Temporal Dynamics of Spontaneous Ca2+ Transients, ERBB4, vGLUT1, GAD1, Connexin, and Pannexin Genes in Early Stages of Human Stem Cell Neurodifferentiation

Abstract

Spontaneous Ca2+ transients drive stem cell proliferation and neurodifferentiation. Deciphering the relationship between neuronal and glial human genes on one side and spontaneous Ca2+ activity on the other side is essential for our understanding of normal brain development, and for insights into the pathogenesis of neurodegenerative and neurodevelopmental disorders. In the present study, forebrain neurons were derived from human embryonic and induced pluripotent stem cells (hESC-H9 and iPSC-15; 22q11.2 deletion) over a period of 21 days in vitro (DIV). Every 1–2 days, multisite optical imaging technique was applied to detect populations of cells with spontaneous Ca2+ transients. The expression levels of 14 genes of interest were analyzed by quantitative polymerase chain reaction (qPCR) on the same biological samples where physiological recordings were performed. The genes analyzed include: the schizophrenia candidate gene ERBB4, connexin (Cx) genes Cx26, Cx36, Cx43, Cx45, Cx47, pannexin-1 (PNX1), neuronal markers PAX6, vGLUT1, GAD1, TUBB3, glial lineage markers BLBP, GFAP, and housekeeping gene ACTB. We found that Ca2+ signals decrease in amplitude, decrease in duration, and increase in frequency during the first 21 days of human neurodifferentiation. The expression levels of ERBB4, PAX6, GAD1, vGLUT1, BLBP, Cx36, Cx45, and PNX1 were found to be strongly positively correlated with the percentage of cells exhibiting spontaneous Ca2+ transients (“Active Cells”). While expression of BLBP, Cx45, ERBB4, GAD1, PAX6, PNX1, and vGLUT1 were correlated with short-duration and long-amplitude Ca2+ transients, Cx43, TUBB3, and Cx47 were better correlated with long-duration and short-amplitude transients. The expression dynamics of Cx26 was unrelated to any aspect of spontaneous Ca2+ activity. Four genes showed an exponential time course with a distinct onset on a given DIV. The onset of PNX1, ERBB4, and vGLUT1 occurred before, while the onset of Cx36 occurred after the first action potentials were detected in early differentiating human neurons.

1. Introduction

Deciphering human brain development requires the understanding of physiological processes of neuron progenitors and differentiating neurons [1–3]. Although electrical and optical recordings from postmitotic human neurons can be achieved using acute human fetal brain slices [4,5], the overwhelming majority of studies on the physiology of postmitotic human neurons use human embryonic or induced pluripotent stem cells [6–12]. The implementation of human stem cell technologies brings two advantages:

1. Culturing of human neurons alleviates severe ethical and technical limitations associated with physiological recording in fetal brain slices.

2. In vitro differentiation of human neurons underlies state-of-the-art experimental models of neurodegenerative disorders [13–15].

Much of the current research effort on stem cell-derived human neurons is focused on the gene expression levels, but it is also necessary to elucidate physiological signals that either drive genes, or are driven by genes. Among these signals are transient changes in intracellular Ca²⁺ concentration [16–19].

Spontaneous Ca²⁺ transients are detected in the developing cortex before the input of sensory information [20]. In the earliest stages of neurogenesis, Ca²⁺ transients control fundamental cell processes such as proliferation, migration, and differentiation of neural progenitors into neurons [2,21,22]. Spontaneous Ca²⁺ transients differ in characteristics such as duration, frequency, and amplitude, encoding information that can be decoded by proteins within the cell, thus allowing the latter to exert control over various processes. High-frequency Ca²⁺ oscillations trigger events such as the release of neurotransmitters, whereas low-frequency activity is indicative of slower processes such as gene transcription [23].

Given the interdependence between Ca²⁺ signals and the gene transcription process, we hypothesized that the expression levels of a particular human gene may correlate with the onset of a particular type of Ca²⁺ activity in differentiating cells. This working hypothesis is bidirectional. On one hand, a gene may code for proteins that bring Ca²⁺ into the cells, in which case an increase in gene expression changes the pattern of ongoing Ca²⁺ transients. On the other hand, if Ca²⁺ activity activates gene transcription, mRNA levels will rise in the cytosol in response to a certain amplitude, duration, or frequency of spontaneous Ca²⁺ transients. In order to test for possible correlations between Ca²⁺ activity and gene expression levels, forebrain neurons were derived from human stem cells over 21 days in vitro (DIV). At regular time intervals (every 2 days), the patterns of spontaneous Ca²⁺ signaling were recorded optically using a Ca²⁺-sensitive dye. Multisite Ca²⁺ imaging allowed us to determine the duration, frequency, and amplitude of spontaneous Ca²⁺ transients at each DIV. The physiological measurements were followed by gene expression analysis on the same in vitro samples (cell cultures) in which physiological recordings were made. We experimented with hESC and iPSC technologies in order to study the physiological activity and gene expression during the earliest stages of human neurodifferentiation, when pluripotent cells transition into neuron progenitors, which in turn transition further to postmitotic neurons. The following 14 genes were examined: the schizophrenia candidate gene ERBB4; neuronal markers: PAX6, vGLUT1, GAD1, TUBB3; the glial lineage markers BLBP and GFAP; several connexins (Cx26, Cx36, Cx43, Cx45, Cx47); a pannexin (PNX1); and the housekeeping gene ACTB. We found that Ca²⁺ transients with specific temporal properties are significantly correlated to the expression of a select group of genes during these early stages of neuronal development.

2. Materials and method

2.1. Cell lines

Human embryonic stem cell line (hESC-H9) was obtained from the University of Connecticut Stem Cell Core. The induced pluripotent stem cell line (iPSC-15, 22q11.2 deletion, passages 11–14, in [24] called SZ22del15-6) was created at the Albert Einstein College of Medicine (Bronx, New York). Both cell lines have also been used in a previous report [8], where it has been demonstrated that there is no difference in spontaneous activity or expression of selected genes between these two lines.

2.1.1. Differentiation protocol

Stem cell lines were differentiated using a five-stage protocol (Figure 1A) with defined media constituents [6, 25]. This protocol does not use typical brain region morphogens in Stages IV–V (besides those present in B27), and so is associated with the default production of forebrain neurons [26]. Briefly, stem cell colonies (Stage I) were dissociated by collagenase, and stem cell aggregates were incubated for 4 days in stem cell (SC) media without bFGF on ultralow adherence plates (Costar, Wilkes Barre, PA) (Stage II). 5 μM dorsomorphin (Chemdea, Ridgewood, NJ) and 5 μM SB431542 (Ascent, Princeton, NJ) were added to the media from days 1–8. Stem cell aggregates were then seeded on dishes coated with 1:100 Geltrex and allowed to expand for 4–8 days in NEP (neuroepithelial)–basal medium (Stage III) until neuroepithelial colonies appeared. NEP-basal medium consisted of DMEM/F12, 1 mg/ml BSA (Sigma), 1× N2, 1× B27 supplements without retinoic acid, and 1× penicillin/streptomycin/antimycotic. Thirty minutes before selecting the resultant neuroepithelial colonies, 0.66 mg/ml ROCK inhibitor (Y27632, Wako USA, Richmond, VA) was added to the media. Colonies with neuroepithelial morphology were removed by trituration and seeded with 0.66 mg/ml ROCK inhibitor on glass cover slips coated with 1:100 Geltrex. Cells were grown in NEP-basal medium with 20 ng/ml basic fibroblast growth factor (bFGF) for 7 days (Stage IV). Cells were then maintained in NEP-basal medium in the presence of 10 ng/ml BDNF (brain-derived neurotrophic factor) (Peprotech, Rocky Hill, NJ) and 10 ng/ml GDNF (glial cell line-derived neurotrophic factor) (Peprotech) (Differentiation Stage V and maintenance). Media were changed every other day.

Figure 1.

