Functional Maturation of Human Stem Cell-Derived Neurons in Long-Term Cultures

Differentiated neurons can be rapidly acquired, within days, by inducing stem cells to express neurogenic transcription factors. We developed a protocol to maintain long-term cultures of human neurons, called iNGNs, which are obtained by inducing Neurogenin-1 and Neurogenin-2 expression in induced pluripotent stem cells. We followed the functional development of iNGNs over months and they showed many hallmark properties for neuronal maturation, including robust electrical and synaptic activity. Using iNGNs expressing a variant of channelrhodopsin-2, called CatCh, we could control iNGN activity with blue light stimulation. In combination with optogenetic tools, iNGNs offer opportunities for studies that require precise spatial and temporal resolution. iNGNs developed spontaneous network activity, and these networks had excitatory glutamatergic synapses, which we characterized with single-cell synaptic recordings. AMPA glutamatergic receptor activity was especially dominant in postsynaptic recordings, whereas NMDA glutamatergic receptor activity was absent from postsynaptic recordings but present in extrasynaptic recordings. Our results on long-term cultures of iNGNs could help in future studies elucidating mechanisms of human synaptogenesis and neurotransmission, along with the ability to scale-up the size of the cultures.


Introduction
Human neurons derived from stem cells offer a multitude of possibilities, from studying developmental or functional properties of large populations of human neurons, to understanding a specific gene's effects in a single neuron, to producing cells and tissues for cell therapy and regenerative medicine applications. Acquiring human neurons from precious tissue samples is a slow process that is limited by healthy tissue availability. The ability to repetitively and rapidly produce a large amount of healthy human neurons for in vitro studies offers many advantages, and this was not possible before the advent of human embryonic stem cells (ESCs) [1] and induced pluripotent stem cells (iPSCs) in 2006 [2,3]. Typical methods for deriving differentiated and functional human neurons from stem cells require months [4], with multiple steps, and a resulting population of mixed neural cells, including neurons. The resulting cell heterogeneity and batch-to-batch variability make thorough characterization of the differentiated neurons difficult. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Faster methods for obtaining differentiated neurons, by inducible expression of specific neurogenic transcription factors in a single step, cause a rapid transformation of the stem cells into differentiated neurons [5,6]. The induction of Neurogenin-1 and Neurogenin-2 expression in human iPSCs was shown to differentiate neurons, called iNGNs, within only 4 days and produce a nearly pure population of neurons [7]. Induced expression of single transcription factors alone, such as Neurogenin-1 [5], Neurogenin-2 [8,9], ASCL1 [10], and NeuroD1 [11,12], has also been used to obtain differentiated neurons from stem cells.
Although iNGN cells have been extensively studied on the molecular level during neurogenesis and at early postmitotic stages ( [7] and ENCODE consortium), we were interested in how iNGN cells functionally mature over time. In this study, we characterized long-term cultures of iNGNs grown in specifically defined conditions. We confirmed that within 14 days iNGNs become electrically active neurons, as previously described with current injection studies [7]. However, it was unclear when and if iNGN cells would become robustly intrinsically functional. Therefore, our aim was to study their functional development at specific time points over the course of months. iNGN cells could survive well beyond 100 days under the specifically defined culture conditions. Patch-clamp electrophysiology recordings were used to assess measures of neuronal function over time, including: cell size, resting membrane potential, voltage-gated channel function, action potential firing properties, spontaneous synaptic activity, postsynaptic and extrasynaptic glutamatergic receptor function. We characterized suitable methods to get iNGNs to express the optogenetic tool channelrhodopsin-2 (ChR2). We also used immunocytochemistry to assess the expression of certain synaptic proteins. In many experiments, we compared data between iNGNs and primary rat hippocampal neurons cultured in the same conditions, to determine the neuronal properties and maturity of iNGNs both functionally and morphologically.

Materials and Methods
Human induced pluripotent stem cell culture and neuron differentiation The modified human induced pluripotent stem cells (iPSCs), called iNGN stem cells, were a kind gift from the lab of Dr. George Church (Harvard Medical School, Boston). The iNGN stem cells are modified PGP1 cells (GM23338, Coriell Institute for Medical Research, Camden, USA) [13], which are iPSCs derived from healthy adult male skin fibroblasts (GM23248, Coriell Institute for Medical Research) through reprogramming using the four Yamanaka transcription factors (Oct-4, Sox2, Klf4, and c-Myc). Lentivirus generation and modifications of PGP1 cells to produce iNGNs were previously published [7], resulting in doxycycline-inducible gene expression of the neuronal transcription factors Neurogenin-1 and Neurogenin-2 in iNGN stem cells. The ENCODE (Encyclopedia of DNA Elements) Consortium [14] has tested iNGN cells according to their standard procedures. The data are freely available (https://www.encodeproject.org, accession number: ENCBS369AAA). The iNGN stem cells were maintained in feeder-free cultures with mTeSR1 medium (Stemcell Technologies, Grenoble, France), without antibiotics, on 6-well plastic plates (Corning Costar, Amsterdam, The Netherlands) coated with Matrigel matrix (Corning). Medium was changed daily or every other day, including gentle phosphatebuffered saline (PBS) (Life Technologies, Darmstadt, Germany) washes, and passaging was done with TrypLE Express (Life Technologies). Cryopreservation of iNGN stem cell stocks was done using mFreSR (Stemcell Technologies). To induce neurogenesis iNGN stem cells were dissociated into single-cell suspension in TrypLE Express, and 6.3x10 4 cells were plated in neuronal growth medium (NGM = Neurobasal A medium + B27 + GlutaMAX-1; all from Life Technologies), supplemented with 5 μM InSolution Y-27632 (Rho-associated kinase inhibitor (ROCKi); Merck, Darmstadt, Germany) and 0.5 μg/ml doxycycline (Dox; Sigma-Aldrich, Taufkirchen, Germany), onto 14 mm Ø poly-D-lysine-coated (PDL, 100 μg/ml; Merck) glass coverslips (Menzel #1, Thermo Fisher Scientific, Darmstadt, Germany) with rat astrocyte cultures (see below for details). Therefore, the iNGNs were directly differentiated in NGM upon seeding onto coverslips. Viable cell counting was done with the Trypan blue exclusion method using a Luna-FL cell counter (Logos Biosystems, Anyang-si, South Korea). The next day the medium was changed to NGM supplemented with 0.5 μg/ml Dox, and on day 5 the medium was changed to NGM alone, without antibiotics. Half of the NGM was replaced once or twice per week. Neural induction and differentiation of iNGN stem cells was by Dox addition (considered day (d) 0) and the resulting neurons are referred to as 'iNGNs' throughout the paper. iNGN stem cells were used between passages 41-52, and our data comprise independent samples of cells that were pooled for analyses, with variability reflected in the error bars.

