Abstract
Nicholls J, Gu J, Chen Z, Liu Z, Antonic-Baker A, Javaid MS, Moore E, Zhu H, Altalhi A, Wright DK, Sumer H, O'Brien TJ, Kwan P, Rollo B. Epilepsia. 2025 Nov;66(11):4554–4572. doi:10.1111/epi.18560. Epub 2025 Jul 28. PMID: 40719753; PMCID: PMC12661287. Objective: Current preclinical epilepsy drug screening relies on animal models that poorly reflect human neurophysiology, leading to high failure rates in clinical translation. We aimed to establish a human in vitro model using human-induced pluripotent stem cell (hiPSC)-derived cortical neurons cultured on multielectrode arrays (MEAs), capable of generating precisely controlled after-discharges (ADs) through electrical stimulation. We optimized stimulation parameters to evoke epileptiform-like hypersynchronous events and validated the model using six approved antiseizure medications (ASMs). Methods: hiPSCs were rapidly differentiated into NGN2 cortical neurons and co-cultured with astrocytes on a 12-electrode, 24-well MEA. Network activity was tracked weekly. Upon maturation, biphasic voltage stimuli (400-2000 mV, 10 pulses at 100 Hz, 100 μs phase width) were applied in 100 ms trains to induce ADs. Stimulation intensity was increased until a maximum spike count per burst was reached. The timing of the stimulating interburst interval (IBI) was shortened from 10 to 1 s. We tested six ASMs with distinct mechanisms of action for their ability to attenuate induced ADs, as measured by the area under the curve (AUC) of spikes within bursts. Results: A ±1000 mV stimulus was sufficient to evoke robust ADs; higher voltages caused network instability without enhancing response strength. The maximum hypersynchronous bursting rate was observed with 2 s IBIs, whereas attempts to induce more frequent events using 1 s IBIs led to desynchronization and a reduction in burst frequency below baseline. Phenytoin, perampanel, clonazepam, and lamotrigine significantly reduced AUC within 5 min in a concentration-dependent manner. Vigabatrin and levetiracetam required longer preincubations: AUC was reduced after 6 h for levetiracetam and at 24 h for vigabatrin. Significance: We present a novel hiPSC-derived, electrically induced in vitro model for screening ASM candidates. This approach captures human-relevant epileptiform dynamics, allows fine control over stimulation parameters, and enables testing of diverse drug mechanisms. Its compatibility with high-throughput platforms makes it a promising tool for ASM discovery and personalized treatment strategies.
Commentary
Introduction
The maximal electroshock (MES) test is a model of electrically evoked acute seizures, which was first introduced in cats by Merritt and Putnam in 1937. Almost 90 years later, MES tests are still widely used (in rodents) and have been included in the US Epilepsy Therapy Screening Program's panel of preclinical tests for ASM candidates. 1 It is commonly accepted that the MES test models human generalized tonic-clonic seizures. However, this view is overly simplistic; for example, the test failed to predict the efficacy of levetiracetam in treating generalized tonic-clonic seizures in patients with idiopathic generalized epilepsy.
Over the past two decades, greater recognition of the Three R's (Replacement, Reduction, and Refinement) as guidelines for animal research, better understanding of the limitations of animal models, and the costs of animal breeding, have all led to the integration of human-derived preparations into drug development. Among these, human-induced pluripotent stem cells (hiPSCs) recapitulate an individual's clinical phenotype and drug responsiveness in a dish and have the capacity for almost unlimited proliferation. The use of hiPSC technology in ASM development is relatively limited. Exceptions include research into the effects of mTOR inhibition in tuberous sclerosis models. 2
The featured paper by Nicholls and colleagues 3 describes the development of an in vitro MES system (iMES) and its validation with six ASMs. The system consists of hiPSC-derived cortical neurons and primary human astrocytes of fibroblast origin. hiPSCs have previously been used as in vitro models of seizures (eg, by Odawara et al 4 ), but “epileptoform” activity was induced by adding chemoconvulsants to the culture medium, or ASMs were not systematically screened. In the present study, the cells were co-cultured on multielectrode arrays that generate controlled after-discharges (ADs) via electrical stimulation. The investigators first optimized the timing of establishing stable connectivity using synchronized network bursts and stimulation parameters, then tested the ability of ASMs, at three concentrations each, to suppress the induced ADs. The duration of incubation was defined by the ASM putative mechanism of action: five minutes for phenytoin, lamotrigine, perampanel, and clonazepam, whose main putative mechanisms of action involve direct modulation of ion channels or postsynaptic glutamate receptors; and longer periods for vigabatrin and levetiracetam, which are considered to act more indirectly. The measured parameter was the area under the curve (AUC) of spikes within bursts, which allowed the establishment of dose–response profiles. Efficacy was assessed by the ASM's ability to attenuate induced ADs.
