The sarin surrogate nitrophenyl isopropyl methylphosphonate (NIMP) was synthesized according to previously documented methods (Meek et al., 2012). For concentration-response experiments, 50% of the culture media was removed and replaced with a 2X concentrated NIMP solution for a final concentration of 0.4, 4, 40, and 100 μM, concentrations selected from a pilot study and within range of previous work (Thinschmidt et al., 2022).

Human iPSC-derived glutamatergic and GABAergic neurons (at a ratio of 70:30) and primary human astrocytes (Neucyte, San Jose, CA, USA) were co-cultured, at a ratio of 75:25 neurons to astrocytes (total 2,835 cells/mm²). The ratio was informed by the vendor’s recommendation in previous studies conducted by our group (Lam et al., 2023) and others (Sasaki et al., 2019; Wang et al., 2023). As described previously (Lam et al., 2022), 6-well layout MEA devices (MEA200/30iR-ITO, Multi-channel systems, Reutlingen, Germany) were plasma-treated (PDC-001-HP, Harrick Plasma) then soaked in PBS overnight at 30°C. The MEA devices were washed with sterile DI water (4X) and then air-dried before autoclaving at 121°C. Both MEA devices and 96 flat bottom-well plates were coated with 0.1% PEI (prepared in borate buffer) for overnight incubation at 37°C, washed with sterile DI water (4X), then coated with 20 μg/ml of laminin for 2 h at 37°C. Laminin was removed before cells were seeded. Specific ratios for neuronal subtypes (e.g., glutamatergic and GABAergic neurons) and cell types (e.g., neurons to astrocytes) were obtained from purified stocks. Cells were resuspended in seeding media (Neucyte) based on live cell concentrations then introduced into the MEA device or 96-flat bottom-well plate. Cultures were maintained in a humidified incubator (37°C, 5% CO₂). After 24 h, media was replaced with short term media (Neucyte) for 1 week before switched to long term media (Neucyte) for the duration of the experiment. For culture maintenance, 50% of media was replaced every 2–3 days.

A 256-channel MEA2100 recording system (Multichannel Systems, Reutlingen, Germany) was used to record electrophysiology activity for 30 min at a sampling frequency of 10 kHz and bandpass filtered between 4 and 4,000 Hz, as before by our group (Enright et al., 2020; Lam et al., 2019; Soscia et al., 2017; Soscia et al., 2020) and others (Novellino et al., 2011; Pastore et al., 2018). This involved the placement of the 6-well MEA device within a 5% CO₂-regulated chamber on the heated stage (37°C), and electrophysiology recording started after a standard 5-min equilibration time. An action potential spike was defined by a lower limit threshold, set at 6.5x the standard deviation of baseline noise, for each electrode. Devices were recorded for 30 min once a week from 7 days in vitro (DIV) and onward to monitor the development and maturation of neural networks. Qualitative counts of the percentage of active electrodes for each device were monitored by the experimenter on the day of recording. By ∼25 DIV, a time point in which the number of active electrodes has stabilized (Supplementary Figure 1), baseline activity of the 6-well MEA device was recorded and then wells were randomly assigned for NIMP treatment (0, 0.4, 4, 40 and/or 100 μM). To our knowledge, NIMP exposure studies in in vitro models are scarce. The concentrations of NIMP used in the present study were determined from a pilot in vitro study that had used 0.4 and 4 μM of NIMP. We extended the concentration range to include 10X and 25X the highest concentration and included 40 and 100 μM. These concentrations were also in range with recently published work on acute NIMP exposure in brain slices (Thinschmidt et al., 2022) and cell culture exposure to insecticides (van Melis et al., 2023; van Melis et al., 2024). For the 0 μM NIMP condition, culture media was added to the wells to mimic the mechanical perturbation of dosing the cultures. Following treatment with NIMP, recordings were conducted within the first hour of exposure (i.e., 2 × 30 min recordings), and 1-, 2-, and 6 days post-exposure.

