ASCL1- and DLX2-induced GABAergic neurons from hiPSC-derived NPCs

Abstract

Background

Somatic cell reprogramming is routinely used to generate donor-specific human induced pluripotent stem cells (hiPSCs) to facilitate studies of disease in a human context. The directed differentiation of hiPSCs can generate large quantities of patient-derived cells; however, such methodologies frequently yield heterogeneous populations of neurons and glia that require extended timelines to achieve electrophysiological maturity. More recently, transcription factor-based induction protocols have been show to rapidly generate defined neuronal populations from hiPSCs.

New method

In a manner similar to our previous adaption of NGN2-glutamatergic neuronal induction from hiPSC-derived neural progenitor cells (NPCs), we now adapt an established protocol of lentiviral overexpression of ASCL1 and DLX2 to hiPSC-NPCs.

Results

We demonstrate induction of a robust and highly pure population of functional GABAergic neurons (iGANs). Importantly, we successfully applied this technique to hiPSC-NPCs derived from ten donors across two independent laboratories, demonstrating its efficient and highly reproducible approach to generate induced GABAergic neurons.

Comparison with existing methods

Our results show that, like hiPSC-iGANs, NPC-iGANs exhibit increased GABAergic marker expression, electrophysiological maturity, and have distinct transcriptional profiles that distinguish them from other cell-types of the brain. Until donor-matched hiPSCs-iGANs and NPC-iGANs are compared directly, we cannot rule out the possibility that subtle differences in patterning or maturity may exist between these populations; one should always control for cell source in all iGAN experiments.

Conclusions

This methodology, relying upon an easily cultured starting population of hiPSC-NPCs, makes possible the generation of large-scale defined co-cultures of induced glutamatergic and GABAergic neurons for hiPSC-based disease models and precision drug screening.

1. Introduction

gamma-Aminobutyric acid (GABA)ergic interneurons serve a vital role within the brain, helping to maintain a balance between excitation and inhibition. Perturbations in the function of GABAergic interneurons have been implicated in many neuropsychiatric diseases, including schizophrenia (Akbarian et al., 1995; Coyle, 2004), autism spectrum disorder (Robertson et al., 2016; Rubenstein and Merzenich, 2003), and epilepsy (Jacob, 2016); however, it remains unclear whether deficits in GABAergic function are a cause or consequence of disease. Human induced pluripotent stem cells (hiPSCs) provide the opportunity to explore the contribution of GABAergic neurons to disease risk by making it possible to generate, manipulate, and functionally evaluate patient-specific GABAergic neurons in vitro.

Traditional protocols for generating hiPSC-derived GABAergic neurons utilized “directed differentiation”, a process whereby small molecules and growth factors were sequentially administered in an attempt to mimic the in vivo signaling environment. For example, activation of sonic hedgehog (SHH) signaling concomitant with inhibition of WNT signaling patterns hiPSCs towards progenitors of a medial ganglion eminence (MGE) identity (Kelsom and Lu, 2013; Liu et al., 2013; Maroof et al., 2013; Nicholas et al., 2013), from which cortical interneurons originate in the developing brain. This combinatorial signaling cascade activates transcription factors such as NKX2.1, which are required for endogenous GABAergic development (Kim et al., 2014). However, these protocols typically rely on costly small molecules and growth factors, require lengthy time periods for neuronal differentiation, and yield a fairly heterogenous population of neurons.

In contrast, neuronal induction relies on the identification and overexpression of lineage-specific transcription factors. Overexpression of FOXG1, SOX2, ASCL1, DLX5 and LHX6 yields a predominately GABAergic population from either fibroblasts or hiPSCs (Colasante et al., 2015), the efficacy of which could be improved by inclusion of microRNA-9/9* and microRNA-124 (Sun et al., 2016). A recent protocol further streamlined this approach by generating a highly homogenous population of functionally mature GABAergic neurons via the transient overexpression of just two factors: ASCL1 and DLX2 (Yang et al., 2017). This simple and efficient method yields induced GABAergic neurons (iGANs) that achieve a high level of synaptic maturity and successfully integrate following transplantation into the rodent brain.

Achaete-scute family bHLH transcription factor 1 (ASCL1) is a pro-neural gene that is abundantly expressed in progenitors of the ventral telencephalon, which give rise to GABAergic interneurons (Castro et al., 2011; Ross et al., 2003). ASCL1 guides progenitors to exit the cell cycle and enter neural differentiation. Loss of ASCL1 in mouse models results in a deficit of cortical GABAergic interneurons (Castro et al., 2011).