Rat primary hippocampal neuron culture
Rat procedures complied with German Animal Welfare Legislation as well as the guidelines issued by the Max Planck Society. The institutional animal welfare officer approved the hippocampal preparation from rat pups and all sacrificed animals were reported to the local government (Regierungspräsidium Darmstadt). All rats were maintained at the Max Planck Institute for Brain Research animal facility (Frankfurt, Germany). Maternal Sprague-Dawley rats (Crl: CD(SD), Charles River Laboratories, Sulzfeld, Germany) were received at day 18 after plug observation (timed matings at Charles River Laboratories) and were kept in the animal facility until litter delivery at day 23. The maternal rats were housed under specified-pathogen-free conditions in Type IV Makrolon cages at 20-24˚C room temperature, with room lighting set to a 12:12 hour light-dark cycle. Rats received a commercial irradiated diet (ssniff Spezialdiäten, Soest, Germany) and water ad libitum. Cage bedding was a sterilized commercial softwood granulate (Lignocel BK 8-15, J. Rettenmaier & Söhne, Rosenberg, Germany). Dissociated rat hippocampal neurons were prepared from 1-day-old Sprague-Dawley rat pups that were sacrificed, immediately after removal from the mother, by decapitation with sharp scissors. Hippocampi were isolated, pooled, dissociated with papain (Sigma-Aldrich), and 4.0x10 4 cells were plated onto PDL-coated coverslips in 24-well plates and maintained in NGM. No antibiotics were added to any neuronal (iNGN or rat hippocampal) cultures, since antibiotics negatively affect synaptic activity [19]. For AAV transductions, ChR2-containing AAV particles were added to the media after 5 days in vitro (DIV), without the addition of all-trans-retinal. AAV incubation time was between 14-21 days before recordings on 19-26 DIV neurons.
For agonist studies, AMPA, KA, and NMDA were bath applied and then washed, and an agonist response was defined as a response where the current returned to baseline levels after the wash step. AMPA and KA recordings were done in ACSF. For NMDA recordings, low-Mg 2+ ACSF containing 10 μM glycine was used, along with Cs-MeSO 4 pipette solution.
Current-clamp measurements were low-pass-filtered at a frequency of 5 kHz and sampled at 50 kHz (20 μs intervals). Resting membrane potentials were measured immediately after membrane rupture (whole-cell mode), and input resistance was measured as the steady-state voltage level in response to a negative current injection (between -10 and -100 pA).
Light stimulation of ChR2-expressing neurons was with a 473 nm laser (OptoTech, Wettenberg, Germany) coupled through a 400-μm diameter light fiber attached to a micromanipulator (NMN-21, Narishige, Tokyo, Japan). The light fiber tip was placed 10 μm above the cell body, and before each recording session light power was measured at the fiber tip to determine the light power density. Laser intensity was adjusted with neutral density filters. Laser input to the light fiber was controlled with a shutter driver (VCM-D1, Uniblitz, Rochester, USA) with a 2 mm aperture shutter (LS2, Uniblitz) connected to the digitizer, and light pulses were triggered through the Clampex software.
CNQX and PTX were from Santa Cruz Biotechnology (Dallas, USA), TTX was from Tocris (Bristol, UK), and bath and pipette solution chemicals, NMDA, S-AMPA, kainate, and glycine were from Sigma-Aldrich.

Data analysis
Patch-clamp recordings were analyzed with Clampfit 10.3 and OriginPro 2015G software. Single action potential properties were analyzed from the first AP at the rheobase (the injected current that results in the first AP) for each neuron. Membrane potential was held between -75 to -80 mV. Voltage changes were measured in response to 10 to 20 pA steps of 500 ms current injections, from -40 to +200 pA. Action potentials were defined as spikes that exceeded 0 mV. Amplitude was measured from the baseline to the peak, 'half-width' refers to the width at halfmaximal amplitude, and thresholds were determined from phase-plane plots (dV/dt vs. V) of the first AP at the rheobase.
