-B) [28]. DA neurons derived from older patients showed higher signs of basal stress and increased number of neurons undergoing apoptosis from LRRK2-G2019S mutants than from controls [30]. iPSC-derived DA neurons with LRRK2-G2019S and LRRK2-R1441C mutations are more vulnerable to cell death by cell stressors [21]. Treatment with Coenzyme Q, Rapamycin, or the LRRK2 kinase inhibitor GW5074 reduced cell vulnerability. These studies suggest that LRRK2 inhibitors might have potential as therapeutics for the disease; clinical trials of several LRRK2 inhibitors are ongoing.

LRRK2 mutations have also been shown to affect the structure and function of the nuclei in iPSC-derived neural stem cells (NSCs) [29]. For example, expression of LRRK2-G2019S caused disorganization of the nuclear envelope and abnormal nuclear morphology [37]. NSCs from LRRK2-G2019S mutants were more susceptible to treatment with the proteasome inhibitor MG-132 and differentiated with less efficiency into neurons than NSCs from controls. These defects were caused by LRRK2, which was confirmed by correcting the G2019S mutation in the patient- derived iPSCs. iPSC-derived LRRK2-G2019S DA neurons exhibited collapse of neurites when treated with thapsigargin (THP), a non-competitive inhibitor of the sarco/ER Ca²⁺ ATPase. Baseline ER Ca²⁺ levels were lower in LRRK2-G2019S human DA neurons, as well as in differentiated midbrain DA neurons in vitro. After THP challenge, DA neurons displayed increased Ca²⁺ influx and decreased intracellular Ca²⁺ buffering upon membrane depolarization. These effects were reversed by correcting the LRRK2 mutation by antisense oligonucleotides and gene editing, suggesting that LRRK2-G2019S causes ER calcium-dependent pathogenic effects [37]. Neurite outgrowth velocity was also decreased in DA neurons derived from LRRK2-G2019S mutants, which was also rescued by targeted correction of the LRRK2 mutation [31]. In addition, DA neurons derived from idiopathic PD or from LRRK2-G2019S patients showed alterations in morphology such as reduction in neurite arborization, neurite number, and accumulation of autophagic vacuoles [30].

A number of microRNAs (miRNA) have been shown to be altered in DA neurons derived from iPSCs of PD patients. miR-9-5p and miR-135b-5p were associated with downregulation of transcription factors related to DNA hypermethylation of enhancer elements in PD, suggesting that miRNA changes may be involved in both monogenic LRRK2-PD and sporadic PD and co-occur with epigenetic changes in DA neurons from PD patients [38]. In LRRK2-G2019S DA neurons, the inhibition of miRNA activity by LRRK2-G2019S is directly blocked by TRIM32, an E3 ubiquitin ligase suggesting that TRIM32 may be a therapeutic target for PD [39].

In DA neurons and microglia differentiated from LRRK2-G2019S iPSCs, treatment with interferon-γ enhanced LRRK2-G2019S-dependent negative regulation of AKT phosphorylation and NFAT activation and increased the vulnerability of DA neurons to immune challenge. Activated LRRK2-G2019S microglia caused shortening of DA neurites [40].

In another study, in DA neurons from LRRK2-G2019S iPSCs, increased LRRK2 kinase activity altered autophagy by disrupting axonal transport of autophagosomes [41]. A human α-synuclein transmission model using LRRK2-G2019S iPSC DA neurons has also been developed where treatment of cells with human α-synuclein pre-formed fibrils (PFFs) led to endogenous α-synuclein aggregation in a time-dependent manner, whereas loss of LRRK2 decreased aggregation [42], consistent with the autosomal dominant inheritance of this mutation. And in patient-derived LRRK2-G2019S DA neurons, expression of the transcription factor NR2F1 is decreased compared to control neurons, suggesting a possible pathogenic mechanism in which dopaminergic differentiation is modified via NR2F1 [43].

