Over the past decade, several differentiation protocols have been developed to generate more phenotypically relevant cell models from FRDA iPSCs. These include cell types significantly affected by the disease, such as neurons, cardiomyocytes and pancreatic β cells (Fig. 1).

The first study to generate FRDA iPSCs was performed in 2010, reprogramming 2 FRDA patient primary fibroblast lines to produce FRDA iPSCs [8]. The derived iPSCs lines not only maintained FXN gene repression but also exhibited similar patterns of repeat expansion and contraction across generations, as commonly described in human disease. The results of the study also crucially provided evidence for the role of mismatch repair (MMR) enzyme MSH2 in repeat instability, demonstrating the potential for FRDA iPSCs to elucidate the molecular mechanisms underpinning FRDA pathology.

One year later in 2011, neurons and cardiomyocytes were successfully generated from FRDA iPSCs [9], followed by a study published in 2013 demonstrating that mature cell types derived from FRDA iPSCs displayed a relevant frataxin deficiency phenotype [10]. Since the publication of these pioneering studies, several groups worldwide have developed differentiation protocols to generate various cellular models derived from FRDA iPSCs. These models have been used to investigate the molecular mechanisms driving GAA expansion instability in FRDA and allow for the exploration of novel therapeutic avenues. The following section will detail the development of these iPSC-derived models by cell type, as well as describing their use to study mechanisms underpinning FRDA pathology (see Table 1 for a summary of the studies described).

FRDA predominately results in the localised neurodegeneration of cell types found in both the CNS and the PNS, including the dorsal root ganglia, posterior columns, cortical spinal tracts, peripheral sensory neurons and cerebellar neurons of the dentate nucleus and its connections [26]. Several of these neuronal cell types have been successfully generated from FRDA iPSCs (Fig. 1). Typically, studies have focused on the derivation of “default” central sensory neurons and neural crest cells. This process involves neural induction of iPSC colonies by noggin to form neural stem cells, which are propagated and maintained as neurospheres. These neural precursors can then be induced to differentiate into mature cell types via the addition/abstention of various factors within the culture medium. The development of patient-derived sensory neurons has provided researchers with a useful tool to investigate the phenotypic characteristics and biological processes driving FRDA disease pathology in a relevant neuronal model. One of the earliest studies to utilise this tool characterised pathological mitochondrial phenotypes in FRDA iPSC-derived sensory neurons. When compared to control iPSCs, neurons from FRDA iPSCs demonstrated a reduced mitochondrial membrane potential and a delayed development of electrophysiological functionality following differentiation, confirming a role of mitochondrial pathogenesis in FRDA and providing several cellular phenotypes for further study [10]. Interestingly, in 2014, a contrasting study was published showing no overt differences in the long-term survival and mitochondrial function in FRDA-derived neural progenitors and neurons, respectively, compared to control lines. The FRDA iPSCs were able to differentiate into fully functional neurons, with normal electrophysiological recording comparative to control lines. Additionally, no alteration in mitochondrial volume or function was seen in FRDA neurospheres, which also showed no increase in cell death compared to control iPSC- and human embryonic stem cell-derived neurospheres. This study was also the first to provide an in vivo characterisation of FRDA-derived neurons, showing that FRDA iPSC-derived neurospheres were able to integrate, differentiate and survive in rats [15]. This contrasting inability to detect overt phenotypic characteristics of neurodegeneration may be due to several factors relating to the underlying differences in the iPSC lines utilised, the immature nature of derived neurons and the late-childhood onset of degeneration seen in symptomatic disease. We shall discuss these factors in a later section of this review.

