Human-Derived Brain Models

As the incidences of chronic neurological diseases continue to increase, new methods are needed to screen therapeutic compounds more effectively.¹ Current estimates for the economic burden of neurological diseases are astoundingly high, totaling nearly 800 billion dollars in 2014, of which Alzheimer’s Disease (AD) accounts for at least 243 billion.¹ This cost is only expected to increase as the country’s elderly population will nearly double, from 43.1 million to 83.7 million, by 2050.² To reduce disease and economic burden, better therapies are needed to prevent and treat debilitating neuropsychiatric disorders.

However, drug development also faces economic challenges, with a single new drug costing upward of 800 million dollars.³ Major hurdles affecting the development of new neuropsychiatric drug therapies include the lack of suitable animal disease models, unknown etiologies, patient-specific differences, especially in spectrum disorders such as autism, and unanticipated drug toxicity in humans.^4,5 Positive preclinical results observed in animal models may not be recapitulated in humans due to discrepancies in brain physiology or genetic disease modeling.⁶ All these factors can lead to unexpected failures, with ∼90% failure rate of drugs that enter the clinical pipeline.⁷

Thus, we propose a model that is an ideal alternative to in situ brain tissue and can serve as an important tool in preclinical research and development: neurons and brain organoids developed from human induced pluripotent stem cells (hIPSCs). This review details a proposed pipeline for the use of hIPSC neuronal models in therapeutic drug discovery. We will begin with an introduction to how hIPSC neuronal models are currently used to model complex neuropsychiatric disorders and therapeutic responses. We will then address how hIPSCs can be combined with recent technologies to enable rapid and effective screening of drug therapies (Figure 1).

Abstract

Human-derived neurons and brain organoids have revolutionized our ability to model brain development in a dish. In this review, we discuss the potential for human brain models to advance drug discovery for complex neuropsychiatric disorders. First, we address the advantages of human brain models to screen for new drugs capable of altering CNS activity. Next, we propose an experimental pipeline for using human-derived neurons and brain organoids to rapidly assess drug impact on key events in brain development, including neurite extension, synapse formation, and neural activity. The experimental pipeline begins with automated high content imaging for analysis of neurites, synapses, and neuronal viability. Following morphological examination, multi-well microelectrode array technology examines neural activity in response to drug treatment. These techniques can be combined with high throughput sequencing and mass spectrometry to assess associated transcriptional and proteomic changes. These combined technologies provide a foundation for neuropsychiatric drug discovery and future clinical assessment of patient-specific drug responses.

hIPSC models of neuropsychiatric disorders

Since the discovery of hIPSCs, a wide array of neurological diseases has been modeled, including neurodegenerative diseases, such as AD, and neurodevelopment disorders, such as Autism Spectrum Disorders (ASDs).⁸ hIPSCs have also provided a unique tool to gain insights in rare neurologic disorders, such as Kleefstra Syndrome,⁹ Dravet Syndrome,¹⁰ and Smith Lemli Opitz Syndrome.¹¹ These hIPSC models mimic brain development and pathology more closely than human immortal cancer cell lines and primary animal cell culture; this will likely allow for more accurate predictions of patient responses.¹²

In addition to drug discovery, hIPSCs can be used clinically to evaluate patient-specific drug responses, for example, drug-induced changes in neural activity. These hIPSC-derived models can be combined with high throughput RNA sequencing technologies to define transcriptomes associated with specific patient populations, eventually leading to personalized therapeutic intervention.^13,14 Furthermore, proteomic, metabolomic, and lipidomic signatures can be used for patient classification and analysis of drug efficacy.^15–18 In neurodegenerative disorders, biomarkers will allow for diagnosis before disease onset, and aid in developing therapies that prevent neuronal loss, rather than inhibit further loss after disease onset.

The following sections describe how hIPSCs have been used to model pathology and therapeutic responses in AD and ASDs as representative neurodegenerative and neurodevelopmental disorders, respectively. We will then use these two examples to demonstrate how our proposed experimental strategies can advance drug discovery for complex neuropsychiatric conditions.