Human stem cell neurodifferentiation. (A) Timeline of the differentiation protocol comprising five stages. SC, induced pluripotent or embryonic stem cells; EB, embryoid bodies; NE, neuroepithelial colonies. (B) Dark-field photograph of undifferentiated stem cell colonies. (C) IR DIC photograph of the neuroepithelium with formed rosettes Stage III. (D₁) IR DIC photograph of postmitotic neurons––Stage V. (D₂) The same neurons as in D₁, loaded with OGB1-AM and photographed in the fluorescence channel (excitation, 470 nm). (E) Whole-cell recordings of spontaneous electrical activity from cultured cells on DIV-13 to DIV-24. Resting membrane potential shown under the trace. Action potentials are truncated for clarity.

2.1.2. Electrophysiology

Whole-cell patch recordings were performed as previously described [6]. Briefly, before recordings, cells were transferred to an Olympus (Tokyo, Japan) BX51WI upright microscope (equipped with infrared video microscopy) and incubated in DMEM/F12, 0.075% BSA, 0.25× B27, Pen/Strep/Antimycotic, and 14 mM HEPES, pH 7.4. Individual cells were selected for recordings based on a small round or ovoid cell body (diameters, 5–10 μm) and typically two or more extended processes. Pipettes (10 MΩ) were filled with an intracellular solution containing (in mM) 135 K-glutamate, 10 HEPES, 2 MgCl₂, 3 ATP-Na₂, 0.3 GTPNa₂, and 10 P-creatine Na₂, pH 7.3. Recordings were performed using Multiclamp 700B and Clampex9.2 (Molecular Devices, Union City, CA). In the current clamp configuration, we measured the resting membrane potential, and then a negative holding direct current (in the range −2 pA to −15 pA) was applied to bring the membrane potential to approximately −60 mV. This was done to compare neurons under identical conditions during spontaneous activity. Spontaneous activity was monitored for 5 min continuously. Electrical traces were analyzed using Clampfit 9.2 (Molecular Devices).

2.2. Ca²⁺ imaging

Multisite Ca²⁺ imaging was used to record spontaneous activity in cells cultured on glass cover slips (12 mm). The cells were loaded with a Ca²⁺-sensitive fluorescent dye by incubating the cover slips for 30 min with membrane-permeable Oregon Green 488 Bapta-1-AM (OGB1-AM, Invitrogen 06807) at 37°C. The cells were then washed with warmed media and allowed to recover for additional 30 min in the incubator (37°C). Cover slips were placed in the recording chamber and perfused with modified Krebs-Ringer solution. Images were projected onto an 80 × 80 pixel CCD camera (NeuroCCD-SMQ, RedShirtImaging, Decatur, GA) and sampled at 800 ms/frame. One to four nonoverlapping visual fields were sampled per cover slip. The cells were then frozen and kept at −80°C until RNA extraction. Analysis of optical signals was done off-line using Neuroplex software (RedShirtImaging). The following definitions were applied for the analysis of data obtained by multisite Ca²⁺ imaging:

Active Cells: In each visual field, the number of cells showing at least one Ca²⁺ transient was divided by the total number of cells in that field (%).

Repetitive Activity: The number of cells in the visual field exhibiting four or more Ca²⁺ transients per 5 min of optical recording, divided by the number of “Active Cells” (cells showing at least one Ca²⁺ transient) in that field (%).

Amplitude of a Ca²⁺ transient was measured from the baseline before the signal onset to the signal peak, and expressed as ΔF/F (%).

Duration of the Ca²⁺ transient was measured at half amplitude (half-width) and expressed in seconds. The median of duration of all recorded Ca²⁺ signals (n = 4744) was determined to be at 8 s. Ca²⁺ signals were then separated into two groups, one consisting of “wide” and the other of “narrow” events. Wide: An event endowed with a half-width longer than 8 s. Narrow: An event was considered to be narrow if the duration (measured at half amplitude) was less than or equal to 8 s.

2.3. RNA extraction

The samples were removed from −80°C, and 300 μl of Trizol was added to each well containing a cover slip. Cells were detached from the cover slips by rinsing with Trizol. The cover slips were then discarded and 1 μl of yeast tRNA 10 μg/μl (Sigma) was added to each well. The cells were allowed to sit for 2 min at room temperature; then they were triturated, and the contents from each well were transferred to a separate 0.5 ml microcentrifuge tube. A volume of 60 μl of chloroform was added to each sample and shaken vigorously for 15 s. The tubes were incubated at room temperature for 15 min and were then centrifuged at 12,000 g for 15 min at 4°C. They were then placed on ice, and the clear, aqueous phase from each tube was removed and transferred to a new 0.5 ml microcentrifuge tube.150 μl of 100% isopropanol was then added to each sample. The contents of each tube were mixed, allowed to rest for 10 min at room temperature, and then centrifuged at 12,000 g for 10 min at 4°C. An RNA pellet was located in each tube and the supernatant removed. 0.3 ml of cold 75% ethanol was added to each tube, which was then vortexed and centrifuged at 7,500 g for 5 min at 4°C.The supernatant was again removed and the pellets allowed to air-dry. The pellets were then resuspended in 7 μl of RNase-free water and incubated at 55°C for 10 min, using the PCR machine (Eppendorf Mastercycler Gradient, Hauppauge, NY). In order to remove any genomic DNA contamination, samples were subjected to treatment with DNase according to the manufacturer’s protocol (New England Biolabs).

2.4. cDNA synthesis

The following were mixed: 5 μl RNA, 0.75 μl of mixed OligodT and Random Hexamers 1:1 (Qiagen, Valencia, CA), 1 μl (10 mM each) dNTPs (Promega, Madision, WI), and 3.25 μl RNase-free water (Invitrogen). There was also an NTC (no template control) tube which contained 5 μl of RNase-free water in place of RNA. The reagents in each tube were mixed thoroughly and placed in the PCR machine with a preheated lid of 80°C. The samples were incubated at 65°C for 5 min and then immediately placed on ice (for at least 1 min). 4 μl of 5× SuperScriptIII Buffer (Invitrogen), 1 μl DTT (Invitrogen), 1 μl RNaseOut (Invitrogen), 3 μl DNase-free water, and 1 μl SuperScriptIII (Invitrogen) were added to the 10 μl already present in each tube. 4 μl 5× SuperScriptIII buffer, 1 μl DTT, 1 μl RNaseOut, and 4 μl DNase-free water were added to −RT (No Reverse Transcriptase) tubes. The reagents in each tube were mixed thoroughly and placed in the PCR machine at 25°C for 5 min, 37°C for 10 min, 50°C for 80 min, and 70°C for 15 min. The samples were then stored at −20°C. For positive controls, total RNA of human fetal brain (Clontech, Mountain View, CA) was used.

2.5. Preamplification

Primer pools of long outer primers were made in water, 1 μM each for up to 14 genes in total. The following reagents were mixed for 80 μl pre-amp reactions: 16 μl 5× GoTaq Flexi Buffer Colorless (Promega), 4.8 μl of 25 mM MgCl₂, 1.6 μl of 10 mM dNTPs, 3.2 μl primer pool, 1.28 μl GoTaq DNA polymerase (Promega), 48.12 μl water, and 5 μl cDNA. The lid of the Eppendorf Mastercycler Gradient was preheated to 104°C, and the pre-amplification PCR was run at 94°C––4 min [94°C––30 s, 55°C––30 s, 72°C––40 s] for 10 cycles, and 72°C––10 min.

2.6. Quantitative polymerase chain reaction (qPCR)

qPCR was performed on the genes of interest by mixing the following reagents: 3 μl 5× PCR buffer, 0.9 μl 25 mM MgCl₂, 0.5 μl 10mM dNTPs, 1 μl 10 μM mixed forward and reverse primers (Table 1), 0.75 μl 10 μM FAM BHQ Probe (Sigma), 0.4 μl GoTaq, 5 μl preamplified cDNA, and 13.45 μl water. The final volume in each well was 25 μl. qPCR was performed using the Bio-Rad CFX Connect Real-Time PCR Detection System. Samples were run at 94°C––3 min, [94°C––30 s, 55°C––30 s, 72°C––30 s] for 40 cycles, with the annealing temperature dependent on the primers. Table 1 shows screened nested primers, probe sets, and the annealing temperatures. Gene expression data were normalized using housekeeping gene ACTB as a reference, using the 2-ΔΔCt method.

