Stem Cells http://www.peprotech.com/
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

First published online November 16, 2006
Stem Cells Vol. 25 No. 3 March 2007, pp. 738 -749
doi:10.1634/stemcells.2006-0246; www.StemCells.com
© 2007 AlphaMed Press

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental Data
Right arrow All Versions of this Article:
2006-0246v1
25/3/738    most recent
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow Reprints/Permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Ban, J.
Right arrow Articles by Ruaro, M. E.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Ban, J.
Right arrow Articles by Ruaro, M. E.

EMBRYONIC STEM CELLS

Embryonic Stem Cell-Derived Neurons Form Functional Networks In Vitro

Jelena Bana, Paolo Bonifazia, Giulietta Pinatoa, Frederic D. Broccarda, Lorenz Studerb, Vincent Torrea, Maria Elisabetta Ruaroa

aInternational School for Advanced Studies-via Beirut 2-4, Trieste, Italy;
bLaboratory of Stem Cell and Tumor Biology, Division of Neurosurgery and Developmental Biology Program, Memorial Sloan-Kettering Cancer Center, New York, New York, USA

Key Words. ES cells • In vitro differentiation • Neuronal network • Hippocampus

Correspondence: Maria Elisabetta Ruaro, International School for Advanced Studies, via Beirut 2-4, 34014 Trieste, Italy. Telephone: +39 040 3756510; Fax: +39 040 3756502; e-mail: eruaro{at}sissa.it

Received April 21, 2006; accepted for publication November 9, 2006.
First published online in STEM CELLS EXPRESS   November 16, 2006.


   ABSTRACT
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 Disclosures
 Acknowledgments
 References
 
Embryonic stem (ES) cells provide a flexible and unlimited source for a variety of neuronal types. Because mature neurons establish neuronal networks very easily, we tested whether ES-derived neurons are capable of generating functional networks and whether these networks, generated in vitro, are capable of processing information. Single-cell electrophysiology with pharmacological antagonists demonstrated the presence of both excitatory and inhibitory synaptic connections. Extracellular recording with planar multielectrode arrays showed that spontaneous bursts of electrical activity are present in ES-derived networks with properties remarkably similar to those of hippocampal neurons. When stimulated with extracellular electrodes, ES-derived neurons fired action potentials, and the evoked electrical activity spread throughout the culture. A statistical analysis indicated that ES-derived networks discriminated between stimuli of different intensity at a single trial level, a key feature for an efficient information processing. Thus, ES-derived neurons provide a novel in vitro strategy to create functional networks with defined computational properties.


   INTRODUCTION
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 Disclosures
 Acknowledgments
 References
 
Systems neuroscience aims at understanding how neuronal networks operate in a concerted way and, in particular, how information is processed [1–8]. A fundamental prerequisite to investigate how neuronal networks process information is to monitor simultaneously the electrical activity of a large population of neurons in the networks. In the last decade, the application of new electrophysiological and optical techniques in vivo [9–11] allowed better characterization of the functional properties of neuronal networks, giving the possibility of studying neural coding mechanisms, and distributed representations in different parts of the brain. In this way, it was shown how information can be encoded in the firing rate of an ensemble of neurons or in the synchrony of firing [6, 12], in the relative timing of action potentials (APs) [4, 13, 14], or in the latency of first evoked APs [8, 15]. However, due to the variability of the experimental conditions and the intrinsic complexity of the investigated system, it is very difficult to obtain a detailed analysis in vivo. Many of these problems are bypassed by studying neuronal networks in in vitro cultures, where experimental conditions and properties can be controlled in a more reliable way. Moreover, the chronic monitoring and stimulation of in vitro preparations for long-term periods provides a unique way to investigate developmental and plasticity mechanisms in a variety of systems [16–19].

When neurons are isolated and plated on an appropriate substrate, they readily grow, forming axodendritic arborization extending up to some millimeters and covered by a large number of functional synapses [16, 20, 21]. Distribution and cell types present in cultures are similar to those found in vivo [22–24], and, although they have lost the original connectivity of the intact tissue, they represent a good system to study how neuronal networks operate as a whole under controlled conditions. Multielectrode arrays (MEAs) represent a unique tool to investigate network dynamics, allowing the recording of the electrical activity of neuronal networks in both space and time [25], and have been used widely to characterize the spontaneous and the evoked activity of neuronal networks [1, 16, 19, 21, 26–33].