AChE can exist as both a soluble and membrane-bound form. Previous studies have detected enzymatic activity from the soluble form in the culture supernatant obtained from HEK293 cells (Velan et al., 1991) and neuronal cultures (Dong et al., 2004; Xiang et al., 2014). We have also confirmed the presence of active human AChE in the culture media from the human-relevant MEA system, according to a modified Ellman’s method (Ellman et al., 1961), which included the addition of 0.5 mM tetraisopropyl pyrophosphoramide (iso-OMPA), an inhibitor of butyrylcholinesterase, for 15 min at 37°C. The culture supernatants were diluted 1:20 into the substrates acetylthiocholine (1 mM) and DTNB (1 mM). Enzyme activity was examined by measuring the optical density (OD) at an absorbance of 410 nm at 25°C using Synergy H1 multi-mode microplate reader (BioTek). OD was recorded in 5-min intervals over the course of 2 h. The rate of change in enzymatic activity (based on OD) over 2 h for cultures treated with NIMP was normalized to the rate of change in enzymatic activity in the 0 μM NIMP culture. Total cellular protein for Ellman’s assay was conducted based on a previous method neurons and non-neuronal cells (Thullbery et al., 2005), cultures were washed with PBS (3x) and cells lysed using 1% Triton X-100 in 50 mm Tris (pH 8.0) and 150 mM NaCl. Then the cell lysate was treated with DTNB (final concentration 0.32 mM) and iso-OMPA (final concentration, 0.01 mM) for 10–15 min at 37°C. After the incubation period, ATCH (final concentration, 0.75 mM) was added before samples were read on the plate reader as described for the culture supernatant.

CyQuant™ Lactate Dehydrogenase (LDH) assay (Thermo Fisher Scientific) was performed on culture supernatants collected at the 24-h time point of NIMP exposure and 6- days post-exposure. Briefly, the supernatant was collected from 0 μM or NIMP-treated (0.4, 4, 40, and/or 100 μM) cultures and processed in 96-well plate format, per kit instructions. Absorbances were read at 490 and 680 nm on the Synergy H1 multi-mode microplate reader (BioTek), and absorbance data for samples exposed or having been exposed to NIMP were normalized to the 0 μM NIMP condition.

Cultures were treated with NIMP (0, 4, and 100 μM) for 24 h, and cells were lysed at 6 days post-exposure using RLT buffer containing β-mercaptoethanol; the supernatant was collected at 6 days post-exposure as well. Total RNA was extracted and purified from the collected supernatant using the RNAeasy mini spin columns (Qiagen). Sequencing libraries were prepared using Illumina Stranded mRNA Prep (Illumina, San Diego, CA, United States) and sequenced using an Illumina NextSeq 2000. The quality of sequencing data was checked using FastQC software.¹ The reads were then mapped to the human genome (hg38) using STAR aligner and read counts per gene were determined using “featureCounts” from Rsubread package (Liao et al., 2014). RUVseq was then used to identify and remove factors of unwanted variation (Risso et al., 2014). Differentially expressed genes were then identified using edgeR, controlling for factors of unwanted variation (Robinson et al., 2010). A gene was considered “significantly differentially expressed” when its false discovery rate adjusted p-value was < 0.05 and fold change was > 1.5. Gene ontology (GO) analysis was performed using ToppGene (Chen et al., 2009). Heatmaps were generated using heatmap.2 function in “gplots” R package. Volcano plots were generated using Galaxy Europe.²

Quantified data are expressed as mean ± standard error of the mean (SEM) for the number of wells indicated unless stated differently. For electrophysiology experiments and in vitro assays, the statistical significance was analyzed in GraphPad version 10 (GraphPad Software, San Diego, CA) using unpaired t-test, mixed model repeated measures for one-way or two-way ANOVA with Tukey’s or Dunnett’s post-hoc analysis.

