Uncovering NMDA receptor-mediated high firing activity in excitatory human neurons

Summary

Human iPSC-derived glutamatergic (iGlut) neurons provide a promising platform for studying neuronal function and modeling CNS diseases, but assessing large populations of neurons from multiple donors remains challenging. We developed a protocol that targets N-methyl-D-aspartate receptors (NMDA-Rs) and enhances neuronal activity, revealing functional phenotypes. Using the calcium indicator GCaMP8f, we demonstrate that Mg²⁺-free ACSF significantly increases neuronal activity, and is enhanced by glycine but inhibited by the NMDA-R antagonist AP-V. Multi-electrode array recordings also show robust firing in Mg²⁺-free ACSF. Lastly, patch-clamp electrophysiology confirms the higher firing rates in Mg²⁺-free ACSF across multiple donor lines, uncovering donor-specific firing phenotypes. This protocol facilitates functional analyses of iGlut neurons while preserving single-cell resolution, enabling detailed characterization of iGlut neurons in diverse applications such as CNS disease modeling and drug screening. This protocol establishes a versatile framework for large-scale studies of neuronal network dynamics and individual excitability in iPSC-derived iGlut neurons.

Graphical abstract

Highlights

• •Human iPSC neurons aid in disease modeling, but analyzing large populations is challenging
• •Unmasking NMDA-Rs enhances activity and shows functional types in human glutamate neurons
• •This protocol enables large-scale functional evaluation and improves drug screening for the CNS

Human; Neuroscience; Cell biology; Genomics

Introduction

The development of induced pluripotent stem cell (iPSC) technologies in the last two decades has led to unprecedented advances in human CNS disease modeling and drug discovery.^1,2 Human iPSC-derived neurons provide an excellent in vitro system to assess neuronal function in physiological and pathological conditions. Human iPSC-based models also overcome many of the limitations of animal models, which cannot replicate species-specific mechanisms³ nor incorporate human genetic variants or backgrounds. Directed and transcription factor-based differentiation protocols have been developed to produce relatively pure populations of glutamatergic neurons (iGlut), helping to reduce batch-to-batch variability.^4,5,6 When co-cultured with glial cells, these iGlut neurons become synaptically mature, exhibiting spontaneous activity.⁷ Taking advantage of the iGlut differentiation protocols, several groups have used iPSC-derived NGN2-directed glutamatergic neurons to model different human diseases, from neurodevelopmental⁸ and neurodegenerative diseases,^9,10 to other psychiatric diseases such as schizophrenia^11,12 or alcohol dependence and response.^13,14,15

A key limitation in studying iGlut neurons is the limited scalability, as generating them from iPSCs is slow and labor-intensive, limiting the number of donor lines that can be processed simultaneously.^{8,9,10,11,12,13,14,15} Increasing the number of iPSC donor lines studied would help provide sufficient power for statistical comparisons. This would be further facilitated by the development of high-throughput (HT) assays to monitor neuronal activity. Electrophysiological recordings provide high-content information but are low-throughput, performed on a cell-by-cell basis (Table 1). Automated patch-clamp technologies are available but require harvesting mature cultures of human neurons in preparation for making recordings, which disrupts neuronal morphology and connectivity. Multielectrode array (MEA) recordings, in which neurons are cultured directly in a well on a plate that has embedded recording electrodes, are another scalable approach, but can be limited by the number of electrodes and location of neurons relative to the electrodes, and lack single-cell resolution. On the other hand, Ca imaging with genetically encoded sensors (e.g., GCaMPs) or Ca²⁺ dyes (e.g., Fluo-4, Fura-2) provides an opportunity to conduct HT measurements of neuronal activity at the single-cell level on iPSC-derived neurons. Here, we searched for conditions to improve the intrinsic network activity of neurons to promote a broader range of activity that would reveal potential differences in neuronal activity between donors. While much focus has been on α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPA-Rs) in iGlut neurons due to their role in mediating excitatory synaptic transmission,^16,17,18 less is known about the function of N-methyl-D-aspartate receptors (NMDA-Rs) in human neurons. NMDA-Rs are unique because they are inhibited by extracellular Mg²⁺ under physiological conditions and require membrane depolarization to remove Mg²⁺ blockade.¹⁹

Table 1.

Comparison of three techniques used to assess neuronal activity in iGlut neurons

Characteristics	Ca imaging	MEA	Patch clamp
Increased activity in Mg2+-free	Yes	Yes	Yes
Single-neuron resolution	Yes	No	Yes
Temporal resolution	tens of sec	∼1 msec	∼1 msec
HT-scalable	Yes	Yes	No
Labor demand	Moderate	Low	High

While these techniques provide different levels of cellular and time resolution, all three showed high levels of neuronal activity in Mg²⁺-free ACSF conditions.

NMDA-R-mediated neuronal activity in human neurons is not fully understood, but it is known that NMDA-R-mediated neuronal signaling is important for neuronal function and plasticity.¹⁹ Here, we investigated the role of NMDA-Rs using Ca imaging to assess the activity of a population of human iGlut neurons with single-cell resolution. We have shown in previously published studies that NGN2-differentiated iGlut neurons exhibit robust basal activity.^14,15 Here, we describe an explicit role of NMDA-Rs in increasing neuronal activity in iGlut neurons. NMDA receptors are known to be inhibited by extracellular Mg²⁺ at negative membrane potentials, and are unblocked at positive membrane potentials, where they can conduct an inward current.²⁰ When Mg²⁺ is removed from the extracellular solution, the NMDA receptor is also unblocked. To probe this property of NMDA receptors, we used a modified ACSF in which we simply removed extracellular MgCl₂ to produce Mg²⁺-free ACSF. Removing extracellular Mg²⁺ revealed high-frequency neuronal activity in iGlut neurons, which was highly synchronized and mediated by NMDA receptors. We characterized this behavior in multiple iPSC donor lines, replicated the firing behavior with MEA recordings, and confirmed with single-cell whole-cell patch-clamp recordings. Interestingly, we observed unique neuronal firing patterns in Mg²⁺-free ACSF that align with iPSC donor line-specific activity. These results highlight the power of Ca imaging for large-scale neuronal population measurements when performed under Mg²⁺-free conditions. This approach will enable more efficient population-based measurements of neuronal network activity on human iPSC-derived excitatory neurons with single-cell resolution, which can be scaled up to multiple donor lines and implemented for drug screening pipelines and phenotypic characterization of iPSC-derived neuronal models of CNS disease.

Results

Human induced pluripotent stem cell-derived glutamatergic neurons show high frequency and synchronized activity in Mg²⁺-free conditions

To generate a nearly pure population of human iGlut neurons, we used the previously characterized neurogenin 2 (NGN2) induction protocol^15,21 with a control donor line (10884) obtained from the Collaborative Study on the Genetics of Alcoholism (COGA) cohort, available through the NIAAA-COGA Sharing Repository.¹⁴ NGN2-derived 10884 neurons were co-cultured with mouse glial cells to promote neuronal maturation and synapse formation.^22,23 To investigate the activity of these neurons with single-cell resolution while monitoring a relatively large population, we utilized Ca imaging for which we expressed lentiviral-transduced GCaMP8f selectively in neurons (human synapsin promoter) at day 28 post differentiation (Figure 1A). At approximately 7–10 days post-lentiviral transduction, green fluorescence could be observed in multiple neurons due to GCaMP8f expression (Figure 1B). We measured fluorescent Ca²⁺ transients in multiple neurons simultaneously using a CMOS camera at 6.1 frames/s with neurons older than 10 weeks post-induction, to allow for GCaMP8f expression and iGlut neuron maturation.

Figure 1.

Mg²⁺-free ACSF reveals high firing rates in mature iGlut neurons

(A) Schematic shows induction and experimental workflow using the NGN2 induction protocol,^4,21 with some modifications (see STAR Methods). GCaMP8f transduction was done on days 28–30 post induction, Ca imaging was performed on days 75–85, patch-clamp electrophysiology on days 30–90, and MEA recording on days 40–45 post induction.