Alzheimer’s Disease

AD is the most common cause of dementia, accounting for up to 80% of all dementia cases.¹⁹ In humans, AD results in extracellular amyloid plaques, intracellular neurofibrillary hyperphosphorylated tau tangles, and synapse loss preceding neurodegeneration.²⁰ There is difficulty recapitulating these key pathological features in mouse models, which require expression of human amyloid precursor protein (APP)^21,22 and human tau²³ for formation of disease-associated amyloid plaques and neurofibrillary tangles. Thus, a human model has the potential to greatly advance therapeutic drug discovery for AD.

Several hIPSC models have been developed to model both sporadic and familial forms of AD (sAD and fAD).^24,25 Beginning in 2011, hIPSC-derived neurons were used to model early onset fAD in patients with presenilin mutations.²⁶ Presenilin-1 and -2 are the major components of γ-secretase, which cleaves APP into amyloid-β (Aβ) peptides.²⁷ fAD-associated presenilin mutations increase the ratio of the amyloidogenic Aβ-42 peptide to the shorter Aβ-40 peptide.^28,29 This study demonstrated that hIPSC-derived neurons from fAD patients secrete more Aβ-42, and that γ-secretase inhibitors (GSIs) decreased the levels of secreted Aβ-42.²⁶

In 2011, another study similarly demonstrated the ability of γ- and β-secretase inhibitors to lower Aβ production in control hIPSC-derived neurons.³⁰ β-secretase functions in the first step of APP processing, preceding γ-secretase, in the production of amyloidogenic Aβ peptides.³¹ These findings were extended to hIPSC-derived neurons from a sAD patient and fAD patients with APP duplications, all of which exhibited disease-associated increases in Aβ peptide, active GSK-3β, and phosphorylated tau.³² While both β- and γ-secretase inhibitors decreased secreted Aβ, only β-secretase inhibitors decreased GSK-3β and phosphorylated tau.³² Notably, hIPSC-derived neurons from another sAD patient did not exhibit disease-associated phenotypes, highlighting the benefit of hIPSCs to reveal patient-specific differences and identify patients likely to benefit from a given therapy.³²

In addition to confirming the ability of secretase inhibitors to reverse AD phenotypes, hIPSC-derived neurons have also been used to overcome the limitations of cell lines for secretase inhibitor optimization. For example, GSIs decrease Aβ peptide production in APP-overexpressing cell lines, but have failed in human clinical trials.³³ This failure prompted researchers to pursue more physiologically relevant AD models, fibroblasts and hIPSC-derived neural progenitor cells and neurons from FAD patients with presenilin-1 mutations.⁴ This study revealed differential drug efficacy in specific cell types, with the greatest Aβ reduction observed in hIPSC-derived neurons.⁴ Importantly, this study also revealed that the IC₅₀ of the GSI, semagacestat, was five times higher in hIPSC-derived neurons than APP-overexpressing cell lines, perhaps contributing to its clinical failure.⁴

Additionally, hIPSC-derived neurons have been used to identify new Aβ lowering compounds through a pharmaceutical library screen and chemiluminescent measurement of the resulting Aβ levels.³⁴ This study led to the development of an Aβ lowering cocktail that significantly reduced Aβ levels in hIPSC-derived neurons from fAD patients with presenilin mutations and to a lesser extent, fAD patients with APP mutations and sAD patients.³⁴ This study used a rapid neuronal induction protocol to screen for Aβ levels within 10 days of neuronal differentiation, thus accelerating drug screening and identification of promising therapies.³⁴ They also ensured that reduced Aβ levels did not result from increased neuronal cell death.³⁴ However, the authors note that this cocktail still cannot move to clinical trials as IPSC-derived neurons alone do not address whether the drugs cross the blood-brain barrier (BBB).³⁴ Thus, cultures that more closely mimic the brain environment, for example, by incorporating vascular endothelium,^35–37 have the potential to advance therapies from drug discovery to clinical trials.