The consequence of OPNA can have widespread effects on the brain. Thus, in the present study, we modeled the human-relevant MEA system after the cortex, containing glutamatergic (excitatory) neurons (70–80%) and GABAergic (inhibitory) neurons (20–30%) (Markram et al., 2004; DeFelipe and Farinas, 1992). Based on our previous work in a 3D MEA study (Lam et al., 2023), in other 2D MEA studies (Sasaki et al., 2019; Wang et al., 2023), and recommended by the vendor, the seeding ratio of 70:30 human iPSC-derived glutamatergic to -derived GABAergic neurons was co-cultured with primary human astrocytes (at a ratio of 25:75 astrocytes to neurons) on the planar 6-well MEA device. As the number of active electrodes plateaued by 25 DIV (Supplementary Figure 1), spiking and bursting activity appeared coordinated (or synchronized) across a pair and/or multiple electrodes, as shown by the representative raster plot (Figure 1a) that displays spiking and bursting activity before NIMP exposure (i.e., baseline). The human-relevant MEA system (∼25 DIV) was selected at random and treated with one of five concentrations of NIMP (0, 0.4, 4, 40, and 100 μM) for 24 h. Representative raster plots (Figure 1a) display spiking and bursting activity during the period of exposure to NIMP, specifically at 60 min and 24 h during NIMP exposure, and 6 days after NIMP has been removed relative to age-matched 0 μM NIMP cultures. Qualitative observations from the raster plots suggest that cultures exposed to a low concentration of NIMP (4 μM) exhibited more uncoordinated spiking activity across electrodes during NIMP exposure (e.g., 60 min and 24 h) that appeared to be resolved once NIMP was removed (e.g., 6 days post-exposure). Cultures exposed to a high concentration of NIMP (100 μM) displayed greater coordination of spiking and bursting activity across multiple electrodes at 60 min of exposure that became less coordinated by 24 h of exposure. By 6 days post-exposure, less activity was observed in cultures treated with a high concentration of NIMP compared to age-matched 0 μM NIMP cultures.

We quantitatively assessed the features of spiking (Figure 1b, left column) and bursting activity (Figure 1b, right column) across NIMP concentrations, and time points were quantified and normalized to baseline activity for each treatment-matched culture to minimize the effect of mechanical perturbation by the addition of NIMP. Then, the fold change difference between age-matched treatment wells was calculated. A value of “0” indicates no difference in activity of the NIMP-treated culture relative to the 0 μM NIMP culture. Time series analysis of neural activity revealed that NIMP altered only two features of spiking activity and had no effect on the features of bursting activity for human-iPSC neurons within the cell culture system (Figure 1 and Supplementary Figure 2). For cultures treated with a low concentration of NIMP (0.4 or 4 μM), significant increases in the firing rate (Figure 1b) and total number of spikes (Supplementary Figure 2) were observed for the duration of NIMP exposure, as early as 60 min for 0.4 μM (p < 0.05) and 30 min for 4 μM NIMP (p < 0.05). Increased excitability persisted at 24 h of NIMP exposure for 4 μM NIMP (p < 0.05); this excitability returned to levels comparable to the 0 μM NIMP cultures when NIMP was immediately removed and was sustained at 6 days post-exposure for both low concentrations of NIMP. Cultures treated with a high concentration of NIMP (40 or 100 μM), however, showed a significant decline in the firing rate and total number of spikes at 24 h of NIMP exposure (p < 0.05) that persisted at 2 days post-exposure (p < 0.05) when compared to age-matched cultures treated with a low concentration of NIMP. This decline in activity was not significantly different when compared to age-matched 0 μM NIMP cultures. Differences between a low versus high concentration of NIMP were observed or trending toward statistical significance for specific time points. For example, a significant decrease in the percentage of active electrodes was detected at 6 days post-exposure (p < 0.05, Supplementary Figure 2) or interspike interval at 60 min (p = 0.08) and burst duration at 6 days post-exposure (p = 0.06) showed opposite regulation of a feature of spiking and bursting activity (Figure 1b). No detectable differences were observed for features of bursting activity (e.g., burst per minute in Figure 1 or the total number of bursts and interburst interval in Supplementary Figure 2). Notably, the concentration-dependent effect observed for the firing rate and the total number of spikes were similar in neuronal excitability for cultures treated with 0.4 and 4 μM NIMP and similarly the level of activity was reduced for cultures treated with 40 and 100 μM. In this paper, here on out, we have categorized cultures exposed to 0.4 and 4 μM as “low” versus 40 and 100 μM NIMP as “high” concentrations of NIMP.

