Although numerous protocols have been developed for differentiation of neurons from a variety of pluripotent stem cells, most have concentrated on being able to specify effectively appropriate neuronal subtypes and few have been designed to enhance or accelerate functional maturity. Of those that have, most employ time courses of functional maturation that are rather protracted, and none have fully characterized all aspects of neuronal function, from spontaneous action potential generation through to postsynaptic receptor maturation. Here, we describe a simple protocol that employs the sequential addition of just two supplemented media that have been formulated to separate the two key phases of neural differentiation, the neurogenesis and synaptogenesis, each characterized by different signaling requirements. Employing these media, this new protocol synchronized neurogenesis and enhanced the rate of maturation of pluripotent stem cell-derived neural precursors. Neurons differentiated using this protocol exhibited large cell capacitance with relatively hyperpolarized resting membrane potentials; moreover, they exhibited augmented: 1) spontaneous electrical activity; 2) regenerative induced action potential train activity; 3) Na+ current availability, and 4) synaptic currents. This was accomplished by rapid and uniform development of a mature, inhibitory GABAA receptor phenotype that was demonstrated by Ca2+ imaging and the ability of GABAA receptor blockers to evoke seizurogenic network activity in multielectrode array recordings. Furthermore, since this protocol can exploit expanded and frozen prepatterned neural progenitors to deliver mature neurons within 21 days, it is both scalable and transferable to high-throughput platforms for the use in functional screens.
- neural differentiation
- induced pluripotent stem cells
- embryonic stem cells
- patch clamp
- neuronal maturation
the ability to derive induced pluripotent stem cell (iPSC) lines directly from human patients and healthy subjects has enormous implications for biomedical research (37). Of particular interest is the prospect of using patient-derived stem cell lines to derive human in vitro models of disease, to use these to research disease mechanisms in the context of human disease genetics, and to develop assays with the highest disease and human relevance for drug discovery (9). iPSC disease modeling has become especially prominent in the field of neuroscience. First, patient-derived iPSCs have been used to model neurodevelopmental disorders (including Down's, Rett, and Fragile X syndromes; Refs. 11, 15, 16), where the effects of disease on the proper differentiation and maturation of neural lineages have been studied. Second, there has been extensive modeling of neurodegenerative disorders (including Alzheimer's disease, Parkinson's disease, Huntington's disease, and amyloid lateral sclerosis) (5, 9, 21a, 23, 27, 33). Models have been validated by phenotyping a spectrum of cell, molecular, and metabolic features associated with neurodegeneration. Most significantly for models of neurodegeneration, disease-associated phenotypes have been seen to manifest in weeks in the context of in vitro cultures, in spite of the fact that the diseases may take decades to manifest in vivo. Third, there is growing interest in iPSC models of neuropsychiatric disorders, such as schizophrenia, bipolar, and autism spectrum disorders (8, 22, 31, 36, 39), where there is even greater emphasis on synaptic function and physiological endpoints.
While reprogramming technologies have enabled significant resources of patient-derived iPSC lines to be derived and banked, for effective disease modeling their utility is entirely dependent on the quality of the protocols used to differentiate them into the somatic cells of disease relevance. In the case of neural differentiation, numerous protocols for neural induction and neural progenitor patterning, or neural subtype specification, have been developed and many succeed in generating neurons with characteristic morphology and the appropriate expression of multiple neuronal markers (13). However, in the majority of cases, the extent of functional neuronal maturation with respect to physiological endpoints is poorly described. Most commonly, examples of induced action potential recordings are shown, and examples of spontaneous networked activity are only reported from long-term cultures differentiated and maintained for many weeks, typically involving coculture of human pluripotent cell (hPSC)-derived neurons with mouse astrocytes or astrocyte-conditioned medium (21, 35, 45, 47). Moreover, problems associated with culture heterogeneity and time-in-culture severely compromise the use of iPSC-derived neuronal models for reproducible higher content and higher throughput phenotyping (29, 42, 48, 49).
Here, we have addressed the shortcomings of current neuronal differentiation protocols and developed new defined media that generate highly mature neurons from iPSC-derived neural progenitors, with high efficiency and reproducibility, within only 21 days of plating and without the use of astrocyte coculture or astrocyte conditioned medium. Critical pathways that control neurogenesis and synaptogenesis include regulated cell cycle exit and Notch pathway inhibition, with GABAA, CREB, and WNT pathway activation. These have been manipulated using small molecules to produce two chemically defined culture media, which, when applied sequentially, synchronize progenitor cell neurogenesis to promote and accelerate synaptogenesis and neuronal maturation within functional networks.
MATERIALS AND METHODS
hiPSC Culture and Neural Differentiation
Studies were performed using two feeder-free human iPSC lines: 34D6, derived using integrating reprogramming vectors (gift from S. Chandran, Edinburgh, UK; Ref. 6), and 33Qn1, derived using nonintegrating reprogramming vectors (gift from C. N. Svendsen, Cedar Sinai, Los Angeles, CA; Ref. 33). iPSCs were cultured on Matrigel-coated plates (BD Biosciences, Oxford, UK) in mTeSR1 medium and expanded by passaging using dispase following the manufacturer's protocols (Stem Cell Technologies, Cambridge, UK). For neural induction, iPSC cultures were grown to 70% confluence and washed three times with phosphate-buffered saline (PBS; Invitrogen Life Technologies, Paisley, UK) and the medium was exchanged for the neural induction SLI medium, which contained Advanced DMEM:F-12 (with Glutamax), 1% penicillin/streptomycin (all from Life Technologies), 10 μM SB431542 (Abcam, Cambridge, UK), 1 μM LDN 193189 (Stemgent, Cambridge, MA), 1.5 μM IWR1 (Tocris Bioscience, Abingdon, UK), and 2% NeuroBrew-21 without retinoic acid (RA) (Miltenyi Biotec, Bisley, UK). On day (D)4 of differentiation, confluent cultures were treated with 10 μM Y-27632 (Abcam) for 1 h before cell dissociation with Accutase (GE Healthcare Life Sciences, Little Chalfont, UK) and passaged onto fresh Matrigel-coated plates, with a split ratio of 1:2. On D8, cultures were passaged 1:2 and cultured in LI medium, which contained Advanced DMEM:F-12, 2 mM l-glutamine, 1% penicillin/streptomycin, 200 nM LDN 193189, 1.5 μM IWR1, and 2% NeuroBrew-21 without RA. Cultures received daily medium changes. D16 iPSC-derived neural progenitor cells (NPCs) were either used directly for neuronal differentiation (see below), frozen in Cryostor CS10 (Stem Cell Technologies) for later differentiations, or expanded in Matrigel-coated flasks in ADF-F medium, which contained Advanced DMEM:F-12 (with Glutamax), 1% penicillin/streptomycin, 2% NeuroBrew-21 with RA, and 20 ng/ml FGF2 (Miltenyi Biotec).
For neuronal differentiation, NPCs were dissociated using accutase and plated at a density of 1 × 105 cells/13-mm glass coverslips. Before use, coverslips were cleaned with 70% nitric acid, washed with deionized water and then with absolute ethanol, oven sterilized, and sequentially coated with 100 μg/ml poly-d-lysine (PDL in borate buffer; Sigma-Aldrich, Poole, UK). They were then washed three times in PBS before being coated with 1:100 Matrigel. D16 NPCs were then plated onto the coated coverslips and were differentiated for up to a further 35 days postplate-down (dpp), although the standard time course lasted for only 21 dpp (i.e., a total of 37 days from iPSC to functional neuron). The proliferation inhibitors N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester (DAPT; Sigma-Aldrich) and PD0332991 (Selleckchem, Newmarket, UK) were tested in neural differentiation medium that comprised Advanced DMEM:F-12 (with Glutamax), 1% penicillin/streptomycin, 2% NeuroBrew-21 with RA (Miltenyi Biotec), 200 μM ascorbic acid (Sigma-Aldrich), and 10 ng/ml brain-derived neurotrophic factor (BDNF; MACS; Miltenyi Biotec). For differentiation studies, three protocols were compared in this study: 1) SFDM4, 2) SCM-Base, and 3) SCM1/2. In all cases, other than switching from SCM1 to SCM2, half-medium changes were made every 2 days and cells were cultured in a humidified incubator in 5% CO2-95% air at 37°C.