Members of the DLX family encode for homeodomain-containing transcription factors. The DLX genes expressed within the telencephalon (DLX1/2/5/6) facilitate in the differentiation and migration of GABAergic interneurons (Poitras et al., 2007). Interestingly, ASCL1 is an upstream regulator of DLX1 and DLX2 (Castro et al., 2011; Poitras et al., 2007). Particularly, DLX2 directly modulates expression of essential GABAergic markers, such as GAD1/2 and VGAT (Pla et al., 2018).

In theory, using NPCs as the starting point for differentiation could minimize variability as hiPSC-NPCs are pre-patterned to a forebrain neural identity (Brennand et al., 2015). Furthermore, NPCs offer a highly scalable option as they proliferate quickly and are relatively easy to handle. As such, our lab has previously generated functional NGN2-glutamatergic neurons from hiPSC-derived NPCs (Ho et al., 2016). In this study, we demonstrate that overexpression of ASCL1 and DLX2 in NPCs is sufficient to generate GABAergic neurons. Our results indicate that transient overexpression of ASCL1 and DLX2 in NPCs induces neuronal morphology, GABAergic gene expression, and synaptic maturity.

2. Methods and Materials

2.1. Tissue Culture

2.1.1. Thawing NPCs

Before thawing cells, 6-well tissue plates were coated with Matrigel (BD biosciences, #354230). One mg of Matrigel was re-suspended in 24 mL of cold DMEM (ThermoFisher Scientific, #10566–016). Each well received 2 mL of the Matrigel-DMEM solution before incubating overnight at 37°.

Neural progenitor cells were generated as previously reported (Hoffman et al., 2017). Stocks of hiPSCs and NPCs are stored in liquid nitrogen at −180°C. Vials of cells were acquired from liquid nitrogen and thawed at room temperature. Thawed cells were transferred to a 15 mL Falcon tube containing 5 mL of room temperature DMEM and centrifuged for 5 minutes at 1,000G. During centrifugation, the Matrigel-coated 6-well tissue plates were removed from the incubator. The 2 mL Matrigel-DMEM solution was aspirated and replaced with 1 mL of room temperature NPC media (Table 1) per well.

Table 1:

Media Recipes

Media	Reagents
NPC	DMEM/F12 + Glutamax (ThermoFisher, #10565018) 1 × Antibiotic-antimycotic (ThermoFisher, #15240062) 1 × N2 (ThermoFisher, #17502–048) 1 × B27-RA (ThermoFisher, #12587–010) 20 ng/mL FGF2 (R&D, #233-FB-10)
Neuron	Neurobasal (ThermoFisher, #21103049) 1 × Antibiotic-antimycotic (ThermoFisher, #15240062) 1 × N2 (ThermoFisher, #17502–048) 1 × B27-RA (ThermoFisher, #12587–010) 1 × GlutaMAX (ThermoFisher, #35050061) 1 mg/mL Natural Mouse Laminin (ThermoFisher, #23017–015) 20 ng/mL BDNF (Peprotech, #450–02) 20 ng/mL GDNF (Peprotech, #450–10) 500 μg/mL cAMP (Sigma, #D0627) 200 nM L-Ascorbic acid (Sigma, #A0278)
HA	Astrocyte media (Sciencell, #1801) AGS (Sciencell, #1852) P/S (Sciencell, #0503) FBS (Sciencell, #0010)

Once the NPCs were done spinning, the supernatant was aspirated, leaving behind only the pelleted cells. The NPCs were re-suspended in 1 mL of room temperature NPC media and transferred to a well containing 1 mL of freshly added NPC media. Tissue culture plates were gently shaken to allow for the even distribution of NPCs. Plates were returned to the incubator. The day following a thaw, NPC media was aspirated and replaced with 2 mL of fresh room temperature NPC media. Media changes occur every other day.

2.1.2. Passaging NPCs

NPCs were split once they achieved confluency, roughly 10 to 15 million cells per well of a 6-well tissue plate. Prior to splitting, six-well tissue plates were coated with Matrigel as previously described. NPC media, DMEM, and Accutase (Sigma #A6964) were warmed to room temperature prior to use. NPC media was aspirated from each well and replaced with 1 mL of Accutase. Plates were incubated at 37°C for approximately 5 to 10 minutes. The detached NPCs were transferred to a 15 mL Falcon tube containing 5 mL of room temperature DMEM and centrifuged for 5 minutes at 1000G. The supernatant was aspirated, leaving behind only the pelleted NPCs, which were then re-suspended in 1 mL of NPC media. NPCs were counted using the Countess (ThermoFisher, AMQAX1000). Approximately 2.5 to 3 million NPCs were added to each well of the 6-well tissue plate. Tissue culture plates were gently shaken to allow for the even distribution of NPCs. Plates were returned to the incubator. The day following a split, NPC media was aspirated and replaced with 2 mL of fresh room temperature NPC media. Media changes occur every other day.