The use of ES cells allows the generation of an unlimited number of different cell types, including neurons [34–41]. Several protocols have been developed to derive neurons from ES cells in vitro [36, 38, 42], and some differentiation procedures allow the selective derivation of specific neuron types [38, 43–46]. Electrophysiological data from ES-derived neurons with different protocols validate their functional differentiation [34, 38, 41, 47–51], as well as the formation of synapse between ES-derived neurons [52] or between an ES-derived neuron and a mature neuron in organotypic slices [53] or in vivo [38, 54, 55]. ES cells provide a source of specialized cells for regenerative medicine [56–60], and it has been shown that ES cells or ES-derived neuronal precursors can incorporate into the nervous system and differentiate into neurons and glia [55, 61–64]. However, little is known of the capability of ES-derived neurons to form functional networks and their properties. Because the main function of neurons is communication, we decided to test the ability of ES-derived neurons to form functional networks and, in particular, to process information reliably.

In the present work, by combining intracellular recordings and extracellular recordings obtained with MEAs, we have analyzed how and to what extent ES cells form functional networks in vitro, and compared the properties of these networks with those of dissociated hippocampal neuronal networks [1]. ES-derived networks exhibited spontaneous activity, similar to that observed in primary networks in vitro [16, 19, 31, 65]. When stimulated with extracellular electrodes, ES-derived neurons fired action potentials, and the evoked electrical activity spread throughout the culture. A statistical analysis of the pattern of firing indicates that ES-derived networks have properties similar to those of primary adult neurons and can process information in a reproducible way in different trials; therefore, ES-derived networks behave as reliable computing elements.


   MATERIALS AND METHODS
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 Disclosures
 Acknowledgments
 References
 
Neuronal Culture Preparation
Hippocampal neurons from Wistar rats (P0-P2) were prepared as previously described [33]. Cells were plated on polyornithine/matrigel precoated MEA [33] at a concentration of 8 x 105 cells/cm2 and maintained in minimal essential medium with Earle's salts (Invitrogen Corp., Carlsbad, CA, http://www.invitrogen.com) supplemented with 5% fetal calf serum, 0.5% D-glucose, 14 mM Hepes, 0.1 mg/ml apo-transferrin, 30 µg/ml insulin, 0.1 µg/ml D-biotin, 1 mM vitamin B12, and 2 µg/ml gentamycin. After 48 hours, 5 µM cytosine-β-D-arabinofuranoside (Ara-C) was added to the culture medium to block glial cell proliferation. Half of the medium was changed twice a week. Neuronal cultures were kept in an incubator, providing a controlled level of CO2 (5%), temperature (37°C), and moisture (95%). In all experiments, hippocampal cells were used after 3 weeks in culture.

ES-Derived Neuron Differentiation
ES cells were induced to differentiate into neurons using the protocol for GABAergic neurons described in Barberi et al. [43], only slightly modified. Undifferentiated BF1/lacZ ES cells were plated on mitomycin-treated MS5 cells as single-cell suspension at a density of 250 cells/cm2 in Knockout Serum Replacement (KSR) medium and cultured for 6 days. The ES-derived epithelia structures were then mechanically separated from MS5 cells monolayer, by flushing Hanks' balanced salt solution through the pipette tip near the colony. The detached ES cells colonies were resuspended with KSR medium and plated on polyornithine-fibronectin-coated dishes. After 2–3 hours, the medium was changed with N2 medium containing basic fibroblast growth factor (bFGF) 10 ng/ml and 1 µg/ml fibronectin (amplification medium), and cells were induced to proliferate in presence of bFGF for 4 days. During the last 2 days, 200 ng/ml Sonic Hedgehog (SHH) and 100 ng/ml fibroblast growth factor (FGF)8 were added as patterning factors. Cells were then trypsinized and plated on polyornithine-laminin-coated MEA plates or glass coverslips in N2 medium containing 10 ng/ml brain-derived neurotrophic factor, 10 ng/ml NT4, and 1 µg/ml laminin (differentiation medium). Half of the differentiation medium was then changed twice a week. The postmitotic neurons were maintained in culture up to 10 weeks. All recombinant proteins were from R&D Systems (Minneapolis, http://www.rndsystems.com/).

Immunocytochemistry
Cells were fixed in 4% paraformaldehyde containing 0.15% picric acid (in phosphate-buffered saline [PBS]), saturated with 0.1 M glycine, permeabilized with 0.1% Triton X-100, saturated with 0.5% bovine serum albumin (BSA) in PBS and then incubated for 1 hour with primary antibodies. The primary antibodies were (a) rabbit polyclonal antibodies: against GABA, serotonin, CaMKII (all from Sigma Chemicals, St. Louis, http://www.sigmaaldrich.com/), and tyrosine hydroxylase (TH; Pel Freeze, Rogers, AR, http://www.pelfreez-bio.com/); (b) mouse monoclonal antibodies: TUJ1 (Covance, Berkeley, CA, http://www.crpinc.com), GFAP (Sigma Chemicals), nestin (Chemicon, Temecula, CA, http://www.chemicon.com/); and (c) guinea pig polyclonal antibody against V-GLUT2 (Chemicon).

