Introduction

Human induced pluripotent stem cell (hiPSC)-derived cell types have become critical tools in modeling disease biology and drug discovery, enabling the study of causal genes in the context of human disease relevant cell types. In particular, hiPSC-derived neurons offer unique access to human cellular neurobiology, which has been limited due to the inaccessibility of primary cell types. Development of methods for large scale differentiation of human neurons, CRISPR editing, and the establishment of genetically diverse hiPSC repositories are increasingly enabling modeling of neurological diseases for target discovery and drug screening1,2. hiPSC-derived neurons have been widely used in drug development for in vitro modeling of a variety of neurological diseases, especially neurodevelopmental epilepsies3,4,5,6,7,8,9.

Overexpression of the transcription factor neurogenin-2 (NGN2) can reproducibly generate excitatory cortical neurons2,5. Both viral transduction and gene-edited inducible expression systems now allow routine, large scale generation of human neurons. Moreover, multiple groups have demonstrated that NGN2 neurons develop action potential firing and excitatory synaptic network activities as they mature in vitro1,2.

Access to human neurons at scale has increased the need for higher throughput approaches to measure their maturation, excitability, and synaptic biology. While traditional patch clamp electrophysiology remains the gold standard, such labor-intensive methods do not scale, in particular for measurements of synaptic properties. Multielectrode array (MEA) technology allows recording of extracellular neuronal activity, such as spontaneous action potential firing, in tissue culture plates compatible with routine cell culture. Most widely used MEA systems sample neuronal activity from 8 to 64 electrodes per well, detecting gross patterns of neuronal network with high temporal but low spatial resolution10,11. Conversely, imaging-based systems, recording fluorescence changes related to voltage or intracellular ions such as calcium, may offer cellular spatial resolution but are generally limited in their ability to resolve individual action potential events due to lower temporal resolution12,13,14.

Here, we leverage a hiPSC-derived NGN2 neuronal system lacking STXBP1 to systematically compare multiple approaches to quantify synaptic activity with medium throughput. The STXBP1 gene encodes syntaxin-binding protein 1, or Munc18-1, which is critically involved in synaptic vesicle docking, priming, and fusion in the presynaptic terminal in both rodent and human neurons15,16,17. In this study, we cross validate multiple plate-based physiological assays to show an impairment of presynaptic machinery. Knocking out STXBP1 preferentially disrupts spontaneous network activity and neurotransmitter release, while retaining the activity of individual cells. Restoration of STXBP1 protein demonstrates the sensitivity of assays to detect dose-dependent phenotypic rescue. This work demonstrates the utility of hiPSC neuronal systems and medium throughput assays to study presynaptic targets relevant to human neurological diseases.

Results

NGN2 neuronal system develops synaptic driven network activity

To generate large numbers of homogenous hiPSC-derived neurons, we used a doxycycline-inducible NGN2 transgene engineered into the AAVS1 safe harbor locus of a parental iPS similar to systems described previously2,18. To confirm the development of action potential firing and synaptic transmission, we characterized membrane properties of NGN2 neurons with whole-cell patch clamp recordings. Following three weeks in culture, 77% of neurons fired action potentials when depolarized with current injection, relative to the 95% at 6 weeks (Fig. 1A,B). While action potential properties matured between 3 and 6 weeks (Supplementary Table 1), both 3- and 6-week NGN2 neurons could maintain firing frequencies > 10Hz during a 1s current injection. Synchronized spontaneous firing was increased in 6-week compared with 3-week cultures (Fig. 1C,D), which was accompanied by an increase in spontaneous excitatory postsynaptic current frequency (sEPSC, Fig. 1E,F). Taken together, these observations suggest that while 3-week NGN2 neurons are able to fire action potentials, increased maturation between 3 and 6 weeks, leading to enhanced synaptic transmission, may drive increased network activity.