A study in LRRK2-G2019S mutant iPSC-derived astrocytes showed increased astrocytic levels of α-synuclein, causing non-cell-autonomous degeneration of DA neurons, indicating the importance of PD-related genes in glial cells [44]. iPSC-derived microglia from LRRK2-G2019S patients mistraffic transferrin to lysosomes proximal to the nucleus under proinflammatory conditions [45], suggesting dysregulation of retromer-dependent retrograde endosomal trafficking. Astrocytes derived from LRRK2-G2019S iPSC’s were characterized by decrease in GFAP and SAB 100-positive astrocytic profiles, decrease in astrocyte complexity, decreased mitochondrial activity, abnormal mitochondrial morphology, increased glycolysis, and increased reactive oxygen species, suggesting that astrocytic abnormalities from LRRK2-G2019S mutation can lead to elevated oxidative stress and failed neuroprotection, possibly contributing to neuronal death in PD [46]. Ribosome profiling of DA neurons derived from LRRK2-G2019S iPSC’s showed that mRNAs with complex secondary structure were translated more efficiently in LRRK2-G2019S neurons, implying a link between dysregulated control of translation and Ca²⁺ homeostasis in LRRK2-G2019S human DA neurons, which could contribute to selective dopaminergic neuronal loss in PD [47]

DA neurons generated from LRRK2-G2019S iPSC’s show impaired membrane localization of the nicotinic acetylcholine receptors (nAChR) and the dopamine D3 receptors (D3R) and formation of the D3R-nAChR heteromer, crucial for maintaining neuronal homeostasis and dopaminergic neurons, phenotypes that could be rescued by inhibiting the increased LRRK2 kinase activity [48]. In DA neurons from LRRK2-G2019S PD patient lines with low GBA levels and enzyme activity, treatment with the atypical antipsychotic quetiapine, a nonspecific antagonist of dopamine, serotonin and histamine receptors, was sufficient to increase GBA levels and activity, and also reduced the accumulation of α-synuclein [49].

These iPSC-derived models link LRRK2 to various features of PD by closely phenocopying underlying mechanistic aspects of LRRK2-PD and have been useful not only for studying LRRK2-based disease mechanisms but also for developing and testing new therapeutic approaches.

iPSC models of GBA

Mutations in the glucocerebrosidase (GBA/GCase) gene (GBA1), the gene responsible for the autosomal recessive lysosomal disorder Gaucher disease, are the most common genetic risk factors for idiopathic PD. Heterozygous carriers of many of the mutations also experience increased risk for PD.

DA neurons have been generated from patients with Type 2 (acute neuronopathic) and Type 1 (non-neuronopathic) Gaucher disease [50]. These cells showed a decrease in glucocerebrosidase activity, as well as lower levels of stored glycolipid substrates such as glucosylsphingosine and glucosylceramide. Levels of α-synuclein were elevated in these DA neurons. Treatment with NCGC607, a small-molecule, non-inhibitory glucocerebrosidase pharmacological chaperone, stabilized the normally unstable mutant enzyme, restored GBA activity and protein levels, and reduced glycolipid storage in both iPSC-derived macrophages and in DA neurons. These data indicate that this compound may be a relevant candidate for treating neuronopathic Gaucher disease, including its Parkinsonism sequelae. In addition, α-synuclein levels were reduced in treated DA neurons from PD patients, suggesting that NCGC607 may prove useful even for idiopathic PD [50].

iPSCs were also derived from a set of monozygotic twins harboring the heterozygous glucocerebrosidase mutation N370S, which is closely associated with increased risk for PD [51]. The DA neurons from these iPSCs demonstrated a GBA enzymatic activity of approximately 50%, a 3x increased level of α-synuclein, lower capacity to synthesize and release dopamine, and increased MAO-B expression. Overexpression of WT-GBA and treatment with MAO-B inhibitors normalized both α-synuclein and dopamine levels, suggesting that, although gene therapy is still in its infancy, eventually combination gene/drug therapy may be beneficial. These studies emphasize the utility of iPSCs to model PD with clear genetic association. In other studies, in DA neurons derived from individuals with GBA-N370S mutations, ER stress, autophagic and lysosomal disturbance, as well as the presence of extracellular α-synuclein was observed, recapitulating early PD phenotypes [52]. A small-molecule modulator of GBA called S-181 increased wild-type GBA activity in iPSC-derived DA neurons from sporadic PD patients, and in DA neurons from patients carrying the 84GG insertional mutation in GBA or mutations in DJ-1, LRRK2 or PARKIN, all of which have lower than normal GBA activity. Treatment of these DA neurons with S-181 partly restored lysosomal function and lowered accumulation of glucosylceramide, oxidized dopamine, and α-synuclein [53].