hIPSC-derived neurons have also been used to validate protherapeutic effects of alternative AD treatments in a human model, such as nobiletin and apigenin. Nobiletin is a compound found in citrus peel that exhibits antidementia effects in mice.^38,39 Nobiletin reduced intracellular and secreted Aβ in hIPSC-derived neurons with an fAD-associated presenilin-1 mutation.⁴⁰ The observed protherapeutic effect is due to increased expression of the Aβ-degrading enzyme, neprilysin.⁴⁰Similarly, apigenin, the polyphenolic compound found in celery, parsley, and artichoke, exhibited neuroprotective effects in hIPSC-derived sAD and fAD neurons.⁴¹ Apigenin’s anti-inflammatory properties reduced nitric oxide production, increased neurite length, and rescued cell viability.⁴¹ Together, these studies highlight the use of hIPSC-derived neurons to validate and optimize AD therapies in a human model.

Brain organoids in AD research

While hIPSC-derived neurons show promise for drug discovery, brain organoids capture key pathological features that can only be observed in a three-dimensional (3D) matrix that mimics the complex tissue microenvironment. For example, AD-derived neurons exhibit increased secretion of insoluble Aβ, but fail to form amyloid plaques.^42,43 Amyloid plaques reside in the extracellular matrix, thus requiring a 3D model for retention. Additionally, Aβ plaques are hypothesized to initiate the formation of tau neurofibrillary tangles, eventually leading to cell death.⁴⁴ This is consistent with observations in AD-derived neurons, which fail to retain Aβ plaques. These AD-derived neurons exhibit increased tau phosphorylation, but they do not form intracellular tau neurofibrillary tangles.⁴² Furthermore, cell death has not been documented in 2D AD-derived neurons.

To test the Aβ hypothesis of AD pathogenesis, 3D brain models are needed to capture Aβ plagues. The first 3D human AD model used hydrogels, which promoted Aβ retention and intracellular tau neurofibrillary tangle formation, although cell death was not reported.⁴² Using this model, both γ- and β-secretase inhibitors reduced Aβ plague formation and tau phosphorylation.⁴² However, the GSK-3β inhibitor, 1-Azakenpaullone, specifically decreased tau phosphorylation, but did not affect Aβ production.⁴² These results support a disease model in which GSK-3β phosphorylates tau, but does not phosphorylate APP to increase Aβ production.⁴⁵ By contrast, cerebral organoid models of AD exhibit amyloid plaques, intracellular tau neurofibrillary tangles, and increased cell death.⁴⁶

By capturing Aβ plaque formation, tau neurofibrillary tangles, and neurodegeneration, 3D brain organoid models provide unique opportunities for neurodegenerative disease research and drug discovery. For example, recent data indicate that mutant tau exhibits prion-like properties, leading to stereotyped propagation of tau neurofibrillary tangles with disease progression.⁴⁷ Similarly, in mice, mutant APP results in the spread of amyloid plaques to distal brain regions.⁴⁷ Brain organoids have the unique ability to monitor the spread of insoluble plaques and tangles, and test for drugs that prevent spreading.

To monitor protein aggregation in brain organoids, they can be injected with toxic Aβ or tau mutants. Alternatively, chimeras of unaffected and AD-derived brain organoids can be created. For example, a recent study used the propensity for brain spheroids to fuse with one another to monitor interneuron migration between brain spheroids.⁸ By similarly allowing for fusion of AD and unaffected brain organoids, one could monitor the propagation of pathological features and the resulting cell death. This method can also be used to fuse brain region-specific organoids⁴⁸ and assay for drugs that block the spread of protein aggregates between brain regions. Thus, brain organoids recapitulate key pathological features missing in hIPSC-derived neurons. By more accurately modeling the disease state, brain organoids have the potential to further advance drug discovery.