Fig. 1
figure 1

Properties of NGN2 neurons characterized with whole-cell patch clamp recordings. (A) Both 3-week (n = 21) and 6-week (n = 18) neurons displayed evoked action potential firing with increasing current injection. Left panel: percentage of neurons with evoked action potentials. Right panel: repetitive action potential firing in response to current steps from 0 to 100pA. (B) Example traces of evoked action potential firing. (C) An increase in the percentage of cells showing spontaneous firing out of all cells sampled from week 3 to week 6 demonstrating maturation. (D) Example traces of spontaneous firing in 3-week and 6-week neurons. (E) An increase in sEPSC frequency but not amplitude in 6-week neurons vs. 3-week neurons demonstrating presynaptic maturation. (F) Example traces of sEPSCs. *p < 0.1, Student’s t-test. Error bars represent mean ± SEM.

Full size image

To longitudinally monitor NGN2 network activity with increased throughput, we cultured neurons on 96-well MEA plates and recorded spontaneous activity over 6 weeks. We first asked whether plating density impacted the development of spontaneous activity. We performed weekly recordings with plating densities of 25 k, 50 k, and 100 k cells per well. Neurons plated at the higher densities (50 k and 100 k cells per well) exhibited more robust activity across multiple commonly used metrics of activity including higher spike rates, higher number and magnitude of network bursts, as well as a higher fraction of spikes in bursts (Fig. 2A). Based on these observations, we chose to use 100 k cells per well for future experiments. Between 3 and 6 weeks in vitro, we observed a large increase in spontaneous firing rate, and the emergence of coordinated bursting across the active electrodes in each MEA well (Fig. 2B), reminiscent of the changes observed with patch clamp. To determine whether the activity observed in 6-week cultures was driven by chemical synapses, we applied the postsynaptic neurotransmitter receptor blockers D-APV, CNQX, and Gabazine to block NMDA, AMPA, and GABA receptors, respectively. While DAPV and Gabazine had minimal effects on spontaneous activity, we found that CNQX reduced the spontaneous spike rate by 64%, while network bursting was completely abolished (Fig. 2C,D). These data demonstrate that AMPA-receptor mediated synaptic activity is necessary for network bursting and, conversely, that network bursting recorded by MEA can be used as a proxy for synaptic activity.

Fig. 2
figure 2

Properties of NGN2 neurons characterized with multi-electrode array. (A) Longitudinal MEA recording of mean spike rates, number of network bursts, burst magnitude, and burst fraction, when neurons were plated at 25 k, 50 k, and 100 k per well density. N = 8. (B) Increase in firing and bursting metrics from 3-week to 6-week neurons demonstrates maturation when cells were plated at 100 k/well density. (C) Quantifying normalized spike rates and burst rates with CNQX, DAPV, and Gabazine treatments. (D) Histogram showing well spike counts over 300s recording period; network bursts in 6-week NGN2 neurons are blocked by CNQX but not affected by DAPV or Gabazine. N = 12. **p < 0.01, ****p < 0.0001, one-way ANOVA. Error bars represent mean ± SEM.

Full size image

Spontaneous network bursting requires presynaptic transmission machinery

Previous studies have shown that STXBP1 is necessary for synaptic transmission in a NGN2 neuronal system as assayed by patch clamp15,19. This observation, combined with our, and others observation, that synaptic transmission is necessary for NGN2 network bursting, suggest that MEA recording should be sensitive to loss of STXBP1. To test this hypothesis, we leveraged CRISPR-editing to knock out (KO) STXBP1 in hiPSCs (Fig. 3A,B), harboring an inducible NGN2 expression system. Similar to previous reports, STXBP1 homozygous KO NGN2 neurons lacked miniature excitatory postsynaptic currents (mEPSCs), consistent with an inability to spontaneously release synaptic vesicles (Fig. 3C,D). The complete loss of mEPSCs precluded analysis of their amplitude in STXBP1 KO NGN2 neurons.