(B) Ca imaging of a population of single iGlut neurons from donor line 10884. Representative fluorescent image shows 10884 iGlut neurons expressing GCaMP8f 2 weeks after lentiviral transduction.

(C) Fluorescence traces from three different 10884 neurons show the effects of the buffer conditions indicated by the bar above.

(D) Plot shows average Ca²⁺ transients/neuron/min from N = 115 neurons from two differentiation batches. Note the significant increase in the frequency of the Ca²⁺ transients in Mg²⁺-free ACSF conditions (4.1 ± 0.4 transients/min) compared to ACSF (<1 transient/min, ∗p < 0.05; ∗∗∗∗p < 0.0001, one-way ANOVA Kruskal-Wallis with Dunn’s post hoc test, n = 115/2 batches). Graph shows the mean (bar) and individual points.

(E) Fluorescence traces from two different 10884 neurons show a decrease in Ca²⁺ transients with the NMDA receptor antagonist AP-V (100 μM).

(F) Plot shows average Ca²⁺ transients/neuron/min for two conditions: Mg²⁺-free ACSF, with 30 μM NBQX and 3 μM Gly (first condition); and the same buffer components plus AP-V 100 μM (second condition) (∗∗∗∗p < 0.0001, one-way ANOVA Kruskal-Wallis with Dunn’s post hoc test, n = 94 cells, 2 batches). Graph shows the mean (bar) and individual points.

(G) Correlation plots show high synchrony of neuronal activity observed in both ACSF and Mg²⁺-free ACSF conditions for donor line 10884. The matrices represent the pairwise correlations across one example FOV, both in ACSF and Mg²⁺-free ACSF. Each cell shows the color-coded correlation coefficients (Kendall's rank correlation) between two neurons. To create the matrices, only neurons that had at least 3 transients in both ACSF and Mg²⁺-free ACSF were used.

(H) Heatmap shows the logarithmic fold increase in Ca²⁺ transient firing rate from ACSF to Mg²⁺-free ACSF conditions.

(I) The mean correlation coefficient (MCC) did not change between the ACSF and Mg²⁺-free ACSF conditions (n = 7 FOV/2 batches, p > 0.05 (ns); Paired two-tailed t test). Graph shows the mean (bar) and individual points. See also Figure S1 and Videos S1 and S2.

To investigate possible NMDA-R-dependent activity, we created an extracellular solution of ACSF (see STAR Methods for solution compositions) that lacks Mg²⁺ (Mg²⁺-free ACSF).¹⁹ Under basal conditions with normal ACSF, we observed spontaneously active Ca²⁺ transients in neurons with a frequency of < 1 Ca²⁺ transients/min (Figures 1C and 1D). Remarkably, simply removing extracellular Mg²⁺ (Mg²⁺-free ACSF) dramatically increased the frequency of Ca²⁺ transients (p < 0.0001, one-way ANOVA, n = 115/2 batches) (See also Videos S1 and S2). Ca²⁺ transients increased to a mean of 4.1 ± 0.4 transients/min, with some neurons exhibiting Ca²⁺ transients every 3 s (20 transients/min). To probe the role of NMDA receptors, we examined the effect of external glycine (Gly), a potentiator for the NMDA receptor.²⁴ Adding Gly (3 μM) to Mg²⁺-free ACSF further increased Ca²⁺ transient activity, reaching an average of 5.5 ± 0.6 transients/min, with some neurons exhibiting Ca²⁺ transients every 2 s or less. Interestingly, the application of the AMPA-R antagonist NBQX²⁵ (10 μM) decreased Ca²⁺ transient frequency to 2.4 ± 0.3 transients/min, indicating some contribution of AMPA-Rs to the higher firing rates observed. Ca²⁺ transient activity was reduced further upon the removal of Gly, and the reintroduction of extracellular Mg²⁺ with normal ACSF (Figures 1C and 1D).

The rapid increase in Ca²⁺ transient frequency observed in Mg²⁺-free ACSF with Gly suggested that NMDA-Rs are involved in the high rate of Ca²⁺ transients. To test this hypothesis, we evaluated neuronal activity in Mg²⁺-free ACSF with an NMDA-R antagonist. We first isolated NMDA-R-dependent activity with Mg²⁺-free ACSF plus Gly (3 μM), and NBQX (30 μM) to selectively block AMPA-Rs (Figures 1E and 1F). We then tested the effect of the selective NMDA-R antagonist AP-V (100 μM). In the absence of external glutamate, iGlut neurons exhibited a high frequency of Ca²⁺ transient firing, with an average of 3.6 ± 0.4 transients/min. The addition of AP-V to the Mg²⁺-free ACSF + Gly + NBQX applied during the recording reduced the Ca²⁺ transient frequency to <0.3 ± 0.04 transients/min, directly implicating NMDA-Rs in the high-frequency firing (Figures 1E and 1F). We next performed single-cell RNAseq on iGlut neurons from line 10884, and another two donor lines, 8092 and 9206 (See also Figure S1). The UMAP shows overall similarity between the three donor lines, closely clustering in a cluster of glutamate neurons (see also Figure S1A). Importantly, we detected mRNA for both AMPA and NMDA receptors (GRID2, GRIA2, GluA2, GRIN2B, and GRIA4). See also Figure S1B). We also detected mRNA encoding proteins involved in action potential firing and synaptic transmission, including voltage-gated potassium channels (KCND2, KCNQ3, and KCNB2), synaptic release proteins (SYN3, RAB3C), voltage-gated sodium channels (SCN2A, SCN3A y SCN9A), voltage-gated calcium channels (CACNA1C), and GABA receptors (GABRB3). These results support the presence of functional AMPA and NMDA receptors, along with the expression of neuronal excitability genes that could contribute to donor-specific differences. Furthermore, a dramatic increase in activity Mg²⁺-free ACSF suggests that iGlut neurons might contain “silent synapses,” which are synapses that express NMDA receptors but lack AMPA receptors.^20,26

In addition to the high rates of neuronal activity (Figures 1C–1F), we observed a high degree of synchronized activity. This can be seen as a large number of Ca²⁺ transients that correlate in time in both ACSF and Mg²⁺-free ACSF conditions (Figures 1C and 1E). To quantify this, we performed a correlation analysis (Kendall rank correlation) based on the binned (bin size: 1 s) time stamps of the Ca²⁺ transients and plotted the pairwise correlation coefficients (CCs) of neurons as correlation matrices (10884 is shown in Figure 1G). We constructed a correlation matrix for each recording/field of view (FOV) and calculated the mean correlation coefficient (MCC). Both Mg²⁺-free ACSF and ACSF conditions showed high synchrony, regardless of the large difference in firing fate (Figure 1H). We averaged the MCCs (Figure 1I) and observed a strong positive correlation in both ACSF and Mg²⁺-free conditions (Figure 1G), with most neurons showing a 0.8–1 log-fold increase in activity (Figure 1H). Interestingly, the MCC did not significantly change between ACSF and Mg²⁺-free ACSF (n = 7 FOV, paired t test) (Figure 1I). Taken together, these experiments illustrate that iGlut neurons exhibit NMDA receptor-dependent high activity in external Mg²⁺-free ACSF conditions and are characterized by robust and strongly synchronized firing.