2.2. GABAergic Induction from NPCs

2.2.1. Preparing plates for seeding

For patch-clamp, coverslips were added to 24-well tissue plates 4 days prior to seeding NPCs. Plates were placed in incubator at 37°C overnight to ensure evaporation of any residual ethanol. Tissue plates were coated with Matrigel as described in 2.1.1. The following day, human astrocytes (HA) (Sciencell, #1800) were counted (as described in 2.1.2) and seeded onto the coverslips at a density of 7.5–10 × 10⁴ cells per well and maintained in HA media (Table 1). On DIV 5, neurons were dissociated and seeded (as described in 2.1.2) onto HA at a density of 1.5–3.0 × 10⁵ cells per well.

For immunostaining, coverslips were added to 24-well tissue plates 2 days prior to seeding NPCs. Plates were placed in incubator at 37°C overnight to ensure evaporation of any residual ethanol. Tissue plates were coated with Matrigel as described in 2.1.1. 5.0 × 10⁵ rat astrocytes (Thermofisher, #N7745100) were added onto the neurons at DIV 7.

For RNA harvest, 24-well tissue plates were coated with Matrigel, as described in 2.1.1, one day prior to seeding NPCs.

2.2.2. Lentiviral production

Third-generation HIV-1 lentiviruses were generated by polyethylenimine (PEI, Polysciences #23966–2) –mediated transfection of HEK293T cells and packaged by VSVG as previously described (Tiscornia et al., 2006).

2.2.3. Seeding cells for induction (DIV-2)

NPCs were split and counted as described in 2.1.2. For patch-clamp, immunostaining and RNA harvest, NPCs were seeded at a density of 5.0 × 10⁵ cells per well of a 24-well tissue plate.

2.2.4. Viral transduction (DIV-1)

NPC media was aspirated and replaced with 500 μL of fresh room temperature NPC media per well. Each well received the following lentiviruses: CMV-rtTA (Addgene ID: 19780), TetO-Ascl1-T2A-Puro (Addgene ID: 97329) and TetO-Dlx2-IRES-Hygro (Addgene ID: 97330). NPCs with the lentiviruses incubated for 15 minutes at 37°C. Cells were spinfected at 1,000G for one hour at room temperature to maximize transfection efficiency (O’Doherty et al., 2000). Following spinfection, plates incubated for another 4 to 5 hours at 37°C. Lentiviral-containing NPC media was aspirated and replaced with 500 μL of fresh room temperature NPC media per well.

2.2.5. Doxycycline induction (DIV 0)

NPC media was aspirated and replaced with 500 μL of fresh room temperature NPC media containing 1 μg/mL of doxycycline (Sigma, #D9891). Doxycycline was withdrawn from the media at DIV 14.

2.2.6. Antibiotic selection (DIV 1)

NPC media with 1 μg/mL of doxycycline was aspirated and replaced with 500 μL of fresh room temperature NPC media containing 1 μg/mL of doxycycline, 1 μg/mL of puromycin (Sigma, #P7255) and 250 μg/mL of hygromycin (ThermoFisher, 10687010). On DIV 4, NPC media with doxycycline and antibiotics was aspirated and replaced with 500 μL of fresh room temperature NPC media containing 1 μg/mL of doxycycline, 1 μg/mL of puromycin and 250 μg/mL of hygromycin. This is the final antibiotic selection.

2.2.7. Introduction of neuronal media (DIV 7)

Cells were introduced to neuronal media (Table 1) via half-media changes; 230 μL of NPC media was removed and replaced with 250 μL of neuronal media. Cells received 1μg/mL of doxycycline until DIV 14. To prevent the proliferation of dividing mitotic progenitors, 50 nM of cytosineb-D-arabinofuranoside, also known as Ara-C, (Sigma, #C6645) may be supplemented in the neuronal media. Half-media changes with neuronal media occurred every other day until day of harvest or assay.

2.3. Analysis of ASCL1 and DLX2 induced neurons

2.3.1. Quantitative real time PCR

Cells were harvested on DIV 35 using TRIzol (Invitrogen, #15596026) and quantified using NanoDrop (ThermoFisher, #ND-2000). Quantitative real-time PCR was performed using Power Sybr Green RNA-to-CT 1-Step Kit (ThermoFisher, #4389986). RNA was diluted to 25 ng/μL and 100 nM of the forward and reverse primers (Table 2) were used for each condition.

