Cultured networks of excitatory projection neurons and inhibitory interneurons for studying human cortical neurotoxicity

See allHide authors and affiliations

Science Translational Medicine  06 Apr 2016:
Vol. 8, Issue 333, pp. 333ra48
DOI: 10.1126/scitranslmed.aad0623

Insights into neuronal cell death

The study of the mechanisms of cell death in human cortical neurons has been hampered by the lack of human neuronal cultures that exhibit a balanced network of excitatory and inhibitory synapses. Xu et al. now describe a method to culture human neurons with a representative ratio of both excitatory and inhibitory neurons derived from human embryonic stem cells or human inducible pluripotent stem cells. Using this new method, they show that human cortical neurons die in a nitric oxide– and poly(ADP-ribose) polymerase-1–dependent manner. These cultures can be used to study the mechanisms of neurotoxicity in human disorders that involve the demise of cortical neurons.


Translating neuroprotective treatments from discovery in cell and animal models to the clinic has proven challenging. To reduce the gap between basic studies of neurotoxicity and neuroprotection and clinically relevant therapies, we developed a human cortical neuron culture system from human embryonic stem cells or human inducible pluripotent stem cells that generated both excitatory and inhibitory neuronal networks resembling the composition of the human cortex. This methodology used timed administration of retinoic acid to FOXG1+ neural precursor cells leading to differentiation of neuronal populations representative of the six cortical layers with both excitatory and inhibitory neuronal networks that were functional and homeostatically stable. In human cortical neuronal cultures, excitotoxicity or ischemia due to oxygen and glucose deprivation led to cell death that was dependent on N-methyl-d-aspartate (NMDA) receptors, nitric oxide (NO), and poly(ADP-ribose) polymerase (PARP) (a cell death pathway called parthanatos that is distinct from apoptosis, necroptosis, and other forms of cell death). Neuronal cell death was attenuated by PARP inhibitors that are currently in clinical trials for cancer treatment. This culture system provides a new platform for the study of human cortical neurotoxicity and suggests that PARP inhibitors may be useful for ameliorating excitotoxic and ischemic cell death in human neurons.


The human cerebral cortex is a complex structure with tightly interconnected excitatory and inhibitory neuronal networks that are organized in six layers (1, 2). The dynamic interplay between excitatory pyramidal cells and γ-aminobutyric acid–releasing (GABAergic) interneurons begins at the early stages of neurogenesis (3). Considerable advances in methods that model human cortical development using human embryonic stem cells (ESCs) or human inducible pluripotent stem cells (iPSCs) have allowed the generation of relatively pure populations of excitatory cortical projection neurons (46), forebrain inhibitory progenitors (711), and inhibitory neurons (1214) or poorly characterized mixtures of excitatory and inhibitory neurons (1517). Because no existing protocol produces a balanced network of excitatory and inhibitory neurons observed in the human cerebral cortex (see fig. S1) (4, 1624), we sought to develop a suitable protocol that would yield appropriately balanced neuronal networks in vitro that closely resemble neuronal networks in vivo. This is particularly important when modeling glutamate excitotoxicity because neuronal nitric oxide (NO) synthase (nNOS)–expressing interneurons play a major role in the death of neurons in response to glutamate or ischemia (2528).

The cerebral cortex is derived from forebrain forkhead box G1 (FOXG1)–expressing primordium (29). Here, we describe the production of a highly enriched population of forebrain region–specific FOXG1 neural precursor cells by the isolation of rosette neural aggregates (that is, neural aggregates derived from rosette type neural stem cells or RONAs) from human ESCs or iPSCs. Spontaneous fate determination of dorsal and ventral telencephalic progenitors coupled with timed retinoic acid administration leads to differentiation of neurons into all classes of excitatory and inhibitory neurons in a balanced manner, reflecting the mature human cerebral cortex. These cultures are sensitive to excitotoxic injury and cell death after oxygen-glucose deprivation mediated by the activation of N-methyl-d-aspartate (NMDA) receptors, nNOS, and poly(ADP-ribose) polymerase (PARP). Clinically available PARP inhibitors are effective in blocking human cortical neuronal death. Thus, this method enables the study of excitatory and inhibitory functional networks and processes that are thought to play important roles in synaptic physiology and pathophysiology.


Isolation of FOXG1-positive forebrain progenitor from hESCs or hiPSCs

Human ESCs and iPSCs have intrinsic mechanisms enabling differentiation into different subtypes of neurons in the setting of appropriate cues (5, 3032). We sought to develop a method to induce differentiation of human ESCs or iPSCs into FOXG1 forebrain progenitors to mimic the patterning and neurogenesis necessary to generate excitatory and inhibitory neurons in the same culture (Fig. 1A). First, human ESCs or iPSCs were detached and grown as embyroid body (EB) suspensions (Fig. 1A) (18). On day 2, dual SMAD inhibition (31) with noggin and SB431542 was initiated for 4 days until day 7 to allow the robust induction of rosette formation (fig. S2, A to D). On day 7, the EBs were plated on Matrigel- or laminin-precoated plates. At day 9, rosettes began to form, which were FOXG1- and Nestin (NES)–positive (Fig. 1B). Rosettes showed profound cellular expansion in a defined central area of each colony where they formed RONAs (Fig. 1, C and D). The centers of these RONAs were FOXG1-positive and were surrounded by cytokeratin 8 (CK8)–positive epithelial-like cells. This demarcation was easily identified by phase-contrast microscopy without immunostaining, which allowed manual isolation of FOXG1 precursor cells, minimizing contamination of the nonneural precursors without the need for cell sorting (Fig. 1, E to G). On day 17, RONAs were manually isolated and grown as neurospheres in suspension for 1 day. On day 18, the neurospheres were disassociated into single cells and plated on laminin/poly-d-lysine–coated plates or coverslips where they formed neural precursor cell (NPC) clusters (Fig. 1A). To determine whether this isolation protocol resulted in the full complement of neural precursor subtypes [FOXG1, LIM homeobox 2 (LHX2), paired box 6 (PAX6), genetic-screened homeobox 2 (GSX2), and NK2 homeobox 1 (NKX2.1)], which pattern the human cortical subdivision and specify cortical excitatory and inhibitory neurons (fig. S2E), immunostaining for these markers was performed at day 9 (rosettes), day 16 (RONA), and day 18 (NPC) (Fig. 1, H to L). The rosettes on day 9 were positive for the forebrain and dorsal markers FOXG1, PAX6, and LHX2 but not for the ventral telencephalic markers NKX2.1 and GSX2. On day 16, RONAs expressed all five cortical patterning related transcription factors, with FOXG1, PAX6, and LHX2 predominating and NKX2.1 and GSX2 expressed in patches at the periphery of RONAs (fig. S2, F to I). A comparable pattern was observed in human ESC (H1) and human iPSC (SC1014 and SC2131) lines, with no significant differences observed among the different lines (Fig. 1, H to L, and fig. S2J). RNA sequencing was performed to gain a comprehensive and quantitative view of the transcriptome of isolated FOXG1 RONAs. Markers of pluripotency, mesoderm, and endoderm were not detected. Pan-NPC markers [SOX2, NES, vimentin (VIM), and cadherin 2 (CDH2)] were highly expressed in isolated RONAs (Fig. 1M). The anterior forebrain markers FOXG1, LHX2, and OTX2 (orthodenticle homeobox 2) were highly expressed (Fig. 1N). Dorsal excitatory neuronal lineage markers [PAX6, empty spiracles homeobox 1 (EMX1), and EMX2] and ventral forebrain GABAergic neuronal markers [ASCL1 (achaete-scute homolog 1) and DLX2 (distal-less homeobox 2)] along with medial ganglionic eminence markers (NKX2.1 and LHX8) and caudal ganglionic eminence markers [GSX2 and SP8 (specificity protein 8)] were all expressed in isolated RONAs (Fig. 1N). Diencephalic [RAX (retina and anterior neural fold homeobox)], midbrain [EN1 (engrailed homeobox 1 and MSX2 Msh homeobox 2)], hindbrain [homeobox A2 (HOXA2) and HOXB2], and more posterior markers (HOXA1, HOXB1, HOXC5, HOXB6, and HB9) were expressed at barely detectable levels (Fig. 1N). Complete ontology and enrichment analysis for all genes expressed in FOXG1 progenitors are shown in data file S1. These results together indicate that the isolation of the FOXG1 RONAs provided a robust method for recapitulating the early induction of the human telencephalon (FOXG1) and its subsequent subdivision into dorsal (PAX6 and LHX2) and ventral (NKX2.1) territories. These NPCs were spontaneously differentiated into neurons after isolation from the RONAs and plating (Fig. 1O). These data suggested that the FOXG1-positive cells were being maintained in a precursor state before their isolation, and that they were capable of spontaneous differentiation into neurons when removed from the RONA niche.