High firing rates of induced pluripotent stem cell-derived glutamatergic neurons in Mg²⁺-free solutions are detected with multielectrode array

To see if the increase in neuronal Ca²⁺ transients measured in Mg²⁺-free ACSF with respect to normal ACSF relates directly to changes in electrical activity, we used multi-electrode arrays (MEAs) to electrically measure spiking activity on a faster timescale (Table 1). In addition to 10884, we generated iGlut neurons from the iPSC lines of two other donors, 8092 and 9206, to evaluate and compare their behavior under Mg²⁺-free ACSF. For MEA experiments, iGlut neurons were first generated and then co-cultured with glial cells⁷ on MEA plates (24/48 wells), each equipped with 16 recording electrodes embedded per well. To generate the iGlut neurons, we used the same NGN2 protocol, with minor modifications to adapt the procedure to seeding on MEA plates (see STAR Methods section). We first recorded basal firing in Mg²⁺-free ACSF for 2 min, and then removed the plate, added MgCl₂ to reach a final concentration of 1.3 mM (i.e., to re-create ACSF), and then re-measured the activity in the same plate for another 2 min (Figure 2A). In this way, we could compare the basal spiking in Mg²⁺-free ACSF with that in ACSF in the same well and recording session. In Mg²⁺-free ACSF, iGlut neurons were highly active in all wells of the MEA plate (Figure 2B). After the addition of MgCl₂, the high basal activity dramatically decreased (Figure 2B). Similarly to Ca imaging, we observed an increased, synchronized bursting in Mg²⁺-free ACSF conditions (Figure 2C). The weighted mean firing rate (WMFR – mean firing rate based on the electrodes with firing rate greater than 5 spikes/min, Figure 2D) was 3.27 ± 0.24 Hz in Mg²⁺-free ACSF, and decreased to 0.24 ± 0.11 Hz (n = 48 wells/2 plates, p < 0.0001, Wilcoxon test) after adding MgCl₂ to mimic ACSF. We compared the WMFR, number of network bursts (synchronized bursts across most of the electrodes in a well), and synchrony (described by synchrony index, 0–1) of the three donor lines (10884, 9206, and 8092) with Mg²⁺-free ACSF. The WMFR in Mg²⁺-free ACSF did not significantly differ between the three donor lines (n₁₀₈₈₄ = 48/two 24-well plates, n₉₂₀₆ = 35/one 48-well plate, n₈₀₉₂ = 24/one 24-well plate; Figure 2E, Kruskal-Wallis test). Both 8092 and 10884 showed a significantly higher number of network bursts compared to 9206 (29.9 ± 3.5 for 8092 and 17.7 ± 2.9 for 10884 versus 6.5 ± 0.8 for 9206; n₁₀₈₈₄ = 48/two 24 well plates, n₉₂₀₆ = 48/one 48 well plate, n₈₀₉₂ = 24/one 24 well plate; Figure 2F, ∗p < 0.05, ∗∗∗∗p < 0.0001, Kruskal-Wallis test). Donor 10884 showed significantly higher synchrony compared to 9206 (p < 0.0001) or 8092; the synchrony index was 0.74 ± 0.03 for 10884 compared to 0.53 ± 0.04 and 0.53 ± 0.06 for 9206 and 8092 respectively (n₁₀₈₈₄ = 48/two 24 well plates, n₉₂₀₆ = 39/one 48 well plate, n₈₀₉₂ = 24/one 24 well plate; p < 0.05, Kruskal-Wallis test) in Mg²⁺-free ACSF (Figure 2G). To probe the involvement of NMDA-Rs in the MEA recordings, we examined the effect of AP-V (100 μM) on neuronal activity observed in Mg²⁺-free ACSF in donor line 10884 (See also Figure S2). The WMFR was nearly completely inhibited in the presence of AP-V.

Figure 2.

High frequency of firing detected by MEA in Mg²⁺-free ACSF conditions

(A) Schematic shows the design of the MEA recordings. The activity of iGlut neurons was recorded on days 40–45 post induction, first with Mg²⁺-free ACSF, then again after adding MgCl₂ to increase Mg²⁺ to 1.3 mM (to mimic ACSF).

(B) A gradient map representation of MFR activity of 10884 donor line observed in one 24-well MEA plate with Mg²⁺-free ACSF (left) and then after adding MgCl₂ (right). Note the difference in scale.

(C) Raster plots for one well show the firing activity pattern for each of the 16 electrodes as a function of time for 10884 with Mg²⁺-free ACSF (left) and then after the MgCl₂ addition (right). Average spike traces shown at top. Note the high degree of synchrony.

(D) The weighted mean firing rate (WMFR) is shown for donor line 10884 in Mg²⁺-free ACSF, and after MgCl₂ addition (n = 48 wells/2 plates, ∗∗∗∗p < 0.0001, Mann-Whitney test). Graph shows the mean ± SEM.

(E–G) Plots show electrophysiological activity metrics (WMFR, number of network bursts, and synchrony index) for all three donor lines (n₁₀₈₈₄ = 48/two 24 well plates, n₈₀₉₂ = 24/one 24 well plate, n₉₂₀₆ = 35/one 48 well plate for WMFR, n₉₂₀₆ = 48/one 48 well plate for number of bursts and n₉₂₀₆ = 39/one 48 well plate for synchrony; ∗p < 0.05, ∗∗∗∗p < 0.0001 one-way ANOVA Kruskal-Wallis with Dunn’s post hoc test). Graphs (E-G) show the mean (bar) and individual points. See also Figure S2.

These results demonstrate that the increase in firing rate under Mg²⁺-free ACSF conditions is consistent between two different recording methodologies: Ca imaging and MEA. Moreover, the high frequency of firing measured with MEAs in Mg²⁺-free ACSF revealed more robust information from human iGlut neurons derived from different individuals.

Single-cell electrophysiology confirms high firing behavior in Mg²⁺-free solutions

We next used single-cell whole-cell patch-clamp electrophysiology to further validate the increased firing rates of iGlut neurons in Mg²⁺-free ACSF. Patch-clamp recordings were carried out over a range of days in vitro to validate the increase in neuronal firing with Mg²⁺-free ACSF observed in both MEA and Ca imaging experiments (Figure 1A). To accomplish this, we patched from neurons between 30 and 90 DIV to include the DIV for MEA experiments (∼40 days) and for Ca imaging experiments (∼90 days) (see Figures 3 and 4A). We generated iGlut neurons from the iPSC lines of three different donors, 10884, 8092, and 9206, using the NGN2-directed differentiation protocol and co-culturing with mouse glial cells. The resting membrane potential (RMP), which was corrected for the junction potential, was ∼ −65 mV and did not change significantly over the 30–90 DIV (See also Figure S3). Similarly, the input resistance, cell size (capacitance), and instantaneous firing rate also appeared stable, as seen when plotted as a function of the age of the neuron (DIV, see also Figure S3).

Figure 3.

High firing activity in Mg²⁺-free conditions with both whole-cell patch-clamp electrophysiology and Ca imaging across three iPSC donor lines

(A) Voltage traces from current-clamp recordings show an increase in neuron firing with Mg²⁺-free ACSF in all three donor lines.

(B–E) Graphs show the increase in firing rates measured by electrophysiology (spikes/neuron/min, B, D) and Ca imaging (Ca²⁺ transients/neuron/min, C, E) in Mg²⁺-free ACSF compared to ACSF. Top plots show individual neurons, and bottom plots show the mean ± SEM. There is a significant increase in both spikes/neuron/min and Ca²⁺ transients/neuron/min for all three donor lines upon Mg²⁺ removal (10884, 9206 and 8092) (electrophysiology: n₁₀₈₈₄ = 17/2 batches, n₉₂₀₆ = 7/1 batch, n₈₀₉₂ = 13/1 batch; Ca imaging: n₁₀₈₈₄ = 205/2 batches, n₉₂₀₆ = 275/2 batches, n₈₀₉₂ = 143/2 batches, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001, Paired two-tailed t test). Graphs (D, E) show mean ± SEM. See also Figure S3.

Figure 4.

Functional phenotypes for iGlut neurons in Mg²⁺-free ACSF

(A) Schematic shows the patching timeline for all three donor lines (10884, 9206, 8092). The recordings were carried out on days 30–90 post induction.

(B–D) Representative current-clamp recordings of four different neurons from the indicated donor line showing the effect of switching from ACSF to locally applied Mg²⁺-free ACSF (left traces). A zoom in of action potential spikes is shown on the right. Neurons classified as “bursting” or “tonic” based on cluster analysis (see main text). Scale bars are 40 mV, 20 s, and 2 s

(E) Graphs show the distribution of recordings for little or no activity (0–3 spikes; blue), tonic activity (purple), or bursting activity (pink) in ACSF (top) and Mg²⁺-free ACSF (bottom). Bursts were characterized as two or more consecutive spikes in a 100 ms window with short ISIs.