The differences in phenotypic presentation seen in cellular models may also be due to the heterogeneity of cells derived from FRDA iPSCs. In their studies, both Hick et al. [10] and Bird et al. [15] employed protocols to derive central sensory neurons, similar to immature cortical neurons. Neurodegeneration in FRDA is predominately seen to affect the peripheral sensory neurons (PSNs) which compose the dorsal root ganglion (DRG) of the spinal cord. These PSNs are more susceptible to decreased frataxin levels compared to central neurons, thus being disproportionality affected by FRDA pathology. With this in mind, several groups worldwide have worked to develop differentiation protocols for the derivation of more FRDA-specific neuronal cell models. The first of these models established was a development-based differentiation protocol to specifically generate neural-crest PSNs from FRDA-derived iPSCs. Utilising their novel model, the group was able to demonstrate an early failure to upregulate frataxin expression during sensory differentiation of FRDA PSNs, highlighting a potential developmental element to FRDA pathology [20]. However, it is specifically the proprioceptive neurons of the DRG that are deemed to be of particular interest in FRDA pathology and have recently become subject to further protocol optimisation. The DRG is composed of many PSN cell types, including nociceptors, thermoreceptors and mechanoreceptors. Comprising only 7.5% of PSNs found in the DRG, proprioceptive neurons are known to be particularly vulnerable in FRDA and have been shown to express the majority of the FXN found in the DRG [27,28]. Thus far, almost all protocols that have been developed for the generation of PSNs from iPSCs, have focused on deriving nociceptor-enhanced cultures or mixed neuronal populations with few proprioceptors.

More recently, a novel protocol has described the generation and enrichment of PSN cultures for proprioceptors, starting with FRDA iPSCs. By modifying a pre-existing protocol for the generation of nociceptors, cultures significantly enriched for proprioceptors, shown to represent 50% of end-differentiated neurons, were successfully obtained. Utilising fluorescent-activated cell sorting, it is possible to further purify cultures to derive almost pure populations of proprioceptors [13]. This study describes, at the time of writing, the most phenotypically relevant iPSC-derived 2D neuronal model for FRDA. As technology continues to progress, attention within the iPSC field is now shifting in an attempt to create iPSC-derived 3D organoid models of neurodegenerative diseases. This has just recently been demonstrated by the generation of a 3D DRG organoid model of FRDA, which displays a transcriptional profile comparable to the native DRG and is composed of the representative PSN cell types [12].

In addition to optimising and developing neuronal cell models for the characterisation of FRDA phenotypes, iPSC-derived models have also been used to further investigate FRDA pathology. There is still much to understand as to how the trinucleotide expansion drives the disease phenotype and several studies have been published utilising various FRDA iPSC-derived neurons to better understand mechanisms underpinning the disease. One such mechanism is the potential role of MMR proteins in FXN gene repression. The first study to generate FRDA iPSCs also crucially provided evidence for the potential role of MMR enzyme MSH2 in repeat instability [8]. A follow-up study investigated the expression of several different MMR enzymes in FRDA fibroblasts, FRDA iPSCs and iPSC-derived neural precursors. The study found an increased expression of MSH2, MSH3 and MSH6 enzymes in FRDA iPSCs when compared to fibroblasts and iPSC-derived neural stem cells. Additionally, the authors demonstrated that shRNA silencing of MSH2 and MSH3 hindered the GAA repeat expansion, strongly implicating a pathological active MMR system and an embryonic nature underlying the trinucleotide expansion [19]. This finding was strengthened by reports of high levels of GAA instability in FRDA-derived iPSCs, which corresponded with the upregulation of MMR enzymes MSH2, MSH3 and MSH6 [10]. Interestingly, reports of comparable levels of MHS2 and MSH6 in FRDA iPSC-derived neurospheres compared to FRDA iPSCs in the same study, contrasted reports of lower expression levels of these MMR enzymes in neural crest cells compared to corresponding iPSCs by another group [19]. Thus, while these findings demonstrate the potential role of MMR proteins in FXN gene repression, they also hint at other factors likely driving GAA-repeat instability.

Many other studies have also utilised iPSC-derived neuronal models to probe FRDA pathology and highlight the possibility of potential future therapeutic avenues. These include illustrating the potential of incretin-analogue therapy by investigating the key mediators of frataxin-deficiency-induced apoptosis [22], identification of potential protein targets and mechanistic pathways of interest for epigenetic therapy [21], determination of a previously undescribed differentially methylated region upstream of the GAA repeat expansion [23] and identification of differentially expressed glycolysis pathway genes associated with FRDA, capable of rescue with lentiviral FXN expression [25].