Fig. 1. Generation of forebrain progenitor cells with regional identity from human iPSCs.

(A) Scheme of the RONA differentiation protocol showing generation of forebrain progenitors from human ESCs and iPSCs. Nog, noggin; SB, SB431542. (B) Rosettes were immunopositive for FOXG1 and NES at day 9 after initiation of neural differentiation. DAPI, 4′,6-diamidino-2-phenylindole. (C and D) With the prolongation of neural differentiation, highly proliferative rosettes started to pile up, resulting in three-dimensional (3D) columnar cellular aggregates, which were positive for FOXG1 and NES. (D) Phase-contrast image of RONA. Scale bar, 50 μm (B to D). (E to G) FOXG1 immunoreactive RONAs were located in the center (E) and demarcated clearly with surrounding CK8-positive cells. (F) Cells underneath the RONAs were FOXG1/CK8. (G) The white inset box in (F) is magnified (×2.5). (H to L) Expression and quantification of the early forebrain regionalization markers FOXG1 (H), PAX6 (I), LHX2 (J), NKX2.1 (K), and GSX2 (L) after 9, 16, and 18 days of differentiation. Data are presented as means ± SEM (n = 3). Scale bar, 20 μm. (M and N) RNA sequencing gene expression profiling of RONAs derived from the human ESC H1 line showing the enrichment of forebrain progenitors with regional identity. Data show the average RPKM (reads per kilobase of transcript per million mapped reads) value of two biological replicates of human ESC H1 cell line RNA sequencing experiments. ESRRB, estrogen-related receptor β; SMA, smooth muscle actin; HNF1B, hepatocyte nuclear factor 1 homeobox B; MGE, medial ganglionic eminence; CGE, caudal ganglionic eminence. (O) Spontaneous neuronal differentiation after isolation of RONA neural progenitors without the application of exogenous mitogens. Nuclei were counterstained with DAPI (blue). Colors are indicated in the images. Scale bar, 20 μm. Data are from the human ESC H1 cell line in (B) to (O). DCX, doublecortin.

Induction of forebrain precursors with regional identity and differentiation of inhibitory neurons

In the absence of exogenous factors, the neural precursors expressed a complex combination of brain regional markers, including FOXG1, PAX6, LHX2, NKX2.1, and GSX2, directly after isolation of RONAs (Fig. 1), indicating that these neuronal precursors had the potential capability of differentiating into both excitatory and inhibitory neurons. After isolation from RONAs, neural precursors spontaneously showed dynamic changes in cell identity. At day 19, close to 100% of the cells were FOXG1-positive, 50% were PAX6-positive, and less than 5% were NXK2.1-positive (fig. S3, A to D). At day 24, 80% of the cells were PAX6-positive and 20% were NKX2.1-positive (fig. S3, A to D). After prolonged culture, more than 90% of these cells spontaneously differentiated into excitatory neurons, whereas less than 5 to 7% showed a mature inhibitory neuronal phenotype.

Because the composition of neurons in the human cerebral cortex is about 80% excitatory glutamatergic neurons and 20% GABAergic interneurons (1, 2), a better representative balance of excitatory and inhibitory neurons was needed. Accordingly, we explored whether retinoic acid or sonic hedgehog (SHH) signaling could facilitate the optimal production of excitatory and inhibitory neurons (fig. S3, A to D). Retinoic acid and either SHH or the SHH pathway agonist purmorphamine in various combinations were applied at day 19 and maintained for 6 days or were applied at day 24 and maintained for 6 days (fig. S3A). Extensive optimization revealed that administration of retinoic acid at days 24 to 30 was required for maintaining the identity of both PAX6/FOXG1 and NKX2.1/FOXG1 progenitors and provided an appropriate percentage of excitatory and inhibitory neurons representative of the human cerebral cortex (Fig. 2A).

Fig. 2. Timing of retinoic acid/SHH exposure determines differentiation of excitatory and inhibitory neurons.

(A) Schematic summary of conditions for differentiation of appropriate balanced excitatory and inhibitory neurons. KoSR, knockout serum replacement; E/I, excitatory/inhibitory. (B) Immunocytochemical analysis of neuronal expression of the excitatory marker VGLUT and the inhibitory marker VGAT. Quantification of the percentages of VGLUT- and VGAT-positive cells with or without retinoic acid (RA) exposure. Data are presented as means ± SEM (n = 3). *P < 0.05, analysis of variance (ANOVA) with Tukey-Kramer’s post hoc test. (C) Immunocytochemical analysis of PSD95 (red) and synapsin (SYN; green) after retinoic acid exposure from days 24 to 30 after the initiation of neural differentiation. Quantification of the proportion of PSD95 puncta that were found associated with synapsin puncta. Data are presented as means ± SEM (n = 3). (D) Differentiation of inhibitory neurons expressing subtype markers. Colors are indicated in the images. Scale bar, 20 μm. (E) Quantification of inhibitory neurons with immunostaining analyses over 32 weeks after differentiation. Data are presented as means ± SEM (n = 3). (F) Composition of interneuron subtypes in adult human cortex (41) and the cultured neurons derived from RONAs treated with retinoic acid. Data from cortical neuronal cultures derived from the human ESC H1 cell line are presented as means ± SEM. (G) Developmental expression pattern of PV, SST, CR, nNOS, and CB in human cortical interneurons from mid-fetal stage, late fetal stage, to infant (60). Data from human cortical cultures are from the human ESC H1 cell line in (B) to (F).

