Cortical versus hippocampal network dysfunction in a human brain assembloid model of epilepsy and intellectual disability

Abstract

Neurodevelopmental disorders often impair multiple cognitive domains. For instance, a genetic epilepsy syndrome might cause seizures due to cortical hyperexcitability and present with memory impairments arising from hippocampal dysfunction. This study examines how a single disorder differentially affects distinct brain regions using induced pluripotent stem cell (iPSC)-derived cortical- and hippocampal-ganglionic eminence assembloids to model developmental and epileptic encephalopathy 13, a condition arising from gain-of-function mutations in the SCN8A gene encoding the sodium channel Nav1.6. While cortical assembloids showed network hyperexcitability akin to epileptogenic tissue, hippocampal assembloids did not, and instead displayed network dysregulation patterns similar to in vivo hippocampal recordings from epilepsy patients. Predictive computational modeling, immunohistochemistry, and single-nucleus RNA sequencing revealed changes in excitatory and inhibitory neuron organization that were specific to hippocampal assembloids. These findings highlight the unique impacts of a single pathogenic variant across brain regions and establish hippocampal assembloids as a platform for studying neurodevelopmental disorders.

Keywords: CP: Neuroscience; brain organoids; cellular neuroscience; disease modeling; epilepsy; hippocampal assembloids; neural circuits; neural stem cells.

Figure 1.. Generation and characterization of cortical and hippocampal assembloids

(A) IHC analysis of SCN8A patient iPSC-derived isogenic control (iCtrl) and mutant (Mut) unfused Cx organoids at specified ages shows similar formation and layered segregation of neural progenitors (PAX6 and TBR2) and deep/superficial neurons (CTIP2, TBR1, and SATB2). (B) Organoids from both conditions display comparable levels of GE progenitor and migratory interneuron markers (NKX2–1, OLIG2, DLX1, and DLX2) and GABAergic interneuron marker GAD65 by day 120. (C) Analysis of p.SCN8A iCtrl and Mut unfused hippocampal organoids at days 56 and 120 reveals the expression of hippocampal neuron markers (ZBTBT20 and NRP2) and dentate granule cells (PROX1), with CA3 and dentate granule layers at day 120. (D) PROX1-expressing dentate granule cells are segregated from KA1-expressing CA3 region. (E) Schematic of organoid generation, patterning, and fusion to create assembloids, with GE organoids treated with AAV1-CAG virus prior to fusion. (F and G) At day 84 in vitro, both iCtrl and Mut Cx+GE (F) and Hc+GE (G) show similar migration of GE-derived Tdtomato-labeled cells. (H) Quantification shows no significant differences; n = 4 independently generated assembloids per genotype, four sections per assembloid with ≥ 1,188 cells/sample for Cx+GE and 790 cells/sample for Hc+GE. Individual datapoints along with their mean ± SD are shown. p = 0.6756 for Cx+GE and p = 0.3254 for Hc+GE; Student’s t test. (I) UMAP of day 120 iCtrl Cx+GE and iCtrl Hc+GE assembloids with cell types (ExN, excitatory neurons; IPC, intermediate progenitors; Caj Ret, Caja-Retzius; IN, inhibitory neurons; CyP, cycling progenitors; RG, radial glia; oRG, outer radial glia; Ast, astrocytes; OPC, oligodendrocyte precursors; and ChP, choroid plexus). (J) Analysis of the ExN group demonstrates that Cx+GE and Hc+GE assembloids contain excitatory subtypes (CFuPN, corticofugal projection neurons; CPN, callosal projection neurons; DG, dentate granule-like; CA/S, cornu ammonis/subiculum-like; and Imm, immature) in the expected proportions. See Table S2 for gene expression differences. (K) Compared to publicly available ex vivo human datasets, Cx+GE ExNs express gene markers that are differentially upregulated in the cerebral cortex neurons, and Hc+GE ExNs express gene markers that are differentially upregulated in the hippocampal neurons. (L) Volcano plot and gene list (in order of statistical significance) highlight unique genes distinguishing Cx+GE from Hc+GE assembloids across all cell types (not just ExNs) with published discriminating genes indicated in bold. Scale bars: 100 μm in (A)–(D) and 200 μm in (F) and (G).