To the plated D16 NPCs, SFDM4 medium was added, which contained Advanced DMEM:F-12 (with Glutamax), 1% penicillin/streptomycin, 2% NeuroBrew-21 with RA (Miltenyi Biotec), 200 μM ascorbic acid (Sigma-Aldrich), and 10 ng/ml BDNF (Miltenyi Biotec). DAPT (10 μM; Sigma-Aldrich) was added for the first 7 dpp.
To the plated D16 NPCs, SCM-Base medium was added, which contained Advanced DMEM:F-12 (with Glutamax), 1% penicillin/streptomycin, 2% NeuroBrew21 with RA (Miltenyi Biotec), 2 μM PD0332991 (Selleckchem), 10 μM DAPT(Sigma-Aldrich), 0.6 mM CaCl2 (to give 1.8 mM total CaCl2 in final complete medium; Sigma-Aldrich), 200 mM ascorbic acid (Sigma-Aldrich), 10 ng/ml BDNF, and 1 μM LM22A4 (Tocris Bioscience).
To the plated D16 NPCs, SCM1 medium was added for the first 7 dpp, which contained Advanced DMEM:F-12 (with Glutamax), 1% penicillin/streptomycin, 2% NeuroBrew21 (Miltenyi Biotec), 2 μM PD0332991 (Selleckchem), 10 μM DAPT (Sigma-Aldrich), 0.6 mM CaCl2 (to give 1.8 mM total CaCl2 in final complete medium; Sigma-Aldrich), 200 μM ascorbic acid (Sigma-Aldrich), 10 ng/ml BDNF (Miltenyi Biotec), 1 μM LM22A4 (Tocris Bioscience), 10 μM forskolin (FSK; Tocris Bioscience), 3 μM CHIR 99021 (Tocris Bioscience), and 300 μM GABA (Tocris Bioscience).
From 8 dpp, the NPCs were cultured in SCM2 medium, which contained 1:1 Advanced DMEM/F-12 (with Glutamax):Neurobasal A (Life Technologies), 1% penicillin/streptomycin (Life Technologies), 2% NeuroBrew21 with RA (Miltenyi Biotec), 2 μM PD0332991 (Selleckchem), 3 μM CHIR 99021 (Tocris Bioscience), 0.3 mM CaCl2 (to give 1.8 mM total CaCl2 in final complete medium; Sigma-Aldrich), 200 μM ascorbic acid (Sigma-Aldrich), 10 ng/ml BDNF (Miltenyi Biotec), and 1 μM LM22A4 (Tocris Bioscience). These media form the basis of an accelerated maturation protocol (patent no. PCT/GB2014/053064, Neuronal Stem Cell Differentiation; P. J. Kemp, N. D. Allen, and Cardiff University), available as the SynaptoJuice kit via collaboration.
Western Blot Analysis
For CREB and ERK analysis, cells were rinsed with ice-cold PBS and then harvested by scraping directly into RIPA buffer (Sigma-Aldrich) containing complete EDTA-free protease inhibitors (Roche Life Science, Welwyn Garden City, UK) and PhosSTOP phosphatase inhibitors (Roche Life Science). Cell lysates were centrifuged at 120 g for 15 min at 4°C, and the supernatants were collected and quantified by the Pierce BCA assay (Life Technologies). Twenty micrograms of total protein were loaded per well into 4–12% Bis-Tris Bolt gels (Life Technologies) and electrophoresed at 165 V for ∼1 h in MOPS buffer. Proteins were transferred to nitrocellulose membranes at 100 V for 1 h and membranes were blocked in 5% milk-PBST (PBS plus 0.1% Tween-20; Sigma-Aldrich) for 1 h at room temperature with agitation. Membranes were probed with the appropriate primary antibodies at the dilutions and incubation time/temperatures defined in Table 1. Secondary antibodies (donkey anti-mouse IR800; Lorne Laboratories, Danehill, UK) and goat anti-rabbit 680 (Life Technologies) at 1/10,000 dilution were incubated for 1 h at room temperature before visualization of blots on the Li-Cor Odyssey system (Li-Cor Biotechnology, Cambridge, UK).
Cells were processed on 13-mm glass coverslips. They were fixed with fresh, 4% (wt/vol) paraformaldehyde in PBS for 15 min at 37°C and then washed three times with PBS. For all antibodies, other than those raised against the NMDA receptors, the samples were blocked and permeabilized using PBS containing 1% BSA (wt/vol), 0.1% Tween-20 for 1 h at room temperature. For the NMDA receptor antibodies, cells were blocked and permeabilized using 1% BSA (Sigma-Aldrich), 0.3% Triton X-100 (Sigma-Aldrich), 0.03% sodium-azide (Fluka BioChemike), and 5% normal goat serum (Vector Laboratories) for 45 min at room temperature before being incubated overnight at 4°C in the appropriate primary antibodies at dilutions defined in Table 1. After overnight incubation, primary antibodies were removed and the cells were washed three times for 5 min in PBS before a 1- or 2-h incubation with fluorophore-conjugated secondary antibody [Alexa Flour-488, goat anti-rabbit, Abcam; Cy3, donkey anti-rabbit, Jackson Immuno Research (JIR); Alexa Fluor 488, donkey anti-mouse, JIR no. 715-545-150; or Alexa Fluor 647, goat anti-rabbit; Molecular Probes, Eugene, OR] at room temperature in the dark. Nuclear staining employed Hoechst or DAPI stain at 1:5,000 in PBS. Coverslips were mounted in Vectashield Mounting Medium (Vector Laboratories, Peterborough, UK) and imaged using Olympus BX61 microscope with SIS F-view SSD camera and AnalySIS imaging software (Olympus, Southend-on-Sea, UK); coregistration images were obtained using a Leica TCS SP2 AOBS spectral or Leica SP5 TCS confocal microscopes.
Sholl analysis was performed on cells individually imaged for GFP expression after low efficiency transfection of pmax-GFP using Lipofectamine 2000 (Invitrogen, Life Technologies). Neurons were then fixed at 7 and 21 dpp with 4% (wt/vol) paraformaldehyde for 5 min at room temperature and washed three times with PBS. Coverslips were then mounted in Fluoromount (Sigma-Aldrich). Images of at least 50 neurons were taken for each condition and time point using an Olympus BX61 microscope with SIS F-view SSD camera and AnalySIS imaging software (Olympus). When required, multiple images were taken to capture the extent of single neuron processes and images were stitched together using Microsoft Image Composite Editor (Microsoft, Berkshire, UK). Images were converted to black and white by splitting the RGB channels of each image, and the green channel was selected in order for axons and neurites to be traced using the simple neurite tracer plugin of Fiji software (http://fiji.sc/Simple_Neurite_Tracer). With the use of this plugin, a line stack of the traces was created for subsequent Sholl analysis. Sholl analysis was performed using the Sholl analysis plugin (http://fiji.sc/Sholl_Analysis) with a starting radius of 5 μm and final radius at the furthest point of the cell in increments of 5 μm. Data relating to neurite length and intersection were also obtained using the simple neurite tracer plugin. Data are presented as scatterplots but also show mean ± SE. Normalized branching data (intersections per μm) are presented as bar graphs showing mean ± SE.
Scanning Electronic Microscopy
Dendritic spines were visualized by scanning electron microscopy (SEM). D16 NPCs (1 × 105) were plated on 13-mm plastic coverslips (Thermanox plastic coverslips; Nunc), which were fixed and processed for SEM imaging at 21 dpp. Samples were processed at the Electron Microscopy Unit of the Scientific and Technological Centers of the University of Barcelona, Spain. Images were taken using a JOEL Field Emission Scanning Electron Microscope (JSM-7001F).
