Over the past decade, the development of induced pluripotent stem cell (iPSC) technology has provided a novel means to generate disease-relevant cells for in vitro investigation of the molecular and cellular mechanisms underpinning human complex disorders [1,2]. Diseases affecting the central nervous system (CNS) have been the most frequent targets, given the limited access to vital human brain cells and to the lack of or adequacy of existing animal models [3]. iPSCs are undifferentiated cells with self-renewal capability and can be directly generated from somatic cells (isolated from the skin, blood, or urine of an individual) by expressing the reprogramming factors Oct4, Sox2, Klf4, and c-Myc [4]. In principle, the resulting patient-specific iPSCs can subsequently differentiate into disease-relevant cell types, including multiple subclasses of neurons, astrocytes, and microglia, thus providing an unlimited source of cells that harbor the patient’s precise genome associated with the pathogenesis in the appropriate microenvironment. Therefore, the use of iPSCs offers opportunity to study CNS disorders overcoming the ethical limits of human embryonic stem cells (hESCs) or limited durability of primary cultures [5]. Recapitulating disease features, these cells can serve as large-scale drug screening platforms for novel therapeutic targets and have the great potential to be used in cell therapy and personalized medicine as compared to different sources of stem cell, such as hESCs or mesenchymal stem cells (MSCs). Main advantages and disadvantages of these stem cell types are briefly summarized in Table 1.

Autism spectrum disorder (ASD) encompasses a broad range of complex polygenic and multifactorial neurodevelopmental diseases diagnosed in early childhood, affecting social interaction, communication, interests, and behavior [8]. The worldwide prevalence of ASD is just under 1% in children with higher estimates in high-income countries and has markedly increased in the last decade probably due to rising awareness of the condition and improved criteria for diagnosis [9]. The core symptoms of ASD vary in degree of severity and are often accompanied by other concomitant diseases and conditions, including epilepsy, anxiety, obsessive compulsive disorder, intellectual disability, schizophrenia (SZ), and attention-deficit hyperactivity disorder (ADHD) [10,11]. Because of the heterogeneity of symptoms and severity of ASD, the disorder may be diagnosed in children at different ages [12]. Most children who are diagnosed with ASD at less than 3 years have retained their diagnosis, while approximately 9% of children who are diagnosed with ASD in early childhood may not meet criteria for ASD by young adulthood [12]. While its exact etiology remains unclear, ASD is known to have a strong genetic background, with family and twins studies suggesting a sizable heritability ranging from 50 to 90% [13,14,15]. The clinical heterogeneity of ASD is due to several factors: the complexity of its genetic profile (the involvement of multiple genes and the existence of different inheritance patterns); the existence of gene-environment interactions; the diverse nature genetic variants involved (ranging from common, rare, de novo and inherited in both coding and non-coding regions of the genome) [16,17]. ASD has been typically classified into non-syndromic (idiopathic, unknown genetic etiology) and syndromic (caused by a known genetic defect), based on a lack of or association to clinical manifestations outside of the autistic features, respectively [18,19]. In the “omics” era, the terms “syndromic” versus “non-syndromic” ASD fare slightly better under careful scrutiny and could be used with appropriate caveats. The use of this classification based on clinical observation demonstrated that it could led to erroneous conclusions that might be corrected retrospectively by means of “omic” studies. For example, rare syndromic mutations were reported in individuals with ASD who were thought not to have syndromic features based on a clinical exam, and, conversely, syndromic features became apparent only by a re-categorization of non-syndromic cohorts after genetic evaluation. Recently, it has become clear that a considerable overlap exists between molecular pathways and biological mechanisms detected in syndromic versus non-syndromic ASD. The term syndromic would be better substituted with “associated with known medical or genetic condition”. Nonetheless, in the description of studies on iPSCs and organoids, we will use the classical classification in syndromic and non-syndromic considering that the cited literature is based on these two groups.

Several autism-susceptibility or genetic risk factors identified in syndromic forms of ASDs involved genes with a role in different processes: the regulation of synaptic plasticity, cell adhesion, neural connectivity, dendritic trafficking, transcription, gene imprinting, chromatin remodeling, and neurotransmission [20,21,22,23,24]. Moreover, modifying risk factors contribute in delineating the observed phenotype. These factors include several pre- and postnatal environmental factors, including maternal smoking, parental age of birth, infection, alcohol consumption, pollution, pesticide exposure, and gestational complications, such as bleeding or diabetes [25,26,27,28,29].