Cardiomyocytes present another mature cell type particularly susceptible to the complex interplay of frataxin deficiency, aberrant iron homeostasis and mitochondrial dysfunction, resulting in the heart being a significant site of clinical pathology in FRDA. The majority of FRDA patients develop cardiomyopathy during their lifetime, with subsequent cardiac failure representing their most common cause of death [29]. Thus, alongside neurons, cardiomyocytes are a second major cell type of focus for studies investigating FRDA pathology. The first report of FRDA iPSC-derived cardiomyocytes was published in the same 2011 paper which also first detailed the generation of FRDA iPSC-derived neurons [9]. These cardiomyocytes were derived using a 2-step process, the first of which was the formation of embryoid bodies (EB) from pluripotent cells. Eight-day-old EBs were then cultured in specific conditions to induce cardiac differentiation, with cardiomyocytes being discernible by beating areas. Although there have been further developments in cardiomyocyte differentiation protocols beyond the initial 2-step process since this publication, progress has been less extensive as compared to neuronal models. This is likely due to the lack of heterogeneity of cardiac cell types susceptible to FRDA pathology alongside already well-established cardiomyocyte differentiation protocols. However, similar to the development of a 3D organoid model of the DRG, there has been one recent report detailing the development of a 3D cardiomimetic model. Utilising FRDA iPSC-derived ventricular cardiomyocytes, coupled with advanced engineered tissue platforms, Wong et al. were able to develop a 3D FRDA ventricular model to investigate the contractile and electrophysiological properties of the disease. When comparing their pathological models to healthy control constructs, they found electrophysiological defects consistent with clinical disease, as well as observing a positive correlation between FXN expression and contractility [11].

FRDA iPSC-derived cardiomyocytes have been utilised in several studies in an attempt to better understand the biological mechanisms that drive FRDA pathology. One of the first studies investigating disease phenotypes in an FRDA iPSC-derived cardiomyocyte model found that when compared to control lines, FRDA cardiomyocytes displayed characteristics of FRDA pathology, such as an increased frequency of compromised mitochondrial homeostasis and a larger abundance of dark respiration-compromised mitochondria [10]. Further characterisation of electrophysiological dysfunction reported that FRDA-derived cardiomyocytes exhibit an increase in beat rate variability in comparison to control lines. These disease phenotypes were reversed when cells were treated with nifedipine (an L-type Ca2+ channel inhibitor), implicating a Ca2+ handling deficiency in the electrophysiological impairment underpinning cardiac FRDA pathology [17]. A more recent study has utilised a novel isogenic model to investigate FRDA cardiomyopathy, by knocking down frataxin in iPSC-derived isogenic cardiomyocytes post-differentiation. Using transcriptomic analysis, it was demonstrated that unsurprisingly mitochondrial pathways were amongst the most affected by frataxin deficiency, further evidenced by the observed mitochondrial electron transport chain dysfunction. Interestingly, it was also shown that a type 1 interferon activation response was associated with frataxin knockdown, potentially mediated by an increase in cytosolic mitochondrial DNA [24]. Another recent study utilising transcriptome profiling and differential gene expression analysis has also identified genes involved with extracellular matrix functionality as significantly affected by FRDA pathology in iPSC-derived cardiomyocytes [25].

Another area of interest has been utilising cardiomyocyte models to investigate the pathological role of aberrant iron homeostasis, which is thought to be a key driver of clinical cardiac pathology seen in the disease. FRDA iPSC-derived cardiomyocytes display an accelerated cardiomyopathy phenotype when subjected to iron overloading conditions, with reports of hypertrophic changes, decreased ATP production and aberrant calcium handling in FRDA cardiomyocytes cultured in the presence of excessive iron supplement [14]. More recently, further insights have been provided into cardiac iron homeostasis in FRDA, with findings implicating a dysfunction of the hepcidin-ferroportin axis in diseased cells [18].