The secondary antimouse fluorescein isothiocyanate (FITC) and anti-rabbit-tetramethylrhodamine isothiocyanate (TRITC) antibodies were from Sigma (Sigma Chemicals), goat anti-mouse immunoglobulin (Ig) G1-FITC and IgG2a-TRITC were from Southern Biotech (Birmingham, AL, http://www.southernbiotech.com/), anti-guinea pig-488 Alexa (Molecular Probes). Total nuclei were stained with 2 µg/ml in PBS Hoechst 33342 (Sigma Chemicals).

Electrophysiological Recordings
Dissociated hippocampal and ES-derived neuronal cultures were transferred in a recording chamber, perfused in Ringer's solution (145 mM NaCl, 3 mM KCl, 1.5 mM CaCl2, 1 mM MgCl2, 5 mM glucose, 10 mM Hepes; adjusted to pH 7.3 with NaOH) and visualized with an upright microscope (Olympus) with differential interference contrast optics. Patch-clamp recordings were performed with an Axoclamp 2-B amplifier (Axon Instruments/Molecular Devices Corp., Union City, CA, http://www.moleculardevices.com). Experiments were performed at room temperature (20°C–22°C). Electrodes were pulled (Narishige, Tokyo, http://www.narishige.co.jp/main.htm) and filled with an intracellular solution containing 120 mM potassium-gluconate, 10 mM sodium-gluconate, 10 mM Hepes, 10 mM sodium-phosphocreatine, 4 mM MgATP, 4 mM NaCl, 2 mM Na2ATP and 0.3 mM Na3GTP (adjusted to pH 7.3 with KOH); in these conditions, the electrode resistance was 15–20 M{Omega}. To enhance the driving force for chloride currents, some experiment were performed in symmetrical chloride conditions by substituting potassium-gluconate with KCl in the pipette solution; in these conditions, the resistance of the electrodes was 5–10 M{Omega}. The data were digitized at 20 kHz (Digidata 1200, Axon Instruments) and analyzed using pClamp9 software (Axon Instruments). Values of membrane potentials were corrected for the effects of liquid junction potential during seal formation.

Pharmacological identification of postsynaptic responses was performed by application of the following synaptic blockers: 30 µM D-AP5, 20 µM 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX); 10 µM SR-95531 (gabazine). All reagents were purchased from Tocris (Bristol, United Kingdom, http://www.tocris.com/).

MEA Electrical Recordings and Electrode Stimulation
ES-derived neurons cultured on MEAs, were kept in an incubator with a controlled level of CO2 (5%), temperature (37°C), and moisture (95%). Before electrical recordings, dishes were sealed with a cap distributed by MultiChannel Systems (Reutlingen, Germany, http://www.multichannelsystems.com/) to reduce gas exchange and eliminate evaporation and contamination and transferred from the culture incubator to a different incubator with controlled CO2 (5%), and temperature (37°C) where the electrical recording system was placed. Before starting the recordings, the neuronal culture was allowed to settle for approximately 30 minutes. After termination of the experiment, usually after 2–3 hours, the medium was changed, and the dish was moved back to the incubator.

MultiChannel Systems commercially supplied the multielectrode array system used for electrophysiology. MEA dishes had 10 x 6 TiN electrodes with an interelectrode spacing of 500 µm, and each metal electrode had a diameter of 30 µm. The MEA is connected to a 60-channel, 10–3-kHz bandwidth preamplifier/filter-amplifier (MEA 1,060 AMP), which redirects the signals toward a further electronic processing (i.e., amplification and analog to digital [AD] conversion), operated by a high-performance computer. Signal acquisitions are managed under software control, and each channel was sampled at a frequency of 20 kHz. One electrode was used as ground. Sample data were transferred in real time to the hard disk for off-line analysis. Each metal electrode could be either used for recording or for stimulation. The voltage stimulation used consisted of bipolar pulses lasting 100 µs at each polarity, of amplitude varying from 200 to 900 mV, injected through a single channel of the STG1004 Stimulus Generator [66]. The voltage pulse generated by the STG1004 was applied in parallel with the set of electrodes manually selected for stimulation (simultaneous multisite stimulation). An artifact lasting 5–20 ms, caused by the electrical stimulation, was induced on the recording electrodes but was removed from the electrical recordings during data analysis [33, 67]. For each stimulus, the culture was stimulated for 100 trials every 4 seconds.