As the multigenic origin of idiopathic ASD makes it difficult to model, iPSCs-based technologies from individuals with genetically complex etiologies have enabled the investigation of pathological mechanisms in disease-relevant human brain cell types and therefore offer a promising platform to test drugs in a safe environment with human genetic background and relevance [30]. Herein, we report advances in modeling ASD with iPSCs-derived technologies and their use in the drug discovery and the development of potential new treatments, as well as their relevance in clinical applications and precision medicine approaches (Figure 1).

The iPSC technology has led to insight into relevant cell types involved in disease mechanisms of ASD, offering a ground-breaking platform to study abnormalities in neurodevelopmental processes that can functionally alter neural circuitry [31]. The majority of iPSCs disease modeling studies use the conventional 2D monolayer culture models to recreate disease-relevant cell types. Two different main strategies are used for producing neural cell starting form iPSCs: the direct differentiation or the reprogramming. In the first case, iPSCs are firstly induced to neural progenitor cells (NPCs) and then differentiated by exposition to trophic factors (e.g., cAMP, BDNF, NT3, and GDNF) [32]. In the reprogramming strategy, iPSCs are transduced with lentiviruses carrying specific transcription factors and an antibiotic selection cassette for efficient conversion and selection [33]. Different iPSCs-derived phenotypes in NPCs, neurons and glial cells (mainly astrocytes and oligodendrocytes) have been already described for both non-syndromic and syndromic forms of ASD.

Three-dimensional self-organized neural cultures rely on the ability of iPSCs or, more in general, stem cell to self-organize in 3D structures, frequently named mini-brain, providing also required self-patterning factors for morphogenesis and differentiation [117]. These organoids are able to sum up sequential neurogenesis, migration, gliogenesis, and synaptogenesis of cortical development that cannot be studied in 2D cell cultures [118]. Being able to be maintained in vitro for up to two years [119], they could reach later stages of neurodevelopment compared to those that can be studied with 2D cultures and, thus, produce later-born neural cells, including astrocytes and oligodendrocytes [119,120].

One of the first study on brain organoid deriving from cells of patients affected by idiopathic ASD and characterized by macrocephaly [69] reported abnormalities in cell-cycle and synaptic growth. In particular, they noted an increment in the expression of the transcription factor FOXG1 that correlated with an increased production of inhibitory neurons. The study highlights the role of this gene and suggests it as a potential therapeutic target [69]. More recently, Schafer and colleagues [121] performed a large study, based on a selected cohort of macrocephalic ASD patients, in which they combined direct iPSCs-to neuron conversion, transcriptomic analysis and cerebral organoids generation. This strategy made it possible to clarify that ASD-related neurodevelopmental abnormalities are triggered by a specific temporal sequence of gene networks dysregulation. The study showed different results in forebrain organoids than in iPSC-derived neurons. In particular, ASD alterations were well recapitulated in organoids, while the direct conversion of iPSCs to mature neurons abolished ASD-associated phenotypes [121]. Owing to the fact that CHD8, a gene encoding for a chromatin-remodeling factor, is mutated in a subgroup of patients with ASD and schizophrenia (SZ), some authors performed an RNA-seq analysis of cerebral organoids derived from iPSCs that are heterozygous for a CHD8 knockout allele, and from isogenic controls [122]. The study showed that CHD8 regulates the expression of other genes implicated in ASD and SZ, in particular TCF4 and AUTS2, and a large overlap was observed for DEGs found by Mariani et al. [69], especially for genes involved in GABAergic interneuron development. Pathway analysis of DEGs revealed an enrichment of genes involved in regulating Wnt/β-catenin signaling, a druggable pathway [122].

Organoids were used also to study RTT pathogenesis. In particular, Mellios and colleagues [123] analyzed alterations in different neuronal processes, such as neurogenesis, differentiation, and migration, in MECP2-deficient and patient-derived cerebral organoids, while, more recently, Xiang et al. [124] performed a transcriptome analyses revealing a cell-type-specific impairment in specific regions of brain organoids carrying MECP2 mutations.

In a recent study performed by combining CRISPR/Cas9 genome editing and brain organoids, Blair et al. [125], investigating the “two hit hypothesis” model of cortical tuber formation, demonstrated that homozygous, but not heterozygous, loss of either TSC1 or TSC2 impairs the developmental suppression of mTORC1 signaling. Moreover, they demonstrated that the development of TSC phenotype in organoid could be prevented by treating them with rapamycin [125,126].