2.6.1. Data sets

In the present study, spontaneous Ca²⁺ activity was recorded in 25 cultures containing developing human neurons at various ages (differentiation days: 7, 8, 9, 10, 11, 14, 15,16, 17, 18, and 21). Ten cultures were from iPSC-15 and 15 cultures contained hESC-H9 cells (Table 2A). The total number of spontaneous Ca²⁺ signals analyzed from Day 7 to Day 21 is 4744. All 25 cultures were saved after the recording session and subjected to gene expression analysis. In addition to these 25 cultures, the identical gene analysis was performed on 23 cultures in which recordings were not performed (Table 2B). No significant differences in gene expression levels of 14 genes (see list above) were detected between recorded and nonrecorded cultures (data not shown); hence, the gene expression data from recorded and nonrecorded cultures were combined. The gene expression data from DIV 8 to DIV 21 were used in the “gene-to-gene” correlation analyses (N = 45 cultures/cover slips). The physiological and gene expression data from DIV 7 to DIV 21 were used in the activity-to-gene correlation analyses (n = 4744 calcium transients, N = 48 cultures).

Gene	Outer primers 5′–3′ forward; reverse (annealing temperature)
	Inner primers 5′–3′, forward; reverse (annealing temperature)
	Probe
ACTB	CCTCGCCTTTGCCGATCC; GATGCCGTGCTCGATGGGGT (55°C)
	CCTCGCCTTTGCCGATCC; GCGAAGCCGGCCTTGCACAT (55°C)
	famATGATATCGCCGCGCTCGTCGTCGAbhq
TUBB3	ACTCCCTTGAACAGGGACAGGGA; GTTCCGGGTTCCAGGTCCACCA (55°C)
	CACAGGCGTCCACAGTTCT; CGAGTCGCCCACGTAGTTG (55°C)
	famCATCAGTGATGAGCATGGCATCGACCbhq
ERBB4	GGAGTACTTGGTCCCTCAGGCT; GGGTGCCACTGGCTTGCGTA (55°C)
	GCAAGAATTGACTCGAATA; CTGGAATTGTGCTAGTTG (50°C)
	famCAGACACTCCTTGTTCAGCAGCbhq
GAD1	GATGATGGGCGTGCTGTTGC; GCGGCTGGGTTGGAGATGAC (55°C)
	CGACACCGGGGACAAGGCAATT; CCGTCGTTGAGGGCTGTCTGG (55°C)
	famCAATGGCGAGCCTGAGCACACAAACGTCTGbhq
vGLUT1	CGCTACATTATCGCCATCATGAG; GGTGGGGCCCATTTGCTCCA (55°C)
	CAGGAGGATTTATCTGTCAAAAAT; GGGTATGTGACCCCCTCTACCAAC (55°C)
	famATGCTGATCCCCTCAGCTGCCCGCbhq
PAX6	AGCCCAGTATAAGCGGGAGTGC; TCCCCCTCCTTCCTGTTGCTGG (55°C)
	TCTTTGCTTGGGAAATCCG; CTGCCCGTTCAACATCCTTAG (55°C)
	famTCATACATGCCGTCTGCGCCCATCTGbhq
PNX1	CGCTGTTTGTTCCATTCCGAC; TCTCTGTCGGGCATTCTTCTC (55°C)
	GGAGAATATTAAGAGCAGTGGTC; CTGTCTATGTTCATACCTTGGAG (45°C)
	famCGACCCAATGCTACTCCTGACAAACCTbhq
CX26	TCCCGACGCAGAGCAAAC; GCAGGGTGTTGCAGACAAAG (55°C)
	TCCCGACGCAGAGCAAAC; AGGGTGTTGCAGACAAAGTC (55°C)
	famCCCACACCTCCTTTGCAGCCACAbhq
CX36	GCAGCACTCCACTATGATCGG; GGAAGACCCAGTAACGTATGTG (55°C)
	ACTCCACTATGATCGGGAGGATCC; CTGCAGGGTGTTGCACAC (53°C)
	famCTGCTCATCATCGTACACCGTCTCCCCbhq
CX43	AGGAGTTCAATCACTTGGCG; GTACTGACAGCCACACCTTC (55°C)
	CTTGGCGTGACTTCACTAC; GCAGTTGAGTAGGCTTGAAC (55°C)
	famCGCTCCAGTCACCCATGTTGCCTGbhq
CX45	TGGACAACAGGGCATACC; GCAAACGCATCATAACAGAC (55°C)
	TGGACAACAGGGCATACC; CGGAAGACAATCAGAACAGTG (55°C)
	famAGTTGGAGCTTCCTGACTCGCCTGCbhq
CX47	CCCAGGACAATCAGGGATTC; CAGGCCAGAAAATTCCAGTC (55°C)
	CAGGACAATCAGGGATTC; CCAGAAAATTCCAGTCTCC (51°C)
	famCTGCTTGGATGCTGGTCCTCCbhq
BLBP	GGTGGGAAATGTGACCAAAC; TGTCTCCATCCAGGCTAACA (55°C)
	CAACGGTAATTATCAGTCAAG; GGCTAACAACAGACTTACA (51°C)
	famTCATCAGGACTCTCAGCACATTCAAbhq
GFAP	CTGCGGCTCGATCAACTCA; GTAGTCGTTGGCTTCGTGCT (55°C)
	GCCTATAGACAGGAAGCA; GTGGATCTTCCTCAAGAAC (49°C)
	famCCTCCAGCGACTCAATCTTCCTbhq

Table 1.

Screened nested primer and probe sets designed to amplify mRNA without amplifying genomic DNA.

Annealing temperatures for standard three-step PCR are included.

2.7. Statistical analysis

Statistical analysis was performed using SigmaStat 2.03 (Systat, San Jose, CA). Data are presented as mean ± SEM. Statistical differences between the groups were evaluated with Student’s t-test for paired data. Pearson’s product–moment correlation coefficient was used to determine the correlation between gene expression levels and spontaneous activity. Unavailable values from DIV 7 to DIV 21 were interpolated from the original data set using the method of Least Squares Fit (n = 15, where n = number of DIV; df = 13, where df = n – 2). Only values with p < 0.001 (r ≥ 0.76) were considered statistically significant. For “gene-to-gene” correlation analyses, n = 12, where n = number of DIV, and df = 10, where df = n − 2). DIV 7 was not included in these correlation analyses.

3. Results

3.1. Stem cell-derived human neurons

Two human stem cell lines (hESC-H9 and iPSC-15) were subjected to a five-stage neurodifferentiation protocol over a period of 21 days (Figure 1A). The first stage (Stage I) was used to grow and maintain undifferentiated stem cell colonies (Figure 1B). Day 0 (DIV 0) marks the Seeding of embryoid bodies. Upon seeding, the cells from the embryoid bodies begin to form neuroepithelial structures (Figure 1C, Rosettes). The neuroepithelial cells were expanded in a bFGF-rich media for 4 days (Stage IV). The expanded NEs were lifted on DIV 12 and re-seeded on glass cover slips in BDNF/GDNF-based neurodifferentiation media (Stage V). This protocol produced cells with neuronal morphologies (Figure 1D₁) and neuronal electrical properties, namely the ability to fire action potentials. Whole-cell recordings were performed on cells with a neuronal appearance––cells with processes visible in IR DIC microscopy (Figure 1D₁). That is to say that sampling of cells for electrical recordings was biased by visual appearance. The first action potentials were detected on DIV 14, n = 3 (n = number of cells) (Figure 1E). During the fifth stage (Stage V), whole-cell recordings revealed that some, not all, stem cell-derived neurons were spontaneously producing bursts of action potentials.

In our previous study, stem cell-derived human neurons were grown for 88 days, characterized electrophysiologically, and individual cells were captured for gene expression analysis [8]. The previous study, which used the same two stem cell lines under an identical five-stage differentiation protocol and systematic whole-cell recordings (n = 229 human cells from Day 13 to Day 88) produced detailed distributions of electrophysiological properties [8, supplemental figure S1]. In the present study, the focus is on the first 21 days of stem cell neurodifferentiation, when pluripotent stem cells transition to neural progenitors and when first postmitotic neurons are born. Also, in the present study, the gene analysis was performed on the whole population of cells, unbiased by visual appearance or physiological recordings; hence, all cell types present in the culture were included in the RNA samples.