Data Analysis
Acquired data were analyzed using MATLAB (The Mathworks, Inc., Natick, MA, http://www.mathworks.com/). For each individual electrode, we computed the SD ({sigma}) of the noise, which ranged from 3 to 6 µV, and only signals crossing the threshold of –5{sigma} were counted as APs and used for data analysis. AP sorting was obtained by using principal component analysis and open source toolboxes for the analysis of multielectrode data [68] with MATLAB. To quantify the amount of small amplitude spikes, a fixed threshold of –50 µV was used, and the number of APs recorded was then compared with that obtained with a threshold of –5{sigma}. The firing rate (FR(t)) of Figure 3B and the probability distribution of the number of spikes per bin of Figure 4A and 4B were computed by counting the number of spikes recorded by the whole MEA in time bins of 250 ms. To compute the average firing rate (AFR) of the neurons, peristimulus time histograms (PSTHs) were calculated for the sorted neurons (Fig. 6A, 6C) using a 10-ms time bin, where time 0 ms corresponds to the delivery of the stimulation. Similarly, when APs recorded by the whole array of electrodes were pooled, the array PSTHs (APSTHs) were calculated (Fig. 6B, 6D). The coefficient of variation (CV) of any variable analyzed is the SD over the mean of the variable. To test the difference of firing between the spontaneous activity and the activity after the extracellular stimulation, we considered the number of APs, respectively, in the 50 ms preceding and in the 50 ms after the stimulus. To test the significant difference in the variation between the two quantities, we used one-way analysis of variance.

Calculation of the Mutual Information
With the aim of decoding the stimulus intensity, we considered the array response (ARt). We used information theory [69] and, in particular, mutual information to estimate the amount of information that can be decoded in different time bins (i.e., varying t) and for different extents of pooling (i.e., different numbers of electrodes). In particular, the mutual information was calculated as follows:


Formula

where


Formula

It quantifies in bits the amount of information that a single response, r, (i.e., ARt,) provides about the intensity of the stimulus, s. pt(r) is the total probability of observing the response r considering the time bin 0 to t ms after the stimulus, averaged over all stimuli. In our case, all stimuli occurred with equal probability, p(s). In order to minimize the effects of finite sample size on our estimates of information, the real response r has been binned into different intervals, following the methods of Panzeri and Treves [70].

Calculation of Correlation
The degree of correlation of firing in the network was measured comparing the multiunit recordings of pairs of electrodes. The spontaneous activity was recorded for approximately 30 minutes and, to convert it into a time series (SA1, ..., SAn), it was binned into firing rate with a bin width {Delta}t. For each pair of electrodes (i, j), the cross-correlation {rho}i,j between the time series (SAi1, ..., SAin; SAj1, ..., SAjn) was calculated according to the equation:


Formula

and the average cross correlation <{rho}> over all the possible pairs of electrodes was calculated. The cross-correlation analysis for the spontaneous activity shown in Figure 4 was obtained varying the size of the bin width {Delta}t.


   RESULTS
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 Disclosures
 Acknowledgments
 References
 
Differentiation of ES Cells
Neuronal differentiation was induced by coculturing ES cells on MS5 murine bone marrow-derived stromal cells as described in Barberi et al. [43]. ES cells seeded at low density on this feeder layer and exposed to serum replacement medium formed small epithelial structures, so that after 6 days many cells in the structure were positive for nestin, a marker for precursors (Fig. 1A; Barberi et al. [43]). For further propagation and selection, these structures were detached from the MS5 feeders, transferred, disaggregated, plated on polyornithine-fibronectin coated dishes, and propagated in the presence of bFGF. After a few days, almost all of the cells were positive for nestin (Fig. 1A, right panel).


Figure 1
View larger version (76K):
[in this window]
[in a new window]

  		Figure 1. Cellular composition of ES-derived and hippocampal cultures. (A): Embryonic stem (ES) cells grown on MS5 cells feeder in KSR medium at day 4 (left) and day 6 (middle); nestin-positive neural precursors (right). (B): ES-derived neurons at 3 weeks of differentiation stained with anti-β-tubulin III (TUJ1), glial cells (GFAP positive), GABAergic neurons (GABA positive), glutamatergic neurons (CaMKII positive), dopaminergic neurons (TH positive), and serotonergic neurons (5-HT positive). (C): Immunofluorescence analysis as in B for the hippocampal neurons kept 3 weeks in culture. Glutamatergic neurons are marked with anti-VGLUT2 antibody. (D, E): Double staining of neurons (TUJ1) and glia (GFAP) with total nuclei marked by Hoechst 33342 in ES-derived neurons and hippocampal neurons respectively. Scale bar = 50 µm. Abbreviations: ES, embryonic stem; GFAP, glial fibrillary acidic protein; 5-HT, 5-hydroxytryptophan; TH, tyrosine hydroxylase.