Electrophysiological Recordings and Analyses
Voltage and current recordings were made using conventional patch clamp in the whole cell configuration (20) employing either Axopatch 200B or Multiclamp 700A amplifiers interfaced to a computer running pClamp 9 using a Digidata 1322A A/D interface (Molecular Devices, Sunnyvale, CA). All electrophysiological studies were performed at a controlled room temperature of 22 ± 0.5°C. Recordings were digitized at 10 kHz and low-pass filtered at 2 or 5 kHz using an 8-pole Bessel filter. The standard bath solution contained the following (in mM): 135 NaCl (Fisher Scientific, Loughborough, UK), 5 KCl (Fisher), 1.2 MgCl2 (Sigma-Aldrich), 1.25 CaCl2 (Sigma-Aldrich), 10 d-glucose (Fisher), 5 HEPES (VWR International); pH was adjusted to 7.4 using 5 M NaOH. The standard pipette solution contained the following (in mM): 117 KCl, 10 NaCl, 11 HEPES, 2 Na2-ATP (Sigma-Aldrich), 2 Na-GTP (Sigma-Aldrich), 1.2 Na2-phosphocreatine (Sigma-Aldrich), 2 MgCl2, 1 CaCl2, and 11 EGTA (Fisher); pH was adjusted to 7.2 with KOH.
Mean resting membrane potential (Vm) and spontaneous action potential (sAP) characteristics of the neurons were determined during 120-s gap-free recording periods in current-clamp mode (I = 0 pA). Neurons were then coded according to the type of sAP activity that they demonstrated, defined as follows: 1) full spontaneous action potentials (sAP-full, at least 1 excursion that overshoots 0 mV); 2) attempted spontaneous action potentials (sAP-attempted, significant excursions from resting Vm that did not reach 0 mV); or 3) quiescent (sAP-none, no significant excursions from resting Vm). This then allowed neurons to be categorized into the three separate groups for further detailed analysis. Once Vm and sAP activity had been recorded, current was injected to hyperpolarize Vm to approximately −70 mV before 1-s current injection steps were imposed from −10 to +180 pA in 10-pA increments to record induced action potential (iAP) activity, which was coded as iAP-none (no significant excursions from baseline during injection); iAP-attempted single (significant excursions from baseline that do not reach 0 mV); iAP-single (a single excursion that overshoots 0 mV); iAP-attempted train (several excursions but only 1 overshoots 0 mV); and iAP-train (several excursions, at least 2 of that overshoot 0 mV). Where iAP trains were recorded, a spike frequency analysis was performed. Input resistance was measured from the voltage difference induced by the −10-pA current step. Spike analysis was performed on the first iAP using Clampfit 9; threshold was determined as the peak of the third differential of voltage with respect to time during the upstroke of the action potential and all other parameters are as defined extensively elsewhere (4).
Na+ currents were recorded using a standard voltage-step protocol (holding potential of −70 mV followed by 80-ms steps from −120 to +80 mV in increments of 10 mV). For Na+ current inactivation curves, cells were stepped for 200 ms to prepulse voltages of between −120 and +80 mV in 5-mV increments before being stepped for 200 ms to the test potential of 0 mV. Note that pharmacological dissection of the voltage-activated currents was not attempted for logistical reasons but that cross contamination of maximal inward Na+ current by outward K+ current was negligible at the voltage at which they were measured (−30 and −20 mV); similarly, contamination of K+ current by Na+ current was negligible at the time and voltage at which they were measured (+60 mV between 180 and 200 ms), since Na+ currents were already inactivated (see Figs. 7 and 8). Cell capacitance and series resistance were measured and compensated; series resistance was compensated 60–90%. Pipette resistances were 8–10 MΩ when filled with the pipette solutions.
For activation and inactivation curves, conductance (G) was calculated by dividing current by the appropriate driving force (Vc − ENa), where Vc = command potential and ENa = +66.7 mV. G/Gmax was plotted against voltage and fitted with a Boltzmann equation using an iterative fitting routine in Microcal Origin: where Gmax is the extrapolated maximum conductance, V0.5 is the voltage corresponding to half the maximum conductance, and k is the slope factor.
Spontaneous miniature excitatory and inhibitory currents, and postsynaptic GABA and NMDA currents, were recorded using solutions modified form (17). The extracellular solution contained the following (in mM): 127 NaCl, 20 CsCl, 5 BaCl2, 2 CaCl2, 12 glucose, 10 HEPES, 0.0005 tetrodotoxin (TTX), and 0.02 glycine at pH 7.3 with NaOH. The intracellular solution contained the following (in mM): 175 N-methyl-d-glucamine (NMDG), 40 HEPES, 2 MgCl2, 10 EGTA, 12 phosphocreatine, 2 Na2ATP, 0.2 Na3GTP, and 0.1 leupeptin at pH 7.3 with H2SO4. These solutions set the Ecat and ECl at positions where at −40-mV inward cationic currents (NMDA) were detected as downward deflections and inward chloride currents were detected as upward deflections. Spontaneous miniature excitatory and inhibitory currents were recorded for at least 2 min. Evoked postsynaptic currents were induced by three 5-s pulses of either 100 μM GABA or 100 μM NMDA at 90-s intervals delivered locally to neurons by a gravity-driven rapid perfusion changer (RSC18; Biologic, Cedex, France). Data were analyzed offline using MiniAnalysis (Synaptosoft), Clampfit 10.3 (Molecular Devices), and GraphPad Prism 6 (GraphPad Software).
D16 NPCs were plated at a density of 150 × 103 cells/well onto poly-d-lysine coated, 24-well multielectrode arrays (MEAs) and cultured sequentially using the three differentiation protocols as described above. At 21 dpp, the medium was switched for a HEPES-buffered physiological solution containing the following (in mM): 130 NaCl, 3 KCl, 1 MgCl2 2 CaCl2, 10 glucose, and 10 HEPES-NaOH at pH 7.4, and the MEAs were placed onto the temperature-controlled (37°C) platform of a commercially available MEA workstation (Multichannel Systems, Reutlingen, Germany). Each well contained 12 electrodes, the activities of which were recorded simultaneously before, during, and following manual addition of agonists, antagonists, and toxins, as defined in the text and legends. Data were band-pass filtered between 100 and 2,000 Hz and finally sampled at 20 KHz using the acquisition software provided with the device. During recordings this software detected single unit activity online and live displays of raw data, spikes per second histograms, and detected spike waveforms were available. For analysis, raw data were converted to a HDF5 file format, which was imported into the MatLab environment in which custom-written analysis routines were coded. Spikes were redetected and sorted, and time stamps were extracted allowing the various plots shown here to be produced. Root mean square recording noise levels were ∼1.7 μV. Spikes were detected using a threshold method (usually 5 SD of the mean) that detected either rapidly rising or falling events. Amplitudes of clearly resolvable spikes uncontaminated by significant levels of false detections varied between 10 and 100 μV. The major source of this amplitude variation is likely to be the physical distance of each detected neuron from the recording site. Typically, spontaneous spiking activity could be detected on multiple electrodes within a single well and in 24-well plates.
The ratiometric Ca2+ sensitive dye fura-2 (Molecular Probes) was used for Ca2+ imaging experiments. Neurons on glass coverslips were incubated in 4 μM fura-2-AM (dissolved in cell culture medium) for 30 min at 37°C in a humidified incubator gassed with 5% CO2/95% air. Coverslips were placed in a perfusion chamber mounted upon an Olympus IX71 inverted microscope equipped with a Cairn monochromator-based epifluorescence system (Cairn Instruments, Faversham, UK) and were continuously superfused with standard extracellular solution (see above for composition). Imaging was conducted at a controlled room temperature of 22 ± 0.5°C. Fura-2 was alternately excited with light of 340 and 380 nm and images were captured at 510 nm using an Orca CCD camera (Hamamatsu Photonics, Welwyn Garden City, UK). Solutions, agonists, and antagonists were locally applied to the neurons using a rapid solution changer. Following background subtraction of the emission intensities evoked by each excitation wavelength, emission ratios (340/380) were calculated offline. A cell was deemed to have responded to any stimulus if the emission ratio change was more than 10 times the standard deviation of the fura-2 signal in its preapplication period. Peak intensity changes from baseline in response to applied agonists were calculated for each individual cell. Averaged values obtained from independent experiments were used to calculate the mean intensity.
All summarized data are expressed as mean ± SE. Statistical comparisons of the means were performed using one-way or two-way analysis of ANOVA with Tukey post hoc test, or Student's t-tests, as appropriate and as stated in text and figure legends; differences were considered significant at P < 0.05.