Next, the neural precursors were allowed to develop and differentiate after withdrawal of retinoic acid at day 30 while being maintained in neuronal differentiation medium. Cultures were assessed at different time points for markers of excitatory and inhibitory neurons, as well as synaptogenesis. Two weeks after withdrawal of retinoic acid, the cultures were composed primarily of β-tubulin III (TUJ1)–positive neuronal cells (>90%) with about 5 to 10% glial fibrillary acidic protein (GFAP)–positive astrocytes (Fig. 1O and fig. S3E). Four weeks after the withdrawal of retinoic acid, the neurons were composed of about 15 to 20% inhibitory neurons as assessed by vesicular GABA transporter (VGAT)/glutamic acid decarboxylase 67 (GAD67) immunostaining and 80 to 85% excitatory neurons as assessed by vesicular glutamate transporter (VGLUT)/calmodulin-dependent protein kinase II (CAMKII) immunostaining (Fig. 2B and fig. S3F). Two weeks after the retinoic acid withdrawal, these neurons started to express the presynaptic marker synapsin and the postsynaptic marker postsynaptic density protein 95 (PSD95) (Fig. 2C). These cultures expressed a variety of inhibitory neuronal markers that included calretinin (CR), calbindin (CB), parvalbumin (PV), nNOS, somatostatin (SST), neuropeptide Y (NPY), and vasointestinal polypeptide (VIP). Thirty-two weeks after differentiation, the microtubule-associated protein 2 (MAP2)–positive neurons comprised about 3.2% CR-, 1% CB-, 3.68% PV-, 2.5% nNOS-, 5.67% SST-, 1.53% NPY-, and 1.39% VIP-positive neurons (Fig. 2, D and E). A comparable composition was observed in cultures of the human iPSC SC1014 cell line (fig. S3G). The composition of interneuron subtypes in adult human cortex and neuronal cells derived from RONAs was compared. The RONA culture showed comparable composition of PV, SST, CR, VIP, and nNOS interneurons to human adult cortex (Fig. 2F) and similar developmental patterns for PV, SST, CR, nNOS, and CB interneurons when compared to human brain tissues (Fig. 2, E and G). Furthermore, these cultures also expressed the NMDA receptor subunit NR1, the AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) receptor subunit glutamate A1 (GluA1), and nNOS in a mature punctate pattern (fig. S3, H to J). These data demonstrated that timed exposure to retinoic acid signaling directed the differentiation of FOXG1 forebrain progenitor cells derived from human ESCs or iPSCs into excitatory neurons and a diverse repertoire of inhibitory neuronal subtypes with high efficiency.

Differentiation of layer-specific cortical excitatory neurons from FOXG1+ forebrain progenitors

To determine whether the neurons expressed cortical layer–specific markers 90 days after differentiation, immunostaining was performed for a set of neuronal specific transcription factors (Fig. 3A). The neuronal cultures expressed the cortical layer I, V, and VI marker T-box brain protein 1 (TBR1); the layer II to IV markers brain-2 (BRN2) and special AT-rich sequence-binding protein 2 (SATB2); and the layer V and VI marker chicken ovalbumin upstream promoter-transcription factor interacting protein 2 (CTIP2) (Fig. 3A). The proportion of upper superficial and deep layer cortical projection neurons was roughly the same in the differentiated cortical system across human ESC H1, H9, and human iPSC SC1014 cell lines (Fig. 3B). The human cerebral cortex is characterized by an organized six-layered complex structure. In a modified 3D basement membrane culture method (3335) (Fig. 3C), immunostaining of cross sections indicated that these neuronal cultures attempted to assemble into cortical layers with the deep layer cortical marker CTIP2 and the superficial cortical layer marker SATB2 tending to segregate into different orientations (Fig. 3, D and E).

Fig. 3. Generation of layer-specific human cortical neurons from FOXG1+ forebrain progenitor cells.

(A) Immunocytochemical analysis of the cortical layer–specific markers TBR1, CTIP2, BRN2, and SATB2 in human neuronal culture differentiated from FOXG1+ neural progenitors derived from the human ESC H1 cell line. (B) Quantification of the percentages of TBR1, CTIP2, BRN2, and SATB2 in neuronal culture. Data are presented as means ± SEM (n = 3). Data are from the human ESC H1 andH9 cell lines and the human iPSC line SC1014. (C) Diagram of the 3D human cortical assembly assay. (D and E) Cross sections from 3D human cortical assemblies were immunostained with the cortical neuron layer markers TBR1, CTIP2, BRN2, and SATB2. Nuclei were counterstained with DAPI (blue). The boxed area is magnified (×2.5) in the fifth column of (E). Colors are indicated in the images. Scale bar, 50 μm. Data in (C) to (E) are from the human ESC H1 cell line.

Functional human cortical excitatory and inhibitory networks show homeostasis

Eight weeks after differentiation, the electrical properties of human ESC–derived cortical neurons were measured. Whole-cell patch-clamp recording of the neurons revealed the presence of voltage-gated sodium currents that were blocked by tetrodotoxin (Fig. 4A and fig. S4A). In addition, there were voltage-gated potassium currents that were sensitive to the potassium channel blocker 4-aminopyridine (Fig. 4B). Human ESC–derived cortical neurons fired action potentials when they were depolarized but not when they were held at low membrane potential (Fig. 4C). To test whether the excitatory and inhibitory synapses (see Fig. 2, B and C) were functional under physiological conditions, miniature excitatory postsynaptic currents (mEPSCs) and miniature iPSCs (mIPSCs) were detected by whole-cell patch-clamp recording. To probe the dynamics of mIPSCs during the development of the neuronal network, the properties of mIPSCs in amplitudes and frequencies were recorded and analyzed over time, revealing a kinetic change over the time measured (fig. S4B). At 8 weeks after induction of neuronal differentiation, consistent with the presence of GABAergic synaptic inputs, whole-cell patch-clamp analysis demonstrated that cells readily received inhibitory postsynaptic currents that could be reversibly blocked by a γ-aminobutyric acid type A (GABAA) receptor inhibitor picrotoxin in the presence of the NMDA receptor antagonist MK801 or the AMPA receptor antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) (Fig. 4D). The mEPSCs were frequently detected in the presence of picrotoxin with 5 to 15 pA of amplitude. The amplitude and frequency of mEPSCs of these cortical neurons increased over time (fig. S4C).

Fig. 4. Development of a functional human cortical excitatory and inhibitory neuronal network.