As frataxin is an essential protein constitutively expressed by all cells within the body, organ systems outside the CNS and the heart are also commonly susceptible to FRDA pathology. One such clinical manifestation is the high prevalence of diabetes in FRDA patients when compared to an aged-matched control population. Both impaired insulin secretion and insulin resistance have been reported in FRDA, with 40% of FRDA patients exhibiting impaired glucose tolerance [30] and around 20% of patients going on to develop type 2 diabetes [31,32]. Pancreatic β cell dysfunction is a key driver of the development of diabetes in FRDA, with β cells being particularly susceptible to mitochondrial dysfunction and metabolic stress [33]. For this reason, β cells present another mature cell type of interest for researchers to investigate FRDA pathology. However, until recently the generation of β cells from an iPSC source had remained a largely challenging proposition [34]. Thus far, only one group has successfully generated and utilised FRDA iPSC-derived β cells within the FRDA field. An FRDA β cell model that expresses differentiation markers consistent with β cell development and comparable to EndoC-βH1 β cells and human islet cells, has been successfully derived by employing a 7-stage differentiation protocol. When treated with the glucagon-like-peptide-1 analogue exenatide, FRDA iPSC-β cells upregulate frataxin expression, and show improved metabolic function and increased mitochondrial activity, demonstrating that iPSC-β cells can be utilised to investigate FRDA cellular phenotypes and potentially develop novel therapeutics [35].

There are clear clinical implications for the visual system in FRDA, with several studies describing the presentation of ophthalmic manifestations in FRDA patients [36,37,38,39]. Optic neuropathy represents the most common manifestation, yet interestingly FRDA differs in presentation from other classical mitochondrial optic neuropathies. While optic nerve degeneration [36,37,38,40] and increased sensitivity of retinal ganglion cells to oxidative stress [39] have been implicated in symptom development, there is still little understanding of the pathophysiology driving visual defects in FRDA. One particular cell type of interest is the retinal pigment epithelium (RPE) layer, the pigmented cell layer responsible for photoreceptor nourishment and associated with iron accumulation and retinal degeneration. Due to the dysfunctional nature of iron homeostasis in FRDA, the role the RPE potentially plays in FRDA optic pathology has been studied. Utilising FRDA-derived iPSCs, one study has successfully generated RPE cells, which were then investigated for a number of phenotypic characteristics. Interestingly, these iPSC-RPEs displayed no significant differences in oxidative phosphorylation activity and exhibited normal phagocytosis in comparison to controls [16]. While these findings implicate that RPE pathology may not be of significance in the development of FRDA optic neuropathy, the ability to generate mature ophthalmic cell types from FRDA iPSCs provides hope for a better understanding of the pathological causes of visual defects in FRDA.

Despite the shared and significant efforts of doctors, scientists, patients and advocacy groups, FRDA currently remains without a cure. Despite the recent FDA approval of Skyclarys (omveloxolone) for the treatment of FRDA [41], the disease remains in desperate need of novel therapies, with existing approaches focusing solely on symptomatic relief and maintaining functionality. The major difficulty of developing efficacious disease-modifying drugs for FRDA is likely due to two barriers. The first barrier is an incomplete understanding of the biological mechanisms that drive the trinucleotide expansion and underpin disease progression. As our understanding of FRDA improves, our ability may improve to identify novel and promising pharmacological targets. The second major obstacle to pharmaceutical development is the enormous discrepancy in translating preclinical targets from bench to bedside. This failure is, in part, due to challenges surrounding the phenotypic relevancy of traditionally used preclinical models, such as patient fibroblast lines, which display limited susceptibility to frataxin deficiency.

The recent adoption of iPSC technology in the field of FRDA research has assisted in tackling both of these problems. Scientists can now derive a range of FRDA cell types of interest and use these preclinical models to study the biological mechanisms underpinning FRDA and to identify novel therapeutic targets. The advancement of differentiation protocols also allows access to the most phenotypically relevant cell types for preclinical testing, enabling iPSC-derived models to be used in the high-throughput screening of novel drugs. This section will explore how FRDA-iPSCs and their derived models have been used in several studies to accelerate the development of novel therapies for FRDA (Fig. 2), focusing on two major avenues: pharmacological approaches (Table 2) and gene therapy (Table 3).