Table 2:

Primers

Primers for RT-PCR
B-ACTIN	Forward	TGTCCCCCAACTTGAGATGT
	Reverse	TGTGCACTTTTATTCAACTGGTC
DLX5	Forward	ACAGAGACTTCACGACTCCCAG
	Reverse	TGTGGGGCTGCTCTGGTCTA
GAD65	Forward	CTATGACACTGGAGACAAGGC
	Reverse	CAAACATTTATCAACATGCGCTTC
GAD67	Forward	TTGCACCAGTGTTTGTCCTCATGG
	Reverse	CCGGGAAGTACTTGTAGCGAGCAG
MAP2	Forward	AAACTGCTCTTCCGCTCAGACACC
	Reverse	GTTCACTTGGGCAGGTCTCCACAA
SOX6	Forward	AGTTCCTCTGGCTGACTCTATCTGT
	Reverse	AAACCTCACTGCTTCCCACCC
VGAT	Forward	CACGACAAGCCCAAAATCAC
	Reverse	CGGCGAAGATGATGAGAAACAAC

2.3.2. Immunostaining

For immunostaining, 5.0 × 10⁵ rat astrocytes were added onto the neurons at DIV7. At DIV 35, media was aspirated from wells containing coverslips and cells were fixed by adding 250 μL of 4% PFA (Electron Microscopy Sciences, #15710) per well and incubating at room temperature for 10 to 15 minutes. Following the removal of 4% PFA, cells were washed three times with 500 μL of PBS (Invitrogen, #14190). Cells were blocked in 250 μL of 5% donkey serum (Jackson, #017–000-121) and 0.1% Triton-X (Sigma, #T8787) in PBS for one hour at room temperature. In the meantime, the primary solution was prepared. Primary antibodies (Table 3) were spun down at 4°C, 10,000G for 5 minutes to minimize any aggregates. Proper primary dilutions were added to 5% donkey serum and 0.1% Tween-20 (Boston BioProducts, #IBB-181X) in PBS. The blocking solution was aspirated and replaced with 250 μL of primary antibody solution and incubated overnight at 4°C.

Table 3.

Antibodies

Antibody	Species	Dilution	Company (Catalog #)
Anti-Nuclei	Clone 235–1	1:500	Sigma (#MAB1281)
GABA	Rabbit	1:500–1000	Sigma (A2052)
GAD67	Mouse	1:1000	Millipore (MAB5406)
GAD1	Mouse	1:500	Synaptic Systems (198211)
GAD2	Mouse	1:500	Synaptic Systems (198111)
MAP2	Mouse	1:500	Synaptic Systems (188011)
MAP2	Rabbit	1:700	Synaptic Systems (188003)
MAP2	Chicken	1:1000	ABcam (ab5392)
Alexa 488 anti-mouse	Donkey	1:1000	Thermofisher (A-21202)
Alexa 488 anti-mouse	Donkey	1:500	Jackson Immuno. (715–545-151)
Alexa 568 anti-mouse	Donkey	1:500	Invitrogen (A10037)
Alexa 594 anti-mouse	Donkey	1:1000	Thermofisher (A-21203)
Alexa 488 anti-rabbit	Donkey	1:1000	Thermofisher (A-21206)
Alexa 568 anti-rabbit	Donkey	1:500	Invitrogen (A10042)
Alexa 594 anti-rabbit	Donkey	1:1000	Thermofisher (A-21207)
Alexa 647 anti-chicken	Donkey	1:500	Jackson Immuno. (703–605-155)

The next day cells were washed three times with 500 μL of PBS. Secondary antibodies (Table 3) were spun down at 4°C, 10,000G for 5 minutes to minimize any fluorophore aggregation. Proper secondary antibody dilutions were added to PBS; 250 μL of the secondary antibody solution was added per well and incubated in a dark place for two hours. In the last 10 minutes of incubation with secondary antibodies, 0.5 μg/mL DAPI (Sigma, #D9542) was added to each well to stain nuclei. Cells were washed three times with 500 μL of PBS. Coverslips were carefully transferred to glass slides (Fisher, #12–544-7) containing AquaPolymount (Polysciences Inc., #18606–20). Freshly mounted slides were stored in 4°C overnight prior to imaging.