(A and B) Voltage-gated sodium (A) and potassium channels (B) present in human ESC H1 line–derived cortical neurons (n = 10 for sodium current and n = 12 for potassium current). 4-AP, 4-aminopyridine. (C) Evoked action potentials (whole-cell recording, current clamping) generated by human ESC H1 line–derived cortical neurons after 8 weeks of differentiation (n = 9). (D) Measurement of mIPSCs indicated the formation of a functional inhibitory network in human cortical neuronal culture. With 40 mM chloride ion in the pipette solution, the mIPSC current is inward (n = 11 for each group). (E) Homeostatic scaling of human ESC H1 cell line–derived cortical neurons. Measurement of mEPSCs from cultured human cortical neurons under control conditions (control), conditions of activity blockade [tetrodotoxin (TTX)], or conditions of activity enhancement with bicuculline (Bic). Forty-eight hours of activity blockade increased the amplitude of mEPSCs, whereas 48 hours of enhanced activity decreased the mEPSC amplitude. *P < 0.05, Mann-Whitney test, Control, n = 17; TTX, n = 21; Bic, n = 18. All data shown are from the human ESC H1 cell line.

Synaptic scaling homeostatically regulates the stability of network activity by balancing excitation and inhibition (36). To examine whether the existing functional network could undergo synaptic scaling, homeostatic scaling was assessed in human ESC–derived cortical neurons treated for 36 to 48 hours with tetrodotoxin (1 μM), which blocked all evoked neuronal activity, or with bicuculline (20 μM), which blocked inhibitory neurotransmission mediated by GABAA receptors and increased neuronal firing (37). There was a robust increase in the amplitude of mEPSCs after tetrodotoxin treatment (Fig. 4E), which was present 4 weeks after differentiation and further increased after 8 weeks (Fig. 4E and fig. S4, D and E). A rightward shift of the cumulative probability distributions indicated that the increase in amplitude was distributed across the range of recorded events. In contrast to the increase in mEPSCs after tetrodotoxin treatment, there was a robust decrease in the amplitude of mEPSCs after bicuculline treatment (Fig. 4E), which began at 4 weeks after differentiation and was more robust after 8 weeks (Fig. 4E and fig. S4, D and E). There was a leftward shift of the cumulative probability distributions, indicating that the decrease in amplitude was distributed across the range of recorded events (Fig. 4E). These results show that the differentiated neurons formed a functionally balanced excitatory and inhibitory network, and neuronal activity could be homeostatically regulated to maintain network stability.

Human cortical neurons are sensitive to NMDA- and oxygen-glucose deprivation–mediated neurotoxicity

To ascertain the susceptibility of human cortical neurons to excitotoxicity, human cortical neurons derived from the human ESC H1 cell line were exposed to 500 μM glutamate for 5, 15, 30, or 60 min. More than 60% of neurons died after a 30-min exposure to 500 μM glutamate as assessed 24 hours later. Exposure to 500 μM glutamate for 5 or 15 min resulted in lower neurotoxicity (Fig. 5A). A dose response to increasing concentrations of glutamate indicated that human cortical neurons were sensitive to a graded concentration of glutamate, with greater than 60% toxicity achieved with 500 μM glutamate for 30 min (Fig. 5B). Glutamate neurotoxicity was markedly attenuated by the NMDA receptor antagonist MK801, but only a modest protection was observed with the AMPA receptor antagonist NBQX (2,3-dihydroxy-6-nitro-7-sulfamoylbenzo[f]quinoxaline) (Fig. 5C). Because glutamate neurotoxicity in human neurons occurred predominantly through NMDA receptors, subsequent studies focused on NMDA receptor stimulation using NMDA to activate the NMDA glutamate receptor subtype. A time course of 500 μM NMDA was performed and revealed a similar pattern of neurotoxicity with about 60% neuronal cell death after 30 min of exposure and no or modest toxicity after 5 or 15 min of exposure to 500 μM NMDA assessed 24 hours later (fig. S5A). A dose response to increasing concentrations of NMDA indicated that human cortical neurons derived from the human ESC H1 cell line were sensitive to a graded concentration of NMDA, with greater than 60% toxicity achieved with 500 μM NMDA for 30 min (Fig. 5D). Calcium ion influx triggered by exposure to NMDA (100 or 500 μM) was almost fully abolished by the addition of MK801 (Fig. 5E). Human cortical neurons were also sensitive to oxygen-glucose deprivation, with cell death observed after 60 min of oxygen-glucose deprivation and with greater than 60% cell death observed after 120 min of oxygen-glucose deprivation assessed 24 hours later (Fig. 5F). Human cortical neuron cultures derived by a similar protocol but without retinoic acid treatment were challenged with glutamate or NMDA at different doses and exposure times or with oxygen-glucose deprivation at different exposure times to compare their sensitivity to excitotoxicity in the presence or absence of retinoic acid treatment (fig. S5, B to F). The absence of retinoic acid treatment led to less sensitivity of human neuronal cultures to neurotoxicity mediated by glutamate, NMDA, or oxygen-glucose deprivation (fig. S5, B to F).

Fig. 5. Characterizing human cortical neurons as a model for excitotoxicity.

(A) Percentages of neuron death were assessed by propidium iodide (PI)/Hoechst staining 24 hours after different exposure times to 500 μM glutamate (Glut) and were quantified and expressed as means ± SEM (n = 3). (B) Percentages of neuron death were assessed by PI/Hoechst staining 24 hours after 30 min of glutamate exposure at different doses and were quantified and expressed as means ± SEM (n = 3). (C) MK801 (NMDA receptor antagonist) but not NBQX (antagonist of AMPA receptors and kainate receptors) abolished 500 μM glutamate–induced human cortical neuron death. Percentages of neuron death were quantified and expressed as means ± SEM (n = 3). Glutamate significantly killed neurons (***P < 0.001), and MK801 significantly protected the neurons against glutamate excitotoxicity (***P < 0.001), whereas NBQX had no effect on glutamate excitotoxicity. (D) Thirty minutes of treatment with NMDA at different concentrations. Percentages of neuron death were assessed by PI/Hoechst staining, quantified, and expressed as means ± SEM (n = 9). (E) Calcium imaging with Fura-2 of human cortical neuron cultures stimulated with 100 or 500 μM NMDA and 10 μM glycine, followed by MK801. Plots show the means ± SEM of 30 to 31 neurons. (F) Human ESC H1 cell line–derived cortical neurons were subjected to different lengths of oxygen-glucose deprivation. Percentages of neuron death were assessed by PI/Hoechst staining, quantified, and expressed as means ± SEM (n = 9). **P < 0.01 and ***P < 0.001, ANOVA with Bonferroni’s posttest. NT, not treated.

