2.1 Protein structure prediction

Wild-type ACTL6B and mutant ACTL6B p.(Val421_Cys425del) was modeled onto the predicted protein structures of human paralog ACTB (PDB P60709) using SWISS-MODEL (https://swissmodel.expasy.org/, RRID:SCR_013032). Briefly, a homology search for ACTL6B was performed using NCBI-BLAST of SWISS-MODEL. ACTB was selected based on GMQE (Global Model Quality Estimation) value of 0.63 and conservation of sequence over the domain containing the variant. Swiss-PdbViewer (https://spdbv.vital-it.ch/, RRID:SCR_013295) was used to create 3D rendering.

2.2 Reprogramming of fibroblasts to induced pluripotent stem cells

Fibroblasts were isolated from skin biopsy explants from the proband harboring a biallelic ACTL6B p.(Val421_Cys425del) variant and a sex-matched unaffected familial carrier (biological mother of the affected patients) described in Yüksel et al. (2019). Fibroblasts were cultured in Minimum Essential Media (Gibco), supplemented with 20% (v/v) fetal bovine serum (Gemini) and 1% (v/v) antibiotics (Pen/Strep 10,000 U/ml). Low passage fibroblasts were reprogrammed to iPSCs using Epi5 Episomal iPSC Reprogramming Kit (ThermoFisher) with the following modifications: fibroblasts were electroporated and plated onto Matrigel (Corning)-coated plates. Reprogramming vectors pCXLE-hOCT3/4-shp53 (Addgene, 27077, RRID:Addgene_27077), pCXLE-hSK (Addgene, 27078, RRID: Addgene_27078), pCXLE-hUL (Addgene, 27080, RRID:Addgene_27080), and pCXWB-EBNA1 (Addgene, 37624, RRID:Addgene_37624) were gifts from Shinya Yamanaka and prepared in-house (Okita et al., 2013). The medium was changed to StemFlex (ThermoFisher) on day 15. iPSC colonies were manually collected and individual clones were expanded in feeder-free conditions on vitronectin-coated plates (ThermoFisher) for karyotyping (Karyostat; Invitrogen) and characterization of pluripotency.

2.3 Spontaneous embryoid body (EB) formation

Embryoid bodies were generated from iPSCs and cultured with human embryonic stem cell media without FGF consisting of Dulbecco's Modified Eagle Media (DMEM)/F12 with l-glutamine (Gibco), supplemented with 20% KnockOut Serum Replacement (KSR, Gibco), 0.2% β-mercaptoethanol (Gibco), and 1% Non-Essential Amino Acids (NEAA, Gibco). After 7 days under constant rotation at 95 RPM in uncoated plates, the EBs were collected for RT-PCR analysis.

2.4 Neural differentiation

Neural progenitor cells (NPCs) were generated as previously described, with minor modifications (Chailangkarn et al., 2016). Briefly, iPSC media were removed and switched to DMEM/F12 (Gibco) with 1 × N2 Supplement, 1 µM Dorsomorphin (Calbiochem), and 10 µM SB431542 (Biogems). On the following day, the colonies were mechanically dissociated and kept under rotation for 7 days at 37°C to form EBs. The EBs were plated onto dishes coated with Matrigel (Corning) in NPC medium (DMEM/F12, 0.5 × N2 (Gibco), 0.5 × B27 (Thermo Fisher), and 100 ng/ml FGF2 (Thermo Fisher)) to promote neural rosettes outgrowth after 7 days. Manually collected neural rosettes were enzymatically dissociated (Accutase, Innovative Cell Technologies) and plated on poly-l-ornithine/laminin (Sigma/Invitrogen)-coated dishes in NPC medium. Neurons were differentiated from patient-specific iPSC-derived NPCs and familial controls. The NPCs were plated onto poly-l-ornithine/laminin (Sigma/Invitrogen)-coated dishes and cultured in neuronal media (DMEM/F12 with l-glutamine and 15 mM HEPES (Gibco)), 0.5 × N2 Supplement (Gibco), 0.5 × B27 Supplement (Gibco), 1% pen-strep (Gibco), 20 ng/ml BDNF (Peprotech), 20 ng/ml GDNF (Peprotech), and 5 µM of Y-27632 Dihydrochloride (Biogems) for up to 31 days.

2.5 RT-PCR analyses

Total RNA was extracted using TriZOL (Thermo Fisher Scientific) according to the manufacturer's recommendation. Reverse transcription of 1 μg of RNA to complementary DNA (cDNA) was performed with SuperScript III First Strand Synthesis System (Invitrogen). Primers were designed with Primer-Blast NCBI tool or obtained from prior publications (Supplementary Table). All primer pairs showed ≥90% amplification efficiency and single melt curves in RT-PCR.