 
Removal of bFGF and addition of specific neurotrophins allowed differentiation to occur. Figure 1B compares the expression of different markers in ES-derived neurons at 3 weeks of differentiation, with dissociated hippocampal cells cultivated for 3 weeks. ES-derived neurons expressed the postmitotic neuronal marker β-tubulin III starting from 3 days after differentiation induction, and this marker was present throughout the observation period (TUJ1-positive). Of these postmitotic neurons, approximately 30% were GABAergic (positive to GABA), less than 5% were dopaminergic (positive for tyrosine hydroxylase) and very few (less than 1%) serotonergic (positive for 5-HT). The remaining neurons were glutamatergic (positive CaMKII or vesicular glutamate transporter 2 [VGLUT2]). Very few astrocytes (cells positive for GFAP) were observed during the first week, whereas their percentage progressively increased, reaching approximately 50% of the population at 4 weeks of differentiation. An increase in glial cells reflected the prolonged period of in vitro differentiation required for network formation. Increased numbers of GFAP+ cells over time is in agreement with in vivo developmental progression and with neural stem cell differentiation studies in vitro [69, 70]. In the hippocampal culture, the fraction of glial cells was approximately 20% during the observation period. Approximately 10% of the neurons were GABAergic, and approximately 90% were glutamatergic (positive to VGLUT2); no serotonin or dopaminergic neurons were detected. (Fig. 1 and [1]).

To confirm that differentiation of ES cells produced functional neurons, intracellular recordings with patch pipettes were performed at 7, 14, 21, and 28 days after differentiation induction (41 cells). The resting membrane potential was –43 ± 4 mV after 1 week (n = 8), and increased to –68 ± 7.8 mV after 3 weeks (n = 13). In the control hippocampal cells, the resting membrane potential was –63 ± 5.9 (n = 18). The input resistance was 2 ± 1.3 G{Omega} during the first week, and decreased subsequently to 1.4 ± 0.6 G{Omega} in the ES-derived neurons. In the control hippocampal cells, the input resistance was 0.4 ± 0.1 G{Omega}. After 1 week, a depolarizing current pulse evoked, at most, one AP with mean amplitude of 12.9 ± 5 mV SD (n = 8). After 2 weeks of differentiation, ES-derived neurons produced APs with amplitude of about 50 mV, but a tonic discharge of APs was evoked only in one of eight cells analyzed. After 3 weeks, 4 of 12 cells fired trains of APs, and after 4 weeks, 9 of 13 cells discharged APs in a sustained way, but neurons with a transient firing were still observed.

Examples of current-clamp recordings of ES-derived neurons at 7, 14, 21, and 28 days of differentiation are shown in supplemental Figure 1. After 4 weeks of differentiation, in some cells, a depolarizing current pulse did not evoke APs, and voltage clamp recordings from these cells indicated the presence of voltage gated K+ currents but not of Na+ current (not shown). As the percentage of glial cells in the culture increased with time, these recordings were presumably obtained from glial cells.

The basis of network connectivity is the formation of functional synapses among different components. Therefore, we analyzed pharmacologically the different transmitters contribution to spontaneous postsynaptic events.

Starting from 14 days of differentiation, in most cultures, intracellular recordings showed both excitatory and inhibitory spontaneous synaptic activity (Fig. 2). The identification of inhibitory synaptic contributions was performed by current-clamp whole-cell patch-clamp recordings. To enhance the contribution of chloride currents, some experiments were performed in symmetrical chloride conditions, where chloride mediated synaptic events result as depolarizing potentials. Addition of glutamate receptor blockers D-AP5 (30 µM) and CNQX (20 µM) allowed isolating inhibitory contributions. At resting potential, the cells showed strong spontaneous depolarizing events that could trigger action potential initiation (Fig. 2A, left trace). Further addition of the GABA-A antagonist SR-95531 (gabazine, 10 µM) completely canceled this spontaneous activity (Fig. 2A, right trace; n = 3). These results demonstrate the contribution of GABAergic transmission in the spontaneous activity of ES-derived neurons. The averaged membrane potential before application of gabazine (Fig. 2A, left trace) calculated in a time window without occurrence of APs was –56 ± 5.4 mV, whereas after the application of gabazine, the averaged value in a time window of the same duration was –70 ± 1.1 mV. The decrease in the SD reflected the block of chloride-mediated depolarization after gabazine treatment.


