Modeling Epilepsies with Human Brain Organoids: Recent Advances and Ongoing Challenges Authors: Miranda P Walker, Jack M Parent Published: 2026 ABSTRACT AbstractEpilepsy is a common and heterogeneous group of neurologic disorders characterized by recurrent unprovoked seizures. With a diverse array of genetic and acquired etiologies, the mechanisms linking insults to epileptogenesis and network hyperexcitability remain poorly understood. Although animal models have provided critical insights into seizure generation and disease mechanisms, differences in human cortical development and limitations in recapitulating spontaneous seizure phenotypes observed in patients highlight the need to develop complementary human-based models. Human pluripotent stem cell-derived brain organoids, including patient-derived and CRISPR-engineered models, offer an increasingly powerful human-specific platform to investigate epilepsy-associated disruptions in neuronal production and differentiation, interneuron migration, and network activity. In this review, we discuss recent advances and ongoing challenges of human cortical organoid models of genetic and acquired epilepsies. Together with established animal models, organoid-based approaches hold promise for advancing mechanistic understanding of epilepsy and enabling the development of more precise therapeutic strategies.Keywords: seizures, iPSCs, neural differentiation, genetic epilepsies, assembloids Introduction Epilepsy is a chronic neurologic disorder characterized by recurrent, unprovoked seizures.1–3 Notably, epilepsy is not a singular disease, but rather, a heterogeneous group of syndromes with diverse clinical features, risk factors, comorbidities, and trajectories—many of which are dynamic and change over time. 3 Adding to this complexity, the etiologies of epilepsy are also highly diverse. In addition to acquired causes such as stroke, traumatic brain injury (TBI), hypoxia, and infection, a rapidly expanding number of genetic variants have been implicated.3,4 Largely driven by advances in next-generation sequencing, more than 200 such genes have been identified over the past decade.4,5 These genes play roles in a broad range of biological functions, including ion channel function, synaptic signaling, metabolism, epigenetic regulation, and cell adhesion.3–5 Despite advances in our identification of genetic and acquired causes of epilepsy, the mechanisms by which these insults drive epileptogenesis and network hyperexcitability remain poorly understood. Historically, mechanistic studies of seizures and epilepsy have predominantly relied upon rodent models, although zebrafish, dog, cat, fruit fly, and nonhuman primate models have also been established. Gene-editing approaches are often used to generate transgenic epilepsy models with disease-associated genetic variants. To model acquired epilepsies, electrical stimulation or chemoconvulsants can be used to induce acute or chronic seizures/status epilepticus in otherwise healthy animals. 6 Although animal studies have provided profound insights into disease mechanisms, these models often have high mortality, variability in seizure frequencies, and fail to replicate spontaneous seizure phenotypes observed in patients. 7 Additionally, differences in cortical development between mouse and human may limit the direct translatability of mouse models for certain human developmental epilepsies.8,9 For example, key differences arise between human and murine cortical development, including differences in the stem cell niche, cell-type diversity, transcriptomic profiles, the timing of neurogenesis, and overall brain cytoarchitecture.10–15 Human pluripotent stem cell (hPSC)-derived brain organoid models help address these challenges by offering in vitro platforms for investigating human-specific mechanisms of cortical development and disease. The increasing availability of CRISPR-engineered and patient-derived induced PSC (iPSC) lines, together with the emergence of region-specific organoid and assembloid systems and recent advancements in in vitro electrophysiology techniques 16 have strengthened these models over the last decade. As a result, a wide range of epilepsy-associated phenotypes can be examined, including altered neurodevelopment, disruptions in excitatory-inhibitory balance and the emergence of network hyperexcitability. In this review, we highlight recent developments in human cortical organoid (hCO) models of epilepsy, discuss current limitations and challenges of these technologies, and propose future directions for their improvement. Brain Organoid Approaches for Modeling Human Cortical Development Although methods for generation of brain organoids from hPSCs vary (Figure 1A), each recapitulate key features of human embryonic cortical development, including the self-organized formation of neural tube-like rosette structures and stratification of ventricular zone-like regions of proliferative progenitors, inner and outer subventricular zone-like domains, and cortical plate-like regions containing both upper- and lower-layer neurons (Figure 1B). Unguided hCO protocols17–19 allow for the cell-autonomous generation of multiple brain regions within the same organoid but are limited by the associated variability. In contrast, guided protocols 20 utilize small molecule morphogens and