Autism Spectrum Disorders

ASD is a heterogeneous, neurodevelopmental condition that is characterized by a complex behavioral phenotype. Key features of the disorder include deficits in social communication and restricted, repetitive patterns of behavior that exist on a continuum of severity.⁴⁹ ASD is one of the most common developmental disorders in the United States, affecting 1 in 59 children, with a male to female diagnosis ratio of 4:1.⁵⁰ Due to the disabling nature of the disorder, children often require special accommodations. Estimates for the lifetime cost of supporting a child who has ASD with an intellectual disability are over 2 million dollars per child, citing educational services, productivity loss from parents, and higher frequency of health care office visits and prescriptions.^51,52

Unlike AD, ASD does not have a defining cellular pathology. Research is currently focused on discovering convergent mechanisms that govern synaptic changes seen in ASD. Postmortem brain samples from idiopathic ASD cases exhibit synaptic alterations, including changes to the density of dendritic spines, the primary sites of excitatory synaptogenesis.^53,54 Much research focuses on syndromic, monogenetic ASD disorders, for example, fragile X syndrome (FXS) and Rett Syndrome (RTT).⁵⁵ FXS results in increased excitatory synapses,⁵⁶ while excitatory synapse formation in RTT depends on MeCP2 (methyl-CpG-binding protein 2) gene dosage, with MECP2 deletion reducing excitatory synapse formation and MECP2 duplication increasing excitatory synapse formation.⁵⁷

Drug development for ASDs is particularly challenging due to this heterogenous nature and future research must be aimed at identifying the molecular pathways that cause different synaptic alterations.⁵⁸ A particular benefit of hIPSC brain models is the ability to classify patients based on phenotypic presentation and develop therapies for specific patient populations within the spectrum.

Using an hIPSC-FXS model of ASD, researchers screened over 5,000 drugs for compounds that increase FMR1 gene expression.⁵⁹ In FXS patients, cognitive disability is caused by the loss of fragile X mental retardation protein (FMRP) via silencing of the FMR1 gene^59,60 In this study, an assay was developed that allowed for FMRP protein detection in hIPSCs derived from four patients who had a completely silenced FMR1 allele; this factor is crucial because positive hits (increased FMRP) indicate FMR1 gene reactivation rather than increases in translation.⁵⁹ This study used a time-resolved FRET assay to measure FMRP levels of lysed cells in a 1,536-well plate.⁵⁹ Of the >5,000 compounds, only 4 compounds (Protoporphyrin IX, SB216763, Geliomycin, and Tibrofan) produced a dose-dependent increase in FMRP.⁵⁹

Future studies are needed to determine the mechanism by which these compounds reactivate FMR1 gene expression, although SB216763 is a GSK3β inhibitor known to improve hippocampus-dependent learning and neurogenesis in the FMR1knockout mouse, where it is not possible to reactivate FMR1 expression.^59,61 Another GSK3β inhibitor, lithium chloride, did not reactivate FMR1 expression, suggesting that SB216763 reactivates FMR1 independent of GSK3β inhibition.⁵⁹ Two of these drugs, Geliomycin and Tibrofan, are FDA-approved as an antibiotic and disinfectant respectively, demonstrating the use of hIPSCs to identify novel and repurposed drugs for the treatment of neuropsychiatric disorders.⁵⁹

A similar study evaluated the effects of 50,000 compounds on hIPSC-derived neurons from FXS patients to assess FMR1gene reactivation.⁶⁰ Positive hits in the assay were defined as drugs that increased FMRP cytoplasmic protein levels three or more standard deviations above the negative control; this procedure was done via high throughput imaging and analysis techniques.⁶⁰ Since DNA hypermethylation silences FMR1 gene expression, this study used 5-aza-2′-deoxycytidine, a known inhibitor of DNA methyltransferase DNMT1, as a positive control for evaluating FRMP increase.⁶⁰ After the primary exploration, 790 compounds were chosen to undergo dose–response curve experiments; only a few compounds showed increased FMRP expression before cytotoxicity, and all identified compounds were more toxic than the positive control.⁶⁰While this study was not successful in finding a lead compound, it did establish a high content image-based assay for drug screening in a population of patients who do not yet have an FDA-approved treatment option.