Fig. 3
figure 3

Patch clamp and MEA recording demonstrating synaptic phenotype of STXBP1 KO NGN2 neurons. (A) Illustration of editing on the STXBP1 gene to generate STXBP1 knockout hiPSC line. (B) Quantification of STXBP1 protein with automated western blot system. (C) Whole cell voltage clamp recording of miniature EPSC frequency in STXBP1 ISO control (n = 13) vs. KO NGN2 neurons (n = 5). (D) Example traces of mEPSCs in STXBP1 ISO control and KO NGN2 neurons. (E) MEA recording of STXBP1 ISO control and KO NGN2 action potential firing from DIV7 to 35. Left panel: time course of spike rate for both genotypes. Right panel: significantly reduced spike rate on DIV28 in STXBP1 KO NGN2 neurons vs. ISO control. N = 20. (F) MEA recording of STXBP1 ISO control and KO NGN2 network bursting from DIV7 to 35-. Left panel: time course network bursts for both genotypes. Right panel: lack of network bursts on DIV28 in STXBP1 KO NGN2 neurons vs. ISO control. N = 20.

Full size image

Next, we used MEA recordings to assess the development of spontaneous network activity of STXBP1 KO neurons. We observed that spontaneous firing was reduced across 5 weeks of maturation (Fig. 3E); in addition, we found a complete lack of network bursting consistent with the absence of synaptic currents in whole cell recordings (Fig. 3F). Interestingly, we noted a steady increase in the spontaneous firing rate of KO neurons. Despite increasing firing rates, we did not observe network bursts at any timepoint (Fig. 3F), highlighting that the burst phenotypes observed are unlikely due to an inability of KO neurons to fire action potentials.

Previous studies of STXBP1 in hiPSC neurons have found a 50% survival deficit in STXBP1 KO NGN2 neurons15. To determine whether a survival deficit confounds our network activity results, we measured the viability of STXBP1 KO and isogenic control neurons with CellTiter-Glo. At DIV42, we found approximately 62% ATP level in KO relative to WT neurons (Fig. 4A), suggesting a survival deficit less pronounced than previously reported. To directly assess whether a reduced number of active neurons was contributing to the network bursting phenotype, we turned to a calcium imaging assay that retains cellular resolution. Following transduction of NGN2 neurons with a genetically encoded calcium indicator, we recorded a steady increase in neurons with calcium transients, and the emergence of coordinated activity over 6 weeks, comparable to the development of network activity recorded on the MEA. Calcium imaging also revealed a large reduction in the number of active neurons in KO cultures (Fig. 4B), suggesting the low firing rate in MEA is associated with a lower number of active neurons.

Fig. 4
figure 4

Calcium imaging demonstrating synaptic phenotype of STXBP1 KO NGN2 neurons. (A) Survival deficit in STXBP1 KO neurons quantified with CellTiter Glo. (B) Measurement of active cell counts in NGN2 neurons and NGN2 neurons co-cultured with human primary astrocytes for both genotypes. Left panel: time course quantifying active cell count from from DIV18 to DIV42. Right panel: active cell counts on DIV28. N = 10. (C) Measurement of coordinated network activity in NGN2 monoculture and co-culture with human primary astrocytes for both genotypes. Left panel: time course from DIV18 to DIV42. Right panel: quantifying coordinated network bursting on DIV28. N = 10. (D) Representative images from Incucyte Neuronal activity assay. Fluorescent images displaying the range in fluorescence for each pixel. *p < 0.1, **p < 0.01, ****p < 0.0001, two-way ANOVA. ****p < 0.0001, Student’s t-test. Error bars represent mean ± SEM.

Full size image

To conclusively rule out a role of reduced active cells in the network activity phenotype, we co-cultured WT and KO NGN2 neurons on primary human astrocytes to provide extra trophic support for their survival and maturation. As expected, we observed a 25-fold increase in the numbers of active STXBP1 KO neurons compared to NGN2 neurons alone confirming that astrocyte co-culture supported the survival and/or maturation of STXBP1 KO neurons (Fig. 4B,D). Despite the increase in active neurons, STXBP1 KO neurons failed to generate coordinated activity at any timepoint (Fig. 4C). Taken together, our characterization of plating density in MEA and calcium imaging suggests the lack of network bursting in STXBP1 KO NGN2 neurons is due to the loss of synaptic transmission, with minimal contribution of cell survival.