growth factors to generate more homogeneous, region-specific organoids, such as dorsal or ventral forebrain, thalamus, cerebellum, or hippocampus. Generation of brain organoid models from human pluripotent stem cells. (A) hPSCs can be differentiated into organoids using either unguided or guided methods. (B) Organoids recapitulate the cytoarchitectural organization of the developing human cortex. (C) The number of neural tube-like rosette structures within human cortical organoids varies between protocols with most being multirosette and some designed to generate single rosette organoids. (D) Region-specific organoids can be fused together to form assembloids. Abbreviations: SMADi, SMAD-inhibition; MZ, marginal zone; CP, cortical plate; SVZ, subventricular zone; VZ, ventricular zone; Cereb, cerebellum; GE, ganglionic eminence; Th, thalamus; Hc, hippocampus; Ctx, cortex; hPSC, human pluripotent stem cell. Most guided organoid protocols begin by aggregating hPSCs into embryoid bodies, yielding organoids with multiple neural rosettes. These protocols are scalable but have high variability in the number, size, and organization of rosette structures (Figure 1C). The multiple rosettes often encroach on one another as cortical layers expand, interfering with laminar formation and potentially hindering the identification of early developmental structural alterations. 21 Alternatively, single rosette brain organoid approaches have recently been developed and display improved laminar organization, uniformity, and reproducibility between organoids but at the expense of scalability (Figure 1C).21–27 Notably, region-specific organoids can be fused together to form assembloids, enabling the study of cross-region neuronal migration and circuit integration.28–31 Of most relevance for epilepsy are assembloids involving hCOs fused with ganglionic eminence (GE)30,31 or thalamic32,33 patterned organoids, and hippocampal-GE assembloids 34 (Figure 1D). Additionally, microfluidics and other bioengineered devices have been used to establish morphogen gradients and spatial patterning within individual organoids, potentially enabling dorsal–ventral and/or rostral–caudal organization without the need to fuse distinct region-specific organoids.35–37 However, the size limitations of the devices typically preclude long-term culture. Nonetheless, these advancements bring organoid models closer to recapitulating in vivo human brain development and enhance their utility for modeling genetic and acquired forms of epilepsy, as described below. Common methods for analyzing disease phenotypes across various organoid models include immunostaining, transcriptomic profiling, migration analyses, and electrophysiology or optical imaging assays at both the single cell (eg, whole-cell patch clamp) and network (calcium imaging, local field potential recording, and multielectrode array recordings) levels (Figure 2). Notably, these organoids can be used not only for disease modeling, but also for screening genetic variants of unknown significance, drug development and testing, personalized medicine, and as a source of cells for transplantation therapies (Figure 2). Common approaches for modeling and studying epilepsy disease mechanisms with brain organoid models, and some of their key applications. Applications for Epilepsy Tuberous Sclerosis Complex (TSC) Tuberous sclerosis complex is a drug-resistant epilepsy caused by mutations in TSC1 or TSC2 that lead to hyperactivation of the mechanistic target of rapamycin (mTOR) signaling. 38 Patients develop infantile seizures, benign brain tumors and dysmorphic neuron-containing cortical tubers. 38 Building on their previous TSC organoid studies 39 showing that overproliferation of caudal late interneuron progenitor cells yields tuber-causing dysmorphic GABAergic interneurons, Wong et al, recently show that unguided cerebral organoids generated from TSC patient iPSCs recapitulate hyperexcitable network dynamics and exhibit epileptic-like high frequency oscillations that resemble pathological network activity recorded from TSC patients. 38 Treatment with the ErbB inhibitor Afatinib partially rescued both the electrophysiological abnormalities and the quantity of dysmorphic interneurons. Together, these findings highlight the capacity of organoids to faithfully recapitulate human disease and associated network dynamics, and their utility for screening potential antiepileptogenic treatments. TSC organoid models also implicate glial cells in disease pathogenesis. For example, Blair et al 40 show that mosaic biallelic loss of TSC1 or TSC2 during neural progenitor expansion leads to increased production of glia, many of which were enlarged and dysmorphic. 40 This group also found that mTORC1 hyperactivity in a TSC organoid model of mosaic TSC2 loss-of-function promotes astrocyte proliferation and reactivity, 41 further suggesting that mTOR dysregulation affects glia. Given that TSC is thought to originate from somatic loss of heterozygosity in TSC1/2, 39 further work comparing heterozygous, homozygous and mosaic loss-of-function organoids would help to identify converging or diverging alterations of cortical development between the different TSC models. Focal Cortical Dysplasia Type II Recent brain organoid models of focal cortical dysplasia (FCD) provided important insight into how biallelic somatic mutations in mTOR pathway genes can drive focal epileptogenesis. 