In a proof-of-concept pharmacogenomic study, Darville et al. demonstrated how hIPSC-derived neurons can be used to find new treatment options for ASD patients using a SHANK3 haploinsufficiency model.⁶² Loss-of-function mutations of the SHANK3 gene affect ∼2% of ASD patients who present with moderate to severe intellectual disability.⁶³ The SHANK3 protein is a scaffolding molecule localized to the postsynaptic density of excitatory synapses; it mediates the interaction between various glutamate receptors and the actin cytoskeleton, indicating an important role for regulation of synaptic plasticity in disease pathogenesis.⁶⁴

Since SHANK3 mutations in ASD patients only affect one allele, transcription of the second allele can be upregulated and SHANK3 mRNA levels can be increased.⁶² The authors screened 202 compounds in 4 patient-derived hIPSC neuron lines using automated high throughput mRNA quantification and then followed up with high content image analysis of 16 selected compounds that demonstrated a dose–response curve.⁶² Lithium and valproic acid, two FDA-approved compounds, increased SHANK3 at both the mRNA and protein level in SHANK3 haploinsufficient hIPSC-derived neurons.⁶² The authors were also able to confirm the clinical efficacy of lithium in one of the ASD patients who donated their hIPSC line,⁶² demonstrating the feasibility of using hIPSC models for personalized therapeutics.

RTT is also used as a genetic model of ASD in hIPSC-derived neurons. This neurological disorder is caused by loss-of-function mutations in the X-linked gene that encodes the epigenetic regulator protein MeCP2.⁶⁵ Neurons derived from RTT patients with an MeCP2 deletion had fewer excitatory synapses and altered electrophysiological profiles, including decreased frequency and amplitude of excitatory postsynaptic currents.⁶⁶ Insulin-like growth factor 1 (IGF1) is currently being explored as a potential therapeutic for RTT patients; it has been shown to increase the number of glutamatergic synapses and increase neurite length back to baseline in RTT-hIPSC derived neurons.^66,67 Clinical studies support the safety and efficacy of recombinant human IFG1 treatment in RTT patients; treatment is associated with significant improvements in disease severity including improved social and cognitive measures.⁶⁸ IGF1 has also been explored in clinical trials for the treatment of other monogenetic neurodevelopmental disorders, including promising phase 2 trials in both FXS and Phelan-McDermid Syndrome patients.⁶⁹ Thus, hIPSC models of ASD are already being used to develop therapies for specific patient subgroups.

Brain organoids in autism research

Similar to AD, brain organoids are providing further insights into the temporal development of autism pathology and unique opportunities for drug discovery. Brain organoids develop on similar timescales to the fetal brain in utero, with synaptogenesis occurring during mid-fetal gestation.⁸ Likewise, excitatory synapses shift to specialized dendritic spines after ∼8 months of organoid culture, at stages that resemble perinatal brain development.⁸ Transcriptional correspondence between brain organoids and the human fetal brain after similar developmental times make brain organoids an ideal system to study both transcriptional and cellular changes associated with specific periods of brain development.⁸

For example, IPSC-derived telencephalic organoids from ASD patients with increased brain size have elevated RNA levels of the transcription factor, FOXG1, resulting in increased production of inhibitory neurons and synapses.⁷⁰ Increased GABAergic cell fate correlated with symptom severity, suggesting that FOXG1 may be an early driver of altered neural circuitry associated with autism pathology.⁷⁰ Dysregulation of GABAergic interneurons has also been seen in other ASD brain organoid models, notably in the heterozygous knockout of chromodomain helicase DNA-binding protein 8 (CHD8), a gene commonly mutated in ASD patients.⁷¹ Brain organoids with the CHD8 heterozygous knockout displayed increased expression of two genes that regulate GABAergic interneuron development.⁷¹ Intriguingly, patients with CHD8 mutations exhibit macrocephaly, suggesting that increased GABAergic production may be a common feature of ASD patients with larger-than-average head size.