NMDA-mediated excitotoxicity and oxygen-glucose deprivation–mediated neurotoxicity depend on NO and the parthanatos pathway

To evaluate the mechanisms involved in NMDA-mediated or oxygen-glucose deprivation–mediated neurotoxicity in human cortical neurons derived from the human ESC H1 cell line, cell death was assessed after exposure to NMDA or oxygen-glucose deprivation in the presence of antagonists to known cell death signals. Only the selective nNOS inhibitor Nω-propyl-l-arginine (NPLA) and the PARP inhibitor DPQ blocked NMDA- or oxygen-glucose deprivation–mediated neurotoxicity (Fig. 6, A and B). Because nNOS neurons play a critical role in excitotoxicity, we quantified and compared the percentage of nNOS neurons in retinoic acid–induced and non–retinoic acid–induced human neuronal cultures after 2 months of differentiation. We observed fewer nNOS-positive neurons in non–retinoic acid–treated cultures compared to retinoic acid–treated cultures, which may partly explain the reduced sensitivity to excitotoxicity in non–retinoic acid–treated cultures (fig. S5G). The broad-spectrum caspase inhibitor Z-VAD, the necroptosis inhibitor necrostatin-1 (NEC-1), or the autophagy inhibitor 3-methyladenine (3-MA) was not effective in altering NMDA- or oxygen-glucose deprivation–mediated neurotoxicity (Fig. 6, A to B). NMDA-induced or oxygen-glucose deprivation–induced NO formation was measured with the Measure-iT High-Sensitivity Nitrite Assay Kit (38). The selective nNOS inhibitor NPLA attenuated NO production consistent with its neuroprotective effect, whereas the PARP inhibitor had no effect (Fig. 6, C and D). NMDA-induced or oxygen-glucose deprivation–induced PARP activity was assessed by a PAR immunoblot assay. The PARP inhibitor DPQ and the NOS inhibitor l-NAME or the selective nNOS inhibitor NPLA attenuated PAR formation (Fig. 6, E and F). To confirm that activation of PARP was critical to NMDA- or oxygen-glucose deprivation–mediated neurotoxicity, PARP-1 was knocked out by CRISPR/Cas9 using two different single-guide RNAs (sgRNAs) (fig. S6A). Knockout of PARP-1 by either sgRNA substantially attenuated NMDA- or oxygen-glucose deprivation–mediated neurotoxicity (Fig. 6, G and H). Furthermore, four PARP inhibitors that are currently being evaluated in clinical trials for cancer were evaluated for neuroprotective actions against NMDA- or oxygen-glucose deprivation–mediated neurotoxicity. AG-014699, ABT-888, olaparib, and BMN 673 all markedly attenuated NMDA- or oxygen-glucose deprivation–mediated neurotoxicity in human cortical neurons (Fig. 6, I to J). In rodent models, questions have been raised regarding the role of PARP activation in male versus female mice (39). Thus, human cortical neurons were derived from the human ESC H9 cell line, which is female, for comparison to male human ESC H1 cell line–derived cortical cultures. Under identical conditions, we observed that NMDA or oxygen-glucose deprivation elicited identical neurotoxicity in the H1 cell line– versus the H9 cell line–derived cortical cultures, which was attenuated by PARP inhibitors equally in both cortical cultures (fig. S6, B to G).

Fig. 6. Pathways involved in human cortical neuronal death induced by NMDA or oxygen-glucose deprivation.

(A and B) NPLA (selective nNOS inhibitor) and DPQ (PARP inhibitor), but not Z-VAD (caspase inhibitor), NEC-1 (necroptosis inhibitor), or 3-MA (autophagy inhibitor), protected human ESC H1 cell line–derived cortical neuronal cultures from cell death due to (A) 30 min of exposure to 500 μM NMDA or (B) 2 hours of oxygen-glucose deprivation. Quantitative data from the PI/Hoechst-positive cell number ratio are shown as means ± SEM. For NMDA-treated neurons (A), n = 8 for the dimethyl sulfoxide (DMSO) control group and the NMDA plus DMSO group; for oxygen-glucose deprivation–treated neurons (B), n = 3 for the DMSO control group and the DMSO + oxygen-glucose deprivation (OGD) group. For the NMDA and oxygen-glucose deprivation neuron groups, n = 3 for Z-VAD, NEC-1, 3-MA, NPLA, and DPQ treatments. (C and D) An nNOS inhibitor (NPLA), but not a PARP-1 inhibitor (DPQ), inhibited NO production in human ESC H1 cell line–derived cortical neuronal cultures after 30 min of treatment with 500 μM NMDA or 2 hours of oxygen-glucose deprivation. Data are shown as means ± SEM of NO units (fluorescence measured at excitation/emission wavelengths of 365 nm/450 nm) normalized to medium-only wells (n = 4). (E and F) Representative blots showing that treatment with a NOS inhibitor (NPLA or l-NAME) or a PARP inhibitor (DPQ) blocked PAR polymer formation (the product generated by PARP) after 30 min of treatment with 500 μM NMDA or 2 hours of oxygen-glucose deprivation in human ESC H1 cell line–derived cortical neuronal cultures; means ± SEM of PAR polymer/actin band density ratio (n = 3). (G and H) Lentiviral vector–based PARP-1 CRISPR/Cas9 sgRNAs protect human ESC H1 line–derived cortical cultures from NMDA- or oxygen-glucose deprivation–induced cell death. Data are shown as means ± SEM (n = 3). (I and J) PARP-1 inhibitors protect human ESC H1 line–derived cortical neuronal cultures from NMDA- or oxygen-glucose deprivation–induced cell death. Data are shown as means ± SEM (n = 4). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 (ANOVA with Bonferroni’s posttest).


This study presents an efficient method for differentiation of human ESCs or iPSCs into excitatory and inhibitory neurons by isolation of highly enriched FOXG1 neural precursors from human ESC or iPSC RONAs. In particular, timed administration of retinoic acid led to balanced excitatory and inhibitory neuronal networks that exhibited all six layers of cortical projection neurons and a rich diversity of GABAergic interneurons. This method enabled the formation of functional excitatory and inhibitory networks that exhibited homeostatic scaling and susceptibility to excitotoxicity and oxygen-glucose deprivation in a NO- and PARP-dependent manner.

FOXG1 is one of the earliest expressed transcription factors that direct the development of the telencephalon (29). In the mouse brain, FOXG1 is required for the development of both the ventral telencephalon and the anterior neocortex, thus providing a source of both excitatory and inhibitory neurons (29, 40). After efficient neuroectoderm conversion of human ESCs or iPSCs expressing high levels of FOXG1, these FOXG1 precursors expressed markers of the dorsal forebrain (PAX6 and LHX2) and ventral forebrain (NKX2.1 and GSX2). The RONA culture method described here does not require cell sorting, is cost-effective using minimal numbers of expensive growth factors, and should be a widely applicable method. It supports the formation of cortical dorsal and ventral subdivisions, reminiscent of forebrain patterning along the dorsoventral axis. This method results in about 20% inhibitory neurons, which is reflective of the known abundance of inhibitory neurons in both mouse and human brain tissue (1, 2). The expression of nNOS neurons in the human brain and RONA-derived neuron culture is equivalent to that observed in mouse and human brain (41, 42). The composition of PV, SST, CR, VIP, and nNOS in the interneuronal culture derived from RONAs is similar to those in human adult cortex (41), which indicates that the percentage of these interneuronal populations is more reflective of human than of mouse cortex. A review of other human cortical neuron differentiation and culture methods indicates that the method described here provides a more balanced representation of both excitatory and inhibitory neurons than previously reported methods (fig. S1). There is a modest underrepresentation of nNOS, CR, and CB interneurons (Fig. 2 and fig. S1), and future studies and optimization will be required to provide a more accurate representation of the interneuron composition of the human cortex. The composition of PV, SST, and VIP interneurons shared by RONA-derived neurons and the human cortex differs from the mouse brain; thus, RONA-derived human neurons may be more suitable for studying neurological diseases involving human excitatory and inhibitory neuronal networks. In addition to the highly enriched neuron culture 8 to 12 weeks after induction of neural differentiation, we observed that about 5 to 10% of the cells were GFAP-positive astrocytes, which is consistent with the astrogenesis pattern shown in 3D cultures of laminated cerebral cortex–like structures (43).