(F) Graphs show the cell capacitance (pF), resting membrane potential (mV), and instantaneous firing rate (IFR, Hz) for all recordings in the three donor lines. Graphs show the mean (bar) and individual points. No significant differences were observed between donor lines (n₁₀₈₈₄ = 17/2 batches, n₉₂₀₆ = 7/1 batch, n₈₀₉₂ = 13/1 batch, p > 0.05, Kruskal-Wallis test). See also Figure S4.

In current-clamp, we first recorded basal firing at the resting potential in ACSF for 1 min, and then again in the presence of locally applied Mg²⁺-free ACSF for 2 min. We observed a significant increase in the number of action potentials (APs) in Mg²⁺-free ACSF in iGlut neurons from all three donor lines (Figure 3A). The mean AP firing rate increased from ∼5 spikes/min in ACSF to ∼12 spikes/min in Mg²⁺-free ACSF in all three donors' iGlut neurons (Figures 3B and 3D). Donor line 10884 showed an average of <5 spikes/min in ACSF but >10 spikes/min in Mg²⁺-free ACSF (Figure 3D). Interestingly, while donor lines 10884 and 9206 showed consistent increases in firing with Mg²⁺-free ACSF, 8092 showed both an increase and a decrease (∼23%) in firing rate in different neurons in Mg²⁺-free ACSF (Figure 3B). Nonetheless, all three lines showed an increase in firing with Mg²⁺-free ACSF (donor 10884: 3.29 ± 1.41 spikes/min to 12.38 ± 2.26 (N = 17); donor 9206: 4.29 ± 2.90 to 11.43 ± 3.44 (N = 7); donor 8092: 6.31 ± 2.47 to 10.96 ± 3.15 (N = 13) (Figure 3D).

We compared whole-cell patch-clamp recordings with neuronal activity measured with Ca imaging from the same three donor lines. We observed a consistent increase in firing across all three lines in Mg²⁺-free ACSF (Figures 3C and 3E). In Ca imaging, the neurons from all three donors showed Ca²⁺ transient rates ranging from 0.43 ± 0.04 transients/min for donor line 9206 to 1.20 ± 0.07 transients/min for donor line 10884 in ACSF conditions. In Mg²⁺-free ACSF, the Ca²⁺ transient frequency significantly increased in the three donor lines: 10884: 1.20 ± 0.07 to 6.26 ± 0.33 transients/min (N = 205); 9206: 0.43 ± 0.04 to 3.66 ± 0.18 (N = 275); 8092: 1.12 ± 0.11 to 4.21 ± 0.34 (N = 143) (Figure 3E). iGlut neurons from donor line 10884 exhibited the largest increase in Mg²⁺-free ACSF in both Ca imaging and single-cell electrophysiology. Although there are differences in the temporal resolution between patch-clamp electrophysiology (i.e., ms) and Ca imaging (i.e., seconds; Table 1), there is robust concordance between the changes in activity in Mg²⁺-free ACSF in all three donor lines. The similarity in the results across both experimental approaches also indicates that Ca imaging provides a good proxy for measuring neuronal activity in Mg²⁺-free ACSF and is therefore suitable for HT studies of neuronal function.

Electrophysiological characterization of neuronal activity reveals functional phenotypes

In addition to changes in spiking frequency in Mg²⁺-free ACSF for all three donor lines, whole-cell patch-clamp electrophysiology revealed unique firing patterns of iGlut neurons in Mg²⁺-free ACSF (Figure 4). We analyzed the AP firing pattern of individual iGlut neurons from the three iPSC donor lines (10884, 8092, and 9206). We observed that some neurons fired tonically while other neurons exhibited bursts of spikes (two or more consecutive spikes in 100 ms). Interestingly, the distribution of these firing types varied between the donor lines (Figures 4B–4E). To classify whether a neuron exhibited predominantly bursting or tonic firing activity, we used the distribution of their instantaneous firing rates (IFRs) and K-means clustering (see also Figure S4). Each of the three donor lines showed different proportions of neurons that fell into three categories: bursting, tonic, or little or no firing (0–3 spikes) neurons (Figure 4E). The distributions of firing types also changed between ACSF and Mg²⁺-free ACSF. In general, there were more low-firing neurons in ACSF for all three donor lines, with an increase in bursting or tonic firing in Mg²⁺-free ACSF. Donor line 10884 showed the highest proportion of bursting neurons, while line 8092 exhibited the highest proportion of tonic neurons. Line 9206 was overall less active but showed a large increase in tonic firing in Mg²⁺-free ACSF. These differences in firing behavior between lines were not due to differences in intrinsic membrane properties, as there were no significant differences in cell capacitance or resting membrane potential in all three donor lines (p > 0.05, Kruskal-Wallis test, Figure 4F).

These results show that single-cell patch-clamp recordings can provide an additional level of functional characterization when conducted in Mg²⁺-free ACSF, suggesting possible phenotypic differences in iGlut neurons derived from different individuals.

Discussion

Numerous laboratories and pharmaceutical companies are using iPSC-derived human neurons to study the etiology of various neurological and psychiatric disorders as well as to develop new therapies.²⁷ The ability to assess functional differences in human neurons is essential for the successful development of new and more efficient therapies, as well as addressing the need to scale up studies to look at larger cohorts. Evaluating larger sample sizes in functional studies on iPSC-derived neurons in combination with high-throughput experimental approaches would allow the analysis of multiple iPSC donor lines more efficiently. Here, we describe a protocol consisting of removing extracellular Mg²⁺ in the extracellular ACSF that addresses these limitations in iGlut neurons, and supports population-based functional studies of complex neuropsychiatric diseases, such as those having a polygenic contribution.^13,28,29 We show high rates of neuronal activity in Mg²⁺-free ACSF in three different assays, Ca imaging, MEA, and single-cell electrophysiology (Table 1), and in iGlut neurons from three different donors. Importantly, the high-frequency and synchronized firing observed is NMDA-R-dependent, since it was completely silenced in the presence of the NMDA-R selective antagonist AP-V, and is promoted by the NMDA-R potentiator glycine. The increase in neuronal activity with Mg²⁺-free ACSF suggests that iGlut neurons may contain what are referred to as “silent synapses”; these are synapses that contain only NMDA receptors.^20,26 Removal of extracellular Mg²⁺ allows these glutamate synapses to become active and thus access NMDA-R-dependent pathways that would otherwise remain silent in normal ACSF. NMDA-Rs play a key physiological role in plasticity and long-term potentiation, and thus enabling their evaluation under Mg²⁺-free conditions will provide valuable insights into in vitro models of plasticity.

The use of Mg²⁺-free ACSF also uncovered potential unique functional phenotypes for iGlut neurons when assayed with patch-clamp electrophysiology, though this technique is low-throughput as compared to MEA or Ca imaging (Table 1). Notably, we found donor-specific differences across multiple levels of our measurements. In MEA, both the number of network bursts and the degree of synchrony differed significantly between the three donor lines, with donor line 8092 showing the highest number of network bursts and donor line 10884 the highest synchrony. Whole-cell patch-clamp and Ca imaging revealed additional single-cell level differences that MEA could not detect. Interestingly, in both patch clamp and Ca imaging, some neurons displayed decreased activity in Mg²⁺-Free ACSF compared to ACSF. Further, the three examined donor lines exhibited different distributions of bursting and tonic phenotypes in patch-clamp recordings. All these differences proved to be donor line-specific, suggesting a possible genetic contribution. Single-cell RNAseq confirmed glutamate receptor expression in all three lines but revealed subtle differences in the expression levels of several major receptor subunit genes (Figure S1). This could contribute to the observed functional differences, and we speculate that the distinct firing patterns arise from intrinsic donor-specific neuronal properties, potentially involving proteins that regulate membrane excitability (e.g., voltage-gated K⁺ channels). Although elucidating the biological background of these differences was beyond the scope of this study, we plan to investigate this further in the future.