To assess cell lineage markers during neural differentiation, RT-PCR was performed for AFP, MSX1, OCT4, PODLX, SOX10, FOXD3, NESTIN, and GAPDH and visualized in a 1.5% agarose gel by electrophoresis. Quantitative (q)RT-PCR was performed using Power SYBR Green PCR Master Mix (Applied Biosystems), according to the manufacturer's instructions. qRT-PCR was performed three times, in triplicate, for ESPNP, MAP2K3, SOX8, TPPP, KCNJ12, and GAPDH. Ct values for each gene was normalized to GAPDH, as a loading control. Fold expression was calculated by normalizing gene expression of affected neurons to that of unaffected neurons, set to 1.

2.6 Western blot analysis and ACTL6B antibody validation

Cells were lysed with radioimmunoprecipitation assay buffer (50 mM Tris-HCl pH 7.4, 1% NP-40, 025% Na-deoxycholate, 150 mM NaCl, 1 mM EDTA) containing 1× protease inhibitor cocktail (Roche). Lysed samples were sonicated for 7.5 min using the bioruptor PICO (Diagnode), and were centrifuged for 20 min at 20,000 RCF, at 4°C to remove cell debris.

Protein concentration was calculated using BCA Protein Assay Kit (ThermoFisher), and 20 μg protein was loaded on 10% SDS-PAGE gels. Gels were run with TGS buffer (250 mM Tris, 1.9 M Glycine and 1% SDS) and subsequently electro-blotted to PVDF membranes (Biorad) using the Trans-Blot Turbo transfer system and buffer (BioRad). Membranes were blocked with Tris-buffered saline with Tween detergent (TBST) (1 M Tris pH 7.5, 1.5 M NaCl, 1% Tween) containing 5% nonfat dry milk and 1% donkey or goat serum, for 1 hr at room temperature. Next, membranes were probed with rabbit anti-ACTL6B (Proteintech, 27215-1-AP, RRID:AB_2857952, 1:1,000), mouse anti-GAPDH (Thermofisher, AM4300, RRID:AB_2536381, 1:1,000), mouse anti-TUJ1 (Millipore, MAB1637, RRID:AB_2210524, 1:1,000), anti-Histone H3 (Cell Signaling Technology, #3638, RRID:AB_1642229, 1:5,000), or anti-HA-Tag (Cell Signaling Technology, #2367, RRID: AB_10691311, 1:3,000) followed by incubation with goat anti-mouse (Thermo Fisher; G-21040, RRID:AB_2536527, 1:20,000), goat anti-rabbit (Abcam, ab205718, RRID:AB_2819160, 1:20,000), donkey anti-mouse secondary (LI-COR, 926-32212, RRID:AB_621847, 1:15,000), or donkey anti-rabbit secondary antibodies (LI-COR, 926-68073, RRID:AB_10954442, 1:15,000).

Membranes were washed with TBST (1 M Tris pH 7.5, 1.5 M NaCl, 1% Tween) three times for 15 min. For chemiluminescence western blot detection, membranes were incubated in SuperSignal West Dura (Thermo Fisher, 34075) for 5 min prior to imaging. Membranes were imaged using Odyssey Fc Infrared Imaging System (LI-COR), and the densitometry analysis was performed using Odyssey Image Studio (LI-COR) Software (https://www.licor.com/bio/image-studio-lite/, RRID:SCR_013715).

To validate the anti-ACTL6B antibody (Proteintech, 27215-1-AP, RRID:AB_2857952, 1:1,000), human ACTL6B was cloned into pcDNA5/FRT/TO (Invitrogen) with a 2xHA N-terminal epitope tag. The expression construct was sequence verified and transfected into HEK293T cells, which do not endogenously express ACTL6B. Doxycycline (50 ng/ml) was added 24 hr later and lysates were collected at 72 hr post-transfection for western blot analysis of HA and ACTL6B signal overlap.

2.7 Immunocytochemistry

For neuronal characterization, NPCs were plated on poly-l-ornithine/laminin-coated 10-cm plates and cultured until they reached 80% confluency. Neuronal differentiation was carried out as described above, with the addition of CultureOne supplement (ThermoFisher) on day 2 of differentiation to eliminate the growth of dividing cells and increase the purity of the neural cultures. Neurons were cultured until day 23 and were dissociated using 1:1 accutase/accumax (Gibco) at 37°C for 30 min. Cells were passed through a 40-μm cell strainer and counted. Neurons were cryopreserved and subsequently used for immunocytochemistry at a cell density of 1.25 × 10⁵ cells/cm².