Re-constitution of presynaptic proteins can rescue deficits in synaptic transmission and network activity

Based on the observed synaptic and network activity deficits in STXBP1 homozygous KO NGN2 neurons, we attempted to restore synaptic function by re-expressing STXBP1 protein with AAV transduction (Fig. 5A). We first recorded mEPSCs from STXBP1 KO neurons treated with STXBP1 or control AAV at 150k MOI. We found that NGN2 neurons treated with STXBP1 AAV rescued mEPSC frequency to 71% of ISO control neurons (Fig. 5B,D), with 79% of recorded neurons exhibiting mEPSCs compared to 0% in the control-treated STXBP1 KO neurons (Fig. 5E). Likewise, the amplitude of mEPSCs in AAV-treated STXBP1 KO neurons was similar to WT (Fig. 5B).

Fig. 5
figure 5

Patch clamp recording of STXBP1 KO NGN2 neurons following STXBP1 re-constitution. (A) Illustration of AAV.PHPB capsid with STXBP1 construct. (B) Whole-cell voltage clamp recording of mEPSCs in STXBP1 ISO vs. KO neurons with control AAV and STXBP1-expressing AAV treatments. Left: quantifying mEPSC frequency. Right: quantifying mEPSC amplitude. (C) Example patch recording traces of mEPSCs. (D) Frequency distribution of mEPSC frequency. (E) Quantifying percentage of patched neurons with mEPSC present.

Full size image

Since MEA recordings allowed us to efficiently examine multiple treatment conditions longitudinally, NGN2 neurons were cultured as before on MEA plates and treated with either an mCherry-expressing control AAV or an AAV encoding STXBP1, following 3 days in culture. As expected, control treated STXBP1 KO neurons had lower spike rates and lacked network bursting (Fig. 6A,B). Comparing multiple MEA metrics, we found that treatment with 75k MOI AAV-STXBP1 was sufficient to restore bursting activity similar to WT at DIV28, although the onset of bursting activity was delayed relative to ISO control neurons (Fig. 6B), potentially reflecting the time course for AAV transgene expression. Further analysis on DIV35 revealed a viral dose-dependent pattern of MEA activity in STXBP1 KO NGN2 neurons. This was particularly pronounced in metrics related to network bursting, where we observed dose-dependent increases in the number of network bursts, burst magnitude, and the number of spikes in bursts (Fig. 6B,C). While in this experiment, spike rates were comparable across all groups at DIV35 (Fig. 6A), control treated KO neurons lacked network bursting. When we compared MEA activity associated with STXBP1 protein rescue, we found that 25k MOI treatment already significantly increased network activity and that 75k MOI treatment, which led to about 50% WT expression of STXBP1, was sufficient to rescue network bursting to WT level (Fig. 6D,E), suggesting that NGN2 MEA network activity does not require a complete rescue of STXBP1.

Fig. 6
figure 6

MEA characterization of STXBP1 KO NGN2 neurons following STXBP1 re-constitution. (A) Longitudinal MEA recording for mean firing rate demonstrating STXBP1 KO phenotype and dose-dependent rescue. Left panel: time course from 1 to 6 weeks. Right panel: mean firing rates on DIV35 are comparable across treatment groups. (B) Longitudinal MEA recording for number of network bursts demonstrating STXBP1 KO phenotype and dose-dependent rescue. Left panel: time course from 1 to 6 weeks. Right panel: number of network bursts, burst magnitude, and burst fraction on DIV35 across treatment groups. N = 20. ****p < 0.0001, one-way ANOVA. (C) Histogram of well spike counts over 20min recording demonstrating dose-dependent rescue of network bursts. (D) Relative STXBP1 protein expression in STXBP1 KO NGN2s following AAV transduction to WT NGN2 neurons at DIV21, 28, and 35. (E) Correlation between MEA activity parameters relative to STXBP1 protein expression on DIV35. *p < 0.1; ****p < 0.0001; ns not significant; one-way ANOVA. Error bars represent mean ± SEM.

Full size image

Detecting neurotransmitter release as a direct measure of presynaptic function

While both MEA and calcium imaging enable longitudinal measurements with relatively high throughput, each of these technologies is an indirect measurement of presynaptic function, specifically synaptic vesicle release. We therefore explored whether measurements of glutamate released into the extracellular media could provide a direct, plate-based measurement of presynaptic function.