It is particularly worth noting that the higher neuronal activity observed under Mg²⁺-free ACSF is consistent across the three different techniques used to interrogate neuronal activity. The temporal resolution of monitoring neuronal activity with Ca imaging is relatively slow compared to MEA or patch-clamp (Table 1) but nonetheless allows for simultaneous recordings of activity from a large number of neurons at the single-cell level. Automated patch-clamp electrophysiology provides higher fidelity recordings and can overcome the low throughput of classic single-cell patch-clamp electrophysiology. Recent advances in automated patch-clamp have accelerated the study of ion channels and ionotropic receptors expressed heterologously in non-neuronal cells (e.g., HEK-293) for drug screening.³⁰ However, there are several disadvantages to automated patch-clamp recordings. First, it is not readily implemented with mature human neurons. As these are chip-based recordings, neurons need to be harvested and resuspended, which removes synaptic connections, as well as axons and dendrites.³¹ Automated patch-clamp recordings also require uniform cell suspensions and purification steps (i.e., cell sorting) to avoid confounding results, which can further disrupt cell structure and viability.³¹ Maintaining neuronal morphology and circuitry in iPSC-derived neuronal cultures is essential for an accurate assessment of neuron functionality. Another population-based electrophysiological approach is to use MEAs, which enable the evaluation of neuronal network activity at different time points without disrupting synaptic connections.³² MEAs have been successfully used to assess iPSC-derived neuronal activity in a variety of diseases.^33,34,35,36 However, MEAs do not capture the function of individual neurons in the population. More sophisticated approaches, such as the Optopatch platform,³⁷ combine channel rhodopsin-based actuators to elicit neuronal activity through photonic stimulation with genetically encoded voltage indicators (GEVIs) to monitor action potential firing, leading to a spatially resolved all-optical electrophysiology. The Optopatch technology has been successfully implemented in iPSC-derived neurons,^38,39 including in the study of amyotrophic lateral sclerosis.⁴⁰ However, the complexity of this technique makes implementation on a large scale less feasible.

On the other hand, Ca imaging provides an excellent approach to studying the neuronal activity of large populations of neurons, allowing large-scale studies in which functional differences between multiple donor lines need to be evaluated (e.g., when modeling complex, polygenic CNS diseases). Genetically encoded calcium indicators (GECIs), such as GCaMPs, are well-suited for studying neuronal activity at single-cell resolution⁴¹ and can be selectively expressed in subtypes of iPSC-derived cell types using a cell-specific promoter, offering extended applicability on mixed-cell cultures and organoids. GEVIs, such as QuasArs, and GECIs have been extensively used in the study of neuronal function as a direct measure of action potential firing (GEVIs)^42,43 or indirectly (GECIs).^44,45,46

Implementation of a protocol that uses Mg²⁺-free ACSF with large-scale measurements affords several advantages. First, the higher neuronal activity we measure with Mg²⁺-free ACSF occurs under basal conditions without stimulating the neurons; that is, the spontaneous activity in the cultures is dramatically enhanced. Second, Ca imaging can be used to monitor hundreds of neurons simultaneously, including in mixed cultures with other cell types. Ca imaging can assess the activity of iPSC-derived iGlut neurons with single-cell resolution within a population, while maintaining crucial neuronal connections. Third, MEA measurements in Mg²⁺-free ACSF align well with those obtained using Ca imaging. Although it does not offer single-cell resolution, recording neuronal activity in Mg²⁺-free ACSF is straightforward for MEA and can provide another dataset for studying neuronal activity of iGlut neurons. Fourth, in addition to GECIs such as GCaMP, chemical Ca²⁺ dyes (e.g., Fluo-4) can also be used to monitor Ca²⁺ transients in Mg²⁺-free ACSF. However, the ability to specifically target neurons with chemical dyes is not possible in mixed cultures, since these indicators are not cell-type specific such as their GECI counterparts. Lastly, although it provides slower throughput, single-cell patch-clamp electrophysiology reveals a rich panoply of firing behaviors for iGlut neurons across different donor lines, including intrinsic and subthreshold membrane potential changes as well as synaptic activities. Some disadvantages of the Ca imaging approach are that large imaging datasets are generated, which require automated Ca²⁺ transient detection software. Lastly, the temporal resolution of Ca imaging is slower than patch-clamp electrophysiology, although we show good agreement between the three methods used in Mg²⁺-free ACSF (Table 1).

In summary, this study establishes a scalable methodology to enhance the functional characterization of human iPSC-derived excitatory neurons under Mg²⁺-free conditions. By leveraging Ca imaging, we demonstrate that Mg²⁺-free ACSF robustly increases neuronal activity in iGlut neurons compared to ACSF, is reversible, and is suitable for population network analyses of neuronal activity across multiple iPSC donor lines. This approach not only overcomes challenges associated with limited sample sizes in iPSC-derived neurons studies but could also reveal donor-specific functional phenotypes, highlighting its utility in personalized CNS disease modeling and drug screening. These findings provide a foundation for integrating functional studies into large-scale iPSC-based research efforts. For example, alcohol (i.e., ethanol) blocks NMDA receptor activity,⁴⁷ which contributes to both acute and long-term effects of alcohol as well as the development of alcohol use disorder, and our findings could have important implications for understanding alcohol addiction and other related disorders.

Limitations of the study

In the current study, we demonstrate using three different recording techniques (i.e., Ca imaging, MEA recordings, single-cell patch-clamp electrophysiology) that switching to a Mg²⁺-free condition increases the activity of iGlut neurons through active NMDA receptors. However, there are some potential limitations with this approach. We primarily focused on NGN2-derived iGlut neurons due to their abundance of glutamate neurons that release glutamate and express glutamate receptors. The robust increase in activity we observed in Mg²⁺-free conditions may not fully translate to other in vitro models of iPSC-derived neurons, such as co-cultures of iGlut and iGABA neurons, growth-factor-induced cortical neurons, or cortical organoids. Although this was not the aim of the current study, it will be important to evaluate the generalizability of this approach to other iPSC-derived neuronal systems. Another limitation we encountered was with longitudinal studies of firing activity. In MEA experiments, switching from Neurobasal to ACSF or Mg²⁺-free ACSF solutions appeared to stress neurons, making it difficult record from neurons on MEA plates repeatedly over time during maturation. One workaround to this would be to develop a version of BrainPhys or Neurobasal that lacks Mg²⁺, mitigating the harm to neurons with media exchanges. Longitudinal experiments using Ca imaging to follow the same neuron over time are also a limitation. Re-imaging the same neurons over time is challenging because it is not always possible to identify the same imaging area and neurons on different days. However, a workaround to this limitation would be to use new MEA plates or new coverslips for each time point in the time course. Thus, one would design an experiment with neurons growing on either MEA plates or cover slips that matured in parallel but were then used once for the recording.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Paul Slesinger (paul.slesinger@mssm.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •Data: Single-cell RNA-seq data have been deposited at GEO and are publicly available as of the date of publication. Accession numbers are listed in the key resources table.
• •Code: All original code has been deposited at Zenodo and is publicly available as of the date of publication. DOIs are listed in the key resources table.
• •Additional information: Any additional information required to reanalyze the data reported in this article is available from the lead contact upon request.