Several different pharmacological approaches have been investigated for the potential treatment of FRDA. The majority of preclinical in vitro studies which have utilised FRDA iPSCs and their derived models to test therapeutic agents have focused on epigenetic-based therapy (Table 2); however, several other exciting avenues have been explored.

Another promising therapeutic avenue for the treatment of FRDA is gene therapy. Unlike some of the previously explored pharmacological methods, a gene therapy approach aims to cure patients directly, rather than acting as a chronic treatment. This is achieved by modulating the expression of key genes to restore function, either by direct excision or more commonly, by delivery of a wild-type transgene to patient cells using adeno-associated viruses (AAVs), retroviruses or herpes simplex virus type 1. Recently, several non-viral vector approaches of delivery have also been explored, including electroporation, zinc-finger nucleases (ZFNs) and layer-by-layer (LbL) particles [67,68,69]. Gene therapy has already been investigated as a promising avenue for many other neurodegenerative diseases including Parkinson’s, Alzheimer’s and Huntington’s, with encouraging results [70,71,72,73]. More recently, the field has also attempted to tackle rarer genetic disorders such as FRDA. As FRDA is caused by a dysfunction in a single gene, it presents a promising and conceptually simple candidate disorder for gene therapy. It is thought that delivery and expression of the FXN gene will complement the loss of FXN expression from the mutant alleles and potentially reverse FRDA phenotypes. Several preclinical gene therapy studies for FRDA have been performed, with current approaches focusing on the delivery of vectors driving FXN gene expression[27,74,75,76,77,78,79]. Some of these studies have returned promising results, demonstrating an improvement of FRDA phenotypes in cell and animal models and reinforcing the prospect of gene therapy as a potential long-term cure for FRDA. The advent of iPSC technology has been instrumental in developing the field of gene therapy for FRDA, with studies utilising FRDA iPSC-derived models to investigate potential gene targets as well as exploring the effectiveness of various vectors of delivery (see Table 3).

One particular area of contention is the ability of FRDA iPSCs and their derived models to exhibit overt and measurable phenotypes representative of FRDA pathology. Several studies have reported various empirical phenotypic traits in FRDA iPSC-derived neurons and cardiomyocytes. These measurable phenotypes provide a predictive model which can be utilised for in vitro drug discovery. Interestingly, however, several reported FRDA phenotypes are seen to be relatively mild when compared to control cell lines. Additionally, other studies have reported a failure to observe characteristic FRDA phenotypes in their iPSC-derived models. For example, in an aforementioned study, neural derivatives generated from FRDA iPSCs were found to give rise to functional neurons, with no apparent signs of FRDA pathology. When FRDA iPSC-generated neurons were compared with their controls, no significant differences in the levels of cell death or mitochondrial volume and function were seen. Further examination of the in vivo differentiation potential of these FRDA iPSC-derived neural progenitors, revealed an ability to integrate and survive within the cerebellum of adult mice, giving rise to both neurons and glial cells [15]. Additionally, FRDA iPSC-derived cardiomyocytes have been shown to be comparable in size to wild-type controls, demonstrating no signs of dysfunctional ATP production or calcium handling. It was only after these cells were cultured with an additional exogenous Fe2+ supplement of 200 µM, that any discernible phenotypic differences emerged [51]. Notably, such high levels of endogenous iron-overloading are highly unlikely to be seen in healthy cardiomyocytes, bringing the physiological relevance of these findings into question.