In vitro differentiation and maturation protocols of human neurons were developed using D16 NPCs derived from the iPSC line 34D6 and further validated using the 33Qn1 line. NPCs from each line were derived by culture in defined medium with dual SMAD and WNT inhibition to enrich for NPCs with forebrain fate potential (30). At D8, cultures were characterized by high expression of SOX2, PAX6, and NESTIN, and FOXG1 with a high incidence of neural rosette formation identified by foci of zona occludens-1 (ZO-1) immunoreactivity at D16 (data not illustrated).
With the use of a standard protocol for neuronal differentiation, D16-dissociated iPSC-derived NPCs plated on PDL/Matrigel-coated coverslips and cultured in SFDM4 (25) showed extensive neurite outgrowth and expression of β-III-tubulin after 21 dpp. However, as commonly observed, after 21 dpp, cultures were seen to increase in complexity with significant clustering of cells, with apparent cell overgrowth, often compromising cultures for use in developmental or electrophysiological and imaging studies (data not illustrated).
Inhibitors of Cell Cycle and Notch Signaling Force Cell Cycle Exit and Synchronize Neurogenesis
To synchronize neuronal differentiation we first determined whether blockade of Notch signaling and the cell cycle checkpoint cyclin-dependent kinases CDK4 and CDK6 would promote cell cycle exit. After D16 34D6 NPCs were plated, blockade of Notch signaling with the γ-secretase inhibitor DAPT (10 μM) significantly reduced the proportion of cells that stained for MKI67, a marker of cell proliferation, from 21.8 ± 1.5 to 12.9 ± 1.8% at 3 dpp (P < 0.05), and from 47.2 ± 3.6 to 25.7 ± 10% at 7 dpp (P < 0.001; Fig. 1, A and B). Treatment with 2 μM PD0332991, a selective CDK4/6 inhibitor (50), resulted in a greater loss of MKI67 immunopositive cells at 3 dpp (to 0.9 ± 0.6%; P < 0.001) and 7 dpp (to 7.9 ± 4.7%; P < 0.0001, Fig. 1, A and B). In combination, γ-secretase and CDK4/6 inhibition resulted in a significant and synchronous sustained cell cycle exit at 7 dpp (to 4.2 ± 0.5%; P < 0.0001; Fig. 1, A and B). Treatment with DAPT and PD0332991 also enhanced neural differentiation, as evidenced by an increase in the proportion of cells expressing the neurofilament protein β-III-tubulin at 7 dpp (control: 64.3 ± 1.6%; DAPT/PD0332991: 84.6 ± 3.6%; P < 0.001) and a marked reduction in the level of expression of the neural progenitor marker NESTIN (control: 92.9 ± 1.9%; DAPT/PD0332991: 32.1 ± 3.7%; P < 0.0001), compared with NPCs cultured in the absence of DAPT and PD0332991 (Fig. 1C). Thus SCM1 medium was formulated to include these two small molecules.
SCM1 Enhances Neurogenesis
In search of the cellular basis for the well-documented benefit of astrocyte coculture to neuronal differentiation (19, 26, 40, 51), we have previously identified excitatory GABAA-dependent depolarization and consequent voltage-activated Ca2+ entry as critical to the process (45). GABAA receptor activation leads to Ca2+-dependent CREB and ERK phosphorylation, which in turn plays a central role in regulating neurogenic gene expression (34, 52). Functional expression of neuronal voltage-gated Na+ and K+ channels during differentiation and maturation can also be regulated by agents such as the adenylyl cyclase inhibitor FSK that directly elevate cAMP (1, 43). Thus SCM1 was formulated to include a higher total extracellular concentration of Ca2+ ([Ca2+]o) of 1.8 mM and 300 μM GABA to drive this GABAA-dependent signaling pathway and 10 μM FSK to augment tonically the intracellular cAMP concentration.
Consistent with developmental regulation of CREB during in vivo neural development (34), Western blot analysis of 34D6 cultures differentiated for 7 days showed high levels of CREB expression in NPCs that decreased with differentiation in all media (P < 0.01; Fig. 2, A and B). Differentiation was associated with CREB phosphorylation, and significantly neurons differentiated in SCM1 showed the greatest increase in the phosphorylation of available CREB (phosphoCREB:CREB: P < 0.05; Fig. 2, A and C). Differentiation did not affect total ERK1/2 levels (Fig. 2D); however, greater levels of sustained and selective phosphorylation of ERK2 were also seen in SCM1 cultures (P < 0.05; Fig. 2, A and E). Together, these data show a greater activation of CREB and ERK2 signaling in neurons differentiated with SCM1.
SCM1 and SCM2 Support Ongoing Neuronal Maturation
As GABA stimulation, leading to sustained CREB phosphorylation, has a transient developmental role in neurogenesis (24), GABA and FSK were removed from the second medium, SCM2, to allow the progression of neuronal maturation during the second phase of the protocol. Since the proneurogenic effects of Notch inhibition are required only for the early phase of neurogenesis and continued γ-secretase inhibition could impair neuronal function (46), DAPT was also removed from SCM2.
To identify further supplements, we performed a directed screen of growth factors and small molecules for their ability to enhance neuronal excitability as early as 14 dpp, measuring the proportion of cells demonstrating each type of induced action potential (iAP) activity (see Electrophysiology Recordings and Analysis). Surprisingly, little effect on iAP activity was seen in response to treatments with BDNF, FGF2, IGF1, or Activin A. However, the screen showed that addition of 3 μM of the GSK3β antagonist CHIR99021 increased the percentage of neurons firing single iAPs from 60 to 100% and increased those firing attempted iAP trains from 0 to 30% (data not illustrated). Therefore, 3 μM CHIR99021 was incorporated into the SCM1 and SCM2 media for further analysis.
Following differentiation for 21 dpp, SFDM4 cultures showed characteristic cell clumping and neural progenitor overgrowth that contained 64 ± 7% NESTIN+ cells and retained a significant 12 ± 1.4% population of MIK67+ proliferative cells, with just 32 ± 3.7% showing expression of the postmitotic neurofilament MAP2. In contrast, cultures differentiated in SCM-Base or SCM1/2 media maintained a more monodispersed neuronal morphology, with very few MKI67+ cells (<0.5%) indicating robust cell cycle exit, and the majority of cells expressed MAP2 (>95%) and the neurotransmiter GABA (Fig. 2, F and G). Interestingly, residual NESTIN expression was nevertheless seen at 21 dpp in SCM-Base-differentiated neurons (70.5 ± 10.1%) and was localized more to the soma than to neurites suggesting an immature neuronal phenotype. NESTIN staining was rarely seen in the SCM1/2 cultures suggesting greater developmental maturation using the SCM1/2 protocol.
To characterize further the morphology of neurons synchronized in development by differentiation in the SCM-Base and SCM1/2 conditions, individual cells were marked for analysis by low-efficiency GFP transfection and expression (Fig. 2D). At 7 dpp, neurons differentiating in SCM-Base (which lacked FSK, GABA, and CHIR99021 but retained BDNF/LM22A for neurotrophic support) showed extensive neurite outgrowth characteristic of early primary neuritogenesis (3). In contrast, even at 7 dpp, neurons differentiated in SCM1 already exhibited consolidation of primary neurites for axonal elongation, demonstrating an accelerated developmental program in this medium. Specifically, neurons differentiated in SCM1 for 7 dpp exhibited significantly longer neurite outgrowth (P < 0.001) with lower number of intersections (P < 0.01), reduced primary neurite complexity close to the soma (Sholl plot of first 100 μm), and consequently lower normalized intersections per micrometer across the entire neuron (P < 0.0001). By 21 dpp, neurons differentiated in SCM-Base had also acquired the more mature phenotype (Fig. 2E).