Cultured iPSCs, NPCs, and neurons were plated onto 8-well chambered slides (Thermo Fisher) and fixed for 15 min in 4% paraformaldehyde and permeabilized with 0.25% (v/v) Triton-X100 in PBS for 15 min, then washed three times with PBS. The slides were blocked for 1 hr in 1% (v/v) goat serum and 0.1% (v/v) Triton-X100 in PBS. Cells were incubated with primary antibodies: mouse anti-Nanog (Santa Cruz, SC293121, RRID:AB_2665475, 1:200), rat anti-SOX2 (ThermoFisher, #14-9811-82, RRID:AB_11219471, 1:200), mouse anti-Nestin (Thermo Fisher, MA1110, RRID:AB_2536821, 1:200), rabbit anti-PAX6 (Biolegend, #901301, RRID:AB_2565003, 1:200), chicken anti-MAP2 (Abcam, #ab5392, RRID:AB_2138153, 1:20,000), mouse anti-TUJ1 (Biolegend, #801201, RRID:AB_2313773, 1:2,000), and chicken anti-GFAP (Abcam, #ab4674, RRID:AB_304558, 1:1,000) overnight at 4°C and washed three times with PBS. For detection, Alexa fluor [488, 555, and 647]-conjugated goat anti-rabbit, donkey anti-rabbit, goat anti-mouse, goat anti-rat, or goat anti chicken secondary antibodies were used (Thermo Fisher, RRID:AB_138404, RRID:AB_2576217, RRID:AB_141778, RRID:AB_2534096, RRID: AB_2535805, RRID:AB_2535792).

2.8 Protein stability assay

Equal numbers of day 31 EIEE-76 patient and unaffected control iPSC-derived neurons were treated with 100 μg/mL of cycloheximide (Sigma) in neuronal media for 12, 24, 48, 72, or 96 hr. The neurons were washed before the nuclear protein fractions were collected using the NE-PER Kit (ThermoFisher), according to the manufacturer's recommendations. An equivalent amount of nuclear lysate was used for western blot analysis. ACTL6B expression was normalized to Histone H3 as a loading control and then normalized to time 0 to calculate the amount of ACTL6B remaining and define protein half-life. The assay was performed with clones CW066 A5 and CW067 U5.

2.9 Multielectrode array analysis

To record population-wide neuronal activity EIEE-76 patient (CW066 A4 and CW066 A5) and unaffected controls (CW067 U5 and CW067 U9), iPSC-derived NPCs were plated at a density of 1.2 × 10⁵ cells/cm² onto 48-well MEA plates from Axion Biosystems. Each well contains 16 electrodes and was pretreated with poly-l-ornithine/laminin (Sigma/Invitrogen). The MEA system enables single electrode measurements of activity from all nearby neurons and each electrode is positioned to assure no overlap between signals. Neural differentiation was induced 24 hr after NPCs were plated as described above. On day 2 of differentiation, CultureOne Supplement (Gibco) was added to the media to prevent overgrowth of any nonneuronal progenitor cells. Recordings were collected every other day with the Maestro MEA system (Axion Biosystems) at 37°C and 5% CO₂, for 5 min using the AxIS Software Spontaneous Neural Configuration (Axion Biosystems, https://www.axionbiosystems.com/products/axis-software, RRID:SCR_016308). The plate was equilibrated in the machine for 2 min prior to each 5 min recording. Cells were collected on days 17 and 27 of neural differentiation and total cell count was used to normalize the neuronal activity data. Axion Biosystems's Neural Metrics Tool classifies electrodes with at least five spikes per minute as active. Bursts were identified using an interspike interval (ISI) threshold requiring a 5-spike minimum and 100-ms maximum ISI.