2.3.3. Electrophysiology

For whole-cell patch-clamp recordings, 7.5–10 × 10³ human astrocytes were first seeded onto Matrigel-coated 12-mm glass coverslips in 24-well plates, and then seeded with 1.5–3.0 × 10⁵ neurons after ~5 days. Neurons were recorded at 7–8 weeks following doxinduction, with media exchange every 3–4 days. Cells were visualized on a Nikon inverted microscope equipped with fluorescence and Hoffman optics. Neurons were recorded with an Axopatch 200B amplifier (Molecular Devices), digitized using a Digidata 1320a (Molecular Devices) and filtered between 1–10 kHz, using Clampex 10 software (Molecular Devices). Series resistance compensation was applied (70–100%). Patch pipettes were pulled from borosilicate glass electrodes (Warner Instruments) to a final tip resistance of 3–5M Ω using a vertical gravity puller (Narishige). Neurons were bathed in artificial cerebral spinal fluid (ACSF) containing (in mM): NaCl, 119, CaCl₂ 2; KCl, 2.5; MgCl₂, 1.3; d-glucose, 11; NaHCO₃, 26.2; NaPO₄, 1, at a pH of 7.4. The internal patch solution contained (in mM): K-d-gluconate, 140; NaCl, 4; MgCl₂, 2; EGTA, 1.1; HEPES, 5; Na₂ATP, 2; sodium creatine phosphate, 5; Na₃GTP, 0.6, at a pH of 7.4. Osmolarity was 290–295 mOsm. Voltages were corrected for junction potential of 15.5 mV. All chemicals were purchased from Sigma-Aldrich Co. (St. Louis, MO). Neurons were chosen at random using DIC. Current-clamp recordings were used for measuring evoked (current injected to hyperpolarize to −65 mV) and spontaneous activity. In voltage-clamp recordings, voltage steps were applied from −80 mV to +50 mV (10 mV increments) to elicit voltage-gated ionic currents. All recordings were made at room temperature (~22 C). Values are reported as mean ± SEM.

2.3.4. RNA-Sequencing preparation and analysis

RNA-seq of Sinai site generated libraries was performed using the Kapa Total RNA library prep kit with ribo-depletion and strand specific cDNA library construction (Kappa Biosystems). Paired-end sequencing (125 bp reads) was performed using the Illumina HiSeq2500 platform (New York Genome Center). RNA-seq of Northshore site was performed by Novogene (Sacramento, CA) at a depth of 20 M reads (2 × 150 bp reads) per sample.

All RNA-seq data analysis was performed in R Studio using the limma (Ritchie et al., 2015) and edgeR (McCarthy et al., 2012; Robinson et al., 2010) packages. Graphs were generated with ggplot2. Variance normalization across sample libraries was performed using the normalization method by trimmed mean of M-values (TMM) (Smid et al., 2018) via the calcNormFactors function in edgeR.

For Fig. 4A the select marker genes were chosen based upon KEGG (Kanehisa et al., 2019; Ogata et al., 1999) Synaptic Pathways and NeuroExpresso (Mancarci et al., 2017) “marker genes” for cell types of interest. RNA-seq samples were grouped according to “Cell Type” (i.e., iPSC, NPC, FB, GLUT, GABA), and the average Log(CPM) of each gene within each cell type group is represented by color-scale.

Figure 4: RNA-Seq analysis of NPC-iGANs exhibits their distinct transcriptional profiles.

(A) RNA samples were grouped by “Cell Type” (i.e., iPSC, NPC, FB, GLUT, GABA) along Y-axis. The average Log(CPM) of each gene within each cell type group is represented by color-scale. Marker genes are shown on the X-axis and grouped into marker gene categories by cell types (top). (B-D) The averaged Log(CPM) of each sample across neuronal and synaptic vesicle genes (B), regional patterning genes (C), and GABAergic function genes (D) are shown as individual data points.

For Fig. 4B,C,D boxplots with individual datapoints from the 14 NPC-iGANs sequenced are shown (SI Table 1). In 4C, genes within each category designation of “Embryonic GABAergic,” “Medium Spiny Neuron,” and “Hypothalamic GABAergic Neuron” were selected on the basis of having top specificity values within the corresponding KI level 1 cell type as calculated from single-cell RNA-sequencing in Skene et al., 2018. Genes within the “GABAergic Neurotransmission,” “Ganglionic Eminence,” “Classic Interneuron Markers,” and “GABAergic Receptors” categories were selected on the basis of extant literature (SI Table 2).

2.4. Co-culturing induced glutamatergic and GABAergic neurons

2.4.1. Generating monocultures

For NGN2-neurons, 2.0 × 10⁶ NPCs were seeded onto each well of a Matrigel-coated 6-well tissue plate and spinfected with TetO-NGN2-BFP2-PuroR and CMVrtTA (Addgene ID: 19780) as described in 2.1.1 and 2.2.4, respectively. NPCs were induced as previously described (Ho et al., 2016).

For NPC-iGANs, 2.0 × 10⁶ NPCs were seeded onto each well of a Matrigel-coated 6-well tissue plate and spinfected with lentiviruses listed in 2.2.4 in addition to TetO-FUW-eGFP (Addgene ID: 30130).