To estimate the number of ChR2(L132C) channels in iNGN membranes the following equation was used: N = (current density x specific neuronal membrane capacitance)/(single channel conductance x holding potential x channel open probability). Where N is the number of channels/μm 2 , the specific neuronal membrane capacitance is 0.9 μF/cm 2 [20], the single channel conductance is 63 fS [16], the holding potential is -75 mV, the open probability is 0.4 [16], and the current density is the measured average photocurrent density.
For spontaneous currents, continuous 4-minute recordings of currents were at -75 mV. The amplitude cutoff value was 5 pA in ACSF containing 1.3 mM Mg 2+ . 10-90% rise times, decay time constants (τ), and charge movement ΔQ (pAÁms) were determined using Clampfit, with decay τ's determined from a single exponential fitting of the decay portion of the mEPSC. Averaged mEPSCs for each neuron and condition were obtained by aligning the data at 0 mV. Data are reported as averages (means) ± standard error of the mean (SEM). For data sets containing only two groups, statistical comparisons were performed using a paired or unpaired Student's ttest, as appropriate. For data sets containing more than two groups, statistical comparisons were performed using one-way ANOVAs, followed by unpaired two-tailed t-tests using a Bonferroni correction factor for multiple comparisons, and P < 0.05 was considered statistically significant.

Intracellular labeling
iNGNs, grown exactly as for electrophysiology experiments, were filled with Neurobiotin (Vector Laboratories, Burlingame, USA) using a patch-clamp glass pipette containing 0.5% Neurobiotin in the pipette solution, for 20-45 minutes, and 200 ms depolarizing 500 pA current pulses at 1 Hz for 2 min. The pipette was very slowly raised from the cell body and then the coverslip was transferred to a 24-well plate and gently washed in PBS for 5 min before being fixed with 4% paraformaldehyde in PBS containing Mg 2+ and Ca 2+ for 20 min. Coverslips were washed 3x with PBS and then incubated in Elite ABC reagent (Vector Laboratories) for 2 hr. After washing 3x with PBS, the coverslips were incubated in DAB peroxidase substrate solution (Vector Laboratories) for 10 min, followed by one wash with ddH 2 O. The coverslips were mounted with Aqua-Poly/Mount onto slides and stored upright at room temperature.

Microscope imaging
Images of Neurobiotin-labeled iNGNs were taken on the same microscope used for patchclamp recordings using a 10x lens, charge-coupled device camera (uEye UI-2240-C, IDS Imaging Development Systems GmbH, Obersulm, Germany), a 0.5x adapter lens (Zeiss), and uEye 4.4 software. Confocal Z-stack images, compiled from images at 0.5-1 μm intervals, were taken with a 40x water immersion lens on a Zeiss LSM710 microscope using Zen 2009 software. For differential interference contrast (DIC) microscope images, pictures were taken with a 40x lens on a Zeiss Axio Vert.A1 FL-LED microscope using Zen software. Image contrast and brightness were adjusted in Adobe Photoshop CS6.

iNGNs in long-term astrocyte co-cultures
The ability to grow neurons in long-term culture is advantageous for studying fundamental questions of neurogenesis and synaptogenesis. Due to long differentiation time lines, typical functional studies on neurons derived from stem cells are done on relatively young neurons. Primary cultures of neurons from rodents have been grown for up to many months [21,22]. Here we applied primary culture techniques to stem cell-derived neurons to obtain healthy long-term cultures (see methods) and studied their function at the single-cell level.
In this way, long-term iNGN cultures were feasible for over 100 days. The protocol for conversion of iNGN stem cells into neurons for long-term culture included plating rat astrocytes (ACs) onto coverslips first, adding the stem cells and differentiating them with the addition of doxycycline (Dox), and removing Dox from the cultures at day 5 ( Fig 1A). Notably, the cells were directly differentiated in neuronal growth medium that contained no penicillin and streptomycin, which can affect synaptic activity [19]. 25,000 astrocytes and 63,000 iNGNs were seeded onto 14 mm coverslips, and this corresponds to a 1:2.5 ratio of astrocytes:iNGNs. Rat ACs cultured alone, under the same culture conditions as in Fig 1A except no iNGNs were Co-cultured iNGN cells with ACs led to longer and more numerous neural dendritic processes and bifurcations with increasing iNGN age in culture (Fig 1B). Intracellular labeling of single neurons with Neurobiotin elucidated the extent of bifurcations in older iNGNs compared to younger iNGNs (S1 Fig). 7d iNGNs had fewer bifurcations than 35d iNGNs (7d iNGNs, 10.8 ± 2.4 branches, n = 4; 35d iNGNs, 18.3 ± 2.0 branches, n = 7; P = 0.02, Student's t-test). If iNGNs were cultured without ACs they had poor viability, and only survived a maximum of 2 or 3 weeks (n = 7 attempts). iNGNs cultured alone had many dendritic processes that showed deterioration over time, with only remnants of iNGN dendrites and cell bodies visible at 28 days (S1 Fig), and the number of viable 28d iNGNs was 0 ± 0 (n = 4 independent samples). In comparison, in 28d and 70d iNGNs grown with astrocytes, the number of viable iNGNs was 18379 ± 2090 (n = 3 independent samples) and 15671 ± 1968 (n = 5 independent samples), respectively (S1 Fig). In whole-cell patch-clamp recordings, capacitance measurements, which are proportional to the area of cell membrane, are an indicator of cellular outgrowth. Comparing the average cell capacitance of age-matched 21d iNGNs, cultured either alone or with ACs, showed that the iNGNs without ACs have a smaller average cell capacitance (21d iNGNs without ACs: 13.5 ± 0.6 pF, n = 4; 21d iNGNs with ACs: 20.2 ± 0.6 pF, n = 92; P = 0.02, Student's t-test). Our data highlight the beneficial effects of astrocyte cocultures for iNGN long-term viability and cellular outgrowth.