Furthermore, RTT brain organoids have enabled research into prenatal roles of MECP2, namely altered neural progenitor cell proliferation and neurogenesis, whereas previous research has primarily focused on postnatal roles of MECP2.⁷² Specifically, brain organoids allowed the researchers to visualize increased area of ventricles, which form in brain organoids, but not monolayer cultures.⁷² These newly discovered transcriptional and cellular changes can be used in drug-screening applications, for example, to identify drugs that regulate interneuron differentiation or neural progenitor cell proliferation.

The following sections explore how high content imaging and multi-well microelectrode arrays (MEAs) can be used to assess drug-induced changes to neuronal morphology and function in both hIPSC-derived neurons and brain organoids. These drug screening assays will further therapeutic discovery for the treatment of neuropsychiatric disorders, such as AD and ASD.

Drug screening for neuroprotective effects

Automated high content imaging and analysis

Decreased neurite length is a common phenotype of ASD-derived^9,14,73–75 and AD-derived^41,76 neurons. Thus, screening for drugs that promote neurite length could reveal potential therapeutic candidates for both neurodevelopmental and neurodegenerative disorders. High content imaging systems can be used to rapidly screen neurite length following drug treatment.⁷⁷ In this platform, fixed neurons are fluorescently immunostained with early neurite markers, such as β-III tubulin/TUJ1, together with a nuclei marker, such as Hoechst or DAPI.^78–80 Following nuclei detection, the system performs neurite segmentation, providing readouts such as neurite length, number of neurites, and neurite branching.^80,81 Automated analysis can be combined with the high content imaging system^78–81 or performed by separate software/freeware, such as CellProfiler.^82,83

Using high content imaging and analysis, Sherman and Bang screened several bioactive compound libraries to identify positive and negative regulators of neurite outgrowth.⁸⁰ Compounds promoting neurite outgrowth included kinase inhibitors, such as inhibitors of the myosin kinases, ROCK and MLCK, and GSK3β inhibitors.⁸⁰ While ROCK inhibition increases neurite outgrowth, it does not affect electrophysiological maturation in hIPSC-derived neurons,⁸¹ demonstrating the need for a comprehensive therapeutic screen that includes both morphological and functional neuronal characterization through techniques such as MEA. Other compounds that promote neurite outgrowth include regulators of steroid hormone receptors and neurotransmitter receptors.⁸⁰

This study also identified novel regulators of neurite outgrowth, including some fatty acids.⁸⁰ Importantly, some compounds had opposite effects on neurite outgrowth than those reported for immortalized rodent cell lines, including the mouse neuroblastoma Neuro-2A cell line and rat pheochromocytoma PC12 cell line,^80,84 providing further evidence that a physiologically relevant human model is needed for neuropsychiatric drug screens.

Importantly, the neurite-promoting effects of these compounds also need to be validated in disease models. For example, GSK3β inhibitors might have an increased effect on neurite outgrowth in AD models, where there is increased GSK3β activation.¹⁹ Alternatively, specific compounds may only have a therapeutic effect in disease models, as is illustrated by NC009-1-mediated increase in neurite outgrowth in hIPSC-derived neurons from fAD patients with APP mutations but not in control hIPSC-derived neurons.⁷⁶ Similarly, specific patients may exhibit different responses. In the case of ASD patients, a rapid screen of neurite length to identify patients with defective neurite outgrowth may help to identify patients likely to benefit from neurite-promoting therapies.