First, we investigated glutamate released from spontaneous network activity in WT and STXBP1 KO NGN2 neurons. To do this, we performed complete media change at DIV35 and incubated the cells for 48 h before taking media samples for a luminescence-based glutamate measurement. We found that media glutamate levels in STXBP1 KO NGN2 were less than WT neurons by 44% (Fig. 7A). We did not observe any significant difference between glutamate levels in cell lysates between the two genotypes (Fig. 7A), suggesting reduced media glutamate levels were not due to a lack of intracellular glutamate available for release. While promising, this result could be confounded by our previous observation of decreased spontaneous burst and spike activity in our MEA experiments. To examine whether STXBP1 KO neurons release less glutamate when activated similar to WT neurons, we used potassium chloride (KCl) stimulation. When DIV35 neurons were depolarized with artificial cerebrospinal fluid (aCSF) containing 50mM KCl for 5 min, we observed an approximately threefold increase of glutamate release compared with standard aCSF treatment in WT healthy neurons (Fig. 7B), which was reduced to 70.8% in STXBP1 KO neurons, similar to the spontaneous release results. Furthermore, we found KCl-evoked glutamate release was rescued to WT levels when treated with 75 k MOI AAV-STXBP1, with intermediate rescue with 7.5 and 25 k MOI (Fig. 7C).

Fig. 7
figure 7

Spontaneous and evoked glutamate release in DIV35 NGN2 neurons. (A) Extracellular and intracellular glutamate levels in STXBP1 ISO control and KO neurons. Left panel: Spontaneous glutamate release into neuron culture media vs. blank media. Right panel: glutamate levels in neuron lysates. N = 20. (B) Healthy WT NGN2 neurons spontaneously releasing glutamate vs. evoked glutamate release with high K + stimulation for 5 min. (C) Evoked glutamate release with high K + aCSF in STXBP1 ISO control and KO neurons with increasing dosage of STXBP1 AAV or 75k MOI control AAV. n = 5. ***p < 0.001, ****p < 0.0001, One-way ANOVA. Error bars represent mean ± SEM.

Full size image

Discussion

Here, we describe a set of complementary assays suitable to investigate synaptic phenotype and rescue, with medium throughput, in hiPSC-derived neurons. Our data suggest that network activity recorded with MEA or calcium imaging is sensitive to loss of synaptic transmission, which is also reflected in reduced evoked extracellular glutamate levels. Likewise, re-expression of synaptic proteins restores network activity, highlighting utility of these platforms to characterize molecules modulating synaptic properties with significantly improved throughput compared to patch clamp electrophysiology.

Our study builds off previous work characterizing STXBP1 function in hiPSC derived neurons as a test case for assay sensitivity to detect phenotype and rescue. Importantly, the current study was not designed to recapitulate the complex circuit abnormalities that underlie STXBP1 developmental and epileptic encephalopathy (DEE)15,16,17,19,20. That MEA bursting parameters return to normal with ~ 50% WT protein levels suggests that haploinsufficient STXBP1 patient neurons would likely exhibit different phenotypes, if any, in this NGN2 system. For example, van Berkel et al., 2003 demonstrated variable calcium imaging network phenotypes across a set of STXBP1 patient hiPSC-derived neurons. Despite this limitation, in the present work, NGN2 neurons provide a scalable, homogenous population of neurons to examine the function and rescue of disease relevant genes in a human cellular system.

Consistent with previous reports using NGN2 neuronal systems, we find neurons acquire active membrane and synaptic properties over a period of several weeks, culminating in robust synaptic driven network activity11,21. We find that both MEA and calcium imaging provide complementary perspectives on network activity phenotypes. Direct action potential detection in MEA allows quantification of the magnitude and structure of bursting, while calcium imaging supports a direct quantification of cell numbers and their correlations. We have corroborated our MEA and calcium imaging with focused patch clamp studies, suggesting these methods are sensitive to cellular and synaptic properties traditionally quantified with single-cell recordings. In the present study we found that both deficits in presynaptic (via STXBP1 KO) and postsynaptic function via neurotransmitter receptor block (CNQX) led to similar impacts on network activity. In future studies it will be important to identify whether specific network activity signatures are associated with pre- vs post-synaptic dysfunction. Similarly, we would predict that genetic perturbations altering neuronal excitability would also impact network activity (e.g.8,22). While our work highlights the power of network activity recording to scale physiological experiments, these observations emphasize the need to corroborate network activity phenotypes with higher resolution physiological methods like patch clamp.