42 One group generated hCOs from patient-derived heterozygous, mosaic (mixed heterozygous and CRISPR-Cas9 knockout iPSCs), and gene-corrected iPSCs to investigate how “second-hit” somatic mutations in the mTOR repressor DEPDC5 contribute to FCD type II (FCDII) epileptogenesis. 43 Using single-cell RNA-sequencing, they revealed that DEPDC5 loss-of-function alters Notch and Wnt signaling pathways in neural progenitor cells, leading to premature generation of upper-layer neurons. These findings suggest that global increases in mTOR signaling can alter neuronal fate specification and help distinguish general developmental consequences of mTOR dysregulation from focal, biallelic somatic pathology. Notably, only mosaic hCOs recapitulated FCDII-like dysmorphic neurons and network hyperexcitability. Transcriptomic analyses also revealed dysregulation of synaptic and ion channel genes, providing insight into potential mechanisms underlying hyperactivity in DEPDC5-related FCDII. Polyhydramnios, Megalencephaly and Symptomatic Epilepsy Syndrome In contrast to many focal mTORopathies, polyhydramnios, megalencephaly and symptomatic epilepsy (PMSE) is caused by germline, homozygous loss-of-function variants in the STRADA gene. 44 A recent study utilized CRISPR/Cas9 STRADA knockout and isogenic control iPSC-derived organoids to investigate how STRADA loss and resultant mTOR hyperactivity affect both dorsal and ventral human forebrain development. 44 In addition to recapitulating mTOR hyperactivity and macrocephaly phenotypes found in patients, the loss of STRADA increased progenitor populations, delayed neurogenesis, and altered cell fate specification of both cortical and subpallial organoids, with some phenotypes rescued by rapamycin treatment. However, this study was limited to CRISPR-edited iPSC lines, and it remains unclear how the observed early neurodevelopmental changes associated with STRADA loss and mTOR hyperexcitability influence long-term neuronal maturation, migration, or network activity. Nonetheless, these findings suggest altered interneuron lineage development as a potential epilepsy mechanism in some mTORopathies. Comparison across PMSE, FCDII, and TSC organoid models highlights how the same signaling pathway can produce distinct phenotypes and suggests that the timing and spatial extent of mTOR pathway disruption shape disease severity, affected cell types, and the emergence of epileptogenic network activity. Rett Syndrome Formation of mature neural networks in hCO models is essential to fully recapitulate cortical development and model complex epileptic activity. To this end, Samarasinghe et al 30 measured local field potentials and calcium sensor activity in hCO-GE assembloids, and demonstrated the ability of organoid models to form mature oscillatory neural network activity that is altered in disease contexts. Rett syndrome (RTT) is a neurodevelopmental disorder caused by mutations in the X-linked gene methyl-CpG-binding protein 2 (MECP2). 45 Typical females with RTT develop normally until 6-18 months of age, but then lose speech and develop microcephaly, intellectual disability, seizures, autism, ataxia, and respiratory and cardiac abnormalities.45,46 Samarasinghe et al show that while RTT human iPSC-derived hCO-GE assembloids display similar cortical organization and gross cell-type repertoire to isogenic control assembloids, single-cell RNA-sequencing and synaptic staining analyses revealed changes in interneuron development and synaptogenesis that alter excitatory–inhibitory balance. Consistent with this imbalance, mutant assembloids showed hyperexcitability, hypersynchrony, and loss of putatively normal gamma and lower frequency oscillations. Network phenotypes were recapitulated in control cortex + mutant GE, but not mutant cortex + control GE assembloids, suggesting the dysfunction is predominantly driven by MECP2 loss in GE-derived interneurons. Given that RTT females often have random x-inactivation-driven mosaic expression of mutant and wild-type MeCP2, it would be interesting to compare these findings in global knockout to mosaic organoids with both wild-type and mutant cortical and GE-derived cells. Further study is also necessary to uncover the cell-type-specific changes and microcircuit alterations underlying the hyperexcitability phenotypes revealed in this work. Dravet Syndrome and SCN8A Developmental and Epileptic Encephalopathy Dravet syndrome (DS) is a severe DEE with seizures beginning in the first year of life and a high risk for sudden unexpected death in epilepsy. 47 It is primarily attributed to heterozygous de novo variants in the SCN1A gene that result in Nav1.1 voltage-gated sodium channel haploinsufficiency.47,48 Nav1.1 is highly enriched at the axon initial segments of parvalbumin (PV) and other interneurons that regulate network excitability and oscillations. 49 Accordingly, interneuron dysfunction is hypothesized to explain seizure generation in DS. However, disease mechanisms remain incompletely understood, and emerging evidence suggests PV dysfunction alone may not fully explain chronic epilepsy in DS. 50 Because of the difficulty of generating PV interneurons from hPSCs in 2D or organoid cultures (see below), there have not been many informative brain organoid studies in DS. However, one group recently explored the potential role of SCN1A in regulating early cortical development. 