In addition to neurite defects, AD and ASD also present with synaptic abnormalities that can be rapidly analyzed using a high content imaging platform, where co-localization of pre- and postsynaptic markers identifies synapses.⁸² For example, co-localization of the presynaptic vesicle marker, synapsin-1, with the postsynaptic scaffold protein PSD-95 identifies excitatory synapses, whereas co-localization between synapsin-1 and gephyrin identifies inhibitory synapses.⁸²

Synapse loss is observed before neurodegeneration in AD, thus drugs that promote synapse formation and/or prevent synapse loss could be attractive therapies.⁸⁵ In contrast to synapse loss in neurodegeneration, synaptic alterations associated with neurodevelopment are more varied. Postmortem brain cortex from idiopathic ASD patients exhibits increased excitatory synapses.^53,54 However, other ASD-associated mutations, such as RTT MeCP2 deletion, reduce excitatory synapse formation.⁵⁷ Decreased excitatory synapses and increased inhibitory synapses have also been observed in both hIPSC-derived neurons and brain organoids from ASD patients with larger-than-average brains.^66,86 This heterogeneity of synaptic abnormalities demonstrates how hIPSC-derived neurons can be used to assess patient-specific phenotypes and develop personalized therapies.

Finally, high content imaging systems can be catered for rapid analysis of a variety of fluorescent-based, and even brightfield, images. For example, a study aimed at identifying tau-lowering compounds as AD therapies, conducted high content imaging and analysis to detect tau levels relative to β-III tubulin.⁷⁸ After screening the 1,280 compounds in the Library of Pharmaceutically Active Compounds (LOPAC), they identified adrenergic receptor agonists as the top tau-lowering compounds.⁷⁸ Thus, high content imaging and analysis can be catered to assess disease-specific phenotypes, such as tau levels for AD,⁷⁸ FMRP levels in FXS,⁶⁰ and SHANK3 levels in SHANK3 haploinsufficient ASD patients.⁶²

When using high content imaging to analyze drug-induced changes to neurons, there are several important considerations. One primary consideration is to first determine whether the drug negatively impacts cell health and viability, in which case the observed effects to neurons may be caused by cellular stress responses rather than specific drug-induced changes to neuronal physiology. Multiple methods have been used for automated analysis of cell health, including abnormal nuclei,⁸⁰ viability dyes, such as calcein-AM,^79,87 and caspase assays for apoptosis.⁸⁷ Furthermore, 384-well cell culture plates can negatively impact cell viability and consistency across wells due to media loss through evaporation.⁸⁸ To overcome this limitation, researchers developed a ferromagnetic micro-raft array, where an array of 1,600 microrafts are cultured together and then magnetically transferred to a 384-well plate.⁸⁸ This technique improved neuronal viability, allowing for robust and reproducible drug screening.⁸⁸ Additionally, cell and organoid cultures are being optimized for 1,536-well glass-bottom plates, further increasing the throughput of high content imaging systems.^89,90

Multi-well MEA analysis of neural activity

In addition to high content imaging, MEA technology has emerged as a powerful tool to analyze basal and evoked neural activity in currently up to 96 samples. The MEA measures the extracellular field potential corresponding to action potential formation.⁹¹ Notably, this technique requires longer maturation than is needed for assessment of neurite length, with spontaneous action potentials occurring after ∼1–2 months of monolayer hIPSC-derived neuronal culture and ∼3 months or more in hIPSC-derived brain organoids.^8,92,93

However, techniques to rapidly induce physiological maturation can be used to accelerate drug screening. For example, in 2D, forced expression of neuronal transcription factors, NeuroD2 or neuroligin-2, can be used to accelerate neuronal maturation, with neural activity occurring after ∼1 week of neuronal induction.^94,95 Furthermore, optogenetics can be used to increase neural activity and potentially accelerate maturation of hIPSC-derived neurons and brain organoids,^96,97 allowing for more rapid assessment of drug-induced changes to neural activity.