SCM1/2 Protocol Promotes Robust and Rapid Electrophysiological Maturation of iPSC-Derived Neurons
To determine the functional benefits of differentiating human neural progenitors in SCM1/2 media, the electrophysiological characteristics were determined of neurons differentiated from 34D6 NPCs and a line that was derived using a nonintegrating reprogramming protocol (33Qn1). The idea was that comparison of functional readouts of these two lines could be compared, and if they were similar, the more useful nonintegrated lines, 33Qn1, would be then be used for the remaining functional studies. Both iPSC-derived NPCs were each differentiated using the three different protocols, SCM1/2, SCM-Base, and SFDM4, and functional readouts were determined and compared at 21 dpp.
In comparison to the two control media (SFDM4 and SCM-Base), differentiation in SCM1/2 of 34D6 iPSC yielded a much higher proportion of neurons that demonstrated spontaneous electrical activity (Fig. 3, A and B). Thus 34% (25/73) of SCM1/2 differentiated neurons were coded as sAP-full, with a further 36% (26/73) coded as sAP-attempted. In complete contrast, only 5% (1/19; SCM-Base) and 0% (0/19; SFDM4) of control differentiated neurons were coded as sAP-full, with just 5% (1/19; SCM-Base) and 21% (4/19; SFDM4) coded as sAP-attempted. Strikingly, at 28 dpp, 100% of SCM1/2 neurons (17/17) were coded as sAP-full and this high level of activity was maintained through to 35 dpp (6/6); in SCM-Base, the proportion of cells with spontaneous activity remained at 0 from 28 to 35 dpp (Fig. 3A). These data are consistent with the observation that differentiation by each of the protocols generated populations of cells that had mean resting Vm values that were significantly different (one-way ANOVA, P < 0.05) and that SCM1/2 generated neurons with a relatively more hyperpolarized Vm than did SCM-Base (P < 0.05; Fig. 3C). However, rather modest values of resting Vm, and the differences observed between the populations as a whole, mask the real differences which could be observed when the populations were analyzed according to their sAP codes. Independently of protocol, neurons coded as sAP-full were strikingly and significantly more hyperpolarized than those coded as sAP-attempted (P < 0.0001) or sAP-none (P < 0.0001; Fig. 3C). Thus the mean resting Vm for SCM1/2 neurons coded as sAP-full was −46.9 ± 1.6 mV (n = 25), those coded sAP-attempted was −36.8 ± 1.41 (n = 31), and those coded sAP-none was −32.6 ± 1.71 mV (n = 53).
Similarly, 33Qn1 iPSC-derived neurons differentiated using SCM1/2 demonstrated a substantially augmented proportion of neurons showing spontaneous electrical activity than did those differentiated in either SCM-Base or SFDM4 (Fig. 4, A and B). Thus SCM1/2 generated 57% (39/69) of neurons, which were coded as sAP-full, with 16% (11/69) sAP-attempted and 28% (19/69) of sAP-none. Only 7% (2/29; SCM-Base) and 0% (0/21; SFDM4) of neurons were coded as sAP-full, with 38% (11/29; SCM-Base) and 10% (2/21; SFDM4) coded as sAP-attempted. Likewise, differentiation of 33Qn1 NPCs using each of the protocols generated populations of neurons that had mean resting Vm values that were significantly different (one-way ANOVA, P < 0.0001), with SCM1/2 generating neurons with a relatively more hyperpolarized Vm than those differentiated in SCM-Base (P < 0.0001; Fig. 4C). Analyses of 33Qn1 iPSC-derived neurons based on sAP codes demonstrated that sAP-full were significantly more hyperpolarized than those coded as sAP-attempted (P < 0.01) or sAP-none (P < 0.0001; Fig. 4C). Thus, mean resting Vm for SCM1/2 neurons coded as sAP-full was −47.5 ± 1.3 mV (n = 39), those coded sAP-attempted was −37.5 ± 3.4 (n = 11), and those coded sAP-none was −34.5 ± 1.6 mV (n = 19). The data obtained with Q33n1-derived neurons (Fig. 4) consistently mirrored those described above for 34D6 iPSC-derived neurons (Fig. 3), with SCM1/2 supporting an accelerated and more complete transition to functional maturity; a notion supported by the data shown in Figs. 3C and 4C, which plot resting Vm against sAP code for each different differentiation protocol at 21 dpp and demonstrate clearly how the SCM1/2 protocol generates many more sAP-full and sAP-attempted coded neurons than do the other two protocols.
The ability of the SCM1/2 protocol to generate a larger proportion of functionally active neurons than the control protocols was also dramatically reflected in the quantity and quality of iAPs. Once again, the SCM1/2 protocol was seen to promote a more mature neuronal phenotype, as demonstrated by the larger proportion of 34D6-derived neurons that exhibited iAP-train activity in the SCM1/2 protocol (59%, 43/73) than in either the SCM-Base (5%, 1/19) or the SFDM4 (11%, 2/19) protocols (Fig. 5A and exemplified in Fig. 5B); indeed, the modal pattern was iAP-single for SFDM4 (63%, 12/19) and SCM-Base (63% 12/19) but was iAP-train for SCM1/2 (59%, 43/73). Where iAP-trains could be recorded, spike frequency was much higher in 34D6 neurons differentiated in SCM1/2 than in either SCM-Base or SFDM4 (Fig. 5C), although there was only a modest decrease in mean input resistance (Rin) in the neurons differentiated in SCM1/2 media compared with SCM-Base or SFDM4 (data not illustrated). Interestingly, independently of which protocol was employed to differentiate the neurons, there was clear correlation between sAP code and iAP code for each individual neuron but, most importantly, SCM1/2 produced dramatically more sAP-full/iAP-train (100%) than did either SFDM4 or SCM-Base (Table 2).
Examining all neurons within each protocol that fired at least one iAP, deeper analysis of the first spike revealed a significant effect of differentiation protocol on overshoot, afterhyperpolarization, amplitude, depolarization rate, and half-width (P < 0.01 or lower by one-way ANOVA). Specifically, SCM1/2 generated neurons with significantly (P < 0.01) shortened half-width compared with SCM-Base neurons (Table 3). As was seen in the earlier Vm analysis (Fig. 3), some of these differences appeared modest when considering the entire population of neurons differentiated with each protocol. However, once neurons were analyzed according to their sAP code, there was, by one-way ANOVA, a significant effect of sAP code on all of the spike parameters, showing that sAP-full neurons had larger and faster action potentials than those coded sAP-none (Table 3). Since in 34D6-derived neurons SCM1/2 generated the highest proportion of sAP-full neurons, it is clear that the SCM1/2 protocol forced more cells to express a mature neuronal phenotype, synchronizing the maturation across a higher proportion of iPSC-neurons, than did the SCM-Base or SFDM4 protocols.
Again, in agreement with the data from 34D6-derived neurons, differentiating 33Qn1-derived NPCs using the SCM1/2 protocol increased the proportion of functionally active neurons compared with those differentiated using the control protocols. SCM1/2-treated 33Qn1 cells exhibited 90% (37/41) iAP-train activity, which was much higher than observed in SCM-Base (10%, 16/230) or in SFDM4 (37%, 7/19) (Fig. 6A and exemplified in Fig. 6B). Likewise, where iAP-trains could be recorded in 33Qn1-derived neurons, spike frequency was much higher in SCM1/2 than in either SCM-Base or SFDM4 (Fig. 6C). Again, independently of which protocol was employed to differentiate the 33Qn1 NPCs, there was clear correlation between sAP code and iAP code for each individual neuron and as previously seen in the 34D6 neurons, SCM1/2 produced dramatically more sAP-full/iAP-train than SFDM4 and SCM-Base (Table 4). Spike analyses of the neurons (Fig. 6, D and E) according to their sAP code by one-way ANOVA displayed a significant effect of sAP code mostly on amplitude parameters and depolarization rate of iAP, showing that sAP-full neurons had larger and faster action potentials than those coded sAP (Table 5).