2.4.2. Generating co-cultures

On DIV 7 neurons were dissociated and counted (as described in 2.1.2). 1.0 ×10⁶ NGN2-neurons and 1.0 ×10⁶ iGANs were seeded onto same well of a new Matrigel-coated 6-well tissue plate.

3. Results

3.1. Lentiviral overexpression of ASCL1 and DLX2 in NPCs induces iGAN fate

The transient lentiviral overexpression of ASCL1 and DLX2 in NPCs (described in Fig. 1A and B) robustly induced a population of GABAergic neurons. Neuronal morphology was acquired rapidly (Fig. 1D). At 5 weeks, MAP2 levels in NPC-iGANs were examined. Quantitative real-time PCR analysis showed mRNA levels of MAP2 in NPC-iGANs to be significantly increased when compared to both hiPSCs (p-value<0.0001) and NPCs (p-value<0.0092) (Fig. 1E).

Figure 1: Lentiviral overexpression of ASCL1 and DLX2 in NPCs induces GABAergic neuron (iGAN) fate.

(A) Schematic of iGANs starting from NPCs using ASCL1 and DLX2. (B) Timeline of NPC-iGANs induction strategy. (C) Design of constructs. (D) Represenative brightfield images of NPC-derived iGANs. (E) Five-week old NPC-derived iGANs robustly express more MAP2 than hiPSCs (p-value<0.0001) and NPCs (p-value<0.0092) by real-time qPCR. Expression level is normalized to B-ACTIN, and standardized to marker expression in hiPSCs. Scale is log of fold change.

3.2. Five-week old NPC-derived iGANs highly express GABAergic markers

To examine GABAergic marker expression in our 5-week old NPC-iGANs, both real-time qPCR and immunocytochemistry were employed. Since this method of induced GABAergic neurons is already established using hiPSCs, we differentiated both hiPSCs and NPCs into iGANs to compare the efficacy of ASCL1 and DLX2 overexpression in driving differentiation regardless of cell source of origin. We selected the following genes to test: DLX5 (transcription factor that facilitates GABAergic migration and differentiation), GAD65 and GAD67 (enzymes that convert glutamate to GABA), SOX6 (transcription factor found in cortical interneurons), and VGAT (packages and transports GABA to the synapse). The mRNA expression of these markers in hiPSC- and NPC-iGANs was compared to the relative mRNA expression in hiPSCs and NPCs (Fig. 2A and B). Both hiPSC- and NPC-iGANs exhibited a significant increase in GABAergic markers. These results indicate that transient overexpression of ASCL1 and DLX2 in NPCs robustly increases neuronal and GABAergic markers. Immunofluorescence imaging of five-week old NPC-derived iGANs showed positive staining for MAP2, GABA, VGAT, and GAD1/2 proteins (Fig. 2C). Overall, NPC-derived iGANs exhibit GABAergic markers at both the mRNA and protein level.

Figure 2: NPC-iGANs robustly express GABAergic markers.

(A) Five-week old NPC-derived iGANs robustly express more GABAergic markers than hiPSCs by real-time qPCR. Expression level is normalized to B-ACTIN, and standardized to marker expression in hiPSCs. Scale is log of fold change. (* p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001) (B) Five-week old NPC-derived iGANs robustly express more GABAergic markers than NPCs by real-time qPCR. Expression level is normalized to B-ACTIN, and standardized to marker expression in NPCs. Scale is log of fold change. (* p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001) (C) Representative images of GABAergic marker expression in five-week old NPC-derived iGANs. Scale bar is 50 μm.

3.3. NPC-iGANs are excitable and spontaneously active

At seven weeks post-induction, we investigated the electrophysiological properties of NPC-iGANs using whole-cell patch-clamp recordings. NPC-iGANs expressed both voltage-gated potassium and sodium currents (Fig. 3A). Consistent with the presence of these ion channels, current-injection steps elicited action potentials with an induced firing frequency of 12.1± 1.0 Hz (Fig. 3B,C). Evoked action potentials exhibited a mean action potential height (AP_h) of 41.7 mV, and mean threshold (AP_t) of −50.9 mV (Fig. 3C). Approximately one third of NPC-iGANs showed spontaneous activity at the resting membrane potential (RMP = −56.8 ± 3.3 mV) (Fig. 3C–E). iGANs had a mean capacitance of 13.0 ± 2.8 pF and input resistance of 2.3 ± 0.4 GΩ (n=12).

Figure 3: NPC-derived iGANs exhibit neuronal activity and maturity.