We examined the degree of coordination (or synchrony) in spike timing between electrodes for networks within the human-relevant MEA system. As before (Lam et al., 2022; Enright et al., 2020; Lam et al., 2019), the SPIKE distance method (Kreuz et al., 2013) was used to score all possible electrode pairings for a given 30-min recording for each MEA system (Lam et al., 2022; Enright et al., 2020; Lam et al., 2019). A score of “0” indicates that the network has no synchrony and a score closer to “1” indicates a high degree of synchrony. At a device level, synchrony analysis revealed that the average score for networks exposed to all concentrations of NIMP (e.g., 0.4, 4, 40, and 100 μM) were no different from the 0 μM NIMP condition at any of the time points during and after NIMP exposure (Figure 2a), which ranged from 0.24 to 0.31. However, when we evaluated the distribution of synchrony scores for all networks (aggregated across all devices) grouped by NIMP-treated condition and each time point, we observed differences in a subpopulation of networks that were masked when analysis was conducted at a gross level.

The distribution of synchrony scores is reported in a histogram (Supplementary Figure 3) and shown as a probability distribution segregated into quantiles in the quantile-quantile (QQ) plot (Figure 2b). At each time point, to generate the QQ plots we used the data from all wells for each concentration. At baseline, synchrony scores are shown to be variable across multiple devices before being selected for NIMP treatment, having networks with synchrony scores within quantiles that range from 0.1 to 1 (in both x and y axes, Figure 2bi). At 30 and 60 min of NIMP exposure, synchrony scores from human iPSC-derived networks at 4, 40, and 100 μM concentration showed that a subset of networks decreased its synchrony score, as shown by a decrease in the quantiles that range from 0.1 to 0.7 in the y-axis, having scores below the 0 μM NIMP condition (red line). By 24 h, a greater subset of networks exhibited a decrease in synchrony scores in the 40 and 100 μM NIMP concentrations, while the 0.4 and 4 μM concentrations displayed a distribution of synchrony scores comparable to 0 μM concentration.

Interestingly, despite the removal of NIMP and the replacement of fresh media, the distribution of synchrony scores for networks previously treated with NIMP, independent of its concentration, had decreased at 1-day post-exposure relative to age-matched networks treated with 0 μM NIMP. By 2 days post-exposure, concentration differences were observed in the human-relevant MEA system previously treated with 0.4, 4, 40, and 100 μM NIMP. Increased coordinated activity was detected in a subset of networks (within the 0.4–0.8 quantiles) from the human-relevant MEA system exposed to lower NIMP concentrations (at 0.4 and 4 μM) that returned to levels exhibited by the age-matched cultures treated with 0 μM NIMP by 6 days post-exposure. Conversely, for the human-relevant MEA system exposed to high concentrations of NIMP (at 40 and 100 μM), a decrease in coordinated activity was observed for a subset of networks (within the 0.4–0.8 quantiles) observed at 2 and 6 days post-exposure.

For quantitative analysis, we used Kullback–Leibler (KL) divergence to determine the statistical distance from the probability distribution of synchrony scores (from Supplementary Figure 3) obtained from the MEA system exposed to 0, 0.4, 4, 40, and 100 μM NIMP compared to the 0 μM NIMP condition (the reference distribution) (Figure 2c), thereby quantifying the concentration-dependent effect of NIMP at each time point. In using this analysis, we observe that the coordinated activity for a subset of networks is affected for the duration of NIMP exposure for all concentrations of NIMP. At the 24-h time point, the magnitude of NIMP’s effect on network activity is greatest for 100 μM, and there is a gradual concentration-dependent decrease in response. However, at 6 days post-exposure, the 40 and 100 μM NIMP conditions’ distributions start to deviate from 0 μM condition’s distribution, relative to 0.4 and 4 μM NIMP conditions’ distributions. This suggests that a single exposure to a high concentration of NIMP can have lasting effects on subsets of neurons and their networks, detected days after NIMP exposure.