In 34D6 neurons, there was a significant effect of differentiation protocol (P < 0.0001) on mean cell capacitance (Cp), with the SCM1/2 protocol generating neurons with significantly higher Cp, at 17.0 ± 1.1 pF, than those differentiated in either SCM-Base (6.4 ± 0.7 pF; P < 0.0001) or SFDM4 (10.3 ± 1.1 pF; P < 0.001). Surprisingly, neither Na+ nor K+ current densities were significantly affected by the differentiation protocol (Fig. 7, A and B). However, the SCM1/2 protocol generated neurons that displayed significant differences in the activation/inactivation profiles of the voltage-activated Na+ currents, resulting in SCM1/2 neurons exhibiting larger availability windows with high G/Gmax maxima and a higher proportion of neurons with Vm values falling within those windows than did either SCM-Base or SFDM4 neurons (Fig. 7C). Strikingly, neurons differentiated by the SCM1/2 protocol exhibited a very small mean difference between half-activation and half-inhibition voltages (Va50 − Vi50 = 7.1 ± 1.2 mV; an inverse measure of Na+ current availability), which was significantly lower than values for either SCM-Base (23.6 ± 2.4 mV; P < 0.0001) or SFDM4 neurons (17.9 ± 2.9 mV); SCM-Base and SFDM4 were not different.
Similarly, 33Qn1-derived neurons differentiated using any of the three protocols also resulted in no significant differences in the maximal voltage-activated Na+ current densities of the neurons (Fig. 8, A and B). Likewise, the maximal K+ current densities of 33Qn1 neurons differentiated using SCM1/2 and SCM-Base were not significantly different. However, the maximal K+ current densities of 33Qn1 neurons differentiated using SFDM4 was significantly smaller, which may reflect the lower variance observed in these experiments. Differences in voltage-activated current magnitudes of the 33Qn1 neurons notwithstanding, higher excitability in SCM1/2 was, like in the 34D6 cohort, accompanied by an increased percentage of neurons with Vm values fitting within the Na+ current availability window than in SCM-Base and SFDM4. (Fig. 8C).
Taken together, these data suggest that the SCM1/2 protocol enhances functional maturation by two major biophysical enhancements: hyperpolarizing the resting Vm to enable higher spontaneous activity and increasing the Na+ current availability to facilitate regenerative iAP-train activity.
Spontaneous Neural Network Activity Develops by 21 Days of Differentiation in SCM1/2
In addition to the hyperpolarizing influence of SCM1/2 media, sAP activity may also reflect the rate and extent of synaptogenesis and functional neural network activity. To examine whether each protocol was able to support functional synaptogenesis, the generation of spontaneous synaptic currents and the exhibition of evoked postsynaptic currents and/or neurotransmitter-evoked changes in intracellular calcium concentration ([Ca2+]i) were determined in neurons differentiated from the two cell lines 34D6 and 33Qn1.
Neurons differentiated from either line using either the SFDM4 protocol for 21 dpp never exhibited miniature synaptic currents. The SCM-Base protocol only rarely supported the generation of GABAergic minis and then only in Q33n1 neurons (2 of 14 cells). In contrast, between 30 and 75% of neurons differentiated from either cell line by the SCM1/2 protocol exhibited robust GABAergic and modest glutamatergic minis (Fig. 9). Thus the mean amplitude of spontaneous GABAergic synaptic events recorded at −40 mV was 6.3 ± 0.8 pA (n = 3/11) and 13.9 ± 1.4 pA (n = 15/17), with interevent interval values of 3.7 ± 0.6 s (n = 3/11) and 0.5 ± 0.1 s (n = 15/17), for 34D6- and 33Qn1-derived neurons, respectively; these currents were effectively and reversibly blocked by 10 μM bicuculline (Fig. 9A). These observations are consistent with the almost uniform GABA immunostaining, as shown in Fig. 2G. The very low number of glutamatergic events observed in both lines differentiated using any protocol made robust analysis unreliable; this is not surprising since the prepatterning protocol using SLI/SL media was actually designed to generate a preponderance of GABAergic ventral forebrain NPCs.
GABA (100 μM) evoked inhibitory postsynaptic currents in 100% of 33Qn1-derived neurons differentiated in SCM1/SCM2, SCM-Base, or SFDM4 with mean current densities at −40 mV of 50.5 ± 6.7 pA/pF (n = 16/16), 50.6 ± 5.8 pA/pF (n = 13/13), and 29.0 ± 4.8 pA/pF (n = 9/9), respectively (Fig. 9B). Similarly, 100% of 34D6-derived neurons differentiated using SCM1/2 showed inhibitory postsynaptic currents with mean current densities at −40 mV of 41.1 ± 5.3 pA/pF (n = 18/18) (data not shown).
In 33Qn1-derived neurons differentiated using the SCM1/2, SCM-Base, or SFDM4 protocol, 5-s applications of 100 mM NMDA/10 mM glycine evoked postsynaptic currents with mean current densities at −40 mV of −2.4 ± 0.6 pA/pF (n = 15/16), −1.5 ± 0.3 pA/pF (10/13), and −1.1 ± 0.3 pA/pF (4/12), respectively (Fig. 10A). In comparison, after differentiation in SCM1/2, the mean current density of the NMDA-evoked postsynaptic currents in 34D6 neurons was −1.7 ± 0.5 pA/pF (n = 3/4) (data not shown). Further evidence for NMDA receptor expression is shown in Fig. 10B, which shows punctate, presumably synaptic, staining of NR1, NR2A, and NR2B NMDA receptor subunits. The idea that the SCM1/2 protocol is prosynaptogenic was supported by the SEM data that showed highly complex arborizing neuronal networks, demonstrating distinct dendritic spines at high power (Fig. 10C), and by extensive staining of the pre- and postsynaptic markers synaptophysin and PSD95, with evidence of their coregistration indicating apposition of pre- and postsynaptic terminals (Fig. 10D).
SCM1/2 Protocol Promotes Maturation of Calcium Signaling in iPSC-Derived Neurons
Concentrating on the nonintegrating iPSCs, Ca2+ imaging of 33Qn1 neurons that had been differentiated using the three different protocols revealed markedly faster maturation rates in cells differentiated using the SCM1/2 protocol (Fig. 11, A and B). The neurons were imaged at different time points after plating, and the responses to high potassium (60 mM KCl, a depolarizing stimulus), 300 μM glutamate/30 μM glycine, and 300 μM GABA in normal and low (7.5 mM) extracellular chloride were recorded (sample traces for SCM1/2 neurons are presented in Fig. 11C). At the day of the plating, 0 dpp, only 10.9 ± 3.8% of cells responded to high potassium, 5.7 ± 5.7% of cells responded to glutamate, and 1.0 ± 1.3% of cells responded to GABA in low chloride (Fig. 11D, D0, n = 4). Even by the end of week 1 postplating (7 dpp), the vast majority of the cells from all three protocols responded to the high potassium challenge, demonstrating the early functional expression of voltage-gated Ca2+ channels at this time point (n = 6–11; Fig. 11A, top). However, by week 2 the intensity of the responses to high potassium was significantly lower for neurons differentiated using both SFDM4 and SCM-Base neurons than it was for neurons differentiated using SCM1/2 (n = 8–13), and the difference remained significant through to 3 wk (n = 7–14; Fig. 11B, top). The percentage of cells responding to 300 μM glutamate/30 μM glycine was significantly higher for SCM1/2 than for SFDM4 and SCM-Base neurons at weeks 1 and 3 postplating (Fig. 11A, middle), and the intensity of the responses was also significantly higher for SCM1/2 cohort than SFDM4 and SCM-Base at weeks 2 and 3 postplating (Fig. 11B, middle). The percentage of the cells responding to GABA in low extracellular chloride remained significantly lower for SFDM4 and SCM-Base treatments compared with SCM1/2 neurons during the whole differentiation period, and reached only 56.5 ± 12.0% for SFDM4, 72.0 ± 10.6% for SCM-Base neurons, but 98.8 ± 0.7% for SCM1/2 neurons at week 3 (Fig. 11A, bottom). The changes in the intensity of the responses to GABA in low chloride over the maturation period were similar to those seen in response to high potassium, being lower for both SFDM4 and SCM-Base neurons than for SCM1/2 neurons at weeks 2 and 3 postplating (Fig. 11B, bottom).