Proper brain function requires a balance of excitatory and inhibitory neurotransmission (36). In neurological and neuropsychiatric diseases such as autism, Alzheimer’s disease, and schizophrenia, there is evidence that the balance of excitatory and inhibitory neurotransmission is disturbed (44). There has been progress in establishing in vitro systems for studying human disease using populations of mature cortical interneurons, excitatory pyramidal neurons, or mixed neural cultures with limited characterization of the interplay between excitatory and inhibitory networks. The protocol described here provides a more accurate representation of the human cortex, allowing the study of the interplay between excitatory and inhibitory networks in both normal and pathological conditions. This method generates functional neurons with relatively mature electrical properties including excitatory and inhibitory postsynaptic currents, mEPSCs and mIPSCs, which are indicative of the copresence of excitatory and inhibitory networks. In response to neural activity, neuronal networks undergo compensatory synaptic scaling to maintain a proper balance of excitation and inhibition (45). Our cortical cultures exhibit homeostatic scaling because pharmacological perturbation by tetrodotoxin and bicuculline results in adaptive synaptic function. Homeostatic scaling has been widely studied in rodent models, and our data indicate that homeostatic scaling can occur in human cortical neurons as well.

Although cultured human fetal cortical neurons are sensitive to glutamate neurotoxicity (46), the underlying mechanisms of glutamate excitotoxicity have not been explored in human neuronal cultures (47, 48). The dynamic interplay between inhibitory neurons and pyramidal excitatory neurons is particularly important in studying neurotoxicity or neuronal injury. In particular, modeling excitotoxicity in primary neuronal cultures from mice and rats requires the presence of nNOS inhibitory neurons (27, 28, 49, 50). A recent study of NMDA excitotoxicity in human embryonic stem cell–derived excitatory neurons required a 24-hour exposure to NMDA to elicit neurotoxicity (47). The requirement of prolonged exposure to NMDA to elicit excitotoxicity is markedly different from the rapidly triggered delayed death that is characteristic of acute neuronal injury that occurs in stroke or trauma (51). On the other hand, the human neuronal culture method reported here exhibited rapidly triggered delayed death in response to glutamate-mediated and NMDA-mediated excitotoxicity and oxygen-glucose deprivation. The enhanced susceptibility is most likely due to an appropriate balance of excitatory and inhibitory neurons (52), including nNOS neurons. Our method revealed that human neurons are sensitive to NO and parthanatos, similar to rodent neuronal cultures, cortical brain slices, and in vivo animal studies (26, 49, 5358). Because neuronal cell death is dependent on the length and strength of the stimulus, other cell death pathways may be recruited with longer exposure to excitotoxins or oxygen-glucose deprivation. However, in acute injury paradigms, NO and parthanatos predominate much in the same manner as has been reported in rodent model systems (26, 53, 57, 59). These results are consistent with the notion that interfering with parthanatos is a particularly attractive approach to treat acute neuronal injury in humans, particularly because there are clinically available PARP inhibitors (54).

In summary, we report a robust and easily reproducible system for generation of excitatory projection neurons and inhibitory interneurons. These human cortical neurons can be used to study the mechanisms of excitotoxicity and screen for clinically relevant neuroprotective compounds. Furthermore, in future studies, this methodology can be used to generate cultures that will be useful in exploring the genetic basis of neuronal connectivity in neuropsychiatric diseases such as autism and schizophrenia, which are thought to involve imbalances in excitatory and inhibitory neural transmission. Moreover, these cultures may be useful in the study of neurodegenerative diseases relevant to the cortex, including Alzheimer’s disease.


Study design

To ensure adequate power to detect an effect size, the effect size was first calculated on the basis of pilot experiments and then used in power analysis by the software G*Power to determine sample size by given error probability (α = 0.05) and power (1 − β > 0.95). All samples were included for data analysis. Imaging and cell count for neural differentiation and neurotoxicity were performed by either automated software or a blinded observer. National Institutes of Health (NIH) registry human ESC lines H1 and H9 (which are widely distributed and researched) and two characterized human iPSC lines, SC1014 and SC2131 (reprogrammed by non-integrating Sendai virus and lentivirus), were used for the neural differentiation and neurotoxicity assays. The human ESC/iPSC lines used and quantified in all the experiments are described in the individual figures and figure legends.

Culture of human ESCs or iPSCs

Human ESC lines H1 and H9 (WiCell) and human iPSC lines SC1014 and SC2131 (Johns Hopkins Medical Institutions) were maintained on inactivated mouse embryonic fibroblasts (MEFs) according to standard protocols. Briefly, human ESCs or iPSCs were maintained in human ESC medium containing Dulbecco’s modified Eagle’s medium/nutrient mixture F-12 (DMEM/F12; Invitrogen), 20% knockout serum replacement (Invitrogen), fibroblast growth factor 2 (FGF2; 4 ng/ml; PeproTech), 1 mM GlutaMAX (Invitrogen), 100 μM nonessential amino acids (Invitrogen), and 100 μM 2-mercaptoethanol (Invitrogen). Medium was changed daily. Cells were passaged using collagenase (1 mg/ml in DMEM/F12) at a ratio 1:6 to 1:12. All experiments using human ESCs/iPSCs were conducted in accordance with the policy of the Johns Hopkins University (JHU) School of Medicine that research involving human ESCs or iPSCs being conducted by the JHU faculty, staff, or students or involving the use of JHU facilities or resources shall be subject to oversight by the JHU Institutional Stem Cell Research Oversight Committee.