51 Zourray et al generated hCOs and cortical-subpallial forebrain assembloids from 2 DS patient iPSC and CRISPR-CAS9 corrected isogenic control lines, using multiomics approaches to assess consequences of SCN1A haploinsufficiency on human corticogenesis. Surprisingly, they found that SCN1A is expressed not only by excitatory and inhibitory neurons during human forebrain development, but also by neural progenitors, wherein SCN1A haploinsufficiency altered G2/M phase cell-cycle progression. Transcriptomic, proteomic, and phosphoproteomic analyses of early hCOs further supported altered cell cycle dynamics inducing premature neurogenesis. Consistent with these findings, mature DS assembloids showed increased production of upper-layer SATB2+ neurons. Notably, they also found cortical progenitor cell-cycle disruption in DS mice; unlike DS assembloids however, both mice and DS postmortem human tissue showed reduced SATB2+ neurons in the superficial layers of the frontal cortex. Further study is needed to investigate these differences. Disrupted hippocampal network activity is also implicated in epilepsies and their cognitive co-morbidities. A recent study investigated this idea within the context of SCN8A gain-of-function DEE-13, comparing cortical + GE and hippocampal + GE assembloids derived from patient and control iPSCs. 34 GE-derived interneurons successfully migrated into the cortical or hippocampal regions of the assembloids, with no statistically significant differences between control and mutant organoids. Interestingly, DEE-13 cortical + GE, but not hippocampal + GE, assembloids were hyperexcitable in local field potential recordings and calcium imaging, suggesting region-specific circuit differences. DEE-13 hippocampal + GE assembloids showed deficits in theta-gamma coupling without hyperexcitability, and the authors showed similar findings in recordings from temporal lobe epilepsy patients. Computational modeling predicted a loss of slow-spiking, somatostatin-expressing oriens-lacunosum/moleculare-like cells. The authors showed altered numbers of excitatory neurons or a specific interneuron subtype in DEE-13 cortex + GE and hippocampus + GE assembloids, respectively, that may explain these region-specific differences. Collectively, these studies illustrate how organoid models, including different assembloid circuits, can be used to reveal developmental and network phenotypes. CDKL5 Deficiency Disorder Loss-of-function variants in the X-linked gene cyclin-dependent kinase-like 5 (CDKL5) cause CDKL5 deficiency disorder (CDD), a DEE characterized by early onset intractable seizures, motor dysfunction, autism, and neurodevelopmental delay.52,53 Interacting with proteins involved in microtubule assembly and neuronal function, CDKL5 is implicated in neuronal function; however the precise underlying mechanisms of CDD are unknown and phenotypes vary across models. 54 Negraes et al generated hCOs from 6 CDD patient iPSC lines, sex-matched related controls, and isogenic control clones. 55 They found CDKL5 deficiency led to increased neurite lengths, decreased synaptogenesis, and hypersynchronous network activity. Neurons in CDD hCOs also exhibited higher intrinsic excitability due to dysfunction of sodium and calcium channels. 56 These electrophysiological phenotypes were similar to those from RTT organoids, suggesting potentially converging disease mechanisms and shared targets for future drug development. 56 TBI and Risk of Epilepsy TBI is a major environmental risk factor for neurodegenerative diseases, including Alzheimer's disease and chronic traumatic encephalopathy. 57 Although rodent studies have provided insight into mechanisms of neurodegeneration after TBI, translation to effective therapeutic interventions has been limited, and how TBI drives long-term changes to neural networks leading to seizures is not fully understood. To study TBI in human 3D brain organoids, several groups have used mechanical impact or pressure,58–60 high-intensity focused ultrasound, 61 or pressure wave exposure.62,63 These studies recapitulate key features of injury, including neuronal loss, phosphorylated TDP-43 accumulation, reactive astrogliosis, and changes in neuronal network activity. Although this area of study is just emerging, applying injury paradigms to hCOs induces altered network oscillation dynamics 62 and increases calcium signaling that scales with injury severity, 60 associating injury with the emergence of seizure-like phenotypes and demonstrating the applicability of organoids for studying TBI-associated mechanism of epileptogenesis. Additionally, drug screening in organoids has identified promising potential treatments that may mitigate injury-induced neurodegeneration. 61 Hypoxic-Ischemic Encephalopathy Organoid models have been successfully used to investigate hypoxic-ischemic encephalopathy, an acquired brain insult common in the perinatal period that is often associated with epilepsy. For example, Paşca et al show reduced TBR2+ intermediate progenitor cells per proliferative zone in hCOs following oxygen deprivation, which persisted 72-h after re-oxygenation. 