Since SCM1/2 supported much faster neuronal maturation, we performed day-to-day imaging of the responses during the first 10 days and then at 14 and 21 dpp for this cohort. Ca2+ imaging of 33Qn1 cells differentiated using the SCM1/2 protocol revealed that cells responded remarkably consistently to neuronal stimuli, producing an almost homogenous pattern of responses across the entire population by 21 dpp (n = 4–14; Fig. 11C). Specifically, the response to high potassium evoked Ca2+ influx in 36.2 ± 12.2% of cells at 1 dpp; this proportion rose to 97.6 ± 1.6% by 6 dpp and was at 99.8 ± 0.2% by 14 dpp, where it remained thereafter (Fig. 11D). Moreover, the magnitude of the Ca2+ influx also rose during differentiation and reached a plateau at 14 dpp (Fig. 11E). The proportion of cells responding to glutamate/glycine application also increased in a time-dependent manner from 3.50 ± 2.2% at 1 dpp to 96.9 ± 1.5% at 9 dpp (Fig. 11D). The magnitude of glutamate-evoked Ca2+ influx also increased with time to 21 dpp (Fig. 11E). Although the time-dependent changes in responses to GABA application were broadly similar to those evoked by glutamate or high potassium, the analysis of the responses was more complex. In physiological extracellular solution, a GABA-evoked rise in Ca2+ can be due to 1) GABAA-evoked depolarization-dependent opening of voltage-activated Ca2+ channels, and 2) GABAB-evoked Ca2+ release (38). In physiological solution, GABA evoked Ca2+ influx in 28.1 ± 12.6% of cells at 1 dpp, which rose to 95.1 ± 4.1% by 9 dpp. Although the magnitude of the GABA responses increased in a time-dependent manner, it was more than 10-fold lower than when GABA was applied in the presence of low chloride concentration, suggesting a major and increasing role for GABAA. In addition, lastly, despite the percentage of cells responding to GABA in low extracellular chloride remained ∼99% at 14 and 21 dpp, the percentage of cells responding to GABA in normal extracellular chloride dropped from 62 ± 9.4% at 14 dpp to 31.6 ± 9.8% at 21 dpp (P < 0.05), perhaps indicating increased amount of inhibitory GABA receptors signaling towards the end of the SCM1/2 maturation protocol (Fig. 11D).
Evidence that the neurons formed excitatory synaptic networks whose activity was limited by an ongoing inhibitory GABAergic tone, indicative of a mature GABAA phenotype, was provided by experiments using multiwell MEA plates (Fig. 12). Glutamate (3–15 μM) and NMDA (5 μM) both produced transient increases in neuronal firing when applied to neurons differentiated with the SCM1/2 protocol (Fig. 12, B and C, and Table 6). The transient nature was most likely due to depolarizing block. Application of 5 μM gabazine, a pharmacological blocker of GABAA receptors, resulted in a substantial and maintained increase in the rate of neuronal firing (Fig. 12D). Application of TTX (500 nM) eliminated all spiking activity (Fig. 12B and Table 6). In total, increased spiking activity was detected from 23 electrodes (4 wells) in the presence of glutamate, 29 electrodes (12 wells) in the presence of NMDA, and 20 electrodes (5 wells) in the presence of gabazine. Decreased activity was detected from 13 electrodes (5 wells) in the presence of TTX.
To address functional deficits commonly seen in hPSC-derived neurons differentiated under established neural culture conditions, we have established a new protocol that both synchronizes the neurogenesis of iPSC-derived neural progenitors and accelerates their functional maturation. This rapid prosynaptogenic, in vitro differentiation protocol generates authentic, mature synaptic networks of connected forebrain neurons from a variety of iPSCs. Taken together, the data presented herein show that, compared with standard neuronal differentiation methods, the SCM1/2 protocol enhances the rate and extent of synchronized maturation of iPSC-derived neurons (34D6 and 33Qn1), generating neurons with 1) accelerated neurite extension; 2) increased cell capacitance; 3) augmented CREB and ERK pathways activation; 4) relatively hyperpolarized resting Vm; 5) higher spontaneous action potential activity; 6) increased regenerative induced action potential train activity; 7) larger Na+ current availability; 8) augmented spontaneous GABA miniature synaptic currents; 9) large GABA and glutamate postsynaptic currents; and 10) rapid and uniform development of a mature, inhibitory GABAA receptor phenotype, which, when blocked, results in dramatically augmented neuronal network activity. However, although the SCM1/2 protocol provides optimal conditions for the differentiation and maturation of neurons, it does not address their long-term maintenance requirements. Recently, a novel medium named “BrainPhys” has been optimized for the long-term maintenance of neurons in culture (2) and this may usefully provide a third medium for use in conjunction with SCM1 and SCM2.
Since neurogenesis is a multistep process, characterized by sequential phases of differentiation and maturation, we aimed to improve NPC differentiation by developing two media for sequential use. Culture in SCM1 medium for 1 wk was designed to support better the early neurogenic phase of differentiation, after which SCM2 medium supported further neuronal maturation and synaptogenesis. SCM1 used as its base a medium similar to the classical DMEM:F-12 medium with N2 and B27 supplements (12). SCM2 used the same base-medium mixed 1:1 with Neurobasal A. Both SCM1 and SCM2 media included BDNF and a small molecule TrkB pathway agonist, LM22A4 (32), to provide neurotrophic support to differentiating neurons. Preliminary studies indicated poor neuronal survival at 21 dpp without neurotrophic support and that BDNF and LM22A4 promote neurite outgrowth and neuronal survival. Although LM22A4 could substitute for BDNF (data not shown), in our experiments we maintained both compounds in both media throughout the protocol.
To address the stochastic nature of NPC differentiation, and problems in defining neuronal age associated with continued NPC proliferation, we first synchronized NPC neurogenesis. A critical step in the transition from neural progenitor to neuron is the lengthening of the cell cycle, leading to cell cycle exit (28, 41); therefore, we used two small molecules, DAPT and PD0332991, to promote cell cycle exit and block cell cycle progression. Previous studies have used γ-secretase inhibitors (exemplified by DAPT) to block Notch pathway signaling (7). Notch pathway inhibition is known to activate proneural gene expression and indirectly reduce cell proliferation. Plating D16 NPCs in medium with 10 μM DAPT caused a significant reduction in cell proliferation; however, this was insufficient to synchronize cell cycle exit, as seen by the progressive rather than abrupt loss of MKI67 immunostaining over 7 days (Fig. 1A). To complement the action of DAPT we also used PD0332991, a small molecule inhibitor of the cyclin-dependent kinase 4/6 that blocks progression through the G1/S checkpoint and would therefore force cell cycle exit (10, 50).
Besides the suppression of Notch effectors that maintain a proliferative progenitor state, developmental mechanisms promote structural and functional programs of neuronal development and synaptogenesis. We previously analyzed the mechanistic basis for astrocyte-enhanced differentiation and maturation of iPSC-derived NPCs and identified critical roles for GABAA-dependent augmentation of voltage-gated calcium channel function (45). Since the neuronal maturation-promoting properties of astrocyte conditioned medium could be mimicked by raising extracellular Ca2+ in ADF from 1.2 to 1.8 mM, and by providing 300 μM GABA to drive GABAA-dependent (bicuculine-sensitive) signaling, these additions were incorporated into the differentiation medium SCM1. GABAA receptor activation leads to Ca2+-dependent CREB and ERK phosphorylation, which in turn plays a central role in regulating neurogenic gene expression (34, 52). Importantly, the presynaptic phase of neuronal differentiation in vivo is characterized by tonic GABA stimulation and sustained Ser-133 phospho-CREB activity (24). Therefore, to help maintain stable levels of CREB pathway activation during the initial neurogenic phase of the differentiation protocol, SCM1 was also supplemented with 10 μM FSK, a small molecule agonist of the CREB pathway (1, 43). By following a developmental rationale for protocol design and media formulation, there was still scope for further enhancement of in vitro neuronal maturation allowing for further improvement, hence the inclusion of the prosynaptic small molecule antagonist of GSK3β CHIR99021.
In our experiments, neurons were essentially ventral forebrain, due to the use of SLI differentiation media during the first 16 days to rosette formation. However, the SCM1/2 media were not designed as a method by which to specify neuronal subtypes, rather they have been developed for accelerating and enhancing the differentiation of any prepatterned NPCs to produce functionally active neurons in the shortest possible time. In this regard, SCM1/2 protocol results in rapid neuronal differentiation and morphological maturation (Fig. 2, F–L) with expression of pre- and postsynaptic markers and the development of spine-like structures (Fig. 10, B–D).