Neural differentiation of human ESCs or iPSCs

To initiate differentiation, human ESC or iPSC colonies were treated with collagenase (1 mg/ml in DMEM/F12) in an incubator for about 5 to 10 min. The colony borders began to peel away from the plate, and collagenase was washed off the plate with growth medium. While the colony center remained attached, the colonies were selectively detached with the MEFs undisturbed. Detached ESC or iPSC colonies were then grown in suspension in human ESC medium without FGF2 for 2 days in low-attachment six-well plates (Corning). From days 2 to 6, noggin (50 ng/ml; R&D system) or dorsomorphin (1 μM; Tocris) and SB431542 (10 μM; Tocris) were supplied in human ESC medium (without FGF2, defined as KoSR medium). On day 7, free-floating EBs were transferred to Matrigel- or laminin-precoated culture plates to allow the complete attachment of EB aggregates with the supplementation of N2 induction medium containing DMEM/F12 (Invitrogen), 1% N2 supplement (Invitrogen), 100 μM MEM nonessential amino acid solution (Invitrogen), 1 mM GlutaMAX (Invitrogen), and heparin (2 μg/ml; Sigma). Cultures were fed with N2 medium every other day from days 7 to 12. From day 12 onward, N2 induction medium was changed everyday. Attached aggregates broke down to form a monolayer colony on days 8 to 9 with typical neural-specific rosette formation. With the extension of neural induction, highly compact 3D column-like neural aggregates (termed RONAs) formed in the center of attached colonies. RONAs were manually microisolated, taking special care to minimize the contaminating peripheral monolayer of flat cells and the cells underneath RONAs. RONA clusters were collected and maintained as neurospheres in Neurobasal medium (Invitrogen) containing B27 minus VitA (Invitrogen) and 1 mM GlutaMAX (Invitrogen) for 1 day; the next day, the neurospheres were dissociated into single cells and plated on laminin/poly-d-lysine–coated plates for further experiments. For neuronal differentiation, retinoic acid (2 μM), SHH (50 ng/ml), purmorphamine (2 μM), or the combination of retinoic acid, SHH, and purmorphamine was supplemented in neural differentiation medium containing Neurobasal/B27 (NB/B27; Invitrogen), brain-derived neurotrophic factor (BDNF; 20 ng/ml; PeproTech), glial cell line–derived neurotrophic factor (GDNF; 20 ng/ml; PeproTech), ascorbic acid (0.2 mM; Sigma), dibutyryl adenosine 3′,5′-monophosphate (cAMP; 0.5 mM; Sigma) at the indicated time points after neurospheres were dissociated into single cells. For long-term neuronal culture, neural differentiation medium containing rat astrocyte–conditioned Neurobasal medium/B27, BDNF, GDNF, ascorbic acid, and dibutyryl cAMP was used for maintenance.

RNA sequencing

Two biological replicates of RONA colonies were dissected away from the supporting cells, disassociated, and plated. Twenty-four hours after plating, total RNA was harvested from the culture using TRIzol (Invitrogen) following the manufacturer’s protocol. Ribosomal RNAs were subtracted from total RNA using Ribo-Zero magnetic kit (Epicentre). Libraries from ribosomal RNA–subtracted RNA were generated using ScriptSeq 2 (Epicentre) and sequenced on an Illumina HiSeq 2500. Reads were mapped to the human genome (release 19) using TopHat2. Reads were annotated using a custom R script using the UCSC Known Genes table as a reference. Gene ontology (GO) shows output from a enrichment analysis on DAVID (Database for Annotation, Visualization, and Integrated Discovery;, searching the following categories: GO Biological Process, GO Molecular Function, PANTHER (Protein Analysis Through Evolutionary Relationships) Biological Process, PANTHER Molecular Function, and KEGG (Kyoto Encyclopedia of Genes and Genomes) and PANTHER pathway. Enrichment terms and associated genes and statistics presented with additional statistical analyses are also provided. P values presented in the text are corrected for multiple comparisons using the Benjamini method (column L).

Immunocytochemistry, immunohistochemistry, imaging, and quantification

Cultured cells were washed in phosphate-buffered saline (PBS) and fixed in 4% paraformaldehyde for 15 min. Attached RONAs and human cortical assemblies were sectioned (25 μm) using a cryostat (CM3050; Leica) and collected on SuperFrost Plus glass slides (Roth). Cells and sections were blocked with blocking buffer containing 10% (v/v) donkey serum and 0.2% (v/v) Triton X-100 in PBS. Then, they were incubated overnight at 4°C with primary antibodies diluted in blocking buffer, washed three times in blocking buffer, treated with secondary antibody (Invitrogen) for 1 hour, and washed three times in blocking buffer. After staining, coverslips were mounted on glass slides, and sections were coverslipped using ProLong Gold Antifade Reagent (Invitrogen). The primary antibodies used were human-specific NES (Millipore, MAB5326 and MAB5922), PAX6 (Covance, PRB-278P), FOXG1 (Abcam, ab18259), CTIP2 (Abcam, ab18465), SATB2 (Abcam, ab51502), BRN2 (Santa Cruz Biotechnology, sc-6029), TBR1 (Abcam, ab31940), DCX (Santa Cruz Biotechnology, sc-8066), TUJ1 (Millipore, AB1637), MAP2 (Sigma, M2320; Millipore, AB5622), nNOS (Sigma, N2880), VGLUT1 (SYSY, 135311), synapsin I (Millipore, AB1543), PSD95 (Abcam, ab2723), VGAT (SYSY, 131003), parvalbumin (Sigma, P3088), CB (Sigma, C9848), CR (BD, 610908), Somatostatin (Millipore, MAB354), GSX2 (Abcam, ab26255), NKX2.1 (TTF1, Epitomics, 2044-1), VIP (Sigma, V0390), NPY (Abcam, ab30914), prospero homeobox 1 (Abcam, ab37128), CK8 (Santa Cruz Biotechnology, sc-101459), LHX2 (Millipore, AB10557), GluN1 (SYSY, 114011), GluA1 (Epitomics, 1308-1), CAMKII (Cell Signaling Technology, 3362S), and GAD67 (Millipore, MAB5406).The following cyanine 2 (Cy2)–, Cy3-, and Cy5-conjugated secondary antibodies were used to detect primary antibodies: donkey antibody against mouse, donkey antibody against rat, donkey antibody against goat, and donkey antibody against rabbit (Invitrogen).

3D human cortical assembly assay

The cell culture insert was initially coated with 2% Matrigel, forming a gelled bed of basement membrane. Isolated NPCs were seeded onto this bed as a single-cell suspension at a high density of 0.75 × 106 to 1 × 106/cm2 in a neural differentiation medium containing NB/B27 (Invitrogen), 1% Matrigel, BDNF (20 ng/ml), GDNF (20 ng/ml), ascorbic acid (0.2 mM), and dibutyryl cAMP (0.5 mM). The neural differentiation medium was replaced every 3 days.

Oxygen-glucose deprivation

Dissolved O2 was first removed from glucose-free medium by bubbling with oxygen-glucose deprivation gas (5% CO2, 9.8% hydrogen, and 85.2% N2; Airgas Ltd.) for 30 min. Neurons were then washed with glucose-free medium. Oxygen-glucose deprivation was started by addition of glucose-free medium prebubbled with oxygen-glucose deprivation gas after incubation in a hypoxia chamber attached with an O2 sensor/monitor (Biospherix Ltd.) for the indicated times. Neuronal cultures were then refed with normal medium and maintained under normal culture conditions (5% CO2 and 20% O2 at 37°C).

Cell death assessment

Two- to three-month-old human cortical neurons were treated with NMDA or oxygen-glucose deprivation at various indicated time periods and doses. Percent of cell death was determined by staining with 5 μM Hoechst 33342 and 2 μM PI (Invitrogen). Images were taken and counted by a Zeiss microscope equipped with automated computer-assisted software (Axiovision 4.6, Zeiss). Z-VAD (20 μM; Sigma, V116), 20 μM NEC-1 (Sigma, N9037), 500 μM 3-MA (Calbiochem, 189490), 20 μM NPLA (Tocris, 1200), 30 μM DPQ (Enzo, ALX-270-21-M005), 500 nM AG-014699 (Selleckchem, S1098), 10 μM ABT888 (Active Biochem, A-1003), 2 μM olaparib (LC Laboratories, O-9201), and 20 nM BMN 673 (Selleckchem, S7048) were applied to evaluate the effect of different antagonists to known cell death signals.