MEA and calcium imaging each revealed a clear phenotype in STXBP1 KO neurons. Similar to acute block of synaptic transmission, we found that perturbation of presynaptic machinery completely disrupted the coordinated network activity between NGN2 neurons. While our data do not rule out STXBP1 KO altering neuronal excitability, our co-culture calcium imaging experiments highlight that despite large numbers of active neurons, KO cultures are unable to coordinate their firing activity. Taken together, these data demonstrate how analysis of NGN2 network activity patterns can reveal synaptic dysfunctions. Numerous synaptic genes have been associated with neurodevelopmental and pediatric epilepsy disorders, including SNARE associated proteins, like STXBP1, as well as synaptic adhesion, scaffolding genes, and neurotransmitter receptors23. Whether the NGN2 system presented here is sensitive to any given synaptic deficit will need to be determined on a case-by-case basis. For example, our data suggest NGN2 synaptic activity is independent of NMDA receptor mediated transmission. Likewise, our pure excitatory neuron culture system lacks spontaneous GABAergic transmission. Therefore, more complex hiPSC-derived neuronal systems would be required to probe biology around these synaptic systems.

In order to directly examine presynaptic function, we measured glutamate release into culture media due to spontaneous network activity or chemical stimulation. In contrast to imaging-based methods, such as FM dye labeling24,25, this method was scalable to higher-throughput formats, allowing parallel measurement of several conditions with reduced assay complexity. It remains to be determined whether this method is sensitive to more subtle stimuli, such as methods to specifically engage the readily releasable vesicle pool26,27,28.

Our data identify a set of assays sensitive to both the loss and graded rescue of synaptic transmission. We believe the use of medium throughput plate-based physiology assays, as deployed here, will facilitate increased use of human cellular systems to validate drug targets and enable screening of candidate therapeutic molecules.

Methods

Genome editing and NGN2 neuron production

STXBP1−/− and isogenic iPSC-derived NGN2 neurons were produced and provided by iXCell Biotechnologies using CRISPR targeting STXBP1 exon 5 on the Tet-inducible human iPSC background. In short, two sgRNAs and Cas9-GFP vectors were co-transfected into a wildtype parental hiPSCs using lipofectamine transfection. The GFP positive hiPSCs were purified by FACS and seeded onto irradiated mouse embryonic fibroblast pre-cultured overnight in a 6-well plate, single clone was isolated into a well of Matrigel-coated 96-well plate and expanded prior to screening for STXBP1 knockout clones. The genomic DNAs were extracted after recovery of clones for sequence confirmation and tested for mycoplasma by PCR amplification of a ~ 270 bp sequence conserved in most mycoplasma species. hiPSC-derived NGN2 neurons were generated as published with minor modifications18. Briefly, hiPSC’s engineered with an inducible NGN2 transgene were seeded at 5 × 104 cells/cm2 in mTeSR Plus containing 10 uM Y-27632. Next day, the culture medium was changed to differentiation medium containing 2 μg/mL doxycycline to induce NGN2 expression. 48 h later, immature neurons were dissociated with Accutase and cryopreserved in CryoStor CS10 freezing medium (STEMCELL, 07930). Cryopreserved NGN2 neurons were thawed and plated in BrainPhys neuronal media containing SM1 and N2A (StemCell technologies) and further supplemented with BDNF (10 ng/mL), GDNF (10 ng/mL) and 0.2 mM Ascorbic acid. Culture One (Gibco), and Doxycycline (1 μg/mL) was added to media for the first 7 days post plating. This media formulation, subsequently referred to as, BrainPhys media, was used for all NGN2 culture experiments. The day of NGN2 neuron plating was considered DIV0.