3.2. Spontaneous Ca²⁺ transients

Every 1–2 days, glass cover slips containing cultured human cells were removed from the incubator, loaded with a Ca²⁺-sensitive dye OGB1-AM, and positioned in the recording chamber of the electrophysiological imaging station (Figure 1D₂). Multisite optical recordings of unprovoked (spontaneous) Ca²⁺ activity were performed from DIV 7 to DIV 21 in 136 locations (visual fields, Figure 2A₁) distributed across 25 cover slips (one visual field approximately 380 × 380 μm²). Regions of interest (ROIs) were placed on all OGB1-stained cells, indiscriminately (Figure 2A₂). Therefore, the ROIs include both neurons and nonneurons present in the culture. From each visual field, the Ca²⁺ signals were recorded for at least 5 min (Figure 2B). In the majority of visual fields, three successive optical recording trials were performed, extending the total monitoring to 15 min. Long recording sessions were needed, because spontaneous outbursts of electrical activity in developing human neurons are interspersed by silent periods that may last for several minutes [5]. Cells were deemed “active” if at least one Ca²⁺ transient was detected (Figure 2B, upper traces). Some cells showed no activity throughout the entire recording sessions (Figure 2B, lower traces). The percentage of cells displaying spontaneous Ca²⁺ activity (“Active Cells”) was calculated for each visual field by dividing the number of active ROIs (cells with Ca²⁺ signals) with the total number of ROIs in the visual field (Figure 2A₂). The average percentage of cells displaying spontaneous Ca²⁺ activity did not significantly differ between hESC-H9 and iPSC cells (Figure 2C). Therefore, in the remaining analysis (figures), the physiological data from the two cell lines were combined.

Figure 2.

Two human lines (hESC-H9 and iPSC-15) produce cells with similar Ca²⁺ activity and gene expression. (A₁) hESC-H9 cells on DIV-9 were bulk-loaded with OGB1-AM. Image captured by fast data acquisition camera (80 × 80 pixels). (A₂) Same as in A₁, except gray dashes mark the actual pixels used to generate regions of interest (ROI). (B) Spontaneous Ca²⁺ transients from 75 ROIs are shown. Each trace represents 3.5 min of continuous optical monitoring. Cells with spontaneous Ca²⁺ activity, “Active Cells,” are grouped on the top. The majority of ROIs have no spontaneous Ca²⁺ transients. (C) The average “Active Cells” per day of differentiation for two stem cell lines, hESC-H9 (black) and iPSC-15 (gray). “Active Cells”: the number of ROIs with Ca²⁺ transients divided by the total number of ROIs in the visual field. Error bars are s.e.m. The number of visual fields studied by multisite Ca²⁺ imaging in the two stem cell lines on each DIV are listed in Table 2A. On DIV-12, the expanding colonies were lifted and reseeded on fresh cover slips in a new type of a differentiation media (addition of BDNF and GDNF). (D) Gene expression of ACTB measured in the hESC-H9 (black) and iPSC-15 line (white). Each data point represents one cover slip (one culture). Black and gray trend lines are polynomial fits through the hESC-H9 and iPSC-15 data sets, respectively. (E) ERBB4 expression in two stem cell lines normalized to that of ACTB.

(A)	Physiology
	hESC-H9		iPSC-15
	n (visual fields)	N (cover slips)	n (visual fields)	N (cover slips)
DIV-07	4	2	0	0
DIV-08	20	2	0	0
DIV-09	8	1	6	1
DIV-10	8	2	13	2
DIV-11	7	1	3	1
DIV-14	7	1	0	0
DIV-15	6	1	8	1
DIV-16	2	1	5	1
DIV-17	6	1	9	1
DIV-18	12	2	5	2
DIV-21	2	1	5	1
Total	82	15	54	10
(B)	Gene expression analysis
	hESC-H9	iPSC-15	Combined
	N (cover slips)	N (cover slips)	N (cover slips)
DIV-07	3	0	3
DIV-08	3	0	3
DIV-09	5	0	5
DIV-10	1	4	5
DIV-11	3	2	5
DIV-12	0	2	2
DIV-14	1	1	2
DIV-15	2	3	5
DIV-16	3	1	4
DIV-17	2	2	4
DIV-18	3	3	6
DIV-20	1	2	3
DIV-21	1	0	1
Total	28	20	48

Table 2.

(A) Experimental samples used for physiological recordings. N indicates number of cover slips with cultured cells.

n indicates number of visual fields used for optical recordings. Upon physiological recordings cover slips were saved for gene analysis. (B) Experimental samples used for gene analysis. These samples include cover slips saved after physiological recordings (25 cover slips in A), as well as 23 additional cover slips that were saved without any optical recordings.

3.2.1. Active cells

The average percentage of cells displaying spontaneous Ca²⁺ activity increased steadily from DIV-7 through DIV-21 (Figure 3A). On DIV-7, the average number of active cells was 1 ± 0.1%. By DIV-14, the average number of active cells was 24 ± 4.64%, and by DIV-21, approximately 40% of cells were active (40 ± 6.3%). There was a statistically significant increase in the average percentage of active cells after the introduction of neurodifferentiation media (DIV-14 to DIV-21) as compared to that before its introduction (DIV-7 to DIV-11) (p < 0.05). As human neurons mature, they fire repetitive action potentials important to initiate contact with the surrounding cells [27,28]. Belinsky et al. [8] have demonstrated that repetitive activity of immature neurons can be used as a measure of their maturity.

3.2.2. Amplitude

The amplitude of spontaneous Ca²⁺ transients was analyzed over the first 3 weeks of human stem cell neurodifferentiation. Only cells with Ca²⁺ transients were included in the analysis (“Active Cells”) (Figure 2B, upper traces). The average peak amplitude of spontaneous Ca²⁺ transients was calculated for each day in vitro (Figure 3B), and there was a statistically significant decrease in the average percentage of active cells after the introduction of neurodifferentiation media (DIV-14 to DIV-21) as compared to that before its introduction (DIV-8 to DIV-11) (p < 0.05, one-tailed t-test). The difference is significant when the data series DIV-8 to DIV-10 is compared to the data series DIV-11 to DIV-21 (p < 0.05, two-tailed t-test). Larger signal amplitudes were predominantly distributed in the first half of the in vitro neurodifferentiation protocol, when cell cultures primarily consisted of undifferentiated cells and neuron progenitors (Figure 3B).

3.2.3. Repetitive activity

Among active cells, we found a group of cells with more frequent Ca²⁺ signals (repetitive activity). The criterion for repetitive activity was set to four or more spontaneous Ca²⁺ transients in the period of time required to complete one optical recording trial (5 min). The number of cells with repetitive activity was divided by the number of active cells in the same visual field, and expressed as a percentage (Figure 3C). The percentage of cells exhibiting repetitive activity significantly increased with the addition of brain and glial cell-derived neurotrophic factors (BDNF and GDNF) into the cell culture media (DIV-14 to DIV- 21, as compared to DIV-7 to DIV-11; p < 0.005).

3.2.4. Duration

We measured the half-width (duration measured at half amplitude) of all recorded Ca²⁺ signals (n = 4726 Ca²⁺ transients obtained from 132 visual fields in 23 cover slips; DIV-7 not considered) over the first 3 weeks of human stem cell neurodifferentiation. The duration of spontaneous Ca²⁺ transients was calculated for each day of differentiation and plotted as a function of time, day in vitro on the x-axis (Figure 3D). The signal duration in the time period between DIV-8 to DIV-11 was significantly longer compared to the time period DIV-15 to DIV-21 (p < 0.001). The median duration of spontaneous Ca²⁺ transients was determined to be 8 s. The Spitzer laboratory had previously reported rapid (narrow) and slow (wide) Ca²⁺ transients, but the significance of these Ca²⁺ transients at very early stages of human neurodifferentiation has as yet to be unraveled [29,30]. Based on a median duration of 8 s, we separated Ca²⁺ transients into two groups, one consisting of wide signals with duration of more than 8 s (Figure 3E, Wide) and the other of narrow signals with duration equal to, or less than, 8 s (Figure 3E, Narrow). This allowed us to calculate the percentage of narrow and wide signals at each day of differentiation (Figure 3F). The proportion of narrow Ca²⁺ transients increased from DIV-8 to DIV-21 (Figure 3F, black bars). However, at the beginning of the neurodifferentiation protocol, during Stages III and IV, the number of narrow and wide signals was about 1:1, with nearly half of the total number of Ca²⁺ transients falling into the narrow category and another half being categorized as wide. The difference between narrow and wide counts in this period (DIV-8 to DIV-11) was not statistically significant (p < 0.05). From DIV-15 onward, this initial balance was thrown off with narrow Ca²⁺ transients greatly outnumbering wide transients (Figure 3F, DIV-14 to DIV-21). There was a statistically significant difference between wide and narrow signal counts after the introduction of the differentiation factors in the culture medium, the cultures predominantly containing postmitotic neurons (DIV-14 to DIV-21), as compared to DIV-7 to DIV-11 (p < 0.01). Thus, the ratio of wide and narrow (less than 8 s) events can be attributed to the maturation of stem cell-derived neurons.