Figure 2.. DEE-13 cortical assembloids demonstrate network hyperexcitability

(A) Raw trace (top) and time expansion (middle) from an LFP recording of an isogenic control (iCtrl, P1) p.SCN8A cortex assembloid (Cx+GE). Bottom: periodogram from the entire raw trace. (B) Raw trace (top) and three time expansions (middle) from the LFP recording of a Mut P1 (p.R1872>L) p.SCN8A Cx+GE assembloid, with time expansions near the dotted lines in the raw trace. Bottom: periodogram from the entire raw trace. (C) Raw trace (top) and time expansion (middle) from an LFP recording of a control (Ctrl 2) cortex assembloid (Cx+GE). Bottom: periodogram from the entire raw trace. (D) Raw trace (top) and two time expansions (middle) from the LFP recording of a Mut P2 (p.V1592>L) p.SCN8A Cx+GE assembloid, with time expansions near the dotted lines in the raw trace. Bottom: periodogram from the entire raw trace. (E) Number of sustained high amplitude events/1000s in iCtrl, Ctrl 2, and Mut, defined as sustained runs ≥30 s with root mean square amplitude ≥5×greater than the preceding 30 s. Individual datapoints along with their mean ± SD are shown. ***p = 0.0004. (F) Frequency of spikes, sharp waves, and LDDs from iCtrl, Ctrl 2, and Mut Cx+GE LFP recordings, with a representative example of each discharge shown in red. For (E) and (F), n = 6 biologically independent samples (differentiations) for iCtrl, n = 5 for Ctrl 2, and n = 7 for Mut P1 and P2. Individual datapoints along with their mean ± SD are shown. *p = 0.0346, **p = 0.0071 (spikes), and **p = 0.0097 (LDDs); Kruskal-Wallis test. (G) Schematic for calcium transient identification and functional cluster generation after multiphoton imaging with a genetically encoded calcium indicator. (H) Normalized ΔF/F of calcium indicator activity in iCtrl Cx+GE. Each line represents a single neuron’s activity. (I) Representative normalized ΔF/F of calcium indicator activity in iCtrl, Ctrl 2, Mut P1, and Mut P2 Cx+GE. Controls (top) show decorrelated traces with minimal synchrony, whereas Mut Cx+GE has both decorrelated (top) and highly synchronous discharges (bottom). (J) Violin plots show a significant increase in average amplitude of Ca²⁺ transients in iCtrl vs. Mut P1 and Ctrl 2 vs. Mut P2 Cx+GE. Linear mixed effects model. ***p = 0.00024 (iCtrl vs. Mut P1) and *p = 0.0332 (Ctrl 2 vs. Mut P2); each dot represents a separate recording. n = 7 independent differentiations for iCtrl and Mut P1, and n = 5 independent differentiations for Ctrl 2 and Mut P2.

Figure 3.. DEE-13 hippocampal assembloids show a reduction in fast oscillations but lack overt hyperexcitability

(A) Raw traces from LFP recordings of iCtrl (P1, blue), Mut P1 (p.R1872>L, red), Mut P2 (p.V1592> L, brown) p.SCN8A, and Ctrl 2 (black) hippocampal (Hc+GE) assembloids. (B) Corresponding power spectra (periodograms) from recordings in (A). iCtrl, Ctrl 2, and Mut P1 exhibited sustained oscillations lasting ≥ 300 s, whereas Mut P2 also had oscillations but briefer (∼60 s) bouts of rhythmic activity (inset). (C) Time-expanded raw traces highlighting candidate oscillatory events. (D) Gamma-filtered (30–80 Hz) traces from segments in (C), with black arrows indicating putative gamma bursts. (E) Ripple-filtered (80–160 Hz) traces from (C), with high gamma/ripple events observed only in iCtrl and Ctrl 2 (black arrows). Mut P1 and P2 exhibited slower low gamma activity without discrete high gamma/ripple peaks. (F) Morlet wavelet spectrograms from events in (C)–(E), showing gamma and high gamma/ripple power over time. (G) Representative ΔF/F calcium imaging traces from iCtrl (P1), Mut P1, Ctrl 2, and Mut P2 Hc+GE assembloids. Top: decorrelated neuronal activity. Bottom: synchronized network events (red dashed boxes). (H) Quantification of gamma bursts (low gamma) per 1000 s showed no significant differences across groups. Mean ± SD are shown, with each dot representing an individual differentiation. A test for normality was followed by a Kruskal-Wallis test. (I) Quantification of high-gamma/ripple bursts per 1000 s revealed significantly more events in iCtrl compared to Mut P1 (***p = 0.0008) and in Ctrl 2 compared to Mut P2 (*p = 0.0170); mean ± SD are shown with each dot representing an individual differentiation. A test for normality was followed by a Kruskal-Wallis test. (J) Quantification of synchronized calcium transient amplitudes showed no significant change in network synchrony in Mut P1 compared to iCtrl or in Mut P2 compared to Ctrl 2 (linear mixed effects model). Each dot represents an individual recording, and the violin plot shows the distribution of the data; n = 6 independent differentiations for iCtrl, Mut P1, and Mut P2; and n = 5 for Ctrl 2.