As expected, the effects of phenytoin, perampanel, clonazepam, and lamotrigine were rapid, whereas 6 and 24 h were required for the activities of levetiracetam and vigabatrin, respectively. The longitudinal recordings allowed the detection of a decay in levetiracetam's effect by 24 h. The dose–response relationships were significant with phenytoin, clonazepam, perampanel, and lamotrigine.
Strengths of the study include the characterization of cell maturation via both electrophysiological recordings and immunohistochemistry, careful optimization of the stimulation protocol, and testing of ASMs with distinct mechanisms of action at several concentrations each. The authors highlighted the advantages of the electrical stimulation model over those involving pharmacological stimulants; the latter were associated with inconsistent effects of phenytoin, and the changes in network activity after the stimulus required a few days, whereas those observed with electrical stimulation were immediate. A rapid return to baseline is important for repeated testing and high-throughput screening, including for drug combinations and rational polytherapy. Caution is required, though, when valproic acid is considered: a key process in the induction of pluripotency is histone deacetylase inhibition for chromatin decompaction, which was also a part of the cited protocol. Theoretically, valproic acid, also a histone deacetylase inhibitor, may facilitate reprogramming of differentiated cells into induced pluripotent stem cells. This might occur if sodium valproate is used at ≥0.5 mM (166 μg/mL of the sodium salt; slightly above the therapeutic concentrations) and is left in the medium for extended periods.
One issue with the current model is sensitivity: It is not surprising that the model was not sensitive to levetiracetam and that the effects of this ASM were delayed. However, the concentrations of clonazepam that led to a 57% change in the AUC were well above the therapeutic range (1 µM vs 0.063-0.253 µM, respectively), and lamotrigine's effect was statistically significant only at 33 µM (therapeutic range: 8.04-35.45 µM). Alternative solvents may be considered, as DMSO could have confounded the results, and co-cultures that include several neuronal types and additional cells may improve the translational value of the findings.
As highlighted by the authors, the iMES platform offers exciting opportunities. First, the cultures and stimulation protocols can be designed to mimic epileptogenesis by developing spontaneous “epileptic” activity. Second, the cells were obtained from a volunteer with no known epilepsy. However, if fibroblasts are genetically edited to carry genetic mutations or are obtained from people with epilepsy, the system can be used for personalized testing of potential treatments. Epilepsy does not have to be genetically characterized for personalized treatment selection. In addition, some mutations that affect nonneuronal cells, for example, those in the GLUT1 gene, are less relevant to the current system. Third, the cells may be cultured as organoids that contain multiple cell types, instead of 2D cultures, to exhibit tissue-like architecture and form higher-order cell-cell interactions. 5 The current system includes astrocytes, which are important for neuronal maturation and survival, but additional cell types, for example, microglia, can improve the system performance and expand the potential therapeutic targets when screening ASM candidates. The authors proposed adding a blood–brain barrier component to the system, but current models generally reflect the healthy barrier and not the heterogeneity that characterizes the epileptogenic brain.
The ILAE/AES Joint Translational Task Force has recently suggested that in vitro epilepsy systems “must be capable of producing epileptiform activity, which may include seizure-like events, interictal-like patterns, status epilepticus (SE)-like activity, or cellular bursting discharges.” 6 The model developed by Nicolls et al meets this definition. As with all other in vitro models, the iMESwill have to be combined with other in vitro tests and with in vivo models, but it can provide an effective early screening platform. Even in vivo, the efficacy of almost all currently used ASMs was predicted only when MES testing was combined with other seizure models.