Using MEA analysis, reduced spontaneous firing rate has been observed for several idiopathic cases of ASD.^14,98,99 However, ASD has also been associated with increased hyperactivation and epilepsy.^100,101 These differences are consistent with patient-specific synaptic differences in ASD, and emphasize the need to identify drug therapies catered to specific patient populations. MEAs provide a powerful tool to capture epileptiform neural activity and screen for antiseizure medications.¹⁰² MEA recordings of control hIPSC-derived neurons revealed increased epileptiform activity in response to proconvulsant drugs and decreased activity in response to seizure medication.¹⁰³ Thus, using MEAs, hIPSC brain models can be used to monitor activity signatures associated with epilepsy and other neuropsychiatric disorders.¹⁰³

In the case of neurodegenerative disorders, hIPSC-derived neurons gene-edited to model triple tau mutations in AD exhibit increased activity, consistent with AD-associated neuronal hyperactivity.^104,105 Hyperactivation is an early indicator of AD, preceding amyloid plaque formation and neuronal silencing,¹⁰⁵ and can surprisingly be treated with GABA receptor antagonists in mice.¹⁰⁶ It will be necessary to confirm whether GABA receptor antagonists similarly rescue neuronal function in human brain models.

While there are currently no published large-scale MEA drug screens of hIPSC-derived neuronal function, this is partially due to optimization of control hIPSC-derived neurons for accelerated maturation and reduced variability in neuronal function between cultures.^95,107 However, given the ability to record temporal changes in neural activity, MEA technology can assess drug-induced changes in neural activity at specific stages of neurodevelopment and disease progression. This will be particularly useful for identifying therapies that function before synapse loss and those that restore normal activity following synapse loss and subsequent neurodegeneration.

Concluding remarks and important considerations in hIPSC drug screens

While this review focused on neuronal morphology and function, other brain cells including glia, such as astrocytes, microglia, and oligodendrocytes, also contribute to neuropsychiatric disease pathology. A recent study using hIPSC-derived AD-associated astrocytes and microglia demonstrated defective Aβ clearance.¹⁰⁸ Furthermore, microglial activation and neuroinflammation contribute to neurodegenerative and neurodevelopmental pathogenesis.^109,110 A specific advantage of the hIPSC model is the ability to derive specific brain cell types (neurons, astrocytes, and microglia) from either controls or affected individuals to address the contribution to disease pathology. These different hIPSC-derived cells can also be used to address how specific drugs differentially affect specific cell types. Furthermore, whole brain organoids can also be used to address drug impact in a complex tissue environment. For example, a recent study validated potential inhibitors of Zika virus infection in both human brain organoids and the adult mouse brain.¹¹¹

While hIPSC models provide an ideal opportunity to screen for on-target drug effects, they are currently limited in their ability to measure drug distribution and BBB penetrance. However, techniques, such as 3D bioprinting, allow for the introduction of blood vessels to organoid models,³⁷ suggesting that we will soon be able to address BBB drug penetrance in hIPSC brain models. Furthermore, hIPSC-animal chimeras can be used to evaluate drug effects in a complex brain environment that includes disease-associated neuroinflammation, and the BBB to limit drug availability.³⁶ These increasingly more accurate models of human brain development will enable a progression of drug screening techniques, beginning with rapid and high throughput screens to identify therapeutic candidates of interest and proceeding to evaluate their effects in engineered human tissues¹¹² or hIPSC-animal chimeras.³⁶

To read this article in its entirety and access the references click here.

ASSAY & Drug Development Technologies, published by Mary Ann Liebert, Inc., offers a unique combination of original research and reports on the techniques and tools being used in cutting-edge drug development. The above article was first published in ASSAY and Drug Development Technologies with the title “Human-Derived Brain Models: Windows Into Neuropsychiatric Disorders and Drug Therapies”. The views expressed here are those of the authors and are not necessarily those of ASSAY and Drug Development Technologies, Mary Ann Liebert, Inc., publishers, or their affiliates. No endorsement of any entity or technology is implied.