Figure 3.

Spontaneous Ca²⁺ transients during early stages of neurodifferentiation. (A) In this and all remaining figures, the data from two cell lines are pooled together. Average number of “Active Cells” per visual field from DIV-7 to DIV-21. “Active Cells”: the number of ROIs with Ca²⁺ transients divided by the total number of ROIs in that visual field (%). (B) Average amplitude of Ca²⁺ signals ΔF/F per day. The total number of spontaneous Ca²⁺ signals analyzed from DIV-7 to DIV-21 is 4744. (C) Cells with repetitive activity (4 or more Ca²⁺ transients per 5 min period) divided by the number of active cells in that visual field. An active cell is any cell with 1 or more Ca²⁺ transients per 5 min period. (D) Durations (half-widths) of all Ca²⁺ transients recorded in this project are plotted as average values per day. Same data set as in B. (E) Optical traces showing wide and narrow spontaneous Ca²⁺ transients. “Wide” is defined as a signal half-width (width at half amplitude) greater than 8 s. “Narrow” is defined as a signal with half-width ≤8 s. (F) Distributions of Wide (gray) and Narrow (black) spontaneous Ca²⁺ transients per day, expressed as percentage of all transients on that day in vitro. Same data set as in D.

Ca²⁺ release from internal stores has been reported to be especially sensitive to Ca²⁺ influx at early stages of differentiation, leading to the generation of Ca²⁺ transients of large amplitude and duration, with the sensitivity to Ca²⁺ influx decreasing as cells mature [29]. We found that larger amplitudes were predominantly distributed in the first half of our in vitro neurodifferentiation protocol, when cell cultures primarily consisted of undifferentiated cells and neuron progenitors. The amplitude of wide Ca²⁺ transients was not significantly correlated to the expression of any of the genes studied. Our results are consistent with those of Gu et al. [29]. We also observed a statistically significant positive correlation between percentage repetitive activity and the number of narrow events (action potentials; r = 0.874, p < 0.001).

The multisite Ca²⁺imaging data indicate that spontaneous Ca²⁺ transients were present during the proliferation and differentiation phases of the in vitro neurodifferentiation protocol. The temporal patterns of three distinct parameters of Ca²⁺ transients: (a) increasing frequency, (b) decreasing peak amplitude, and (c) decreasing duration, are consistent with the gradual transition between the Ca²⁺ activities of undifferentiated cells (large amplitude–long duration) driven by a release from internal stores [17,31,32] to the spontaneous Ca²⁺ activities of postmitotic human nerve cells driven by spontaneous electrical activity [30,33,34]. The relation between electrical activity and Ca²⁺ transients in human postmitotic neurons was previously determined by simultaneous whole-cell and Ca²⁺ imaging from the same cell [8, Figure 5].

Figure 4.

Gene expression levels during early stages of stem cell neurodifferentiation. The expression levels of 12 human genes (Cx26, Cx43, Cx47, TUBB3, Cx45, BLBP, PAX6, GAD1, PNX1, ERBB4, vGLUT1, and Cx36) normalized to the expression of ACTB over a period of 21 DIVs. Genes are organized in three rows based on the temporal patterns of expression. Top row: genes consistently expressed throughout the 21 DIVs, or with expression slowly rising and then slowly declining. Middle row: gene expression increases steadily throughout the study period (21 days). Bottom row: genes with expression starting very low and then rising exponentially as differentiation progresses. Each of the 48 data points shown in the graph represents one cover slip with cultured cells used in a qPCR reaction. N = number of cover slips. For DIV-7 (N = 3), DIV-8 (N = 3), DIV-9 (N = 5), DIV-10 (N = 5), DIV-11 (N = 5), DIV-12 (N = 2), DIV-14 (N = 2), DIV-15 (N = 5), DIV-16 (N = 4), DIV-17 (N = 4), DIV-18 (N = 6), DIV-20 (N = 3), and DIV-21 (N = 1) (Table 2B). The trend lines are second order polynomials, except gene Cx36 which is fitted with a fourth order polynomial.

3.3. Gene expression during neuronal differentiation

Gene expression analysis was performed on the same biological samples in which physiological recordings were made (Table 2A), in order to determine if there are any correlations between gene onset, gene dynamics, and spontaneous Ca²⁺ activity. Connexin and pannexin gene isoforms are expressed in the human fetal brain at the development stages when spontaneous Ca²⁺ activity is also found [35]; these genes were analyzed in the present study along with several neuronal, glial, and human genes. More precisely, the samples were tested for four classes of genes: (1) connexins (Cx26, Cx36, Cx43, Cx45, Cx47, pannexin-1 (PNX1); (2) neuronal marker genes PAX6, vGLUT1, GAD1, TUBB3; (3) glial lineage markers BLBP, GFAP; (4) schizophrenia-associated gene (ERBB4). Housekeeping gene (ACTB) was used to normalize the gene expression values.

We compared the average expression for each of the genes, in the two cell lines, hESC-H9 and iPSC-15, over the time period DIV-7 to DIV-21 (Table 2B). Our gene expression results failed to detect any principal differences between the hESC-H9 and iPSC-15 cells (p > 0.05; N = 28 cover slips containing hESC-H9, and N = 20 cover slips with iPSC-15 cells). Our results are consistent with the single-cell PCR analysis of stem cell-derived neurons, which also found similar gene expression patterns between HESC-H9 and iPSC-15 lines [8]. Based on similar physiological (Figure 2C) and gene expression results (Figure 2D, E) obtained in two human cell lines, the data from the two cell lines were combined in the remainder of the study.

With the exception of GFAP, all tested genes (total number of genes tested = 14) were detected at some point between DIV-7 and DIV-21. The expression of Cx26 was fairly constant (Figure 4, Cx26). While Cx43, Cx47, and TUBB3 were upregulated in the presence of bFGF (cell proliferation stage, DIV-7 to DIV-12), their levels fell when bFGF was withdrawn on DIV-12 and differentiation factors BDNF and GDNF were added to the culture (Figure 4). The expression of genes Cx45, BLBP, PAX6, and GAD1 did not seem to be influenced by the addition or removal of growth factors from the cell culture media (Figure 4). The expression of ERBB4, Cx36, and vGLUT1 was negligible in the presence of bFGF (Figure 4), but sharply increased after the introduction of differentiation factors BDNF and GDNF to the culture medium on DIV-12.

The temporal patterns of gene expression during the course of early differentiation of human neurons were divided into four general categories (a – d) and characterized as follows:

1. Unchanging, constant expression (Figure 5, line “a”). The example of unchanging constant expression is the temporal pattern of Cx26 (Figure 4).

2. Hyperbolic increase (Figure 5, line “b”). The examples of hyperbolic increase are the temporal patterns of Cx43, Cx47, and TUBB3 (Figure 4).

3. Monotonously rising, linear increase (Figure 5, line “c”). The examples of linearly increasing pattern of expression are PAX6, BLBP, Cx45, and GAD1 (Figure 4).