In DA neurons derived from patients with idiopathic and familial PD, mitochondrial oxidative stress caused oxidized dopamine accumulation, reduced GBA enzymatic activity, dysfunction of lysosomes and accumulation of α-synuclein [24]. DA neurons from iPSC lines from PD patients with heterozygous GBA mutations (N370S, L444P, and RecNcil) showed defective NAD+ metabolism compared to isogenic gene-corrected and unaffected controls [54]. Mitochondrial functions were altered in the mutant GBA-DA neurons, which could be rescued using nicotinamide riboside, an NAD+ precursor. DA neurons from a PD patient line harboring the heterozygous 84GG GBA mutation that results in complete loss of mutant GBA and reduced wild-type protein demonstrated lower GBA levels compared to a CRISPR-corrected isogenic line as control, as expected [49]. This reduction could be partially rescued by treatment with increasing concentrations of the nonspecific receptor antagonist quetiapine. Quetiapine also reduced α-synuclein levels and reduced accumulation of oxidized dopamine in these DA neurons [49].

These studies suggest various GBA-mediated mechanisms of PD and provide valuable models for developing GBA-based therapeutics, a program ongoing in a number of laboratories and companies.

iPSC models of PINK1 and PARK2/PARKIN

Several studies have demonstrated the role of the mitochondrial kinase PINK1 and the E3 ubiquitin ligase PARKIN using iPSC models. Mutations in PARK2/PARKIN and PINK1 are associated with early onset PD [55]. PARKIN and PINK1 have been demonstrated to work together and have been shown to be involved in controlling mitochondrial function, although PARKIN PD patients typically do not show accumulation of α-synuclein aggregates in Lewy bodies. PARKIN mutations increased the susceptibility of DA neurons to oxidative stress [56,57]. The basal levels of reactive oxygen species (ROS) were significantly higher in PARKIN mutant DA neurons than in controls [58]. However, DA neurons from iPSCs harboring a PARKIN mutation did not show any increase in the basal level of oxidative stress [57]. Nevertheless, these neurons were highly susceptible to oxidative stress after treatment with dopamine [57] and also generated more reactive oxygen species than control DA neurons after manganese treatment (manganese exposure is an environmental risk factor for PD) [56]. DA neurons derived from PARKIN and PINK1 iPSCs displayed cell type-specific vulnerability with α-synuclein accumulation, abnormal neurotransmitter homeostasis, and mitochondrial dysfunction [59]. Mitophagy was impaired via formation of S-nitrosylated PINK1 (SNO-PINK1). DA neurons derived from iPSCs transfected with mutant PINK1 (C568A)-GFP showed impaired valinomycin-induced mitophagy, suggesting a direct link between SNO-PINK1 formation, nitrosive stress and mitophagy that could contribute to PD pathogenesis [60].

PARKIN /PINK1 patient-specific iPSCs and isogenic control lines have been used to screen an FDA-approved compound library to rescue disease-related vulnerability to apoptosis, identifying certain calcium channel antagonists as a novel candidate therapeutic agent class [61].

DA neurons derived from iPSCs of PARKIN or PINK1 patients show defective PINK1/PARKIN mediated mitophagy, a quality control mechanism enabling the autophagic degradation of damaged and superfluous mitochondria. The autophagy-activating drug, memantine, increased clearance of damaged mitochondria and might represent an interesting therapeutic strategy in PD [62,63]. In DA neurons derived from iPSCs of patients with PINK1 mutations, treatment with Coenzyme Q10 reduced the vulnerability of the neurons to valinomycin and concanamycin but did not reduce LDH release by the same iPSC-derived DA in response to higher concentrations of valinomycin and concanamycin A [21]. The LRRK2 inhibitor GW5074 reduced release of LDH from neural cells of PINK1 Q456X PD patients exposed to valinomycin but not to concanamycin A. These studies suggested that Coenzyme Q and LRRK2 inhibitors have therapeutic potential for PD [27].