Passive electrical properties of iNGNs
Whole-cell patch-clamp recordings allowed us to quantitate iNGN neuronal properties at the single-cell level. As well as elucidating function, patch-clamp recordings were used to assess passive electrical properties of iNGNs over the course of time, including cell capacitance, resting membrane potential (RMP), and input resistance (R i ). Over the course of 70 days, iNGN cultures showed typical signs of neuronal maturation, including an increase in average cell capacitance (Fig 1C, top graph), which corresponds to an increase in cellular outgrowth. This corresponds to the more numerous dendritic processes and bifurcations seen over time ( Fig  1B). The average cell capacitance values for 7d, 28d, and 56d iNGNs were 16.4 ± 0.6 pF (n = 100), 21.8 ± 0.8 pF (n = 11), and 40.6 ± 2.6 pF (n = 20), respectively. In comparison, the average cell capacitance values for undifferentiated iNGN stem cells and 13-15 DIV primary rat hippocampal neurons were 15.2 ± 2.9 pF (n = 9) and 32.5 ± 6.6 pF (n = 12), respectively. The average RMP (Fig 1C, middle graph) and R i values ( Fig 1C, bottom graph) for iNGNs decreased over 70 days. The average RMP values for 7d, 28d, and 56d iNGNs were -44.4 ± 1.6 mV (n = 17), -64.9 ± 3.8 mV (n = 14), and -76.5 ± 3.9 mV (n = 4), respectively. The average R i values for 7d, 28d, and 56d iNGNs were 1.8 ± 0.1 GO (n = 17), 1.3 ± 0.2 GO (n = 20), and 1.0 ± 0.1 GO (n = 13), respectively. Although the average R i appeared to increase for 70d iNGNs (Fig 1C, bottom graph), the R i of 70d iNGNs was not different compared to any of the other ages (P > 0.05, ANOVA). The larger cell capacitance, more negative RMP, and lower R i over 70 days follow the general trends observed during neuronal maturation [23,24,25]. On the other hand, undifferentiated iNGN stem cells had an average RMP of -42.8 ± 2.7 mV (n = 5) and R i of 3312 ± 786 MO (n = 9). These three parameters together are evidence for iNGN maturation over the course of 70 days.
Over the course of 70 days, iNGNs showed clear indications of functional maturation, including larger Na v and K v channel currents and increased firing frequencies. These are prerequisite properties for iNGNs to be able to form networks for neural communication.

Controlling iNGNs with light
Channelrhodopsin (ChR2) is a cation channel from the alga Chlamydomonas reinhardtii that is maximally activated by~470 nm blue light, resulting in cell depolarization [26], and it is one of the most common light-gated ion channels used as a tool to depolarize neurons [27,28]. Creating iNGNs that express ChR2, and other optogenetic tools, is a powerful combination. With the speed of deriving mature human neurons from iNGNs, which can also be functionally manipulated using light, there are numerous possibilities for using ChR2-expressing iNGNs. We used viral transduction to get iNGNs to express ChR2 in order to depolarize the neurons using blue light.
The viral transduction strategy we used was adeno-associated virus (AAV) transduction of the iNGNs. AAV particles were added to cultures of iNGNs at 1 day and again at 5 days after Dox addition (Fig 1A). When AAVs were added to older iNGNs (7d or 14d iNGNs), there were no fluorescent cells observed 2 weeks after transduction (n = 4 independent transductions). To see if there was an AAV serotype more effective for transducing iNGNs, we first used a "serotype kit" with AAVs 2.1, 2.5, 2.8, or 2.9, which all contain GFP. A minimum of 14 days after AAV addition to the cultures, we observed iNGNs for any green fluorescence. All four serotypes (1, 5, 8, and 9)   Long-Term Cultures of Stem Cell-Derived Human Neurons AAV2.1-ChR2-Venus expressed poorly in the iNGNs, with <1% expression over the course of numerous attempts (n = 10), and any of the few fluorescent iNGNs appeared unhealthy. A few of the fluorescent neurons that could survive patch-clamp recordings for several minutes at -75 mV, had 473 nm (blue) light-induced currents with a relatively low average steady-state photocurrent density of 2.2 pA/pF (n = 2) at a saturating light power density of 18 mW/mm 2 . The AAV2.1-ChR2(L132C)-mKateA and AAV2.2-ChR2(L132C)-2A-eGFP constructs expressed in a small percentage of iNGNs (~2%). When held at -75 mV, 15d -24d iNGNs expressing ChR2(L132C) had an average steady-state photocurrent density of 9.6 ± 0.7 pA/pF (n = 7) in response to saturating blue light power densities (20-40 mW/mm 2 ) (Fig 3A and 3B). This corresponds to an estimated channel density of~46 ChR2(L132C) channels/μm 2 . We also compared ChR2(L132C) function between iNGNs and primary cultures of rat hippocampal neurons. In 13-26 DIV rat hippocampal neurons, average steady-state photocurrent density under the same conditions was 11.3 ± 1.6 pA/pF (n = 9) (Fig 3B). Therefore, although AAV2.1-ChR2(L132C) transduced the iNGNs with much lower efficiency, compared to~80% of rat hippocampal neurons, the cells that did express ChR2(L132C) showed comparable expression levels to rat hippocampal neurons, and average photocurrent densities were not different when compared using a Student's t-test (P = 0.171). In current-clamp recordings, trains of 1 ms pulses of blue light could trigger action potentials with 100% accuracy up to 25 Hz in iNGNs expressing ChR2(L132C) (Fig 3C, left side). In a previous study, similar experiments done in rat hippocampal neurons expressing ChR2(L132C) reliably trigger action potentials up 50 Hz [16] (Fig 3C, right side), which is the maximal firing frequency of these neurons [30]. The maximal firing frequency for light-triggered action potentials is dependent upon the neuron's innate firing capability. We showed that iNGN average maximal firing frequency in response to long (500 ms) duration current injections (Fig 2A and 2C) increases with age, and a 500 ms duration light stimulation also elucidates an iNGN's firing capability (Fig 3A).