Acknowledgments

We want to acknowledge members of the Slesinger lab and the Goate lab for discussions throughout this work, as well as feedback from COGA investigators. This study is supported by NIH Grant U10AA008401 from the National Institute on Alcohol Abuse and Alcoholism (NIAAA). We also gratefully acknowledge the support provided by the Training Program in Stem Cell Biology fellowship from the New York State Department of Health (NYSTEM-C32561GG) to I.G.R. Some images in the figures were drawn with Biorender. The Collaborative Study on the Genetics of Alcoholism (COGA), Principal Investigators B. Porjesz, V. Hesselbrock, A. Agrawal; Scientific Director, A. Agrawal; Translational Director, D. Dick, includes ten different centers: University of Connecticut (V. Hesselbrock); Indiana University (H.J. Edenberg, T. Foroud, Y. Liu, M.H. Plawecki); University of Iowa Carver College of Medicine (S. Kuperman, A. Anderson); SUNY Downstate Health Sciences University (B. Porjesz, J. Meyers); Washington University in St. Louis (L. Bierut, A. Agrawal, S. Hartz); University of California at San Diego (M. Schuckit); Rutgers University (D. Dick, R. Hart, J. Salvatore, J. Tischfield); The Children’s Hospital of Philadelphia, University of Pennsylvania (L. Almasy); Icahn School of Medicine at Mount Sinai (A. Goate, P. Slesinger); and Howard University (D. Scott). Other COGA collaborators include: C. Holzhauer, M. Hesselbrock (University of Connecticut); D. Lai, J. Nurnberger Jr., L. Wetherill, X., Xuei, S. O’Connor, (Indiana University); J. Kramer (University of Iowa), G. Chan (University of Iowa; University of Connecticut); C. Kamarajan, A. Pandey, D.B. Chorlian, P. Barr, S. Kinreich, G. Pandey, Z. Neale, S., C. Chatzinakos, J. Zhang, Saenz deViteri, R. Christian, A. Bingly (SUNY Downstate); G. Pathak (Icahn School of Medicine at Mount Sinai); A. Anokhin, K. Bucholz, F. Dong, A. Hatoum, E. Johnson, V. McCutcheon, J. Rice, S. Saccone (Washington University); F. Aliev, Z. Pang, S. Kuo, S. Brislin, J. Moore (Rutgers University); A. Merikangas (The Children’s Hospital of Philadelphia and University of Pennsylvania); M. Gitik, NIAAA Staff Collaborator. We continue to be inspired by our memories of Henri Begleiter and Theodore Reich, the founding PI and Co-PI of COGA, and also owe a debt of gratitude to other past organizers of COGA, including Ting-Kai Li, P. Michael Conneally, Raymond Crowe, and Wendy Reich, for their critical contributions. Special thanks to the COGA collaborators who collected and classified the samples from the donors that were used in this study.

Author contributions

I.G.R. conceived the study, performed the Ca imaging data acquisition and analysis, and contributed to the writing and critical revision of the article. A.J.T. conceived the study, performed the MEA and patch-clamp data acquisition and analysis, and contributed to the writing and critical revision of the article. I.P. contributed to the initial discussions and the design of the data analysis algorithms. C.K. provided editorial feedback on the whole article. Z.P.P. contributed to discussions and critical revisions of the article. A.M.G. contributed to the critical revision of the article. R.P.H. performed the RNAseq data acquisition and analysis, and contributed to discussions, writing, and critical revisions of the article. P.A.S. conceived and supervised the study and data analyses, and contributed to the writing and critical revisions of the article. All authors have approved the final article.

Declaration of interests

A.M.G. is a member Scientific Review Board for Genentech and has previously served as a consultant for Merck. The rest of the authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological samples
Mouse Glia	C57BL/6 mice, Institutional Animal House	N/A
Chemicals, peptides, and recombinant proteins
StemMACSTM iPS Brew XF	Miltenyi Biotec	#130-104-368
DMSO	Sigma-Aldrich	# D8418
Matrigel	Corning Inc.	#354230
ReLeSRTM	Stemcell Technologies	#100-0483
AccutaseTM	Stemcell Technologies	#07920
ROCK inhibitor Y27632	Miltenyi Biotec	#130-103-922
NeurobasalTM	Thermo Fisher/Gibco	#21103049
CultureOneTM	Thermo Fisher/Gibco	#A3320201
B27TM	Thermo Fisher/Gibco	#17504044
GlutaMAXTM	Thermo Fisher/Gibco	#35050061
Heat Inactivated Fetal Bovine Serum (HI FBS)	Fisher Scientific	#MT35011CV
DMEM	Thermo Fisher	#11965092
Neurobasal plusTM	Thermo Fisher/Gibco	#A3582901
CultureOneTM	Thermo Fisher/Gibco	#A3320201
B27TM Plus	Thermo Fisher/Gibco	#A3582801
GlutaMAXTM	Thermo Fisher/Gibco	#35050061
Papain	Sigma-Aldrich	#A3582801
Cosmic calf serum	Fisher Scientific	#SH3008704
Sodium pyruvate	Thermo Fisher	#11360070
MEM non-essential amino acid solution	Thermo Fisher	#11140050
2-mercaptoethanol	Sigma-Aldrich	#M6250
Critical commercial assays
Chromium Next GEM Single Cell Multiome ATAC + Gene Expression Reagent Bundle	10X Genomics	#1000285
Deposited data
Raw single-cell RNA sequencing data	Present work, deposited at, GEO repository	GSE293298
Experimental models: Cell lines
COGA iPSC line	COGA repository	8092
COGA iPSC line	COGA repository	9206
COGA iPSC line	COGA repository	10884
Recombinant DNA
pTet-O-Ngn2-puro	AddGene	#52047
FUdeltaGW-rtTA	AddGene	#19780
FSW_jGCaMP8f	AddGene	#197034
FUGW	AddGene	#14883
pGP-AAV-syn-jGCaMP8f-WPRE	AddGene	#162376
pMD2.G	AddGene	#12259
pMDLg/pRRE	AddGene	#12251
pRSV-Rev	AddGene	#12253
Software and algorithms
NIS elements AR	Nikon	version 5.21.03
AirPLS	Zhang et al.48	https://github.com/zmzhang/airPLS
Default “signal” package in R	Ligges, U.49	https://CRAN.R-project.org/package=signal
AxIS Navigator	Axion Biosystems	Version 3.7.2
Easy Electrophysiology	Easy Electrophysiology	https://www.easyelectrophysiology.com/
AxIS Metric Plotting tool	Axion Biosystems	Version 2.5.0
Neural Metric Tool	Axion Biosystems	Version 4.0.5
Other
Original Analysis codes	This work, deposited at Github	https://doi.org/10.5281/zenodo.14841953

Experimental model and study participant details

Generation of human iPSCs and iGlut neurons

Subjects were selected from the NIAAA/COGA Sharing Repository of the Collaborative Study on the Genetics of Alcoholism (COGA) (https://cogastudy.org). We used a neurogenin2 (NGN2) protocol to generate glutamatergic neurons (iGlut) from human iPSCs,⁴ with some modifications. iPSCs were selected from three different COGA study donors who were unaffected by AUD: lines 10884 (male, 27 y/o), 8092 (female, 33 y/o), and 9206 (female, 22 y/o). Each donor line was tested for expression of pluripotency markers and for the presence of a euploid karyotype.¹³ Original donor iPSC lines received from the COGA repository had the following passages: P7 for line 9206, P14 for 8092, and P11 for 10884; they were stored with 2-5 additional passages, and used for the experiments described in this manuscript with 4-6 additional passages from the original vial. The human iPSCs donor lines were frozen using culture media (StemMACS^TM iPS Brew XF, human, Miltenyi Biotec, cat. #130-104-368) with 10% DMSO (Sigma-Aldrich, cat. #D8418; 0.5-1∗10⁶ cells/cryotube) and frozen in liquid nitrogen, and were thawed and cultured on Matrigel-coated (1:250 dilution; Corning Inc. cat. #354230) 6-well plates for at least one week before the induction. All donor iPSC lines in culture were checked daily for bacterial and/or fungal contamination (bright field microscope visualization) and tested once a month for mycoplasma contamination. iPSCs donor lines were maintained in StemMACS^TM media until the day of induction (at 37°C, 5% CO₂) and passaged at least once after thawing before starting the differentiation, with a routine splitting rate of 1:6, using ReLeSR^TM as dissociation reagent (Stemcell Technologies, cat. #100-0483).