To study the role of endogenous PINK1, dopamine neurons were differentiated from iPSCs from three PD patients with nonsense (c.1366C>T; p.Q456X) or missense (c.509T>G; p.V170G) mutations in the PINK1 gene [64]. These neurons, upon mitochondrial depolarization, displayed increased mitochondrial copy number, impaired ability to recruit lentiviral-expressed PARKIN to mitochondria, and upregulation of PGC-1α, which is an important regulator of mitochondrial biogenesis. These defects were rescued by lentiviral expression of wild-type PINK1 [64].

These studies provide valuable information on PINK1/PARKIN-mediated mechanisms in PD, on underlying disease mechanisms in general, and on pre-clinical behavior of potential therapeutics.

iPSC models of VPS35

VPS35 is part of the retromer multiprotein complex, a master controller of endolysosomal trafficking and recycling of proteins, including those involved in autophagy, mitophagy, and lysosomal degradation. The D620N mutation in the VPS35 gene causes a rare autosomal-dominant familial PD [65]. Decreased autophagic flux and lysosomal mass associated with accumulation of α-synuclein was shown in patient-derived VPS35-D620N DA neurons compared to controls [66]. The mutant-harboring DA neurons also underwent massive apoptotic cell death. Mitochondrial dysfunction with impaired mitochondrial respiration, decreased membrane potential, and increased production of reactive oxygen species were observed in these mutant DA neurons, as well as slower movement of Rab5a- or Rab7a- positive endosomes. Frequencies of endosome fission and fusion were also lower in neurons from the VPS35-D6020N mutation than in the control group and α-synuclein accumulated in TH positive dopaminergic neurons [66]. These studies are valuable for the study of the retromer complex and the role it plays in dysfunction of the endosomal–lysosomal system in PD, as well as providing a potential model system to evaluate retromer-based therapeutic strategies, which are under development.

iPSC models of Monogenic and Sporadic PD

Most cases of PD are sporadic and idiopathic, although various genetic and environmental risk factors have been identified. More than 15 genes have been identified that cause monogenic forms of PD but over 40 independent loci have been associated with increased risk of sporadic PD. These data indicate a pathophysiological link between monogenic and idiopathic forms of PD [67]. iPSC models of monogenic and sporadic PD have been generated from early-onset sporadic PD patients carrying, e.g., a heterozygous deletion of exon 5 (Ex5del) in PARK2. The neurons demonstrated abnormal accumulation of α-synuclein and downregulation of antioxidative pathways and the proteasome. Treatment with H₂O₂ and proteasome inhibitor MG132 markedly induced cell death, while benzamil, a proteasome enhancer, significantly reduced their susceptibility to cell death [68]. In another study, DA neurons derived from monogenic and sporadic PD patients demonstrated abnormal epigenetic changes [69]. Using iPSCs from monogenic but not fully causal genes as a model for sporadic PD may unravel unknown mechanisms of PD pathogenesis - in our opinion, an important and underutilized potential of iPSC-derived PD models.

3D Cultures and Organoids:

Current in vitro approaches to model PD are primarily based on neuronal cultures grown under 2D conditions. Such cultures lack many complex characteristics like cell-cell interactions and cytoarchitecture that are important, e.g., to predict the effectiveness of in vitro tested compounds in clinical trials [70].

The in vitro human brain organoid technology is a valuable tool to understand complex neurobiological process in a more physiologically relevant system and also enables translational applications [71]. 3D systems can mimic in vivo environments more closely than 2D models, and therefore can accelerate better neuronal differentiation and formation of neural networks in vitro [72,73]. Neurons in 3D cultures have also been shown to express more neuronal genes than neurons derived in 2D cultures [74].