Using AAV transduction, we could get healthy iNGNs expressing ChR2(L132C), with photocurrent densities on par with those from rat hippocampal neurons expressing ChR2(L132C). The major drawback of AAV transduction in iNGNs was the much lower expression efficiency compared to rat hippocampal neurons.

Synaptogenesis of iNGNs
Presynaptic marker expression. The ability to form synapses is central to neuronal function. Cellular outgrowth serves to find and form synaptic connections with other cells for neural communication. In order to study the neuronal maturation of iNGNs, we also studied their ability to form synapses using single-cell synaptic recordings and immunocytochemical analyses. The presynaptic vesicular-associated protein Synapsin-I was expressed in iNGNs from 4 to 70 days, and colocalized with the dendritic protein microtubule-associated protein 2 (MAP2) that was also expressed from 4 to 70 days (Fig 4). For comparison, the same antibody staining was done in 21 DIV rat hippocampal neurons (HCN). Synapsin-I had more punctate expression in older iNGNs, and rat HCNs, compared to younger iNGNs. Synapsin-I expression is an indication of synaptic vesicle assembly and tethering to the cytoskeleton, and suggests the priming of iNGNs for synapse formation.
Spontaneous network activity. Functional synaptogenesis in iNGNs was studied in voltage-clamp recordings to observe any spontaneous postsynaptic currents (sPSCs). A representative trace in a 63d iNGN is shown in Fig 5A on the left side, where the downward deflections are sPSCs. After addition of tetrodotoxin (TTX) and picrotoxin (PTX) to the ACSF bath solution, only the AP-independent miniature excitatory postsynaptic currents (mEPSCs) are observed (Fig 5A, right side). Beginning as early as 14 days, approximately 40% of iNGNs had Long-Term Cultures of Stem Cell-Derived Human Neurons sPSCs (Fig 5B), and by 28 days and older all iNGNs recorded had sPSCs. The average frequency of the sPSCs was~0.1 Hz between 14 and 49 days, and at 56 days and older was~10x higher (Fig 5C). Bursts of sPSCs often observed at the older ages were due to AP-induced currents, since they could be blocked with TTX (n = 3), and they had amplitudes of several hundreds of pA in size (Fig 5A). These data indicate that iNGNs in long-term culture have competent network activity as early as 14 days, as well as increased excitability over the course of 70 days.
Spontaneous glutamatergic activity. Spontaneous vesicle release onto glutamatergic receptors is an important influence on synaptic maturation and homeostasis [31]. Miniature excitatory postsynaptic currents (mEPSCs), which are due to spontaneous presynaptic vesicle release, were recorded at -75 mV and could be recorded as early as 7 days (in one 7d iNGN only, Fig 5D). Between 14,21,28,35, and 42 days, the fraction of iNGNs with mEPSCs went up with increasing age. The average mEPSC frequency showed an increasing trend with increasing iNGN age (Fig 5E), which could be due to release probability changes during this period. The average mEPSC peak amplitudes from 21d iNGNs were larger compared to 28d, 35d, and 42d iNGNs (P = 0.01378, P = 0.00018, P = 0.00006, respectively, ANOVA), but there were no differences between comparisons of any of the other ages ( Fig 6A, Table 1). It could be that dendritic electrotonic filtering, where mEPSCs occurring at longer distances from the cell body have smaller measured amplitudes [32], accounts for older iNGN mEPSC amplitudes that are smaller than at 21d. It could also be possible that there are changes in postsynaptic receptor numbers or receptor clustering between iNGNs 21 days and older. The average mEPSC event charge transfers (ΔQ) from 21d iNGNs were larger compared to 14d and 42d iNGNs (P = 0.03393 and P = 0.00003, respectively, ANOVA), and mEPSC ΔQs from 35d iNGNs were larger compared to 42d iNGNs (P = 0.00232, ANOVA, Table 1). Average mEPSC rise times were not different between any of the ages, and decay τ's from 35d iNGNs were longer compared to 42d iNGNs (P = 0.02688, ANOVA, Table 1). The rise and decay kinetics of the mEPSCs in 14d, 21d, 28d, 35d, and 42d iNGNs had relatively fast rise time and decay τ values ( Fig 6A, Table 1). The kinetics properties suggest the expression of fast postsynaptic AMPA and KA receptors, but not NMDA receptors with much slower kinetics. A small fraction of slower events (rise time >3 ms, decay τ >10 ms) were observed in 21d, 28d, 35d, and 42d iNGNs; however, the majority of events had rise times <3 ms and decay τ's <10 ms (Fig 6A). Averaged mEPSC traces (all events for each age recorded) are shown in Fig 6B. Recordings for mEPSCs were also done at +30 mV and in low-Mg 2+ ACSF bath solution to optimize conditions for observing any possible larger NMDA receptor-mediated mEPSCs. However, at 21, 28, and 35 days (n = 11), any events had rise times <3 ms and decay τ's <15 ms, and there were no detectable NMDA receptor-mediated mEPSCs under these optimized recording conditions.