Exposure to OPNA is known to induce a “cholinergic” crisis in the central and peripheral nervous system (Franjesevic et al., 2019). OPNA-induced inhibition of AChE prevents the hydrolysis of acetylcholine and results in the accumulation of the neurotransmitter in the synapses and the overstimulation of nicotinic and muscarinic acetylcholine receptors (Franjesevic et al., 2019). To evaluate whether the electrophysiological changes during and after exposure to NIMP paralleled the inhibition of AChE activity, we examined the catalytic activity of the soluble form of AChE from the culture supernatant at the 24-h time point of NIMP exposure, and at 2- and 6- days post-exposure (Figure 3ai). All concentrations of NIMP showed reduced AChE catalytic activity (38–52% decrease) within the MEA system at the 24-h time point, but only the 4 μM-NIMP-treated MEA system appeared to show a sustained decrease in AChE activity at 2 days post-exposure, relative to age-matched 0 μM NIMP-treated human-relevant MEA system. By 6 days post-exposure, AChE activity for NIMP-treated cultures were no different to age-matched 0 μM NIMP-treated cultures. We also assessed the catalytic activity of the membrane-bound form of AChE from the total cellular proteins extracted from the MEA system at the 24-h time point for NIMP exposure (Figure 3aii). A greater reduction of AChE activity to 35% for 4 μM and 26% for 100 μM NIMP was observed. Thus, we detected the reduction of AChE activity during NIMP exposure and the recovery of its activity post-exposure using the human-relevant MEA system. However, the time-dependent electrophysiological changes observed in the low versus high concentrations of NIMP during and after exposure suggests that spiking and synchronized network activity are not solely dependent on NIMP-induce AChE inhibition. We also examined whether the observed changes in neural activity from the human-relevant MEA system were attributed to cell health. Cell viability was quantified by the amount of LDH released from the human-relevant MEA system exposed to NIMP at the 24 h time point (Figure 3bi) and 6 days post-exposure (Figure 3bii). LDH is a cytoplasmic enzyme that is released from cells with compromised membrane integrity, a feature of cells undergoing cell death (e.g., apoptosis, necrosis). As a positive control, cultures were treated with staurosporine (1 μM) to induce cell death (Moors et al., 2009). Despite having been treated with NIMP, all concentrations showed no change in cell viability, relative to the 0 μM NIMP condition, at either time point assessed. Collectively, these findings suggest that the functional changes observed in neural and network activity (e.g., features of spiking activity and synchrony) were either environmental or intrinsic changes to the cell.

Having observed differences in the functional responses of neural and network activity from human-iPSC derived glutamatergic and GABAergic neurons exposed to a low versus high concentrations of NIMP, we examined whether transcriptomic changes within the human-relevant MEA system could provide insight on whether and how changes in the cellular environment or its molecular machinery affected neuronal function. Transcriptomic analysis was conducted on human-relevant MEA system at 6 days post-exposure to 0, 4, and 100 μM NIMP. Our analysis identified 227 differentially expressed genes (DEGs) in our human-relevant MEA system exposed to 4 μM NIMP relative to 0 μM NIMP versus 671 genes in our MEA system exposed to 100 μM NIMP relative to the 0 μM NIMP condition. When comparing 100 μM NIMP relative to 4 μM NIMP, 878 genes were found to be differentially expressed. As shown in the Venn diagram (Figure 4a), a greater proportion of genes were differentially expressed and unique to the 100 μM NIMP condition, as indicated by the lack of overlap between the 4 vs. 0 μM NIMP and 100 vs. 4 μM NIMP (green and red circles, respectively).