Calcium imaging

Human cortical neurons were administrated with Fura-2 (Molecular Probes) for 30 min at room temperature. After washing, calcium imaging was conducted at 340- and 380-nm excitation in 2-month-old human cortical neuron cultures. Neurons were stimulated with 100 or 500 μM NMDA followed by MK801 to detect intracellular free calcium.

Measurement of NO production in culture medium

The NO production of 2-month-old human H1 cortical cultures after a 30-min treatment with NMDA or 2 hours of oxygen-glucose deprivation was assessed using the Measure-iT High-Sensitivity Nitrite Assay Kit (Life Technologies).

PARP-1 knockout in human cortical neuronal culture using CRISPR/Cas9

PARP-1 sgRNAs (#2, GTGGCCCACCTTCCAGAAGC; #3, ATACCAAAGAAGGGAGT-AGC) were synthesized and subcloned into lentiCRISPR vector (Addgene, pXPR_001, plasmid 49535). Lentiviral vectors were cotransfected into HEK293FT cells with the lentivirus packaging plasmids pVSVg and psPAX2 using FuGENE HD. Supernatants containing virus were collected 48 and 72 hours after transfection, passed through a nitrocellulose filter (0.45 μm), and applied on cells in culture. At 8 to 12 weeks after differentiation, human cortical neurons were transduced with lentiCRISPR lentivirus carrying control sgRNA or PARP-1 sgRNAs. Cells were then challenged with NMDA or oxygen-glucose deprivation at 5 days after transduction, and cell death was assessed by PI/Hoechst staining 24 hours later. Another batch of cells was collected at 5 days after transduction and analyzed by Western blotting for PARP-1 expression.

Western blot

SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and transfer were performed according to laboratory protocol with slight modifications. In brief, cultured cells were lysed in radioimmunoprecipitation assay buffer containing 1% Triton, 0.5% Na-deoxycholate, 0.1% SDS, 50 mM NaF, 10 mM Na4P2O7, 2 mM Na3VO4, and EDTA-free protease inhibitor mixture (Roche Diagnostics) in PBS (pH 7.4). Protein extracts were separated by 4 to 12% SDS-PAGE. Proteins were transferred at constant voltage of 80 V for 150 min at 4°C from the SDS to polyvinylidene difluoride membranes. The membranes were then blocked with 5% nonfat milk and incubated with primary antibodies overnight at 4°C. The antibodies used were anti-PAR (Trevigen, 4336-APC-050), actin (Cell Signaling, 5125), PARP-1 (BD, 611039), and β-actin (Cell Signaling Technology, #5125S). After washes with TBST (tris-buffered saline with 0.1% Tween 20), the membranes were incubated with horseradish peroxidase–conjugated secondary antibodies for 1 hour at room temperature. Immunoreactive bands were visualized by the enhanced chemiluminescent substrate (Pierce) on x-ray film and quantified using the image software TINA.

Electrophysiological recordings

Whole-cell voltage-clamp recordings were obtained with whole-cell configuration using an EPC-10 amplifier (HEKA Elektronik). Patch-clamp recordings from human ESC-derived neurons were made using recording electrodes of 8 to 10 megohms pulled from Corning Kovar Sealing #7052 glass pipettes (PG52151-4, WPI) by a Flaming-Brown micropipette puller (Sutter Instrument Co.). The extracellular solution contained 125 mM NaCl, 2.5 mM KCl, 25 mM d-glucose, 25 mM NaHCO3, 1.25 mM NaH2PO4, 1 mM CaCl2, and 1 mM MgCl2, and 5% CO2 and 95% O2 bubbled (pH 7.4). Intracellular solutions contained 135 mM KMeSO4, 8 mM NaCl, 2 mM ATP-Mg, 0.1 mM Na3GTP, 0.3 mM EGTA, and 10 mM Hepes (pH 7.2). Action potential recordings were obtained with current clamp configuration. Spontaneous mEPSCs were obtained with voltage-clamp configuration, the membrane potential was held at −70 mV, and 100 μM picrotoxin and 0.5 μM tetrodoxin were in the extracellular solution; for iPSC recording, 10 μM CNQX and 10 μM MK801 were in the extracellular solution. Spontaneous mEPSC and mIPSC recordings were acquired and stored digitally using the data acquisition/analysis package PULSE (HEKA Electronik). Potential recordings were acquired at 6 kHz and filtered at 3 kHz. We used Mini Analysis 6 (Synaptosoft) to detect mEPSC and mIPSC events and create cumulative amplitude histograms of the pooled data.

Statistical analysis

Data are presented as means ± SEM. Statistical analysis was performed using GraphPad Prism and noted in the text and figure legends. Information on sample size (n, given as a number) for each experimental group/condition, statistical methods, and measures are available in all relevant figure legends. Unless otherwise noted, significance was assessed as P < 0.05.


Fig. S1. Comparison of human forebrain neuron differentiation method.

Fig. S2. Neural induction of human iPSCs by dual inhibition of SMAD signaling enhances the efficiency of neural conversion via the RONA method.

Fig. S3. Isolated FOXG1+ forebrain progenitors preserve excitatory and inhibitory neurogenic potential.

Fig. S4. Development regulates the excitatory/inhibitory network formation and synaptic scaling.

Fig. S5. NMDA induces human cortical neuron cell death.

Fig. S6. PARP activation in male versus female human ESCs.

Data file S1 (Excel file)


Funding: This work was supported by grants MSCRFII-0429 and MSCRFII-0125 to V.L.D., grant 2013-MSCRF-0054 to J.-C.X., grant 2013-MSCRF-0028 to J.F., grant 2014-MSCRFE-0758 to S.M.E, NIH/National Institute of Neurological Disorders and Stroke grant NS67525 and NIH/National Institute on Drug Abuse grant DA00266 to T.M.D. and V.L.D. T.M.D. is the Leonard and Madlyn Abramson Professor in Neurodegenerative Diseases. Author contributions: J.-C.X., J.F., T.M.D., and V.L.D. designed the experiments. J.-C.X. and J.F. differentiated cortical neurons from human ESCs and iPSCs. J.-C.X. and J.F. performed immunohistochemistry and confocal microscopy studies and carried out neurotoxicity studies on the neurons derived from human ESCs or iPSCs. X.W. carried out electrophysiology studies. X.W. and X.Y. analyzed electrophysiology data. S.M.E carried out the RNA sequencing experiment and analyzed the RNA sequencing data. T.-I.K., L.C., J.Z., Z.C., H.J., and R.C. analyzed the data. J.-C.X., J.F., T.M.D., and V.L.D. wrote the manuscript. Competing interests: The authors declare that they have no competing interests.

Stay Connected to Science Translational Medicine

Navigate This Article