(A) Example of current traces elicited by voltage steps from −70 mV to 50 mV (10 mV increment) reveal voltage gated Na⁺ (inward; red asterisk) and K⁺ (outward) currents. Holding potential was −80 mV. (B) Current-injection steps (0.02, 0.04 nA) from −80 mV induces action potentials. (C) Dot plot shows the mean (red bar) voltage for AP_h, AP_t, and RMP (i), and the mean rate of induced firing (ii). (D) Example of spontaneous firing activity at resting membrane potential. Mean rate of spontaneous firing was 1.5 ± 0.5 Hz (n=5). (E) Pie chart shows fraction of spontaneously active neurons.

3.4. NPC-iGANs exhibit distinct transcriptional profiles

To validate the transcriptional profile of NPC-iGANs, RNA-seq was performed on various cell types generated from across two independent laboratories: hiPSCs, NPCs, forebrain (FB), NPC-derived NGN2-glutamatergic neurons (GLUT) and NPC-derived iGANs (GABA) (SI Table 1). Expectedly, hiPSCs showed increased expression of hiPSC markers while FB, GLUT and GABA demonstrated increased expression of neuronal markers (Fig 4A). Additionally, GABAergic neurons revealed a decrease in expression of glutamatergic markers and an increase in expression of GABAergic genes (Fig. 4A), confirming that NPC-iGANs acquire a GABAergic transcriptional profile.

We next analyzed 14 NPC-iGANs generated from 10 donors across the two independent laboratories (Fig. 4B). The gene groups examined were either involved in (1) pan-neuronal expression and suggestive of functional maturation or (2) GABAergic neurotransmission, interneuron subtype markers, and regional patterning. All NPC-iGANs robustly expressed pan-neuronal markers, such as MAP2 and SYN1/2/3. Most GABAergic cortical interneurons originate from progenitors within the medial, caudal and lateral ganglion eminences (MGE, CGE, LGE) of the ventral telencephalon (Kelsom and Lu, 2013; Lim et al., 2018). Signaling patterns and transcription factors guide these progenitors towards their distinct interneuron sub-type fate; NPC-iGANs demonstrated increased expression for FOXG1, a transcription factor essential for telencephalon development, and DLX1/2/5/6, a family of transcription factors driving progenitors of the MGE (Manuel et al., 2010). GABAergic subtypes from MGE (SST) and CGE (CALB2) were expressed across all NPC-iGANs (Al-Jaberi et al., 2015), and SOX6, a cortical interneuron marker, was upregulated relative to hiPSCs (but not NPCs). Markers of medium spiny, embryonic and hypothalamic GABAergic neurons showed relatively low expression (Skene et al., 2018). Furthermore, NPC-iGANs demonstrated high expression of markers implicated in inhibitory GABAergic neurotransmission (GABBR2/3, GAD1/2, KCC2, VGAT, etc.). Overall, NPC-iGANs from both laboratories revealed similar trends in the gene groups examined (Fig. 4B), highlighting the reproducibility of our adapted ASCL1/DLX2-GABAergic induction to hiPSC-derived NPCs.

4. Discussion

We adapted ASCL1/DLX2-GABAergic induction to hiPSC-derived NPCs, demonstrating that transient over-expression of the transcription factors ASCL1 and DLX2 was sufficient to induce neuronal differentiation, GABAergic gene expression and electrophysiological maturity. Importantly, we successfully applied this methodology to hiPSC-NPCs derived from ten donors across two independent laboratories, finding it to be a rapid, efficient and highly reproducible approach to generate induced GABAergic neurons.

NPCs offer a highly scalable starting cell type that proliferates quickly and are easy to culture. The combination of exogenous transgene expression, antibiotic selection and Ara-C treatment described here yields a highly pure population of GABAergic neurons. NPC-iGANs show GABAergic marker expression comparable to hiPSC-iGANs by qPCR; 7-week-old NPC-iGANs exhibit spontaneous neuronal activity and electrophysiological maturity. Although GABAergic gene expression is elevated in both NPC-iGANs and hiPSC-iGANs, we do observe a trend in which GABAergic gene expression appears relatively higher in hiPSC-iGANs than NPC-iGANs; how this unexpected observation pertains to the resultant cell type remains unclear. Additional experiments to assess the synaptic maturity of NPC-iGANs remain necessary, notably assays of burst-firing or fast-spiking activity, GABA secretion and inhibitory synaptic transmission. It should be noted that it is well-accepted in the field that hiPSC-derived neurons (whether induced or differentiated) lack the maturity observed in the human brain (Brennand et al., 2015).