2.10 Statistical analysis

One-way ANOVA test was employed to test the significance of ACTL6B protein levels. To evaluate the significance of gene expression data between affected and control, two-tailed t-test was used for each gene. For the analysis of neuronal activity, normalized mean firing rate data were evaluated using two-way ANOVA with Geisser–Greenhouse correction. GraphPad Prism (https://www.graphpad.com/, RRID:SCR_002798) 8.4 was used to perform the analysis. A value of p ≤ 0.05 was considered as statistically significant, with p ≤ 0.05 denoted as *, p ≤ 0.01 denoted as **, p ≤ 0.001 denoted as ***, p ≤ 0.0001 denoted as ****, and ns as not significant. Figure 4b: Unaffected d10 versus Unaffected d20, p = 0.0001; Unaffected d10 versus Affected d10, p = 0.1674; Unaffected d10 versus Affected d20, p = 0.2001; Unaffected d20 versus Affected d10, p = 0.0005; Unaffected d20 versus Affected d20, p = 0.0006; Affected d10 versus Affected d20, p = 0.9991. Figure 4c: Unaffected neurons versus Affected neurons, p = 0.2514. Figure 5a ESPNP comparison, p = 0.0018; MAP2K3 comparison, p = 0.0356; SOX8 comparison, p = 0.0246; TPPP comparison, p = 0.6632; KCNJ12 comparison, p = 0.0026. Figure 5c: Unaffected versus Affected p = 0.0010. Figure 5d: Unaffected versus Affected p = 0.0317.

3.1 p.(Val421_Cys425del) is predicted to disrupt the COOH-terminal domain structure of ACTL6B

Previously reported pathogenic variants in ACTL6B are located throughout the transcript, the majority of which are terminating mutations predicted to elicit nonsense-mediated decay (NMD; Bell et al., 2019; Wenderski et al., 2020). We recently described an extended consanguineous family with six affected children with EIEE-76 caused by a novel biallelic mutation in ACTL6B c.1261_1275del; p.(Val421_Cys425del) (Yüksel et al., 2019). The in-frame deletion of five C-terminal amino acids occurs in the ultimate exon (Exon 14) and is unlikely to elicit NMD (Khajavi et al., 2006). The amino acids are found within a highly conserved sequence across multiple actin-related proteins (Figure 1a). To predict whether the loss of these residues is likely to disrupt ACTL6B function, we modeled the mutation on the crystal structure of the paralogous protein ACTB (UniProt, P60709, https://www.uniprot.org/uniprot/P60709, RRID:SCR_002380) (Galkin et al., 2008). Both ACTL6B and ACTB have actin-related domains that are composed of four conserved subdomains. Residues Val421_Cys425 near the COOH-terminal are part of subdomain 1 that contains residues that interact with ATP (Holmes et al., 1990; Otterbein, 2001). Our in silico modeling revealed that ACTL6B p.(Val421_Cys425del) would disrupt eight inferred hydrogen bonds at the COOH terminus, which may impact protein structure or function (Figure 1b).

3.2 ACTL6B p.(Val421_Cys425del) does not alter induced pluripotent stem cell reprogramming or differentiation of neural progenitor cells to neurons

Prior models of ACTL6B-associated epileptic encephalopathy have been developed from patient-derived cells harboring frameshift variants or engineered as null lines (Bell et al., 2019). To assess the pathogenicity of ACTL6B p.(Val421_Cys425del) and its effect on neuronal function, we reprogrammed primary fibroblasts isolated from an affected EIEE-76 patient harboring a biallelic ACTL6B p.(Val421_Cys425del) mutation as well as a familial unaffected carrier to iPSCs (Figure 2) (Yüksel et al., 2019). The iPSC clones were indistinguishable and expressed classic stem cell markers NANOG, SOX2, OCT4, and PODXL (Figure 2a,b). We performed karyotyping using the sensitive Karyostat SNP analysis and found a grossly normal karyotype for all clones. Unaffected clone CW067 #9 (U9) had 2 Mb duplication on chromosome 18, smaller than the 5–10 Mb resolution of typical G-band karyotyping. The small duplication in U9 contained LINC01541, LOC102724913, CBLN2, NETO1, MIR548AV, and LOC100505797, genes which have not been associated with epilepsy. Spontaneously differentiated EBs generated from iPSC lines expressed markers for the endodermal (AFP) and mesodermal (MSX1) lineage (Figure 2c), while the ectodermal lineage was confirmed by the differentiation and expression of NPC markers (Figure 3b), demonstrating pluripotency.

No differences were observed in iPSC reprogramming and differentiation to NPCs between affected and unaffected cells, as expected. All iPSCs formed EBs and neural rosettes upon neural induction (Figure 3a,b). NPCs were co-positive for NESTIN and PAX6 (Figure 3b) or SOX2 and NESTIN (Figure S1a) and lacked expression of neural crest (SOX10 and FOXD3) and stem cell (OCT4) markers (Figure 3c). To examine molecular pathogenesis of ACTL6B p.(Val421_Cys425del), affected and unaffected-derived NPCs were differentiated to neurons that expressed neuronal markers MAP2 and TUJ1 (Figure 3d) and lacked astrocyte marker glial fibrillary acidic protein (GFAP) (Figure S1b).