Figure 4.. Perturbation of theta-gamma coupling in both DEE-13 hippocampal assembloids and human temporal lobe epilepsy

(A) Example of phase-amplitude coupling (PAC) in an iCtrl Hc+GE assembloid showing higher gamma amplitude coupled to the 0 phase of theta (3–10 Hz). (B–E) Representative heatmaps of theta phase and gamma amplitude coupling in p.SCN8A P1 iCtrl (B), Ctrl 2 (D), and Mut P1 and P2 (C and E) Hc+GE assembloids. Both iCtrl and Ctrl 2 assembloids exhibit monophasic PAC, whereas both Mut assembloids exhibit disordered PAC. (F–I) Quantification of monophasicity (F), theta power (G), theta instability (H), and Ca²⁺ spiking instability (I) in iCtrl vs. Mut P1 and Ctrl 2 vs. P2 Mut Hc+GE assembloids. The monophasicity is significantly reduced in P1 relative to iCtrl (p = 0.0103) with a trending reduction in P2 relative to Ctrl 2 (p = 0.053). No significant differences in theta power were found. P1 Mut dynamics show increased instability relative to iCtrl both at the level of theta oscillations (*p = 0.0361) (H) and Ca²⁺ spiking (****p < 0.0001) (I), while P2 Mut only shows increased instability relative to Ctrl 2 at the level of Ca²⁺ spiking (****p < 0.0001) (I). Violin plots with datapoints (average for each assembloid) are shown in (F)–(H) (LFP data); n = 5–8 independent differentiations/genotype, one-tailed Wilcoxon rank-sum test. Violin plot with datapoints (average for each assembloid) is shown in I (two-photon imaging); n = 5–6 independent differentiations/genotype, linear mixed effects model. (J and K) PAC heatmaps in the healthy (I) and epileptic (J) hippocampus of a single patient. The healthy side shows highly monophasic coupling, whereas the epileptic hippocampus from the same patient shows disordered coupling. (L–N) Monophasicity (L), theta power (M), and theta instability (N) in healthy and epileptic hippocampi of eight temporal lobe epilepsy patients. Monophasicity was reduced in the epileptic hippocampi of each of the eight patients. No differences were noted in theta power/instability. One-tailed Wilcoxon signed rank paired t test, n = 8 patients, **p = 0.0039.

Figure 5.. In silico model predicts loss of O-LM cells

(A) Schematic of the CA3 microcircuit model, incorporating pyramidal cells, parvalbumin-expressing basket cells, and somatostatin-expressing O-LM interneurons. Synaptic connectivity and cell-intrinsic conductance were modeled using Hodgkin-Huxley equations (see STAR Methods). (B) IHC of Hc+GE assembloids shows spatial proximity of vGlut1⁺ pyramidal neurons/excitatory pre-synpatic puncta (red), SST⁺ O-LM cells (green), and PV⁺ basket cells (yellow), supporting the plausibility of local circuit/synaptic interactions between the modeled populations. Additional examples with a full field of view are in Figure S1F; these tripartite ensembles were observed in 41/46 regions of interest from four imaged sections. Scale bars, 5 μm. (C and D) Phase-amplitude coupling plots from iCtrl and Mut simulations show theta-gamma interactions consistent with the in vitro LFP recordings. (E) Predicted impact of the p.SCN8A R1872>L gain-of-function mutation on the hippocampal circuit parameters, including cell-intrinsic persistent sodium currents and population-level changes. The model predicts a substantial loss of O-LM cells, increased pyramidal cell number, and increased persistent Na⁺ current across all modeled cell types.