Our electrophysiological recordings revealed that NPC-iGANs are sufficiently mature: they exhibit a hyperpolarized resting membrane potential, reliably produce action potentials with depolarization, and in some cases, exhibit spontaneous activity. While previous studies have reported similar electrophysiological properties as well as synaptic events (Colasante et al., 2015; Sun et al., 2016; Yang et al., 2017), there may be differences in maturation between ASCL1/DLX2-GABAergic induction from hiPSCs and our NPC-iGANs. Indeed, the cell size (capacitance) appears to be different between hiPSC- and NPC-derived iGANs. Notably, the resting membrane potential for NPC-iGANs was more hyperpolarized than hiPSC-iGANs (Yang et al., 2017), but perhaps not as much mature cortical neurons in rodents (Martínez et al., 2017). It will be important in the future to conduct a side-by-side comparison of donor-matched hiPSC-iGANs and NPC-iGANs to fully elucidate possible differences between these GABA neurons. It is conceivable that a difference between hiPSC-iGANs and NPC-iGANs may be caused by differences in pre-patterning. For example, hiPSCs can be considered as a “blank slate” in that they by definition exist in a genomic state open to any cell lineage, whereas NPCs are pre-patterned to a forebrain neural identity (Brennand et al., 2015).

Critically, as we did not compare donor-matched hiPSC-iGANs and NPC-iGANs in parallel experiments, we cannot definitively resolve the extent to which regional patterning or functional maturity differs between these highly similar methodologies. Although previous studies have reported the electrophysiological and synaptic properties of ASCL1/DLX2-GABAergic induction from hiPSCs, making any reliable conclusions on the functional maturity between hiPSC-iGANs and NPC-iGANs is challenging due to differences in starting cell type and technical differences (seeding densities, medias, etc.). A comparison between donor-matched hiPSCs-iGANs and NPC-iGANs will be needed to assess possible differences in maturation between hiPSC-iGANs and NPC-iGANs. Nonetheless, one should always control for cell source in all iGAN experiments.

It is conceivable that any difference between hiPSC-iGANs and NPC-iGANs may be due to differences in pre-patterning. For example, hiPSCs can be considered as a “blank slate” in that they by definition exist in a genomic state open to any cell lineage, whereas NPCs are pre-patterned to a forebrain neural identity (Brennand et al., 2015). Although it is possible that neuronal induction from hiPSC-NPCs might minimize variability within or between cultures of induced neurons, our data does not suggest that starting from forebrain-patterned hiPSC-NPCs necessarily guarantees that induced ASCL1/DLX2-GABAergic have a cortical identity.

Overall, although overexpression of ASCL1 and DLX2 is a highly reproducible and scalable method sufficient to robustly and rapidly induce GABAergic neurons from both hiPSCs and NPCs, we caution that one should always control from cell-type-of-origin effects in the design of hiPSC-based studies, as neurons induced from hiPSCs and NPCs need not necessarily represent equivalent populations. Thus, our adapted induction may serve of use when (1) NPCs are readily available and/or (2) a direct comparison between NPC, and NPC-derived other cell types such as astrocytes or excitatory neurons is necessary, as NPC-derived GABAergic neuron may represent a cell state that arguably better reflects its intrinsic lineage trajectory.

ASCL1/DLX2-GABAergic and NGN2-glutamatergic neurons can easily be re-plated early during the maturation process, making possible co-culture experiments with other neural cell types, and ultimately, the construction of defined circuit-like population (Fig. 5). Moving towards a more physiologically relevant context will facilitate studies of cell autonomous and non-cell autonomous effects across patient- and control-derived cell types. Moreover, being highly amenable to lentiviral transduction, we anticipate the straightforward application of CRISPR-based genomic engineering approaches (Powell et al., 2017), including techniques such as CRISPR activation and inhibition (CRISPRa/i) that we have already applied to NGN2-neurons (Ho et al., 2017). Ultimately, we expect this will be a useful platform for modeling neuropsychiatric disorders, functional validation of disease-associated variants identified by large-scale genetic studies, and high throughput drug screening.

Figure 5: Co-cultured induced glutamatergic and GABAergic neurons.

NPCs were transduced with NGN2-BFP (left) and GABA-GFP (right) and then re-plated together following doxycycline induction and antibiotic selections.

Highlights.

• hiPSC-NPCs offer an efficient and scalable platform for neuronal differentiation.

• Transient overexpression of ASCL1 and DLX2 in NPCs induces GABAergic fate.

• induced GABAergic neurons express inhibitory genes and are functionally mature.

Abbreviations

hiPSCs

Human induced pluripotent stem cells

NPCs

neural progenitor cells

ASCL1

Achaete-scute family bHLH transcription factor 1

GABA

gamma-Aminobutyric acid

iGANs

induced GABAergic neurons