4. Exponential increase (Figure 5, “d” lines). The examples of exponentially increasing gene levels can be found in the expression patterns of PNX1, vGLUT1, ERBB4, and Cx36 (Figure 4). An interesting aspect of the exponential increase pattern is the day of onset when gene expression becomes detectable. Our measurements showed that this day of onset varied between the genes. For example, the expression onset of Cx36 (~DIV-17) was at least 3 days later than the onset of ERBB4 and vGLUT1 (~DIV-13) or 5 days later than the onset of PNX1 (~DIV-11). Three genes, ERBB4, vGLUT1, and Cx36, exhibited their onsets after DIV-12, followed by a sharp rise in the gene expression level, suggesting that these genes responded to the addition of growth factors (GDNF and BDNF) on DIV-12.

Figure 5.

Characteristic patterns of gene dynamics during early stages of stem cell neurodifferentiation. Differences in differentiation media before and after DIV-12 are indicated by rectangles in the background. The four temporal patterns of gene expression levels (a–d) are based on the polynomial fits obtained from the raw data points shown in (Figure 4 (a) A relatively constant expression level over the entire course of differentiation for Cx26. (b) Hyperbolic pattern––a slow rise followed by a slow decline for genes Cx43, Cx47, and TUBB3. (c) Monotonous linear increase in the expression of PAX6, BLBP, Cx45, and GAD1. (d₁, d₂, and d₃) An exponential increase with a delayed onset. The onset of the exponential rise begins on DIV-11 for PNX1; on DIV-13 for ERBB4 and vGLUT1; and on DIV-17 for Cx36.

Next, we examined gene expression levels in relation to one significant physiological event: the onset of AP (Action Potential) firing capacity. The first APs were detected in young human neurons on DIV-14 using patch clamp recordings (Figure 1E). This event divides the time course of our experiment into two periods: (a) before AP firing activity was present and (b) after AP firing was established in developing human neurons. The experimental data reveals three gene expression trends in relation to the AP firing onset: (i) Negatively correlated; (ii) Unperturbed/Unrelated; and (iii) Positively correlated. In the first group “Negatively correlated”, the gene expression levels were rising before AP firing was present. Upon the onset of AP firing, the gene expression trend reversed from “rising” to “falling”, Three genes (Cx43, Cx47, and TUBB3) showed an apparent decline after the AP firing activity was established on DIV-14 (Figure 4, Top row). In the second group “Unperturbed/Unrelated”, the gene expression trend did not react to the appearance of the first APs in young neurons on DIV-14. These genes are Cx45, BLBP, PAX6, GAD1, and PNX1 (Figure 4). In the third group “Positively correlated,” the gene expression was not detected before the appearance of the first APs. Following the appearance of APs on DIV-14, three genes, ERBB4, vGLUT1, and Cx36, exhibited an exponential increase in expression levels (Figure 4, Bottom row).

3.4. Activity to gene correlations

We next examined if there are any significant correlations between a cell’s physiology, manifested through spontaneous Ca²⁺ activity, and gene expression patterns during early neuronal differentiation. Recall that the multicell calcium imaging data (Figure 2AB) were used to quantify the population of spontaneously active cells (“Active Cells”) on a given DIV (Figure 3A). The same cells, in which calcium imaging was performed, were also subjected to gene analysis (Figure 4). This allowed us to plot physiological data and gene expression analysis data on the same graph.

3.4.1. Active cells

For example, Figure 6A shows similar trends of “Active Cells” (fraction of cells exhibiting spontaneous Ca²⁺ activity) and relative expression levels of the PNX1 gene. The time course of Cx26 expression, on the other hand, was unrelated to the fraction of cells exhibiting spontaneous Ca²⁺ activity (Figure 6B, “Active Cells”). We used Pearson’s product–moment correlation coefficient test to study the relationship between spontaneous Ca²⁺ activity and gene expression. Two groups of genes emerged from the numerical analysis of correlation. One group of genes possessed expression patterns highly correlated with the percentage of spontaneously active cells (Table 3, Row: “Active Cells”; bold numbers, p < 0.001). This group of genes included Cx36, BLBP, Cx45, ERBB4, GAD1, PAX6, PNX1, and vGLUT1. All aforementioned genes had a positive correlation coefficient (CC), indicating that gene expression levels increased with an increase in the number of spontaneously active cells (“Active Cells”). The other group of genes exhibited expression profiles unrelated to the “Active Cells” parameter (Table 3, Row: “Active Cells”; pale numbers; p > 0.001). This group of genes included Cx26, Cx43, TUBB3, and Cx47.

(A)		Cx26	Cx43	Cx36	BLBP	TUBB3	Cx45
1	Active cells (%)	0.5651	0.1844	0.8624	0.9727	−0.0570	0.9998
2	Repetitive activity (%)	0.5477	0.1638	0.8715	0.9677	−0.0779	0.9991
3	Narrow: signal duration < 8 s; all signal amplitudes included	0.7372	0.2599	0.7244	0.9987	0.1152	0.9862
4	Moderate duration: signal duration between 8; and 20 s; signal; amplitude >10%	−0.5069	−0.8902	0.7758	0.1590	−0.9474	0.3678
5	Moderate duration: signal; duration between 8 and; 20 s; signal; amplitude > 5%	0.5534	0.0203	0.8566	0.9574	−0.1268	0.9971
6	Wide: signal duration > 20 s; signal amplitude < 10%	0.8996	0.9935	−0.3271	0.4146	0.9660	0.2090
7	Wide: signal duration > 20 s; signal amplitude > 20%	0.9106	0.9902	−0.3034	0.4380	0.9590	0.2343
(B)		Cx47	ERBB4	GAD1	PAX6	PNX1	vGLUT1
1	Active cells (%)	0.1641	0.9776	0.99998	0.9994	0.9958	0.9753
2	Repetitive activity (%)	0.1434	0.9818	0.9996	0.9999	0.9975	0.9797
3	Narrow: signal duration; <8 s; all signal; amplitudes included	0.3490	0.8972	0.9634	0.9591	0.9448	0.9071
4	Moderate duration: signal; duration between 8; and 20 s; signal; amplitude >10%	−0.8437	0.6190	0.4631	0.4770	0.5175	0.6008
5	Moderate duration: signal; duration between 8 and; 20 s; signal amplit. >5%	0.1137	0.9770	0.9996	0.9990	0.9958	0.9817
6	Wide: signal duration >20 s; signal amplitude <10%	0.9998	−0.0802	0.1055	0.0899	0.0433	−0.0573
7	Wide: signal duration >20 s; signal amplitude >20%	0.9989	−0.0543	0.1313	0.1157	0.0692	−0.0314

Table 3.

Correlation between parameter of spontaneous calcium activity (e.g. Active Cells [%]) and normalized gene expression level (e.g. Cx36).

All numerical values on the right side of the table are correlation coefficients (CC), calculated using Pearson Product–Moment Correlation test. Bold CC values are statistically significant (p < 0.001). Pale CC values are not statistically significant (p > 0.001).

3.4.2. Repetitive activity

We evaluated the relation between repetitive activity and gene expression. “Repetitive Activity” is a parameter extracted from the multicell calcium imaging data. This parameter describes the number of cells in the visual field exhibiting four or more Ca²⁺ transients per 5 min of optical recording, divided by the number of “Active Cells” (cells showing at least one Ca²⁺ transient) in that field (%). We found that the same set of genes positively correlated to “Active Cells” was also positively correlated with “Repetitive Activity” (Table 3, Row: “Repetitive Activity”).

3.4.3. Ca²⁺ signal properties

As many as 4744 spontaneous Ca²⁺ transients were documented in 25 cover slips over the period of DIV-7 to DIV-21. Each spontaneous Ca²⁺ transient was measured in terms of signal peak amplitude and signal duration (half-width), which allowed us to group data points based on signal amplitude and signal duration (Figure 3).

3.5. Narrow signals (<8 s)

3.5.1. GAD1 and vGLUT1

Gu and Spitzer [30] observed that the expression of the neurotransmitter GABA in developing Xenopus spinal neurons is regulated by spontaneous Ca²⁺ spikes (<5 s duration) [30]. Based on the median duration of 8 s, we categorized data points into “Narrow” and “Wide” Ca²⁺ events (Figure 3F). The number of Narrow events (<8 s) from DIV-7 to DIV-21 was significantly correlated to the expression of GAD1 and vGLUT1 (Table 3, Row 3). Since GAD1 is involved in the synthesis of GABA [17,30], these data suggest that in developing human neurons, there is a close relationship between spontaneous Ca²⁺ activity and the differentiation of neurons into GABAergic and glutamatergic neurons.