One potential reason for the discrepancy in the phenotypic presentation of FRDA iPSC-derived cells may be due to the intrinsic differences in their repeat expansion lengths. The severity of disease phenotype is known to be correlated with FXN production, which has been associated with the number of pathological GAA repeats [84,85]. Over 95% of FRDA patients have bi-allelic expansions ranging from 66 to 1700 repeats, with the majority of symptomatic patients possessing 600-900 repeats [86,87]. It is thought that FXN protein expression of approximately ~ 5%-35% of normal levels is required for FRDA patients to display symptoms [88,89,90,91]. Published studies which reported differences in their observed phenotypes, utilised FRDA patient cells differing in the size of their repeat expansions to generate FRDA iPSCs and their derived models. Thus, it is likely that differences in repeat length between patient cell lines influenced levels of FXN protein expression in these models and consequently their phenotypic presentation of FRDA pathology. The problem of line-to-line variability may be addressed with the use of isogenic cell lines, allowing the creation of genetically modified homogenous cell lines which better resemble healthy counterparts.

Additionally, it is important to note that FRDA is a late-onset childhood disorder, with symptoms tending to present between the ages of 5-15. Often neurons and cardiomyocytes derived from FRDA iPSCs and used for experimentation are immature, with a neonatal resemblance, therefore limiting the ability of these models to accurately recapitulate disease phenotypes. This is due to the reprogramming process, which resets the developmental clock of donor patient cells back to a stem cell state. These subsequent iPSC-derived models may not demonstrate overt FRDA phenotypes until they have matured to a later stage in their growth, or have had the required time to be adequately exposed to potential environmental stressors or accumulate the progressive degenerative pathology. Potential solutions for this issue include generating artificially-aged neurons through the direct lineage conversion of patient fibroblasts to create induced neurons (iNs) [92,93], or actively stimulating the rapid ageing of iPSC-derived models via progerin exposure [94]. A recently published study has also described a novel differentiation method using Sfrp2 to induce mature cardiomyocytes from human iPSCs [95].

Neuronal degeneration and phenotypic presentation may also be specific to particular neuronal subpopulations. The majority of studies investigating phenotypes in FRDA iPSC-derived neurons have generated and utilised central sensory neurons, mixed, or non-specific nociceptor predominant cultures of PSNs. As these neuronal subtypes are less vulnerable to frataxin deficiency, their exhibited pathological phenotypes may consequentially be mild and/or difficult to detect. Although it has been shown that proprioceptor-enhanced cultures of FRDA iPSC-derived neurons can be generated [13], further work is required to establish whether this vulnerable neuronal subpopulation displays more pronounced and representative phenotypes of FRDA pathology. Reports on the ability to observe and reverse phenotypes such as defective survival and morphology, impaired Fe-S cluster biogenesis and defective axonal mitochondrial behaviour in a 3D iPSC-derived DRG model of FRDA [12], provide further hope for the use of novel iPSC-derived FRDA neuronal models and their phenotypes for the study and development of potential therapies.

Alongside peripheral nerves and the dorsal root ganglion, the cerebellum presents another major site within the CNS severely affected by the neuropathology of FRDA [96]. The dysfunction and degeneration of neurons of the deep cerebellar nuclei give rise to gait ataxia, progressive loss of motor coordination, dysarthria and saccadic eye movement alterations frequently seen in FRDA patients [97]. Of particular interest are Purkinje cells, which constitute the sole neural output in the cerebellar cortex. Despite numerous studies detailing established protocols for the derivation of cortical and PSNs from FRDA-iPSCs, there have been no reports of the generation of FRDA iPSC-derived neurons of the deep cerebellar nuclei, or Purkinje cells which modulate their activity. Even outside of the field of FRDA research, only a handful of studies have successfully utilised pluripotent cells for the generation of Purkinje cells [98,99,100], with just a single study reporting successful derivation from patient iPSCs [98].

The generation of Purkinje cells has remained largely elusive likely due to their lengthy and challenging differentiation protocols, as well as their complex morphological characteristics making them difficult to differentiate and culture [101,102]. Successfully deriving Purkinje cells from FRDA iPSCs remains a key hurdle the field must overcome. Recent advancements such as the reports of a simplified and more reproducible Purkinje cell differentiation protocol [103] and the ability to mature pluripotent stem cell-derived cerebellar neurons without co-culture [104], mean that the generation of FRDA iPSC-derived Purkinje neurons is one step closer to reality. Such models have the potential to provide insight into cerebellar dysfunction in FRDA and may be of huge utility for the development of novel therapies. If and when it is possible to generate FRDA iPSC-derived Purkinje neurons of the cerebellum, it may also be of particular interest to look at differences between affected and unaffected neurons, to determine which characteristics prompt certain neurons to be susceptible to frataxin deficiency.