Figure 6.. Hippocampal but not cortical DEE-13 assembloids demonstrate reduced inhibitory interneuron numbers

(A) IHC analysis of Hc+GE assembloids reveals a reduction in the pan-inhibitory interneuron marker glutamic acid decarboxylase-65 (GAD65/GAD-2) and in the somatostatin (SST) subtype of interneurons in both Mut P1 and Mut P2 compared to iCtrl or Ctrl 2, respectively. No significant reductions were observed in parvalbumin+ (PV+) interneurons. (B) IHC analysis for excitatory neurons with CAMKII-α revealed a significant increase in both Mut P1 and Mut P2 compared to iCtrl or Ctrl 2, respectively. n = 3–14 independent differentiations per genotype and four sections per assembloid with ≥ 4,280 cells per sample. For Mut P1 and Mut P2 respectively; GAD65 **p = 0.0031 **p = 0.00124; SST **p = 0.0025 and *****p = 2.40E-16; CAMKII-α *p = 0.039 and *p = 0.0149. (C) IHC for both GAD65 and PV in Cx+GE assembloids reveals no significant differences between iCtrl vs. Mut P1 or Ctrl 2 vs. Mut P2. (D) IHC for CAMKII-α reveals a significant increase in both Mut P1 and Mut P2 compared to iCtrl or Ctrl 2, respectively. n = 3–11 independently generated assembloids per genotype, four sections per assembloid with ≥ 2,681 cells per sample. For Mut P1 and Mut P2, respectively, CAMKII-α **p = 0.0028 and *****p = 6.27E-6. For (A–D), all datapoints with their mean ± SD are shown; significance calculations used a linear mixed effects model; and Scale bars, 50 μm.

Figure 7.. DEE-13 differentially affects circuit composition and gene expression in day 120 cortical versus hippocampal assembloids

(A) 2D UMAP of cells from day 120 iCtrl and Mut P1 Cx+GE assembloids revealed excitatory neurons (ExN), intermediate progenitors (IPCs), Cajal-Retzius (Caj Ret), inhibitory neurons (INs), cycling progenitors (CyPs), radial glia (RG), outer radial glia (oRG), astrocytes (Asts), oligodendrocyte precursors (OPCs), and choroid plexus (ChP). (B) The Mut P1 Cx+GE assembloids show more ExNs compared to iCtrl. “Other” cell types (IPC, Caj Ret, CyP, RG, OPC, and ChP) are grouped for clarity. (C) No significant changes in the corticofugal (CFuPN) or callosal (CPN) projection ExN subtypes between Mut P1 and iCtrl Cx+GE. “Other” excitatory neuron subtypes include immature neurons (see Figure S5A). (D) Mut P1 Cx+GE assembloids show more mixed CGE- and LGE-like INs compared to iCtrl. Note that medial ganglionic eminence-like (MGE) INs include MGE types 1–4 described in Figure S5B. “Other” inhibitory interneuron subtypes include all immature and unspecified interneurons (see Figure S5B). (E) UMAP of cells from day 120 iCtrl and Mut P1 Hc+GE assembloids. (F) Mut P1 Hc+GE assembloids show more ExNs, fewer INs, fewer Asts, and more co-clustered oRG/Asts compared to iCtrl. (G) Mut P1 Hc+GE assembloids show more cortex-like CFuPN and CPN ExNs and fewer hippocampal dentate granule-like (DG) and cornu ammonis/subiculum-like (CA/S) ExNs compared to iCtrl. (H) Mut P1 Hc+GE assembloids lack RELN+ INs present in iCtrl Hc+GE. (I and J) IHC shows few cells co-expressing Reelin and GAD65 in cortical assembloids across iCtrl, Mut P1, and Mut P2 lines, whereas hippocampal assembloids show a substantial population in iCtrl that is lost in Mut lines (datapoints are shown with their mean ± SD; Mut P1 ***p = 0.000237, Mut P2 ****p < 0.0001, linear mixed effects model, five iCtrl, four Mut P1 and P2 assembloids, four sections per assembloid, and ≥ 6,324 cells/sample). Scale bars, 50 μm. (K and L) Differential gene analysis volcano plots of day 120 Cx+GE and Hc+GE, respectively. Selected upregulated genes (red) in Mut are involved in glutamatergic pathways and likely promote a hyperexcitable state. Reduced expression of selected downregulated genes (blue) has been associated with epilepsy and intellectual disability. The upregulated and downregulated genes are shown in relative proportions. Gene expression data are provided in Table S3. (M and N) Gene ontology (GO) analysis of Cx+GE and Hc+GE, respectively, shows upregulation of glutamatergic pathways and downregulation of basic cellular processes in the hippocampal Mut P1 assembloids. Circle sizes correspond to the number of GO terms. GO data are provided in Table S4.