3.5.2. Cx45

The number of narrow events (with duration <8 s) was strongly (p < 0.001) correlated to the expression of Cx45 (Figure 6D; Table 3, Row 3). Cx45 is known to influence the spontaneous firing patterns in the developing retina [2].

3.5.3. BLBP

The number of narrow events (duration < 8 s) from DIV-7 to DIV-21 was significantly (p < 0.001) correlated to the expression of BLBP (Table 3, Row 3). BLBP expression has been reported to increase during neuronal differentiation [36], being required to induce morphological changes in radial glia in response to neuronal cues [37,38].

3.6. Wide signals (>20 s)

3.6.1. TUBB3, Cx47, and Cx26

We found that Ca²⁺ transients with duration >20 s and peak amplitude ≤10% ΔF/F not only had a strong positive correlation to the expression of the neuronal marker TUBB3, but also to three connexin isoforms Cx43, Cx47, and Cx26 (p < 0.001, Table 3, Row 6). Connexin isoforms Cx43 and Cx47 were previously found in astrocytes [39,40] and oligodendrocytes [41], respectively. Cx26 has been reported to be located on neurons and astrocytes of the developing brain as well as the leptomeninges (Nagy et al., 2001).

The count of Ca²⁺ transients with duration >20 s and amplitude >20% ΔF/F showed a significant positive correlation (p < 0.001) to the expression levels of four genes: Cx26, Cx43, TUBB3, and Cx47 (Table 3, Rows 7). Interestingly, these were the only four genes that showed no correlation to the parameters “Active Cells” and “Repetitive Activity” (Table 3, Rows 1 and 2). These subsets of Ca²⁺ transients with long duration and large amplitude (duration > 20 s and amplitude > 20% ΔF/F, as well as duration > 20 s and amplitude ≤ 10% ΔF/F) had no significant correlation to the expression of eight genes: Cx36, BLBP, Cx45, ERBB4, GAD1, PAX6, PNX1, and vGLUT1. These are exactly the same eight genes that are highly correlated to the parameters “Active Cells” and “Repetitive Activity” (Table 3, Rows 1 and 2).

Overall, the correlation analyses revealed potential links between the number of cells with spontaneous calcium transients (“Active Cells” and “Repetitive Activity”) and expression patterns of two connexins (Cx36, Cx45), one pannexin (PNX1), and several general neuronal genes, including ERBB4, GAD1, PAX6, and vGLUT1 (Table 3, Rows 1 and 2). These results indicate one of two possibilities: (a) that specific subclasses of spontaneous Ca²⁺ transients could be driven by the expression of specific human genes, (b) specific amplitude and duration of Ca²⁺ transients may selectively promote expression of some human genes at early stages of neuronal differentiation. In the current study, however, we did not perform experiments to explore if expression of a certain human gene affects the incidence of spontaneous Ca²⁺ transients, with certain combinations of amplitudes and durations.

Figure 6.

Correlations between different parameters of spontaneous Ca²⁺ transients and gene expression levels. (A) In this and all the following panels, a specific parameter of the spontaneous Ca²⁺ activity is plotted using the left ordinate axis (green), while the relative gene expression levels are plotted using the right ordinate axis (red). “Active Cells”: the number of ROIs with Ca²⁺ transients divided by the total number of ROIs in that visual field (%). Each green bar represents an average value obtained on a given DIV. Each red bar represents an average relative expression of PNX1 on a given DIV. The trend lines are polynomial fits through spontaneous Ca²⁺ activity (green) or the gene expression levels (red). The two trend lines are strongly correlated; correlation coefficient, CC = 0.99. (B) Same physiological parameter (“Active Cells”) as in A is compared against the relative expression of Cx26. The two trend lines, physiological activity (green) and gene expression (red), are not correlated; CC = 0.56. (C) The percent of cells with repetitive Ca²⁺ activity on each day of in vitro differentiation (green) is in good correlation to the expression levels of GAD1 (red); CC = 0.97. (D) The fraction of narrow Ca²⁺ transients on each day of in vitro differentiation (green) is positively correlated with the expression levels of Cx45 (red); CC = 0.98.

3.7. Gene-to-gene correlations

The time courses of gene expression levels shown in Figure 4 suggest that the expression of all genes in the bottom row––PNX1, ERBB4, vGLUT1, and Cx36 correlate to each other. In addition to this, we found a strong correlation (r = 0.91, p < 0.001) between Cx43 and TUBB3 expression (Figure 4, Top row).

We found that there was a statistically significant correlation between Cx36 and GAD1 (GABA expression; r = 0.79, p < 0.01), as well as a statistically significant correlation between Cx36 and vGLUT1 (r = 0.63, p < 0.05).

These data indicate that systematic sampling of gene expression levels from DIV-7 to DIV-21 during human in vitro neurodifferentiation can be used to detect groups of genes that rise and fall in parallel to each other, suggesting that their expression is driven by the same sequence of cellular and molecular events [42].

4. Discussion

Abnormal fetal brain development is linked to a myriad of neurological and psychiatric diseases. Human fetal gestation is a particularly sensitive stage of brain development, during which aberrant genetic programs or external factors (viruses, pollutants, toxins, and other agents) exert the most devastating impact [43,44]. The etiopathogenesis of mental illnesses, such as autism and schizophrenia, is not exclusively determined by genetic variation, but rather by the interaction of genes with the environment [45,46]. Variations in the duration and amplitude of Ca²⁺ transients and their time of occurrence affect neuronal development and plasticity [3], but these physiological aspects of human neurodifferentiation are largely absent in clinical and laboratory research. Unsurmountable ethical and technical difficulties preclude experiments on human fetuses. One possibility for studying the physiology of young human neurons is to use hESC and iPSC technologies and mimic human neurodifferentiation in vitro [6,11,47,48].

In this experimental project, we have applied embryonic and pluriopotent stem cell technology to study physiological activity and gene expression during the earliest stages of human neurodifferentiation. Using multisite Ca²⁺ imaging and qPCR, we have compared time courses of spontaneous physiological Ca²⁺ activity and gene expression in two human cell lines (hESC-H9 and iPSC-15), as these cells transitioned from undifferentiated stem cells to postmitotic neurons. The iPSC-15 cell line was derived from a patient with schizo-affective disorder with a deletion on 22q11.2 that results in haploinsufficiency of ~ 40 genes, which is one of the most common causes of psychotic disorders [49–51]. We did not observe any significant differences in electrical properties, action potential frequencies, and the expression of genes important for neural development between the two cell lines, demonstrating the presence of intact cellular machinery for basic neuronal function [8].

Spontaneous Ca²⁺ transients were observed in both cell lines from the first day of recording (DIV-7), and increased in occurrence throughout the entire in vitro differentiation period. Ca²⁺ transients act via a variety of signal types, with amplitude, duration, and frequency, each encoding a specific piece of information [52]. The multisite optical imaging method used in the present study provided means for monitoring the critical parameters of Ca²⁺ transients: amplitude, duration, and frequency in up to 80 cells simultaneously. The results suggest that (a) lower amplitude, (b) greater ratio of narrow to wide signals, and (c) increased frequency are all characteristics of maturing nerve cells (Figure 3E, F).

4.1. Correlating spontaneous electrical activity with gene expression

Interestingly, as human cells differentiated from DIV-7 to DIV-21, there was a strong positive correlation between the number of cells exhibiting repetitive spontaneous Ca²⁺ transients and the expression levels of GAD1 and ERBB4 (p < 0.001; Table 3, Row 2). Also, there was a strong positive correlation (p < 0.001) between the number of narrow events and the expression levels of either GAD1 or ERBB4 (Table 3, Row 3). These parallel trends of ERBB4 and GAD1 are not surprising as GABAergic interneurons have been suggested to employ ERBB4 receptor signaling early in the development of the cerebral cortex [53]. In accordance with our present observations, a previous study employing patch clamp recording followed by single-cell PCR revealed that individual neurons expressing GAD1 and ERBB4 are more likely to engage in spontaneous electrical activity than their counterparts which do not express these two genes [8]. As human cells differentiated into neurons and matured, the expression of GAD1 and ERBB4 increased at the same rate as the number of cells showing spontaneous Ca²⁺ activity (Table 3, Row 1) or the number of cells showing repetitive Ca²⁺ transients (Figure 6C). The regulation of the GABAergic neuronal phenotype (GAD1 expression and production of GABA) appears to occur via an activity-dependent mechanism and is Ca²⁺-mediated, as shown by removing extracellular Ca²⁺ or stimulating cultured neurons with different frequencies of Ca²⁺ spikes [30,54].