Early studies on iPSCs have relied on using cells that are aged-matched, unaffected by the pathology and with the same family pedigree as controls. However, despite the best efforts to match patient and control cells, there are differences in genetic background, single nucleotide polymorphisms, and other confounders that make comparisons between them sometimes hard to interpret. The rapid development of genome editing techniques has made possible the creation of isogenic lines. These lines are generated from the same individual and have been engineered to differ only at one specific locus, in this case, the GAA expansion at intron 1 of the FXN gene. External to expansion length, cells are genetically identical, overcoming in this way, inter-individual variabilities in all types of studies including genome-wide studies. For the majority of FRDA isogenic cell lines, expansion length is genetically modified prior to iPSC induction and any potential further differentiation. Interestingly, a recent study has also reported the ability to generate a novel isogenic model of FRDA by directly knocking down frataxin expression in FRDA iPSC-derived cardiomyocytes [24]. By modifying genes after differentiation, this technique further controls for variables caused by the independent processing of isogenic pairs, including passage number and any potential variance which occurs during the differentiation process, such as the cellular composition of obtained cultures. The use of isogenic cell lines has also been fundamental for disease modelling, for the understanding of disease biology and more recently, for testing novel therapeutic agents. The ability to specifically alter the length of the GAA expansion may also provide researchers with a potential tool to theoretically investigate how repeat length influences treatment response to therapy, in otherwise genetically identical cell lines. With time, more studies will only use isogenic cell lines as controls, and these important tools will become the must-have controls in a wide variety of studies.

For decades, cell culture has been central in advancing our understanding of cell biology principles, tissue and cell morphology, mechanisms of disease, and the action of drugs. It is a widely used technique that up until recently, has mainly relied on 2D cultures. There are many advantages of using this type of culture including that they are well established, they are mostly inexpensive to maintain, the data generated with them are vast and allow the comparison between studies and they are easier to understand. However, they do not recapitulate the in vivo environment where they grow, the interaction between different cell types is usually perturbed, and they do not offer a physiologically relevant spatial context. These and other limitations have been key to the fast-growing field of 3D cultures. The capability to model cells in vivo while being cultured in vitro has become one of the key drivers of the adoption of 3D cultures in many laboratories. When speaking about the brain, it is clear why 3D cultures are being increasingly developed and used. The brain being an extremely complex 3D arrangement of cells cannot be modelled properly in 2D. Moreover, certain research questions cannot be addressed using a 2D cell culture system. However, it is important to keep in mind that 3D cultures are much more heterogeneous and complex and they are challenging to standardize and analyse.

3D neuronal cultures are not new in FRDA. Neurospheres, one of the earliest 3D neuronal cultures, have been used by many groups over the years. However, more recently the tendency has been to move to more complex systems such as organoids. Recently, DRG organoids were generated by in vitro differentiation of FRDA iPSCs [12]. These 3D DRG-like structures were generated by culturing single iPSC aggregates in low-adhesion 96-well plates for 16 days in vitro and exposing them to a sequential combination of small molecules that differentiate iPSCs into neural crest neuronal derivatives [12]. Initially, the authors of the study tried to generate a 2D culture of sensory neurons from FRDA patient iPSCs and encountered extensive cell death. They hypothesized that the absence of glial cells and the lack of sufficient cell-to-cell contacts in a 2D system could explain the reduced viability. Their new 3D arrangement of cells proved to be, at the transcriptomic level, similar to primary DRGs, exhibiting a global transcriptional reconfiguration towards sensory neuronal cells, as well as supporting peripheral glial cells. Additionally, they better model the spatial organization and cellular composition when compared to classical 2D differentiated neuronal cultures.