64 Transcriptional analysis suggested this deficit may be due to the unfolded protein response pathway, which was supported by their findings that application of Integrated Stress Response Inhibitor, an unfolded protein response pathway modulator, prevented TBR2 deficits in both organoids and in primary human cortical tissue following oxygen deprivation. Saleem et al showed that oxygen-glucose deprivation in 3-7-month-old hCOs induces hyperexcitability and broadband increases in spectral power that are characteristic of seizure-like activity, which were comparable or greater than organoid responses to common chemoconvulsants. 65 Using human dorsal forebrain organoids, others showed that transient hypoxia leads to loss of outer radial glia, a progenitor population highly abundant in humans but limited in rodents. 66 Furthermore, Umans and Gilad identified hypoxia-evoked gene regulatory transcriptional changes in dorsal telencephalic organoids, highlighting pathways that may contribute to hypoxia-induced cortical development and circuit dysfunction. 67 Current Challenges, Limitations, and Future Directions Although still a relatively new field, the work mentioned above encompasses just a small subset of studies that have successfully used organoid models to investigate disease mechanisms, including acquired and genetic epilepsies. Advances in electrophysiology techniques, as well as the continual emergence and optimization of organoid protocols, have greatly expanded the utility of hPSC-derived brain organoids for modeling human brain development and disease.17,68,69 However, brain organoid models are not without their challenges and limitations. Reproducibility is one of the biggest challenges in organoid models. 70 Differences in organoid derivation protocols, culturing conditions, experimental timing, and extracellular matrices can all lead to variation between research groups; however, providing detailed methods and protocols within publications can help mitigate this and allow for high cross-site consistency. 71 As with other stem cell-derived models, there is often line-to-line and batch-to-batch variability. Generation of standardized quality control and quantification metrics can reduce such variability, as can the inclusion of multiple replicate experiments with appropriate controls, including isogenic controls where possible, and the implementation of protocols with higher consistency, such as single-rosette protocols or newly developed microfluidic methods. 72 In addition, the use of chimeroids allows for the testing of multiple human genomic backgrounds within the same brain organoid culture. 73 Another key challenge for the field is the limited maturation and cell type repertoire within organoids. Although organoids display oscillations and complex network activity, it takes at least 3 to 6 months for these mature electrophysiological features to emerge.30,38 Groups are actively working to address the limited cell-types and lack of an in vivo microenvironment in organoid models by adding missing components such as endothelial cells for vascularization or microglia, building larger multiregion assembloids, or xenografting organoids into a rodent host environment.20,72,74 Progress has also been made by our group 75 and others 76 to generate organoid models that express PV interneurons, a late-maturing MGE-derived cell type that plays key roles in regulating network activity and oscillations, but that has been notoriously difficult to generate in vitro. Importantly, these organoids could provide a valuable source of cortical interneuron progenitors for future epilepsy cell transplantation therapies. In addition to disease modeling and serving as a potential cell source for therapeutic grafting in epilepsy and related disorders, human brain organoid models are emerging as key in vitro platforms for medium-throughput preclinical drug testing, particularly when paired with expanding AI approaches. 77 For example, Turpin-Moreno et al examined organoid network activity responses using machine learning and identified several effective drugs for further investigation in treating DS, 78 while Negraes et al successfully developed high throughput screening methods to test leading CDD candidate drugs. 55 In addition to screening new drugs, organoids can also be used to test for toxicity, including teratogenicity of antiseizure medications, 79 helping to inform clinical treatment and patient safety measures. Advances in the cortical laminar organization and neuronal subtype repertoire of hCOs are also needed, as the current level of microcircuit organization is not sufficient to address many epilepsy disease mechanisms, nor for complete drug efficacy/safety screening. Animal models remain essential for testing behavioral, social, learning, and memory paradigms that are often impaired in various epilepsies. Moreover, the most catastrophic complication of epilepsy, sudden unexpected death in epilepsy (SUDEP), cannot be modeled in brain organoids alone, although progress also has been made in developing human cardiac organoids 80 and one can envision a not-so-distant future in which human cortical/brainstem/cardiac assembloid models are used to probe SUDEP mechanism and biomarkers. At present, combining both in vitro hCO models and in vivo animal work offers a powerful approach for understanding epilepsy mechanisms and developing better therapies.