DA neurons derived from iPSC’s carrying the LRRK2-G2019S mutation cultured in a 3D microfluidics system revealed robust endophenotypes, including increased cell death, decreased differentiation of DA neurons and reduced branching complexity, as well as altered mitochondrial morphology compared to isogenic lines, without the need of stressing the cells [75]. The LRRK2-G2019S-dependent dopaminergic phenotypes were rescued by treating the cells with a LRRK2 inhibitor. These data illustrate the utility and significant potential of 3D models and organoids for studying pathogenic mechanisms in PD.

A decrease in the number and complexity of DA neurons in LRRK2-G2019S compared to control organoids and the floor plate marker was observed in LRRK2-G2019S organoids [76]. FOXA2, required for DA neuron generation, was found elevated in PD patient-derived midbrain organoids, suggesting that there is a neurodevelopmental defect in mid-brain DA neurons harboring LRRK2-G2019S [76].

Organoids grafted in mice brains showed progressive neuronal differentiation, maturation, gliogenesis, integration of microglia, and growth of axons to different regions of the host brain [77]. In vivo two-photon imaging showed functional blood vessels and neuronal networks in the grafts. Functional studies revealed neuronal activity and suggested graft-to-host functional synaptic connectivity [77]. These 3D in vitro models and the implantation of human neural organoids into an in vivo physiological environment such as an animal brain can greatly improve disease modelling for PD and other diseases [78].

Concluding Remarks and Future Perspectives

One of the greatest challenges in the iPSC field is that iPSC lines show much variability among clones that appears unrelated to their genotype, making it difficult to interpret results across different studies and experiments. One way to overcome this problem would be to generate multiple lines from each patient and compare lines that differentiate with similar efficiency [3]. Other strategies could involve cell line pooling approaches, enabling parallel study of up to 30 distinct disease-related genotypes [79]. This approach can reduce variability across lines by forcing them to grow in the same well in a shared microenvironment at identical cell density. We hope to see more studies of these types.

Another challenge, as mentioned earlier, is that the transcriptional and epigenetic profile of iPSC derived neurons may not accurately resemble that of mature or aged neurons. It will be important to develop more efficient reprogramming and differentiation strategies [80-85] as well as strategies to trigger relevant age-related features in iPSC-derived DA neurons [86]. One must have stringent selection criteria for iPSC lines and highly validated and efficient neural differentiation protocols to study PD in a dish.

Most studies using iPSCs focused on familial PD, for reasons already discussed. Some recent studies have tried to model sporadic PD; more are needed. As mentioned earlier, it will be important to study iPSC lines from patients with non-causal PD mutations as well as from sporadic forms to understand the pathogenesis of sporadic PD by finding common phenotypes resulting from different genetic factors [87]. Another approach to understand how gene-environment interactions cause PD in sporadic cases will be to use iPSC-derived neurons from patients carrying PD-associated (causal and risk) mutations who are unaffected by the disease [3,4,27,29,30,32,56-58,88,89]. Table 1 presents a summary of some Parkinson’s disease iPSC lines, and their corresponding phenotypes.

Table 1.

Parkinson’s disease iPSC lines and their corresponding phenotypes

Mutation	Phenotype	Reference
SNCA triplication and α-synuclein A53T mutation	Increase of α synuclein protein levels Increase in oxidative stress Vulnerability to cell stressors	[3] [3,4] [7,14]
LRRK2 mutations	Increase of α synuclein protein levels Increase in oxidative stress Vulnerability to cell stressors Mitochondrial dysfunction Defective autophagy Morphological abnormalities	[30] [26,27] [26] [26,32,33,34,35] [29,31] [28,30]
GBA mutations	Increase of α synuclein protein levels Mitochondrial dysfunction	[49] [53]
PINK1 mutations	Increase in oxidative stress Vulnerability to cell stressors Mitochondrial dysfunction	[55,56] [26] [58,60,61]
VPS35 mutations	Increase in GluR1 cluster intensity Decreased autophagic flux Increased accumulation of α synuclein Mitochondrial dysfunction Increased membrane potential Impaired mitochondrial respiration	[63] [63] [63] [64] [64] [64]
Idiopathic PD	DA neurons cell death Oxidative stress Epigenetic changes	[65] [65] [66]

Variations at multiple genetic loci determine genetic penetrance and influence the risk of developing PD. Therefore stringent selection criteria must be developed when selecting control subjects for iPSC-based studies [1]. More recent studies have used gene editing to introduce PD-related non-coding variants identified by GWAS-studies to dissect their cis-acting defects on gene regulation [90].