Although there have been numerous neuronal differentiation protocols published to date, those that have systematically determined the functional characteristics of the resulting cell type have been surprisingly scarce, and none have integrated the information with MEA-based network analysis. Nevertheless, several robust and carefully performed studies have provided evidence for functional maturation using a variety of cell physiological endpoints (29, 48, 49).
One of the most complete basic electrophysiological analyses of hPSC-derived neurons was performed by Song et al. (48), which elegantly described the time courses of iPSC and excitatory synaptic current (ESC)-derived neuronal differentiation in terms of the many important determinants of neuronal maturation, including passive properties, iAP behavior, basic spike characteristics, voltage-gated currents, sAP behavior, and basic synaptic current properties. Although their data are in broad agreement with those presented herein, there are significant differences, the most important of which centre around the rate, extent, and uniformity of the functional maturation process. Thus, by 21 dpp, the Song protocol produced neurons of which only 18% exhibited repetitive iAPs, more similar to our SFDM4 protocol at 11%, whereas the SCM1/2 protocol generated a population of neurons with 59% (34D6) and 90% (33Qn1) able to fire repetitive iAPs (see iAP-train analysis above and Figs. 5 and 6). Even after a further week of differentiation, the Song protocol was still only able to support repetitive iAPs in 44.5% of the neurons; likely due to the lack of synchronization of the differentiation program by cell cycle exit as employed in the SCM protocol. Moreover, the action potential parameters obtained at 28 dpp in neurons differentiated by the Song protocol were, where available, reminiscent of a more immature neuronal phenotype compared with those of the SCM1/2 protocol at 21 dpp; specifically, they exhibited smaller spike amplitudes, extremely low depolarization rates, and long half-widths (see Tables 3 and 5 vs. Song et al. 2013 Table 1). Again, those spike parameters were more reminiscent of neurons generated at 21 dpp by the SCM4 and SCM-Base protocols than here with our SCM1/2 protocol (Tables 3 and 5). However, the Song neurons were particularly hyperpolarized, even at 14 dpp, an observation that was attributed to high and progressive upregulation of the M current seen during differentiation. Interestingly, although the SCM1/2 protocol was more efficient at hyperpolarizing the resting Vm than was the SFDM4 and tended to “push” a higher proportion of cells in the sAP-full category, our neurons can only be routinely hyperpolarized below −60 mV; this maneuver promotes sAP activity as Na+ current inactivation is removed, and the excitatory synaptic input is able to evoke action potentials. In the Song protocol, the reverse is true, with the 28 dpp neurons already hyperpolarized and what synaptic input is available is not able to induce sAP activity. Such a difference may be more a function of neuronal subtype being generated, since the Song NPC patterning protocol employs purmorphamine to ventralize, whereas our prepatterning employs the inhibitor of WNT response compound IWR1 to promote intermediate progenitor domain specification. Overall, although the Song protocol produces high-quality neurons, they take longer to differentiate, are more heterogeneous in their characteristics and fire spikes, which are still not fully mature at equivalent time points.
The Livesey protocol (29) was focused particularly on maturation and used several important readouts, including GABAA phenotype, immature excitatory vs. inhibitory mature, which is known in rodent models to depend developmentally on GABA-regulated Cl− cotransporters and subsequent [Cl−]i (18). In that study, the GABAA response of ESC and iPSC-derived forebrain neurons shifted from excitatory to inhibitory during the 7-wk differentiation, and this was due to a progressive reduction in ECl as the expression of Na+/K+/2Cl−cotransporter 1 (NKCC1) and K+/Cl− cotransporter 2 (KCC2) was, respectively, reduced and enhanced (44, 53). Unfortunately, no measures were reported between 1 and 5 wk of differentiation. In our study, using voltage-activated Ca2+ entry as an indirect indicator of GABAA receptor activity, there was a progressive increase in both the proportion of cells responding to GABAA, and in the magnitude of the individual responses, during the first 2 wk of differentiation (Fig. 11). By 21 dpp, blockade of GABAergic signaling with gabazine, causing disinhibition, resulted in a dramatic increase in neuronal network activity; all activity was eliminated by the addition of TTX (Fig. 12). In rodent models in vitro the functional GABA switch is dependent on GABA in the medium (18), a situation that is mimicked in the SCM1/2 protocol.
The Telias protocol (49) generated neurons only from ESCs but employed a similar induction and differentiation time course to that employed in the SCM1/2 protocol. Within 3 wk, neurons demonstrated significant maturation properties including a higher proportion of neurons able to generate iAP train activity. However, input resistance was consistently very high, at over 2 GΩ throughout, and iAPs had mean amplitudes of <40 mV, compared with <1 GΩ and ∼70 mV, respectively, using SCM1/2. Unfortunately, no details of Na+ current activation/inactivation were provided, although the K+ current densities appear comparable. Similarly, no indication of sAP activity was presented, although spontaneous synaptic currents are similar, and we have no objective measure of synchronization efficiency. Their use of DAPT would be expected to begin to synchronize the maturation of the derived neurons. However, as we show in Fig. 1, complete blockade of proliferation, at least in iPSC-derived cells, is only possible by the further addition of PD0332991. Nevertheless, with further characterization, the Telias protocol may well be shown to produce high quality, synchronized mature iPSC-derived neurons, but currently there is an incomplete dataset described which has only been performed on ESCs.
In conclusion, the use of SCM1 and SCM2 media separates two key phases of neural differentiation; the neurogenic phase and the synaptogenic phase, each characterized by different signaling requirements. These requirements have been met by formulation of two defined media that when used sequentially on plated PSC-derived NPCs lead to rapid neuronal differentiation and functional maturation. Crucially, this protocol uses expanded and frozen prepatterned neural progenitors to deliver mature neurons within 21 days, making it scalable and transferable to high-throughput platforms.
The CHDI Foundation generously provided funding for the majority of this work. An NC3R's Crack-It Challenge grant funded the MEA component of the work. The FP7 Grant Repair-HD contributed to the electrophysiological synaptic characterization. Grants from the Ministerio de Economia y Competitividad (SAF2012-37417 to J. M. Canals) and from the ISCIII-Subdirección General de Evaluación and European Regional Development Fund (ERDF) [RETICS to J. M. Canals (RD12/0019/0002; Red de Terapia Celular), Spain] funded the analysis of NMDA receptors and spine formation.
No conflicts of interest, financial or otherwise, are declared by the author(s).
Author contributions: V.T., P.L.Y., J.M.C., A.D.R., N.D.A., and P.J.K. conception and design of research; V.T., C.S., P.L.Y., S.Y., P.S., E.C., A.H., B.A.T., C.G., J.M.H., S.J., L.B., N.C.-R., A.C., M.S., G.B., A.D.R., N.D.A., and P.J.K. performed experiments; V.T., C.S., P.L.Y., S.Y., P.S., E.C., J.M.H., J.T.B., J.M.C., A.D.R., N.D.A., and P.J.K. analyzed data; V.T., P.L.Y., A.D.R., N.D.A., and P.J.K. interpreted results of experiments; V.T., C.S., P.L.Y., P.S., A.H., C.G., J.M.H., A.D.R., N.D.A., and P.J.K. prepared figures; V.T., N.D.A., and P.J.K. drafted manuscript; V.T., C.S., P.L.Y., S.Y., P.S., E.C., A.H., B.A.T., N.C.-R., A.C., M.S., G.B., J.M.C., A.D.R., N.D.A., and P.J.K. approved final version of manuscript; A.D.R., N.D.A., and P.J.K. edited and revised manuscript.
We thank Siddharthan Chandran and Dr. Bilada Bilican for providing the 34D6 iPSC and Clive Svendsen and Dr. Virginia Mattis for providing 33Qn1 iPSC line. We also thank Dr. Josep M. Rebled and Blanca Peguera from the Electronic Microscopy Unit at the Scientific and Technological Centers of the University of Barcelona for support and Dr. Adrian Waite for Western blot quantitation.
- Copyright © 2016 the American Physiological Society