The AMPA/KA receptor inhibitor CNQX, but not the NMDA receptor inhibitor AP5 (n = 4), inhibited 100% of the mEPSCs in all cells recorded (Fig 7A and 7B). Upon washout of the CNQX from the ACSF bath solution, there was recovery of the mEPSCs. The mean number of events before CNQX addition was not different between the ages (P > 0.3 for all comparisons, ANOVA). The number of events was different before and after (always resulting in 0 events) addition of CNQX at 21, 28, 35, and 42 days (21d iNGNs: P = 0.0098, n = 7; 28d iNGNs: P = 0.0224, n = 6; 35d iNGNs: P = 0.0403, n = 7; 42d iNGNs: P = 0.0383, n = 6; paired Students t-tests). This confirmed the presence of postsynaptic AMPA/KA, and lack of NMDA, receptors in 21d, 28d, 35d, and 42d iNGNs.   7d, 14d, 21d, 28d, 35d, and 42d iNGNs. Voltage-clamp recordings were done at -75 mV. Decay τ is plotted against rise time for each age and one circle represents one event. n is the number of cells recorded at each age. (B) Averaged event traces in 14d, 21d, 28d, 35d, and 42d iNGNs. n is the number of events recorded at each age. doi:10.1371/journal.pone.0169506.g006 Long-Term Cultures of Stem Cell-Derived Human Neurons (Fig 8A, left and middle traces). 100 μM NMDA, along with its co-agonist 10 μM glycine (Gly), induced smaller, but detectable, outward currents at +30 mV (Fig 8A, right trace). Most of the iNGNs responded to both exogenous AMPA and KA at 14, 21, 28, 35, and 42 days ( Fig  8B, left and middle histograms), and average current densities were~2-4 pA/pF for both AMPA and KA (Fig 8C, left and middle histograms). At 7 days (n = 4), 14 days (n = 11), and 21 days (n = 8), none of the cells responded to exogenous NMDA (Fig 8B, right histogram). At 28 days, 35 days, and 42 days (Fig 8B and 8C, right histograms), iNGNs responded to exogenous NMDA; however, the average current densities were less than those for exogenous AMPA or KA,~0.5-1 pA/pF. This suggests that NMDA receptors are expressed in iNGNs, but are less abundant than AMPA and KA receptors, and are functional by 28 days and older, but   Long-Term Cultures of Stem Cell-Derived Human Neurons AMPA receptor expression. While iNGNs show functional synapses at early ages, their synapses dominantly express AMPA and KA, but not NMDA, receptors from 14 days-42 days, as suggested from the previous mEPSC and extrasynaptic receptor analysis. AMPA receptors are tetramers made up of four subunits (GluR1-4). GluR1 and GluR2 receptor expression was assessed in iNGNs over the course of 70 days using immunocytochemistry. 21 DIV rat hippocampal neurons were also immuno-stained under identical conditions. GluR1 dendritic expression was clearly seen at older ages, when it was colocalized with MAP2 (S4 Fig). GluR2 expression could also be seen in iNGN dendrites (S5 Fig). However, at 7 days and 14 days, GluR2 expression was dominant in the cell body.

Discussion
Most studies showing the functionality of neurons derived from stem cells are performed on a shorter-term time scale, up to a couple of weeks. Inducible expression of neurogenic transcription factors affords a kind of in vitro neuronal "birth date" for iNGNs, not possible in other neuronal differentiation protocols for stem cells. Here, we systematically studied iNGN function at specific time points, up to months in culture after the induction of neurogenesis.
Cultures of pure differentiated neurons derived from stem cells, such as by using induced expression of neurogenic transcription factors, fail to produce healthy, long-living, and functional neurons and neuronal networks. When iNGNs were grown alone in neuronal medium, they thrived poorly at 21 days and did not survive past 28 days. Most culturing conditions for pure stem cell-derived neurons have been based on conditions for primary rodent, primate, and human neurons, and these conditions mostly underperform with stem cell-derived neurons. Different strategies to enable long-term cultures (~100 days) of stem cell or iPSC-derived neurons have been used, such as growing the neurons on micropatterned surfaces [33] and astrocyte co-culture [34]. It has also been shown that adding mouse neurons in addition to astrocytes significantly increases the maturity of induced stem cell-derived neurons [9]. We successfully grew iNGNs in long-term cultures using astrocyte co-cultures and a simple defined protocol that results in functional neuronal networks. The astrocytes could be helping to "micropattern" the single iNGNs in our studies, since we observed no cell clustering, that hinders long-term cultures [33], and in this way act as physical support for the neurons. The astrocytes also exert their support through secreted factors that benefit neuron viability and synaptogenesis [17,18].