AAV vector production

Recombinant AAV viral vector particles were produced and provided by Packgene Biotech Inc. The STXBP1 construct encodes a codon-optimized human STXBP1 sequence driven by a neuronal specific promoter. In the control vector, the CDS for STXBP1 was replaced with a codon-optimized sequence encoding mCherry. Both vectors were packaged in PhP.B capsids by triple transfection in HEK293 cells, purified by iodixanol-based ultracentrifugation and formulated in Dulbecco’s PBS 1x + Pluronic F68 (10 ppm). Genome titer of the viral vectors was re-evaluated internally at Biogen with ddPCR.

NGN2 neuron and human primary astrocyte co-culture

Frozen vials of human primary astrocytes (ScienCell Cat#1800) were thawed and plated in poly-D-lysine-coated T75 vessels containing Astrocyte Medium (ScienCell Cat#1801) at 300,000 cells/T75. Culture medium was refreshed after 2 days to remove unattached cells. Once confluent, astrocytes were dissociated with 0.25% trypsin and resuspended astrocyte media supplemented per manufacturer instructions with FBS and astrocyte growth supplement, and re-plated in PDL-coated (0.1 mg/ml) 96-well plates 40 k/well. Once astrocytes have become confluent and formed a monolayer, remove the astrocyte media completely and add human iPSC derived NGN2 neuron suspension in BrainPhys media (as described above) suspension on top of the astrocyte layer. For monocultures, frozen vials of NGN2 neurons were thawed and plated directly on PDL-coated (0.1 mg/ml) plates without the astrocyte layer. Half-media change was then performed with BrainPhys media every 3–4 days. NGN2 neurons were dosed with AAV on DIV2 for rescue studies.

Sally Sue automated western blot system for protein quantification

Frozen vials of wild-type and STXBP1 homozygous knockout NGN2 neurons were thawed and plated on PDL-coated 12-well plates with 400,000 cells per well. Lysates from NGN2 cultures were collected on ice using phosphate-buffered saline (PBS) wash (3x) followed by cell lysis using RIPA (Thermo Scientific; 89900) with phosphatase and protease inhibitors (Thermo Scientific; 78444). Protein was detected using the Sally Sue Simple Western size-based capillary electrophoresis system (Bio-techne) with the 12–230 kDa separation kit. STXBP1 and actin proteins were detected using primary antibodies for STXBP1 at 1:200 (Sigma; HPA008209) and actin at 1:300 (Cell Signaling Technology; 8H10D10), both diluted in antibody diluent 2 (Bio-techne). Secondary anti-mouse (Bio-techne; NC1178181) and anti-rabbit (Bio-techne; NC1196891) antibodies were used with streptavidin-HRP for protein and ladder detection (Bio-techne; 042-414). Cell lysates were diluted using 0.1 × sample buffer and prepared with 5 × master mix containing DTT, SDS, and fluorescent standards (Bio-techne). Samples were denatured at 95 °C for 5 min and pipetted onto a 384-well assay plate. Antibodies and samples were pipetted into designated wells in a 384-well plate. Sally Sue was performed following manufacturer’s instructions, and signals were detected and analyzed using the Compass for Simple Western Software (Bio-techne, version 5.0.1). STXBP1 protein was normalized to Actin loading control then further normalized to wild-type control.

Multi-electrode array (MEA)

NGN2 neurons were plated on PDL-coated Axion 96-well MEA plates. PDL was diluted to 0.1 mg/mL in PBS and plates were coated overnight at 4 °C, washed 3 × with PBS and allowed to air dry for 30–60 min prior to cell seeding. Neurons were plated at a density of 100k cells/well to ensure sufficient neural activity and synaptic interaction. The day of NGN2 neuron plating was considered DIV 0. Neurons were grown in BrainPhys neuronal media as described above. Half-media change was performed every 3–4 days. Electrical activities were recorded for 20 min with the Maestro Pro system, twice a week before half media change since the initial 7 days of plating. Recording chamber was maintained at 37 °C/95% O2/5% CO2. Signals were sampled at 12.5 kHz frequency and filtered with a high-pass 200Hz IIR and a low-pass 3 kHz Kaiser Window. Network bursts were detected with custom MATLAB scripts. Briefly, well wide spike counts were calculated for 1s bins and network bursts were considered as peaks in activity with over 100 spikes in a bin.