Current differentiation protocols typically lead to the generation of heterogeneous neural populations, with relatively low yields of the neurons of interest. It is critical to improve and standardize dopamine neuron differentiation protocols. Furthermore, it remains controversial whether a disease that can take decades to develop can be effectively modelled in vitro over a period of just a few months. An important goal for future work, therefore, would be to maintain neurons for longer times in vitro [30]. Another major goal, both for better viability and for bringing the models closer to physiology, would be advance co-culture techniques, for instance of neurons together with mouse or human glia [30,44].

Research on transplantation approaches has been progressing and opens up interesting directions for future work. It has been shown that iPSCs from PD patients, when grafted in brains of adult rats, can reduce motor deficits [91]. Midbrain DA neurons from iPSCs derived from cynomolgus monkey (CM) have been analyzed for up to 2 years after autologous transplantation in a PD model in which unilateral engraftment of CM-iPSCs improved motor function on the side in which neurons were transplanted [17]. Postmortem analyses showed significant survival of midbrain-like dopaminergic neurons and extensive growth into the putamen [77]. These studies support the idea that transplanting neurons in vivo can be viable in animal models of PD. Brain organoids and transplantation of organoids into animals can also be a valuable tool for screening drug candidates.

In conclusion, despite many remaining limitations, the iPSC technology has developed into a valuable tool for studying the mechanisms underlying pathogenesis of PD and for carrying out pre-clinical discovery and evaluation of therapeutics.

Outstanding questions:

iPSC lines show substantial variability among clones, primarily due to differences in genotypes. In addition, the transcriptional and epigenetic profiles of iPSC-derived neurons may not resemble those of mature or aged neurons. Will generating multiple lines from the same patients and developing better reprogramming techniques overcome these challenges?

Most of the studies to elucidate underlying mechanisms of PD have focused on familial forms of the disease. Only 10-15% of patients have a positive family history of PD while the rest of PD cases are sporadic and idiopathic. Therefore, it is extremely important to study sporadic forms of the disease. Could expanding the study of familial forms of PD to additional genotypes, along with studying sporadic forms of the disease, elucidate common pathways?

Current protocols of iPSC cell lines generation produce a population of heterogeneous neurons in a span of 30 to 60 days. It would be important to develop induced “aging” protocols with neurons and other cell types such as astrocytes and microglia, grow them for longer, and perhaps transplant them in vivo. Will developing longer differentiation protocols, induced aging protocols and transplantation studies of iPSCs into animal models produce better (i.e., more pathophysiologically relevant) PD models?

3D in vitro systems can recapitulate in vivo environments more closely. They can also accelerate and improve neuronal differentiation, as well as the formation of neural networks. Cells in these systems express more neuronal genes, have more connections, etc., all of which promotes functional signal transduction. Will 3D cultures and organoids improve the understanding of disease mechanisms in vitro, and help address some of the challenges of 2D systems in replicating features of fully developed brains? Can transplantation help in this context as well?

Highlights.

• Generation of iPSC lines and isogeneic controls from sporadic and genetic forms of PD have recapitulated many of the cellular disease phenotypes of PD in cell cultures.

• SNCA-triplicate neurons and neurons derived from patients harboring the A53T mutation, neurons from LRRK2, PARK2/PARKIN and PINK1 iPSCs, and from patients with familial DJ-1 PD and other forms of PD, all show increased accumulation of α-synuclein.

• Mitochondrial defects, increased oxidative stress, lysosomal defects and defects in endolysosomal transport are also observed in neurons derived from PD patient iPSC lines.

• One of the greatest challenges is that iPSC lines show substantial variability among clones. One way to overcome this problem is to generate multiple lines from each patient and compare lines, as well as to develop protocols that better reflect the effects of aging on neurons.