Long-term culture studies of other induced-neurons that could be useful to directly compare our functional data with are not yet available; however, iNGNs showed many hallmark properties for functionally mature neurons. At the younger ages, 21 days or less, we could compare iNGNs with similarly aged ASCL1-induced neurons, where both neuron populations had similar values for resting membrane potential (RMP), cell capacitance, input resistance (R i ), and action potential (AP) threshold; however, AP amplitude was larger in iNGNs than in ASCL1-induced neurons [10]. The passive electrical properties and bifurcations of iNGNs further matured beyond 21 days, and cell capacitance increased to~50 pF by 7 weeks, indicating neurite outgrowth.
iNGNs develop a mature phenotype more quickly than neurons derived from stem cells through multi-step differentiation protocols. 70d iNGNs had a RMP of~-79 mV and R i of 1.3 GO (Fig 1C). In comparison, in two studies of human neurons derived from iPSCs at week 7 and 10 of differentiation, RMP was~-28 mV and R i as~2.7 GO [35] and RMP was -49 mV and R i was~1.7 GO [36], respectively. Similar to young developing rodent neurons [25], iNGN RMP was depolarized (~-45 mV) at 7 days and steadily decreased to~-75 mV by 35 days. Undifferentiated iNGN stem cells had a high R i , whereas over the course of long-term culture iNGN neurons had a decreasing R i . One contribution to the decrease in R i is the corresponding increase in Na v and K v channel expression.
During the first 4 days of iNGN differentiation, Busskamp et al. showed an increase in gene expression of many of the Na v and K v channel subtypes [7]. We observed larger Na v and K v channel current densities with increasing ages (S2 Fig). By 42 days in culture, 100% of all 57 recorded iNGNs could fire trains of APs and older iNGNs could fire at higher firing frequencies (Fig 2B and 2C), up to 48 Hz. This is similar to cultures of rat excitatory pyramidal hippocampal neurons, which have maximal firing frequencies of 40-50 Hz [30]. AP properties matured with increasing iNGN age, such as decreasing half-width, and this facilitates iNGNs to fire at higher frequencies. iNGN AP properties matured more quickly compared to neurons derived from stem cells through multi-step differentiation protocols. 70d iNGNs had APs with a peak amplitude of~114 mV and half-width of~2.3 ms (Fig 2D). In comparison, in one study of human neurons derived from iPSCs at week 7 of differentiation, AP peak amplitude was 39 mV and half-width was~3.2 ms [36]. Upon blue light stimulation, the active light-gated cation channel ChR2 results in cell depolarization. By using AAVs, we could get a fraction of iNGN cells expressing the ChR2 variant ChR2(L132C) [16] and thereby control iNGN activity with blue light stimulation. Although the AAV transduction efficiency was 10x less in iNGNs than in rat hippocampal neurons, the iNGNs that expressed ChR2(L132C) had similar levels of channel expression (Fig 3B). Whether to manipulate a single neuron or whole neuronal networks of iNGNs, ChR2-expressing iNGNs allow the option of controlling iNGN activity with high temporal and spatial precision. In the future, it will be important to develop refined methods where a majority, rather than a minority, of iNGNs express ChR2. Two key improvements would be higher ChR2 expression efficiency and stable ChR2 expression over longer time courses.
Neurons expressing ChR2 offer a viable therapeutic approach for some diseases, such as for restoring visual functions in blind retinas, for example, by expressing ChR2 in ganglion cells [37] or ON bipolar retinal cells [38,39]. In vivo experiments have been done using mouse ESCs expressing ChR2, which were then transplanted and integrated as motor neurons to restore muscle function in denervated muscles in mice [40]. In the future, iNGNs robustly expressing ChR2 could have applications in translational medicine.
Human synaptogenesis is difficult to study from healthy human neural tissue samples, and stem cell-derived neurons offer an easier way gain insight into the development of human synapses. As seen by immunohistochemistry, iNGNs had clear Synapsin-I protein expression as early as 4 days (Fig 4), which was previously seen [7]. Correspondingly, we could detect synaptic activity in~25% of 14d iNGNs (Fig 5), which shows they had the necessary pre-and postsynaptic architecture for functional synapses. By 21 days and older, the majority of iNGNs had functional synapses. We also observed large (hundreds of pA) spontaneous postsynaptic currents in older iNGNs, which were AP-dependent and indicate increased iNGN excitability over time. Our studies on the synaptic properties of iNGNs are not easily comparable to other studies since our culture conditions included only the iNGNs with astrocytes. For example, the ASCL1-neurons had synaptic activity but when cultured together with primary hippocampal neuron cultures [10]. Here we show synaptic activity with iNGNs cultured only with astrocytes, and without an additional neuronal cell type [9,41].
Spontaneous release onto glutamatergic receptors plays a role in synaptic homeostasis [31] and in the regulation of dendritic protein synthesis [42]. Furthermore, spontaneous glutamate release acting on AMPA receptors is important for the maintenance of dendritic spines [43]. Miniature EPSC recordings, measuring postsynaptic currents in response to spontaneous vesicle release, showed a trend for increasing event frequency with age. This could indicate that the number of iNGN synapses increases from 14-42 days and/or there is an increase in vesicle release probability.