Figure 2
View larger version (25K):
[in this window]
[in a new window]

  		Figure 2. Electrophysiological and pharmacological identification of spontaneous synaptic inputs. (A): Whole-cell current-clamp patch-clamp recording at 0 bias current of an embryonic stem cell (ES)-derived neuron in presence of D-AP5 (30 µM) and CNQX (20 µM). The recording was performed in symmetrical chloride conditions in order to enhance the contribution of chloride currents (first trace). Addition of gabazine (10 µM) abolished the spontaneous activity, with a hyperpolarizing effect on the membrane potential (second trace). (B): Spontaneous outward (arrows) and inward currents were observed in voltage clamp recordings at –50 mV with potassium-gluconate-based pipette solution. Traces in left column are representative of control conditions, showing both outward and inward currents. Addition of gabazine completely prevented outward currents, but did not influence inward currents occurrence (representative traces in right column). (C): Spontaneous activity of ES-derived neuron recorded in current clamp at 0 bias with potassium-gluconate-based pipette solution (first trace). Application of gabazine synchronized excitatory synaptic inputs with consequent onset of bursts of action potentials (second trace). Further application of D-AP5 and CNQX completely abolished the occurrence of bursts (third trace). After wash out, bursting activity regenerated (last trace). (D): Hippocampal neuron spontaneous activity recorded in current clamp at 0 bias with potassium-gluconate-based pipette solution (first trace); effect of gabazine (second trace); effect of further addition of D-AP5 and CNQX (third) trace; and wash out (last trace). Traces scale as in C.

 
Further evidence for inhibitory contributions in the spontaneous synaptic activity could be found by recording the cells in voltage clamp conditions with potassium-gluconate based intracellular pipette solution (see Materials and Methods; Fig. 2B). By clamping the cell at –50 mV, spontaneous occurring outward currents were recorded at a frequency of approximately 1 Hz (Fig. 2B, arrows). Subsequent application of gabazine completely prevented the occurrence of such events.

Excitatory synaptic input was also present in the spontaneous activity (Fig. 2C, control). Application of gabazine in potassium-gluconate-based intracellular pipette solution to block inhibition both isolated excitatory contributions in whole-cell patch-clamp recordings and increased the synaptic coupling among the neurons participating to the network. This resulted in synchronous excitatory synaptic inputs in the recorded cells, often triggering bursts of APs (Fig. 2C, gabazine). In these conditions addition of glutamate receptor blockers D-AP5 and CNQX completely abolished the occurrence of such events (Fig. 2C, AP5; CNQX) that reappeared after wash out (Fig. 2C, wash out; n = 3).

The same experiment was also performed in hippocampal cells, and data are shown in Figure 2D (n = 3). These results show the presence of inhibitory and excitatory synaptic contributions for both ES-derived and hippocampal neurons. However, difference in the firing pattern between the two networks was observed. In particular, the duration of the burst in the ES-derived neurons was significantly longer (13.4 ± 5.2 seconds) than the hippocampal bursts (3.5 ± 1.3; Student's t-test p < .0005).

Spontaneous Activity of Maturing ES-Derived Neurons
Spontaneous activity is a common characteristic of developing neuronal networks both in vivo and in vitro, which is believed to play an important role in network development [16, 19, 73–76]. To evaluate the distribution and propagation of the spontaneous activity in ES-derived neuronal culture, neuronal precursors were plated and induced to differentiate directly on the MEA. In parallel, hippocampal networks grown on MEA were analyzed for comparison. After 1 week of differentiation of ES-derived neurons on the MEA, it was possible to record the extracellular voltage signals with a shape very similar to those observed in cultures obtained from neonatal hippocampal rat neurons (supplemental Fig. 2). Extracellularly recorded APs varied in a number of active sites and amplitude of spikes in different cultures because it depends on the quality of the contact of the cells with the electrodes. Figure 3 illustrates the spontaneous activity of ES-derived networks (3A and 3B) and of hippocampal networks (3C and 3D). Simultaneous extracellular recordings are shown in Figure 3A and 3C for ES-derived and hippocampal networks, respectively. The spike trains of ES-derived neurons contained a greater proportion of small spike amplitudes than present in hippocampal ones (Fig. 3A, 3B). In fact, less than 5% (4.22% ± 0.57%) had amplitude higher than 50 µV (see Materials and Methods) although amplitude as large as 100 µV could be observed. In the hippocampus, the percentage of spikes with amplitude higher that 50 µV was more than double (10.14% ± 2.22%) and spikes could reach 200 µV.


Figure 3
View larger version (57K):
[in this window]
[in a new window]

  		Figure 3. Multielectrode recording of spontaneous activity of an embryonic stem cell (ES)-derived network at 25 days of differentiation. (A): Snapshot of 1 second of spontaneous activity at selected electrodes of the multielectrode array (MEA) during a burst. (B): Temporal analysis of 2 minutes of spontaneous activity. Upper panel: 2 minutes of global activity of the network; firing rate (number of action potentials [APs] in a bin width of 250 ms) recorded by all electrodes of the MEA. Lower panel: spontaneous activity recorded by each single electrode in time bins of 250 ms; the intensity of the activity at the electrode in a particular time bin is represented in grayscale (see Materials and Methods). (C) and (D) represent the same analyses as in (A) and (B) for a hippocampal network after 3 weeks in culture.