Engineered human cardiac models of FRDA have also been reported [11]. iPSCs have been used in a pioneering study to generate both 3D human ventricular cardiac anisotropic sheets and ventricular cardiac tissue strips, useful for measuring electrical conduction as a syncytium and evaluating contractility, respectively. These cardiomimetic in vitro model systems possess the ability to produce multicellular readouts such as conduction velocity and magnitude of force generation, not possible to obtain in single-cell models. When it comes to a complex arrangement of cells such as those in different parts of the brain or the heart, it is increasingly useful to understand these systems as a whole, to more accurately translate research to a clinical setting.

Drug discovery and development have been performed for the past few decades using immortalized cell lines, as well as primary cells derived from patients and animal models. The discovery of iPSCs opened a new avenue for the study of otherwise inaccessible human cells in vitro, presenting a novel platform for drug testing. In FRDA, we can now study neurons and cardiomyocytes, two historically inaccessible cell types disproportionately affected by frataxin deficiency. Moreover, differentiation protocols are becoming more specialized, allowing the derivation of precise cellular subpopulations. This particular characteristic is important for addressing specific questions where a mixed culture may create confounders. iPSCs can also provide an extra layer of safety to assess toxicity and teratogencity before advancing drugs to lengthy and expensive clinical trials.

In order to advance potential therapeutic agents to clinical trials, it is vital to understand if these drugs reproduce their desired effect on the cell types most affected by FRDA pathology. Translating in vitro results into a clinical setting is already challenging and presents one of the key features of the poor predictability of drug efficacy during clinical trials. The use of cell types with little relevance for FRDA in the preclinical phase only further increases this uncertainty. This was the case with resveratrol, which performed well in fibroblasts and lymphoblastoid cells, but underwhelmingly in a subsequent clinical trial [105,106]. Once resveratrol was tested on neurons derived from FRDA iPSCs, only a modest (roughly 1.3 fold) increase in FXN mRNA expression was observed 48 h following treatment with 25 µM resveratrol, with the drug failing to significantly upregulate frataxin protein expression [48]. Nicotinamide, another HDACi which had returned promising results in FRDA primary lymphocytes [107], also failed to significantly upregulate frataxin protein expression in FRDA neurons when tested in the same study [48]. With this in mind, we propose an ideal paradigm for drug discovery and development using iPSC-derived specific cell types in the preclinical validation of potential drugs (Fig. 3). This model emphasizes the importance of using iPSC-derived cells for preclinical modelling to select the most promising agents to progress to clinical trials, saving resources and increasing the potential of translational success.

Cell replacement therapy represents a novel and exciting therapeutic avenue currently being explored for FRDA. Transplantation studies have shown that human donor cells survive and functionally integrate into the CNS, and this successful integration can lead to damaged neural tissue repair [108]. However, most of the studies performed address diseases that affect the CNS, and fewer have been reported for the PNS. FRDA iPSC-derived neural progenitors transplanted into adult rat cerebellum not only survive but also integrate, and differentiate within the host tissue [15]. Furthermore, it has recently been demonstrated that FRDA iPSC-derived sensory neural progenitors can mature and integrate within the adult DRG in vivo. These grafted cells express neuronal sensory markers including nociceptors, mechanoreceptors, and proprioceptors [109]. Transplantation of haemopoetic stem and progenitor cells has been shown to rescue deficits in the YG8R mouse model of FRDA [110]. Furthermore, gene-corrected human hematopoetic stem cells have already been reported and provide a basis for autologous transplantation [111]. These studies demonstrate proof-of-principle that cell replacement may be a promising treatment strategy to replenish some of the most severely affected cell types in FRDA. By combining gene-editing technologies and transplantation therapy in the future it may be possible to offer FRDA patients the potential of transplanting sensory neuron or cardiomyocyte grafts where the GAA expansion has been removed. This may be achieved by creating isogenic corrected cell line grafts from FRDA patient iPSCs. While this type of scenario currently remains futuristic, promising and rapid advances in science may give life to this approach sooner rather than later.

Not applicable.

Not applicable.

The authors declare that they have no competing interests.