Mouse glia cultures

We used postnatal day 3 C57BL/6J wild-type mice to produce glial cell cultures for the co-culture with our iGlut neurons. Mice were housed with their breeding pairs of two females and one male in a 12 h light/dark cycle at 22 ± 2^oC with food and water available ad libitum. All mouse procedures were carried out following protocols approved by the Institutional Animal Care and Use Committee at the Icahn School of Medicine at Mount Sinai.

Method details

NGN2 induction

For generating the iGluts we used a doxycycline-inducible tetO-Ngn2-T2A-Puro/rtTA lentiviral induction. Lentiviral plasmids to induce neurons (pTet-O-Ngn2-puro and FUdeltaGW-rtTA) were obtained from AddGene (#52047 and #19780, respectively), sequenced to confirm identity, and packaged using standard packaging plasmids (pMD2.G, #12259; pMDLg/pRRE, #12251; pRSV-Rev, #12253). The hSyn1-driven GCaMP8f virus was constructed from pGP-AAV-syn-jGCaMP8f-WPRE (AddGene #162376) by cloning the hSyn1-gGCaMP8f cassette into a backbone made from FUGW (#14883). The resulting plasmid is available from AddGene with cat. #197034. On the initial day of induction (D0) we dissociated the iPSCs with Accutase^TM (Stemcell Technologies, cat. #07920) and re-plated them in StemMACS^TM, supplemented with the NGN2 and rtTA containing lentiviruses and 5 μM of the ROCK inhibitor Y27632 (RO kinase inhibitor, Miltenyi Biotec cat. #130-103-922), onto another Matrigel-coated (2:250 dilution) plate, in a specific cell density (2.5-3∗10⁵/ml). The cells were counted with a Countess 3 automated cell counter (Invitrogen). On day 1 (D1), we changed the media to Neurobasal^TM (Thermo Fisher/Gibco, cat. # 21103049), supplemented with 1 V/V% CultureOne^TM (Thermo Fisher/Gibco, cat. #A3320201), 2 V/V% B27^TM (Thermo Fisher/Gibco, cat. # 17504044), 1 V/V% GlutaMAX^TM (Thermo Fisher/Gibco, cat. #35050061), and 0.1 V/V% ascorbic acid, with 2 μg/ml Doxycycline to activate the Tet-ON-based NGN2 induction. On D2, we added 1 μg/ml of Puromycin in the culture media for 48 h to select for transduced cells. On D4, we removed the Puromycin but maintained Doxycycline for another 24 h. In parallel, on D4 we plated the mouse glia on acid-edged, Matrigel-coated (4:250 dilution) coverslips (12 mm) in a 24-well plate. Neurobasal was supplemented with 5 V/V% Heat Inactivated Fetal Bovine Serum (HI FBS, Fisher Scientific, cat. #MT35011CV) as a plating media for the mouse glial cells (6.5-7∗10⁴ cells/well). On D5, iGlut neurons were re-plated onto the mouse glia in the 24-well plate. First, iGlut neurons were dissociated with Accutase (6 minutes incubation at 37^oC), harvested and diluted in DMEM (Thermo Fisher, cat. #11965092, 1:4 dilution), then centrifuged for 4 minutes at 0.8 rcf at room temperature. The pellet was resuspended in 1 ml of Neurobasal + 2 V/V% FBS, counted and neurons diluted to 1-1.5∗10⁵ cells/ml in a 24 ml final volume of Neurobasal + 2 V/V% FBS. We gently aspirated the media from mouse glia and plated the iGlut neurons (1 ml of neuron suspension to each well) onto the glia. On D8, we switched the co-culture to Neurobasal plus^TM media (Thermo Fisher/Gibco, cat. #A3582901) supplemented with 1 V/V% CultureOne^TM (Thermo Fisher/Gibco, cat. #A3320201), 2 V/V% B27^TM Plus (Thermo Fisher/Gibco, cat. #A3582801), 1 V/V% GlutaMAX^TM (Thermo Fisher/Gibco, cat. #35050061), 0.1 V/V% ascorbic acid, 2 V/V% FBS and 2 μM Ara-C (to prevent survival of mitotic cells) by doing a half media change with Neurobasal plus + 4 μM Ara-C. On D11, we refreshed the Ara-C with another half media change with 2 μM Ara-C in Neurobasal plus + 2 V/V% FBS. On D15, we started the removal of Ara-C (half media change only with Neurobasal plus + 2 V/V% FBS) and then cultures were maintained with half media changes using Neurobasal plus + 2 V/V% FBS twice a week.

Mouse glia cultures

Dissected brain cortices from 3 pups (at p0-3) were dissociated in a papain-containing (Sigma-Aldrich, cat. #A3582801) solution (19-38 units Papain, 0.5 μM EDTA, 1 μM CaCl₂ in HBSS) at 37°C for 15 minutes, with gentle shaking every 5 minutes. The dissociation solution was removed and treated tissue was washed twice (the media was added then removed with caution) with 10 ml MEF media (88 V/V% DMEM, 10 V/V% calf serum (Fisher Scientific, cat. #SH3008704), 1 V/V% sodium pyruvate (100 mM, Thermo Fisher, cat. #11360070), 1 V/V% 100x MEM non-essential amino acid solution (Thermo Fisher, cat. #11140050), 0.0008 V/V% 2-mercaptoethanol (Sigma-Aldrich, cat. #M6250). The tissue was then triturated in 1 ml of MEF with a pipette until no large tissue chunks were visible. An additional 4 ml MEF media was added, and the mixture was passed through a 0.4 μm cell strainer into a 50 ml falcon tube containing 5 ml MEF media and seeded in T75 flask. The media was changed the next day and then changed again every 3 days until the glial cells became confluent (in approximately 7 days). The cells were passaged at least once using trypsin and MEF media, with the media being changed every 3-4 days until plating on coverslips for the neuronal co-cultures. We never used glia older than 10 days or passaged more than three times (more than P3).

Single-cell RNA sequencing

Single-cell RNA sequencing was performed on induced iGlut neuron cell villages as described previously.^2,3 Cells were matched with subjects using demuxlet¹ and aggregated by subject. Scaled, normalized gene counts were extracted and plotted using the pheatmap function in R⁵⁰ to plot color-scaled expression of genes shown from donor lines 10884, 9206, and 8092; genes were grouped by function.

Ca imaging

Lentivirus expressing GCaMP8f was produced and used to transduce (1∗10⁶ IU/100,000 cells) iGlut neurons at D28-D33, achieving expression levels suitable for Ca imaging around 2 weeks after infection (D42-D47). We measured Ca²⁺ transients at D72-D85 (see Figure 1A) using a Nikon Eclipse TE2000-U microscope equipped with 20x objective, a 480 nm LED (Mic-LED-480A, Prizmatix Ltd.) passing through a HQ480/40x nm excitation filter (Q505LP dichroic mirror) and a HQ535/50m emission filter (Semrock), a sCMOS Zyla 5.5 camera (Oxford Instruments, Andor), and NIS elements AR software (version 5.21.03, Nikon) for data collection and analysis. The images were obtained using 150 ms exposure time, 6.14 fps frame rate, and 4x4 binning, under constant perfusion with different ACSF-based solutions, in a laminar flow diamond-shaped chamber (Model #RC-25; Warner Instruments) at room temperature (RT, ∼20^oC). Different bath solutions were applied using a gravity fed system connected to a ValveBank (AutoMate) that was controlled remotely and provided a TTL time stamp. For all of our experiments, we used a custom-made artificial cerebrospinal fluid (ACSF; 125 mM NaCl, 5 mM KCl, 10 mM D-Glucose, 10 mM HEPES-Na, 3.1 CaCl₂, 1.3 mM MgCl₂,¹⁵ and Mg²⁺-free ACSF (125 mM NaCl, 5 mM KCl, 10 mM D-Glucose, 10 mM HEPES-Na, 3.1 CaCl₂) as base solutions. For all solutions, the pH was adjusted to 7.4 with HCl and the osmolarity to ∼300 mOsm with sucrose. To record only NMDA-R only activity, we used Mg²⁺-free ACSF with 10 μM NBQX (AMPA-R antagonist; Figures 1C and 1D) or 30 μM NBQX (Figures 1E and 1F). In some experiments, 100 μM AP-V (NMDA-R antagonist, Figures 1E and 1F) was also applied to block NMDA-R-dependent activity. We also examined the effect of the NMDA-R potentiator glycine (3 μM Gly).