 
APs were detected on different electrodes, and bursts of APs recorded simultaneously from several electrodes invading the entire network were often observed. The global electrical activity of the network was described by computing the firing rate of the entire network FR(t) (see Materials and Methods) counting the total number of recorded APs in a given bin width. As shown in the upper panels of Figure 3B and 3D, the FR(t) of both ES-derived and hippocampal networks had large peaks corresponding to the simultaneous firing of several neurons, separated by periods where only occasional APs were observed. The global electrical activity of the network can also be visualized by considering raster plots, where a dot represents the occurrence of an AP, of several electrodes. Raster plots from ES-derived and hippocampal networks are shown on the bottom panel of Figure 3B and 3D, respectively. In raster plots, a vertical black line indicates that APs were recorded simultaneously from all, or most, extracellular electrodes of the MEA.

A comparison between the FR(t) and raster plots from ES-derived and hippocampal networks shows that, in both networks, large bursts of simultaneous activity were observed. The size of these bursts was random and did not have any regular structure. These bursts did not have any obvious periodicity, and their occurrence could not be predicted. We observed spontaneous activity in these cultures (n = 5) for more than 2 months and, in one case, up to 3 months. Although the firing pattern of both types of networks was quite similar, the spontaneous activity baseline was slightly higher in ES-derived networks than in hippocampal networks (compare upper panels of Fig. 3B and 3D). This was likely due to the presence of a tonic firing of small amplitude, resulting in rare periods of quiescent activity (compare bottom panels of Fig. 3B and 3D). This tonic firing of small amplitudes spikes (Fig. 3A) was reminiscent of that of immature hippocampal networks (data not shown).

To further investigate the firing pattern of both networks, the firing rate was binned and the probability distribution of the number of spikes and active electrodes present in each bin was computed (see Materials and Methods). Both networks had a probability distribution of the number of spikes per bin fitted by lognormal distributions (Fig. 4A and 4B), indicating the presence of multiplicative effects [77] among the firing pattern of individual electrodes, broadening the range over which large numbers of spikes per bin can be observed.


Figure 4
View larger version (20K):
[in this window]
[in a new window]

  		Figure 4. Correlated firing. (A, B): Probability distribution of the number of spikes per bin (250 ms) in the hippocampal- and embryonic stem cell (ES)-derived networks, respectively. Both networks are fitted by a lognormal distribution (solid line) and not by a Poissonian distribution (dashed line), as expected in the case of random firing among neurons. The skewed lognormal distribution and its long right tail account for the presence of few large bursts, containing several hundreds of spikes, and occurring in both networks. Insets: probability distribution of the number of active electrodes per bin. The dashed line represents a Poissonian fit, indicating that the pattern of active electrodes in each time bin occurs at random. (C): Examples of cross correlation between couple of electrodes 1 mm apart, in the hippocampal- (top panel) and ES-derived networks (bottom panel). (D): Dependency of the network cross correlation on the bin size used for both ES-derived networks (filled circle) and hippocampal networks (open diamond).

 
The lognormal fit, characterized by a long right tail, also excluded a random firing pattern typically described by Poissonian distributions [78]. Moreover, the presence of tonic firing in ES-derived networks was further revealed by the presence of a peak in the lognormal distribution, centered on the number of active electrodes. This means that at least one spike was present in each bin most of the time, in contrast to hippocampal networks, where small numbers of spikes fired in each time bin have the highest probability.

Although the network firing pattern was not random and was dependent of multiplicative interactions among neurons, the number of active electrode in each bin followed a Poissonian distribution (Fig. 4A and 4B, insets), suggesting that the pattern of active electrode in during each time bin occurred at random. In both network types, cooperative effects between neurons recorded at distant electrodes were also indicated by their high cross-correlation coefficient (Fig. 4C).