During the first 4 days of iNGN differentiation, Busskamp et al. showed there was an increase in gene expression of glutamatergic receptors [7]. In our studies, AMPA/KA receptors were present and post-synaptically functional as evident from mEPSC recordings. The mEPSCs had fast kinetics, indicative of AMPA/KA receptors, as opposed to the much slower kinetics of NMDA receptors. An AMPA/KA receptor inhibitor blocked 100% of the mEPSCs (Fig 7). NMDA receptor-mediated mEPSCs were not detectable, even in conditions which remove the extracellular Mg 2+ block according to NMDA receptor IV plots [44]. Extra-synaptic glutamatergic receptor stimulation, using AMPA, KA, or NMDA, showed sizable responses to AMPA and KA, but smaller responses to NMDA. NMDA currents also had a later onset compared to AMPA and KA currents (Fig 8).
AMPA receptors mediate fast excitatory glutamatergic neurotransmission in the mammalian central nervous system. They are tetramers, often heterotetramers, made up of four subunits, called GluR1-4 or GluA1-4 [45]. GluR1 and GluR2 are vital in the normal functioning of the human brain, and also have roles in neurological diseases, such as Alzheimer's disease [46,47] and drug addiction [48,49]. Evidence for AMPA receptor expression in iNGN synapses was confirmed by the kinetic properties of mEPSCs (Fig 6 and Table 1), inhibitor studies (Fig  7), and by immunocytochemistry using antibodies against GluR1 and GluR2 (S4 Fig and S5  Fig). These long-term cultures of iNGNs could be a convenient model to study neurological disorders where GluR1 and GluR2 are implicated. They could also be used for drug screening on specific neural targets, such as the GluR2-containing AMPA receptors.
Human neurons derived from iPSCs are an important tool for convenient brain disease modelling. Disease-specific neurons can be made and studied in vitro, which is an advantage to learn about neuronal and network properties in pathological conditions. Neurons differentiated from iPSCs have been derived from patients with neurodegenerative or neurodevelopmental disorders, such as spinal muscular atrophy [50], Down's syndrome [51,52], familial dysautonomia [53], amyotrophic lateral sclerosis [54,55], and Alzheimer's disease [56,57]. Alternatively, disease model neurons can be created in vitro using genome-editing strategies on healthy stem cells, which in turn can be differentiated into neurons. Fattahi et al. used the CRISPR/Cas9 system [58] on human ESCs to target the EDNRB locus to model Hirschsprung disease in the resulting enteric nervous system cells. They also used optogenetics (ChR2) and light stimulation of the enteric neurons to measure smooth muscle cell contractile responses [59]. A model for epileptic encephalopathy was made using gene-targeted ESCs, also in combination with optogenetics (ChR2/ChiEF) [60]. Together with the above-mentioned methods, modified iNGNs could also be created for disease model studies, along with ChR2 as a tool to precisely activate the iNGNs.
iNGN cultures can also offer a more thorough system to characterize and compare new optogenetic tools. The rodent cortical or hippocampal neuron cultures commonly used to characterize light-gated channel or pump function can complicate conclusions due to the variable neuronal properties, from numerous neuron subtypes, in one culture. These primary cultures typically contain approximately 15% interneurons [61,62], and ChR2 function has variable effects on different neuron subtypes [63]. On the other hand, iNGN cultures offer a more homogenous population of neurons and the ability to control many conditions, such as the age of the neurons.
In summary, we have generated long-term cultures of functional iNGN neuronal networks that show many of the hallmarks of neuronal maturation and form glutamatergic synapses rich in postsynaptic AMPA, but not NMDA, receptors. We also generated iNGNs expressing ChR2, which enable opportunities for manipulating iNGN activity in future studies. iNGNs cultures offer a greater degree of control to study mechanisms of early human synapse formation, and we have the ability to answer new questions about connectivity. Further studies can precisely define iNGN synapse architecture at the molecular level, and the ability to scale-up cultures could produce large iNGN networks for high-throughput studies on neural networks. Currents are shown in response to 700 ms voltage steps from -95 mV to +45 mV, in +10 mV increments, from a holding potential of -75 mV. Ages are labeled below the current traces. Zoomed-in traces below show the fast Na v peak currents (they open and close within several milliseconds) at -25 mV. The rightmost trace shows currents after addition of the Na v inhibitor TTX (1 μM). (B) Average steady-state K v current densities from 7d, 28d, and 56d iNGNs. (C) Average peak Na v current densities in 7d, 28d, and 56d iNGNs. (D) Average Na v channel activation in 7d, 28d, and 56d iNGNs. Na + conductance (G) was determined from G = I/(V-V Rev ), where I is the measured peak current at a specific voltage (V), and V Rev is the extrapolated reversal potential. G was normalized to G max , the maximum recorded conductance for each cell. The lines are curves showing averaged Boltzmann fits to individual recordings. Mean values are: 7d iNGNs V 1/2 = -30.5 ± 1.5 mV, n = 20; 28d iNGNs V 1/2 = -35.7 ± 1.6 mV, n = 9; 56d iNGNs V 1/2 = -44.3 ± 1.2 mV, n = 11; ANOVA showed that V 1/2 values differed between 7d and 56d iNGNs, P = 0.0002. Error bars show SEM and n is the number of neurons recorded. Representative z-stack confocal microscope images of 4d, 7d, 14d, 35d, 49d, and 70d iNGNs, or of 21 DIV rat hippocampal neurons (HCN). Immunocytochemistry was used to observe MAP2 (green) and GluR2 (red) expression and colocalization in iNGNs, along with the nuclear stain DAPI (blue). Ages in days (d) are labeled on the left side of the images, and the white scale bar is 50 μm long. (TIF)