The top 30 genes differentially expressed are shown in the volcano plots [DEGs; abs (Fold Change, FC) > 1.5 and False discovery rate (FDR) < 0.05] for each comparison (Figure 4b). Ontology enrichment analysis of upregulated genes from 4 μM NIMP vs. 0 μM NIMP condition had identified biological processes associated with the extracellular environment and interaction with the cells (e.g., cytoskeleton organization, actin-filament-based process, extracellular matrix organization), nerve signals (e.g., sodium ion homeostasis, regulation of transmission of nerve impulses), and receptor ligand activity and signaling pathways (e.g., cytokines, PKC-activating G protein-coupled receptor) as enriched (Figure 4ci). Genes downregulated in 4 μM NIMP relative to 0 μM NIMP condition did not cluster into known biological processes. For 100 μM NIMP relative to 0 μM NIMP condition, upregulated genes were involved in biological processes that included the regulation of specific cellular functions (e.g., cell motility and inflammatory response), and intracellular signaling (e.g., protein phosphorylation, and second messenger-mediated signaling) (Figure 4cii). Biological processes associated with the downregulated genes in 100 μM NIMP relative to 0 μM NIMP condition included negative regulation cell communication, cell adhesion, positive regulation of cell activation, oxidative stress, cerebral cortex GABAergic interneuron differentiation and signaling receptor regulator activity. When comparing DEG in 100 μM NIMP relative to 4 μM NIMP condition (Figure 4biii), there were only 45 upregulated genes involved in transcription regulator activity, specifically transcription factor binding (e.g., HMGB2, SALL4, RORA, HNF4A) that involved STAT family protein binding (e.g., CEBPA, SPI1), central nervous system differentiation (e.g., NKX2-2, NKX6-1, TAL2, GLI1, HOXA1) (data not shown). Downregulated genes in 100 μM NIMP relative to 4 μM NIMP condition included processes such as negative regulation of cell communication, positive regulation of cell differentiation, negative regulation of the immune system process, positive regulation of monoatomic ion transport, and negative regulation of hormone secretion, cell adhesion and the extracellular matrix.

In Figure 5, we highlighted genes that were involved in processes for the extracellular environment and cell interaction, nerve signals, intracellular signaling, and the inflammatory response. For extracellular environment and cell interaction (Figure 5a), genes relevant to cell morphogenesis (EDN1, USH1C, COL15A1, CXCR4, and STC1) were upregulated in both the 4 and 100 μM NIMP conditions. Several genes involved in collagen fibril organization (e.g., COL3A1, COL12A1, FMOD, and MMP9), cytoskeletal organization (e.g., DOCK2, NOX4, NYH11, ACTBL2, ELN, ANKRD1, MYL2, CAPN3, and ASB2), and nerve signals were upregulated only in the 4 μM NIMP but not 100 μM treatment condition relative to 0 μM. In particular to nerve signals (Figure 5b), upregulated genes in the 4 μM NIMP condition were involved in protein kinase C-activating (e.g., EDN1, CCK, AZU1), phospholipase C-activating G protein-coupled receptor (e.g., EDN1, OPRK1, GPR35) signaling pathways and sodium homeostasis (e.g., EDN1, AVP, TMPRSS3, UTS2). For intracellular signaling (Figure 5c), upregulated genes in only the 100 μM NIMP condition were involved in cGMP-mediated signaling (e.g., GUCA1A and NPPB), cAMP-mediated signaling (e.g., ADCY8, CRHR2, RASD2) and calcium-mediated signaling (e.g., GRIN2C, SELE, ITGAL, MCTP2, P2RX2, CXCR4, ERBB3, LAT2, CCL3L3, PLCG2). In addition, genes involved in the modulation of chemical synaptic transmission (e.g., MCTP2, P2RX2, ADCY8, CRHR2, and GRIN2C) were also upregulated in only the 100 μM NIMP condition. For inflammation (Figure 5d), there was a subset of genes that did not meet the fold change and p-value cutoff but was trending toward upregulation in the 4 μM NIMP condition and was significant in the 100 μM NIMP condition. These genes were involved in the regulation of the immune system process (e.g., ITGAL, FOXF1, CXCR4, UNC13D, and RARRES2). In addition, genes involved in IL-6, type I interferon, and/or tumor necrosis factor production (e.g., IRAK3, ELF4, FLT3, PLCG2, TLR9, NMBR, CCN1, IL6R, CRHR2, KLF2, and TLR5) were upregulated in the 100 μM NIMP condition.

We also observed significant expression changes (p < 0.05) for genes encoding neurotransmitter receptor subunits that did not meet the FC > 1.5 criteria but exhibited at least 1.25 fold difference (i.e., log2FC > 0.322) and may be sufficient to modulate neural and network activity. These included receptor subunits for GABA_A receptors (e.g., GABRR1, GABRE, GABRD) that are upregulated in the low (but not high) concentration of NIMP, and glutamate ionotropic receptor NMDA type subunit (e.g., GRIN2C) and muscarinic acetylcholine receptor subunit (e.g., CHRM5) that is upregulated in both low and high concentrations of NIMP.