Whole-cell patch clamp electrophysiology

NGN2 neurons were plated on 12 mm Corning PDL-coated coverslips (CAT. # 354086), (250,000 cells/ coverslip) and cultured in BrainPhys media as described above. Cultured NGN2s were then transferred to the recording chamber containing bath solution at the time of experiment. At room temperature, membrane potentials held at − 65 mV and synaptic transmission or voltage-dependent currents were recorded in the whole-cell voltage-clamp mode. To record miniature EPSCs (mEPSCs), 1 uM of TTX was added to bath solution to block action potentials and thereby mEPSCs were recorded for 5 min. Spontaneous action potential firing was recorded in gap-free current clamp mode with 0 current injection. Evoked action potential firing was sampled at 10 kHz with step-current injections from − 50 pA with a step of 10 pA and a total of 16 sweeps. The bath solution contains 139 mM NaCl, 4 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 10 mM D-glucose, and 10 mM HEPES. For current clamp, the pipette solution contains 130 mM KMeSO4, 10 mM KCl, 4 mM ATP magnesium, 0.3 mM GTP disodium, 10 mM phosphocreatine disodium, 0.7 mM EGTA, and 10 mM HEPES; for voltage clamp, pipette solution contains cesium methane sulfonate 130 mM, CsCl 10 mM, HEPES 10 mM, ATP Mg salt 4 mM, GTP Na salt 0.3mM, EGTA 0.2 mM, phosphocreatine disodium salt 10 mM.

Calcium imaging

NGN2 neuron monoculture and astrocyte co-cultures were transduced with the Incucyte Neuroburst Orange lentivirus (Sartorius Cat. #4736) on DIV2. This product is a lentiviral-based live-cell neuronal labeling reagent driven by a synapsin promoter, resulting in long-term expression of a genetically encoded mRuby-based calcium indicator in neuronal cells. Frozen virus vial was thawed on ice and 5ul/well was added to cell culture media in 96-well plates. Cell culture plates were then placed in Incucyte for maturation and imaging. Neuronal activity imaging was acquired with the S3 Neuro orange/NIR optimal module at 3Hz and 4× objective. Data was analyzed with Incucyte software and image analysis settings include object size at 60 um, min cell width at 9um, and min burst intensity at 0.15.

Evoked glutamate release assay

NGN2 neurons in culture were allowed to release neurotransmitters either spontaneously in aCSF solution (126 mM NaCl, 2.5 mM KCl, 1.25 mM NaH2PO4, 1.3 mM MgCl2, 2.5 mM CaCl2, 26 mM NaHCO3, 10 mM Glucose) or upon high potassium (K +) stimulation (76 mM NaCl, 52.5 mM KCl, 1.25 mM NaH2PO4, 1.3 mM MgCl2, 2.5 mM CaCl2, 26 mM NaHCO3, 10 mM Glucose) at room temperature. 10 uM of CNQX was added to standard aCSF and high K + aCSF solutions to block AMPA receptors on the postsynaptic neurons. Neurons were incubated in either standard aCSF or high K + aCSF in the incubator for 5 min, then aCSF solutions were transferred to U-bottom plates and stored at − 80 °C until time for glutamate detection assays. Stored solutions were then tested for glutamate concentration with the Glutamate-GloTM assay (Cat. #J7021) following the protocol for ‘cell culture medium’. Luminescence was recorded with the Promega GloMax microplate reader. Luminescence values are correlated with the samples’ glutamate concentration within the standard curve.

Cell viability assay

Cell viability assay was performed with the CellTiter-Glo 3D viability assay reagent (CAT. # G9683) at the terminal timepoint of glutamate release assay and MEA assay. Reagent was thawed at room temperature and added to an equal volume of cell culture medium in each well. After mixing the content vigorously to induce cell lysis, the reagent and cell lysis mix was transferred to a white bottom plate. Luminescence was recorded with the Promega GloMax microplate reader. Luminescence values are correlated with the samples’ ATP content and thus with the number of viable cells.