Narrow calcium events (Ca²⁺ transients with duration < 8 s) fit the description of AP-induced Ca²⁺ signals in postmitotic human neurons obtained by simultaneous patch clamp and optical recordings [8], and as such these short-duration Ca²⁺ signals mark the arrival of newborn neurons in stages IV and V of the current neurodifferentiation protocol (Figure 3F). We, therefore, explored the possibility that the expression of all the other genes that correlated with repetitive activity also had a significant correlation to the number of narrow events. We found that except for Cx36, which is significantly correlated (p < 0.001) to percent repetitive activity, but not to the number of narrow events, the rest of the genes (BLBP, Cx45, ERBB4, GAD1, PAX6, PNX1, and vGLUT1) had a significant positive correlation to both repetitive activity and number of narrow events (Table 3, Rows 2 and 3). Cx36 has been reported to be expressed during the development of retinal ganglion cells. It is thought that Cx36 is capable of influencing spontaneous firing patterns, but is not required for spontaneous retinal wave generation [2].

The addition of growth factors BDNF and GDNF to the cell culture media appeared to have an effect on the expression of ERBB4, vGLUT1, and Cx36. These three genes were present at minimal levels, until the addition of growth factors on DIV-12 caused a sharp increase in their level of expression (Figure 4). Interestingly, the three genes showing exponential rise after DIV-14 are also highly correlated with Ca²⁺ signals of Moderate duration of 8–20 s and amplitude greater than 5 % ΔF/F (Table 3, Row 5). Cx36 is thought to participate in neuronal gap junction formation in vivo. An increased expression of Cx36 has been demonstrated to promote neuronal differentiation by enhancing cell–cell contact between progenitors [55]. In situ hybridization techniques have shown high levels of Cx36 mRNA in olfactory bulbs, pineal gland, inferior olive, hippocampal pyramidal neurons, and in the retina. We found a significant correlation between Cx36 and GAD1 expression patterns (r = 0.79, p < 0.01), suggesting some functional link between these two genes. Cx36 has been reported to be present on GABAergic interneurons in the neocortex and hippocampus [56]. In Cx36-deficient mice, gap junction-coupling among neocortical inhibitory interneurons has been shown to be nearly absent. GAD1 (GAD67) is an enzyme that synthesizes GABA and is itself influenced by Ca²⁺ signaling [17,30]. Thus, Cx36 and GAD1 have a close relationship, which may be a part of the regulatory mechanisms for expression of GABA and/or GABAergic interneuron differentiation.

In addition to having a significant correlation to GAD1, we found that there was a significant correlation between Cx36 and the vGLUT1 expression patterns (r = 0.63, p < 0.05) over the period DIV-8 to DIV-21. The Cx36 gene has been reported to be induced at the time of the surge of the transcription factors that determine β-cell differentiation [57]. Thus, during neuronal differentiation, Cx36 could potentially influence transcription factors that determine the terminal differentiation of neurons to either glutamatergic or GABAergic, with the Cx36 channels forming intercellular pathways for transmission of developmentally relevant molecules. Upregulation of Cx36 expression has also been previously associated with the sequential expression of specific ligand-gated responses (GABA and glutamate), thus heralding terminal differentiation of neurons, with the peak expression of Cx36 being associated with a developmental window in which the neuronal network connections are being developed [58].

In the present study, we found that Cx43, Cx47, and TUBB3 showed a hyperbolic pattern of expression, wherein there was an upregulation of these genes during the cell proliferation stage (in the presence of the proliferation factor bFGF) and a subsequent decrease in their expression. Both the increase and the decrease of Cx43, Cx47, and TUBB3, relative to the characteristic time periods of the differentiation protocol (e.g. replacement of proliferation media on DIV-12; appearance of first APs on DIV-14; sharp onset of Cx36 expression on DIV-17), may become useful for understanding the roles of these three genes in human neurodifferentiation.

We found that the decline of TUBB3 after DIV-14 was paralleled by an upregulation of GAD1 (putative GABAergic neurons) and vGLUT1 (putative glutamatergic neurons) during the differentiation phase (Figure 4, compare TUBB3, GAD1, and vGLUT1). Similar to our results, a decrease in the expression of the neuronal progenitor marker TUBB3 along with a simultaneous increase in GABA and glutamatergic neurons during neuronal differentiation was also observed by [42]. We found a strong correlation between Cx43 and TUBB3 expression (r = 0.91, p < 0.001), which suggests a close relationship between Cx43 expression on astrocytes and TUBB3 expression on neuronal progenitors. Cx43 is essential for the maintenance of neural progenitors in a proliferative state [59, 60] and forms communication channels, between migrating neurons and the glial cells, during neurogenesis [61].

By sampling data points on consecutive days (DIV-7 to DIV-21), we observed that the expression of human GAD1 occurred before the expression of human vGLUT1 (Figure 4), as has been reported in the literature, where the formation of GABAergic synapses in the CNS preceded that of glutamatergic synapses [62]. In summary, as cells exit from the cell cycle and undergo lineage commitment, as shown by the expression of ERBB4, GAD1, and vGLUT1, they demonstrate a changing profile of connexin expression with Cx43 and Cx47 declining, Cx45 and PNX1 increasing, and Cx36 being activated.

Although BLBP was expressed in our cell culture, GFAP was not expressed at detectable levels. Both GFAP and BLBP are expressed on radial glia during early neuronal development, but their expression has been reported to be nonoverlapping [63]. In line with our experimental results, GFAP was observed to be expressed only after the expression of BLBP during in vitro culture of human embryonic stem cells [42]. Thus, culturing cells for a period longer than 21 days is necessary for observing GFAP expression.

When correlation analysis was performed between Ca²⁺ transients with specific characteristics of duration and amplitude, and the expression of specific genes involved in neurodifferentiation, we observed that specific subclasses of Ca²⁺ transients are linked to specific genes. Two possible scenarios can explain the observed positive correlations between spontaneous physiological activity and gene expression. First, specific calcium transients could be driven by the expression of specific human genes. Second, one type of calcium transients (e.g., short duration and small amplitude) may selectively influence the expression of some but not all human genes during early neuronal development.

Spike stimulation mimicking the frequency of spontaneous transients has been shown to be most effective in replicating the effects of spontaneous transients on neurotransmitter expression, for example, GABA expression [30,54]. Thus, it would be quite useful to find calcium spike properties (amplitude, duration, and frequencies) that promote the expression of genes involved in early neuronal differentiation in order to improve current methods for modeling human brain development in culture.

Acknowledgments

This study was supported by Connecticut Innovations grant 09-SCA-UCHC-13 to SDA, NIH grants MH063503 and MH104775 to SDA, HCRAC Institutional grant to SDA, and NIH grants MH087840 and MH099427 to HML. Electrophysiological and optical measurements were performed in the Stem Cell Physiology and Chemistry Core, which is supported by Connecticut Innovations grant SCD-01-2009 to SDA. The LED-based epi-illumination system was designed and built by Dr. Peter Lee, Essel R&D Inc., Toronto, Canada. The authors are grateful to Drs. Steven Crocker and Richard Mains (UConn Health) for help with the qPCR apparatus.

Author contributions

Designed the study: GSB and SDA. Produced iPSCs: EP and HML. Stem cell neurodifferentiation: GSB, PVL, and MLM. Patch electrode recordings and multisite calcium imaging: MBS, KDO, and SDA. Gene expression levels: PVL, GSB, and MLM. Analyzed the experimental data: PVL, MLM, MBS, and SDA. Wrote the paper: PVL, HML, and SDA.

Competing interests

The authors have declared that no competing interests exist.