Brain organoids were also successfully obtained from patient affected by TS in a study that represents the first example of fused organoid approach for the in vitro study neuronal circuits involving different brain regions [127]. In particular, authors studied the migration pattern of inhibitory neurons during brain development using live imaging of TS-forebrain assembloids (fused cerebral organoids resembling the dorsal and ventral forebrain regions). They demonstrated that TS interneurons display a cell-autonomous defect consisting in overall delayed migration from the ventral to the dorsal forebrain [127]. The functional phenotype was rescued by treatment with nimodipine, a drug inhibiting the activity of L-type calcium channel.

Very recently, brain organoids were studied to go deep inside into the pathological mechanisms underlying AS [128]. The study demonstrated that UBE3A suppresses neuronal hyperexcitability via ubiquitin-mediated degradation of calcium and voltage-dependent big potassium (BK) channels. Thus, the use of a BK channels antagonist could normalize neuronal excitability ameliorating also seizure susceptibility [128].

The use of iPSCs as a tool has helped to unveil not only the etiology of ASD, but also aided the progress of the therapeutic areas. Mouse models have routinely been used to test drug candidates, but those identified to date have shown low efficacy in humans, thus limiting their translational potential [132,133]. Moreover, the development of drugs for ASD treatment is still challenging due to the high heterogeneity of the phenotypes, the limited understanding of its pathophysiology, and difficulties in modeling ASD both in vitro and in vivo.

Besides their value for the uncovering of cellular and molecular phenotypes underlying ASD, human iPSCs-based technologies have offered novel possibilities for personalized medicine and for screening drug candidates that can rescue normal phenotype and function. Among them, different iPSCs studies provided evidence that IGF-1 ameliorates some physiological deficits associated with ASD, both in non-syndromic [35,36] and syndromic forms [39,87,105]. In particular, this growth factor was able to restore impaired neuronal network [35], synaptic deficits [38,39,87], or neuronal morphological abnormalities [105] in iPSCs-derived neurons. The evidences coming from iPSCs-based studies in RTT and PMDS modeling have led to a stage II clinical trial on IGF-1 for the treatment of idiopathic and syndromic ASD with results expected in 2022 (https://ClinicalTrials.gov; protocols numbers: NCT01970345; NCT01894958; NCT01777542; NCT01525901; NCT01253317) [134,135]. Furthermore, gentamicin treatment on RTT iPSCs-derived neurons enhanced MECP2 protein levels and increased glutamatergic synapse, further confirming that patient-specific iPSCs can correct neuronal phenotypes [39]. Given MECP2 function at the epigenetic level, some authors planned to screen a library of 43 compounds with defined activity on epigenetic pathways in order to test the reversibility of the alterations observed in iPSCs-derived neurons harboring MECP2 duplicated gene. Among them, the histone deacetylase inhibitor NCH-51 was validated as a potential clinical candidate [41]. Intriguingly, in another study performed on iPSCs from MECP2-mutated neurons, phenotypic abnormalities were reverted by the treatment with selective inhibitors of HDAC6, the main α-tubulin deacetylase [71]. As mentioned above, other promising results were achieved by roscovitine that can reverse altered expression of TH and increased production of norepinephrine and dopamine displayed in iPSCs-derived neurons from TS individuals carrying a mutation in CACNA1C gene [48]. Additionally, the pharmacological intervention with rapamycin, mTOR regulator, successfully recapitulated the increased neuronal activity and morphological changes in TSC iPSCs-derived neurons, suggesting that these cells can also respond to existing drugs in vitro [84,86,125]. In this regard, it should be noted that clinical studies already showed encouraging improvements in TSC patients after treatment with rapamycin [136].

The aforementioned findings are very promising and suggest that disease-specific models based on iPSCs may represent an emerging and scalable platform for developing novel therapeutic agents targeting molecular mechanisms for ASD treatment (Figure 2). Since iPSCs are renewable, they can provide unlimited cellular resources for large-scale high-throughput screening (HTS). In this regard, one HTS was performed in iPSCs-derived cortical neurons from ASD individuals harboring SHANK3 mutations to search for compounds that rescue SHANK3-associated phenotypes by increasing its mRNA levels. Among 202 tested compounds, lithium and valproic acid were identified as agents with the best efficacy to correct SHANK3 haploinsufficiency, demonstrating that iPSCs can offer a novel cellular platform for developing specific disease-modifying treatments [88]. However, HTS platforms for drug discovery are still under development and require further standardization, including a homogeneity of cell types and a speeding up of the neuronal maturation by improving the differentiation protocols. Although the current advances in iPSC-based systems hold great promises for drug discovery, these technologies cannot completely replace animal models for preclinical evaluation of drugs.