Next, we assessed ACTL6B expression in our neuron cultures by western blot, following validation of antibody specificity (Figure 3e). Differentiating neurons expressed TUJ1 by day 10 and ACTL6B by day 3, both of which increased during maturation (Figure 3f). Together, these findings validate our EIEE-76 iPSC model, which recapitulates the developmental increase in ACTL6B expression during neurogenesis (Wu et al., 2007).

3.3 ACTL6B p.(Val421_Cys425del) impairs protein stability

Our 3D protein modeling of ACTL6B p.(Val421_Cys425del) predicted a significant loss of inferred hydrogen bonds in a conserved subdomain necessary for ATP binding (Holmes et al., 1990; Otterbein, 2001) (Figure 1b); therefore, we hypothesized that p.(Val421_Cys425del) may impair ACTL6B expression. To test this, we assessed ACTL6B protein levels at 0, 10, and 20 days of neural differentiation in affected and unaffected neurons. Similar levels were observed at day 10; however, ACTL6B was reduced by 31.4-fold at day 20 in affected neurons compared to unaffected controls (Figure 4a). Notably, ACTL6B expression increased 4.3-fold by day 20 of differentiation in unaffected cortical neurons, while affected neurons maintained a steady, low level of ACTL6B p.(Val421_Cys425del) (Figure 4a).

To examine whether reduced ACTL6B levels in affected neurons was due to impaired protein stability, we measured protein half-life in day 31 neurons using cycloheximide to inhibit protein synthesis. In affected neurons, approximately half (49%) of ACTL6B p.(Val421_Cys425del) was degraded 48 hrs post-cycloheximide treatment, while unaffected neurons had 96% of ACTL6B remaining (Figure 4b). This finding demonstrates the half-life of ACTL6B p.(Val421_Cys425del) is shorter than wild-type ACTL6B. We next confirmed ACTL6B gene expression levels were not changed in affected and unaffected neurons by quantitative RT-PCR for ACTL6B mRNA and found no differences (Figure 4c). Together, these results indicate that decreased ACTL6B protein accumulation in EIEE-76-derived neurons is caused by reduced protein stability rather than co-transcriptional regulation, consistent with our in silico protein modeling predictions.

3.4 ACTL6B p.(Val421_Cys425del) causes increased neural activity

Since ACTL6B is a component of the nBAF chromatin remodeling complex, we next tested whether ACTL6B p.(Val421_Cys425del) impacts the expression of previously described nBAF target genes (Bell et al., 2019). ESPNP, MAP2K3, SOX8, TPPP, and KCNJ12 are differentially expressed in neurons with ACTL6B p.(Ter427AspextTer33) during neurogenesis compared to controls (Bell et al., 2019), and linked to neurological phenotypes (Mertens et al., 2015; Veglianese et al., 2006). We found EIEE-76 patient iPSC-derived affected neurons had elevated expression of KCNJ12 and MAP2K3 and reduced expression of ESPNP and SOX8, while TPPP was unchanged compared to unaffected controls (Figure 5a). These results suggest ACTL6B p.(Val421_Cys425del) alters nBAF function during neurogenesis.

Among the main clinical manifestations of EIEE-76 patients are severe seizures and EEG abnormalities, therefore, we sought to measure neuronal function of patients' iPSC-derived neurons using MEA technology. For this purpose, EIEE-76 iPSC-derived NPCs harboring ACTL6B p.(Val421_Cys425del) and unaffected control NPCs were plated in 8 wells per clone of a 48-well MEA plate at identical densities (Figure 5b). MEA electrophysiology recordings enabled the analysis of extracellular action potentials as a measurement of neural functionality (mean firing rate) and excitability (bursts) (Figure 5b). Spontaneous action potentials were recorded every 2 days, starting on day 7 of neural differentiation and carried out until day 25 (Figure 5c). Data were normalized to the number of cells to control for potential neuronal loss during culture. We observed spontaneous neuronal activity starting at day 13 in both affected and unaffected iPSC-derived neurons, which remained active throughout the duration of the recordings (Figures 5c and S2). We found that EIEE-76 patient-derived neurons have increased mean firing rate, indicative of elevated neuronal activity (Figure 5c). Additionally, we assessed burst frequency on day 21 of neural differentiation, the time point showing reduced ACTL6B accumulation, and discovered a significant increase in EIEE-76 patient-derived neurons compared to controls (Figure 5d), suggesting hyperexcitability. Taken together, these results are consistent with the EIEE-76 clinical presentation of refractory seizures and EEG abnormalities observed in patients.