Electrophysiology

We carried out whole-cell patch-clamp electrophysiology with iGlut (D30-90) neurons as described previously.¹⁵ Briefly, we recorded spikes in I = 0 current-clamp mode from iGlut neurons derived from the three human donor lines, 10884, 9206, and 8092 at room temperature. Borosilicate glass capillary pipets (3″ thin walled, 1.5 OD/1.12 ID, World Precision Instruments) were pulled with a PC-10 puller (Narishige International USA) and had resistances of ∼4 ΜΩs, with K-D-Gluconate internal solution (140 mM K-D-Gluconate, 4 mM NaCl, 2 mM MgCl₂-6H₂O, 1.1 mM EGTA, 5 mM K-HEPES, 2 mM Na₂ATP, 5 mM Na-Creatine-PO₄, 0.6 mM Na₃GTP, pH = 7.4, osmolarity = ∼290 mOsm) and ACSF or Mg²⁺-free ACSF external. Recordings were made with a MultiClamp 700B amplifier (Molecular Devices), low-passed filtered at 2 kHz, digitized at 20 kHz with a Digidata 1440A A/D converter (Molecular Devices) and stored on a laboratory computer. Solutions were applied locally through a perfusion pipet such that only the cell recorded from was surrounded by Mg²⁺-free ACSF. In this way, we could reveal NMDA receptor-dependent synapses on the recorded neuron. Recordings were carried out with pClamp 10 software.

Multielectrode array

MEA data was collected with an Axion Maestro Pro multiwell MEA plate reader with a CytoView MEA24, 24-well plate and CytoView MEA48, 48-well plate (Axion Biosystems). Our iGlut culture protocol on the MEA plate is based on the culture protocol provided by Axion biosystems,⁵¹ but we made several modifications to better fit our experimental approach. Prior to the seeding on the MEA plate, we followed the same protocol detailed above for iGlut neuron generation. On day 5 we coated the plates with Matrigel (4:250) for at least one hour before seeding the neurons. We dissociated the neurons from the 6 well plates following the above-described protocol and prepared a 1.2∗10⁷ neuron/ml suspension. Then we placed a 10 μl droplet of the neuron suspension directly onto the electrodes, and incubated the plates at 37^oC, 5% CO₂ for 1 h. After 1 h incubation, we plated the glial cells onto the neurons (6.5-7∗10⁴ cells/well) in the remaining 500 μl iGlut media (Neurobasal plus + 2% FBS). Throughout the differentiation and iGlut neuron maintenance we used the culture media compositions described above in our protocol above, instead of those suggested in the Axion Cell Culture protocol. From this point, we followed our usual maintenance procedure, with half-media changes twice a week. Recordings were performed under controlled conditions of 37°C and 5% CO₂.

Quantification and statistical analysis

Ca imaging analysis

For Ca imaging data analysis, we used the NIS elements AR software to select the ROIs (placed on the soma of neurons) and exported raw fluorescence traces. We used a baseline-drift correction with penalized least-squares algorithm (AirPLS)⁴⁸ and calculated ΔF/F₀ with the following formula [F(t)-F₀)/F₀], where F₀ was the minimum fluorescence intensity (RFU) in the first 10 s of the recording. Prior to peak detection, we applied a 3rd order Butterworth filter using the default “signal” package in R (version 4.3.0).⁴⁹ Ca²⁺ transients were detected by a custom-made R script. Ca²⁺ transients were identified based on their kinetics: < 6.4 s width, more than 325 ms rise phase, more than 650 ms fall phase, the rise phase duration less than the fall phase duration, the peak height is more than 5∗SD and more than 5∗max background signal.¹⁴

To measure the synchrony, Kendall rank correlation was performed on the binned (bin size: 1s) timestamps of the Ca²⁺ transients. We only used cells/ROIs for this analysis, which had at least 3 spikes in both ACSF and Mg²⁺-free conditions. We counted only the significant correlations (p < 0.05) toward the MCC.

Statistical details for Ca imaging experiments can be found in the corresponding figure legends.

Patch-clamp electrophysiology

To analyze patch-clamp data, we used Easy Electrophysiology software. Before the AP counting, we removed the baseline based on the first 30 s of the recording and a polynomial fitting method to remove the possible baseline fluctuation, which was crucial for the AP thresholding. For spike counting, we used automatic thresholding, which uses the first derivative method, with the rise time 3 mV/ms, fall time 1 mV/ms, and 5 ms AP width.

We classified iGlut neurons as tonic and bursting neurons based on the interspike intervals (ISIs). First, we determined the instantaneous firing rate (IFR = 1/ISI) and then calculated the standard deviation of the IFRs. Cells with tonic firing usually present single APs with relatively equal ISIs between them, while bursting neurons have short ISIs between the spikes in the bursts and long ISIs between the bursts themselves. Based on this, we were expecting a high SD of IFR for bursting and a low SD of IFR for tonic neurons. To classify the recorded population of iGlut neurons, we first plotted the IFR SD values for each cell, then we performed K-mean clustering on the dataset (Figure S4). To calculate the ideal number of clusters we used the elbow method. For a proper analysis, we needed at least 3 APs (two ISIs/IFRs to calculate the SD). We identified the high IFR SD cluster as bursting and the low IFR SD cluster as tonic neurons. The resting membrane potential (RMP) was corrected for a junction potential of -17 mV.

Statistical details for patch-clamp electrophysiology experiments can be found in the corresponding figure legends.

MEA

To analyze the MEA data, we used the AxIS Navigator (Version 3.7.2, Axion Biosystems), to detect spikes and to make a summary of the recordings, which contains the following information: MFR, WMFR, number of network bursts and synchrony, for all wells of the MEA plate. Further, we used AxIS Metric Plotting tool (Version 2.5.0, Axion Biosystem) and Neural Metric Tool (Version 4.0.5, Axion Biosystems) to generate plots and summary reports. The mean firing rate (MFR) was calculated based on all electrodes, including inactive electrodes (firing rate < 5 spikes/min), while the weighted mean firing rate (WMFR) was calculated based on the active electrodes only. The WMFR was only calculated for those wells which had at least one active electrode. The number of network bursts refers to bursts that occurred simultaneously across the majority of electrodes. Synchrony is defined by a synchrony index, which is a unitless value between 0 and 1.^52,53 The time window for synchrony was set as 20 ms. To calculate synchrony, a minimum of 2 spikes in a particular well was required. In Mg²⁺-free ACSF almost all electrodes were active in all wells (15-16 active electrodes out of the 16/well).

Statistical details for MEA experiments can be found in the corresponding figure legends.

Further data analysis

All subsequent data representation, analysis, and statistics were carried out in R, Python, GraphPad Prism, and Microsoft Excel software.

Throughout the analyses, we used paired t-test and ANOVA if the data followed normal distribution, and their non-parametric counterparts, Wilcoxon test, Mann–Whitney U or Kruskal–Wallis test with Dunn’s post hoc, if the data did not draw from the normal distribution. For correlation analysis, we used Kendal rank correlation. Averages are shown as mean ± SEM. Significance was indicated by ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗∗p < 0.0001.

Additional resources

All original code has been deposited at the GitHub repository named “Uncovering NMDA receptor-mediated high firing activity in excitatory human neurons”⁵⁴ and is publicly available at https://doi.org/10.5281/zenodo.14841953 as of the date of publication.

Raw single cell RNA sequencing data is publicly available on the GEO repository, with accession number GSE293298.

Further information and requests for resources, data and reagents should be directed to and will be fulfilled by the lead contact, Paul A. Slesinger (paul.slesinger@mssm.edu).

Published: February 17, 2026