The degree of connectivity within the network was quantified by computing the network correlation, defined as the average cross-correlation <{rho}> among electrical recordings from active electrodes (i.e., exhibiting evident APs; see Materials and Methods). As shown in Figure 4D, <{rho}> increased with the size of the bin width (i.e., the size of the time window over which correlations were considered). For small values of the bin width (i.e., less than 20 ms), the value of <{rho}> was lower than 0.1, indicating that spontaneously occurring APs were almost completely uncorrelated on a short time scale. At larger bin width (around 200 ms), as there is an increasing number of observations per bin, there is an increasing probability of observing correlations, and the value of <{rho}> increased to approximately 0.5 in both networks. This observation indicates a larger correlation of spontaneously occurring APs on a longer time scale. The dependence of <{rho}> on the used bin width is a consequence of the biophysical mechanisms underlying synaptic transmission and the generation of APs [1]: synaptic release and AP initiation in different neurons is statistically independent at the ms level, but becomes correlated in the presence of a common input lasting some hundreds of milliseconds.

Spread of the Evoked Electrical Activity in the Network
The propagation of electrical excitation in a network is a proof of the presence of chemical and electrical synaptic pathways and of functional network formation as has been previously shown in different studies on hippocampal [33] and cortical networks [20, 79, 80].

Therefore, the spread of the evoked activity was studied in ES-derived network cells cultured on MEA. Electrical activity was evoked by delivering brief voltage pulses to the extracellular electrodes of the MEA (see Materials and Methods) These voltage pulses induced a transient depolarization of neurons in good electrical contact with the stimulating electrode, possibly triggering the initiation of an AP. When a bar of electrodes was stimulated (black squares in the grid of Fig. 5), APs were evoked in neurons on nearby electrodes with a latency of some milliseconds. These evoked APs propagated over almost the entire network, with a progressively longer delay (Fig. 5A). The latency of the first evoked AP at electrodes near the stimulating bar, as the one indicated by the number 1 was less than 10 ms. Clear burst of evoked APs at electrodes more than 3 mm from the stimulating electrodes, as electrodes 14, 15, and 16 were observed. The latency of the first evoked APs at these more distant electrodes was around 40 ms. The results shown here are very similar to those previously described for hippocampal cultures [1, 33] and for cortical cultures [17, 21, 81].


Figure 5
View larger version (31K):
[in this window]
[in a new window]

  		Figure 5. Evoked electrical activity in embryonic stem cell (ES)-derived networks. (A): Examples of extracellular voltage recordings after the stimulation of a bar of electrodes (black line in the scheme on the right) with a bipolar voltage pulse of 900 mV. (B): Peristimulus time histogram (PSTH) of the evoked activity from 16 electrodes of the multielectrode array (MEA). The thin vertical line indicates the stimulation artifact. Data averaged over 100 repetitions of the same stimulation.

 
The variability of the evoked electrical response was characterized by computing the peristimulus time histograms, obtained by averaging evoked response over 100 repetitions of the same stimulus (Fig. 5B). The peak of the PSTH of the activity recorded on more distant electrodes occurred at progressively longer times: at a distance of approximately 2,000 µm from the stimulating electrodes, the peak amplitude of the PSTH occurred with a delay of approximately 40 ms from the stimulus. The evoked response was usually composed of a first reproducible AP, occurring with a similar timing in every trial, followed by less reproducible APs (Fig. 5B), as previously described for cortical networks [81]. The standard deviation of the latency of the first evoked AP, usually referred to as its "jitter," was used to measure the reproducibility of the evoked AP. A significant correlation between jitter and latency of the first evoked AP was observed in ES-derived networks ({rho} = .63, n = 4; supplemental Fig. 3A) similarly to what observed in hippocampal networks ({rho} = .8, n = 5; supplemental Fig. 3B and Bonifazi et al. [1]). The average latency increased with the physical distance between the recording electrode and the stimulating electrodes in both ES-derived and in hippocampal networks. The maximal speed of AP propagation was approximately 400 mm/s in both ES-derived and in hippocampal networks (supplemental Fig. 3C, 3D; Bonifazi et al. [1]).

The results shown in Figure 5 and in supplemental Figure 3 indicate that the evoked electrical activity in ES-derived networks spread with properties that are very similar to those observed in hippocampal networks.

Reproducibility of the Response and Information Processing
Neurons under certain network conditions fire APs in a poorly reproducible way, and, in particular, neurons receiving the same stimulation in different trials fire a variable number of APs or the same number of APs but with a different timing [1, 82, 83]. In most trials, the first evoked AP occurred with approximately the same latency, and it is considered reliable [1, 81]. Variability is characterized by repeatedly applying the same stimulation and analyzing the evoked average firing rate and the corresponding coefficient of variation (CV = {sigma}AFR/AFR) defined as the ratio between the standard deviation of the AFR ({sigma}AFR) and the AFR itself. As shown in Figure 6A, for selected individual ES-derived neurons, the CV (black dots) of the AFR (white bars) was rarely lower than 0.5, indicating that the evoked electrical activity was not reproducible.


Figure 6
View larger version (27K):
[in this window]
[in a new window]

  		Figure 6. Reproducibility and coding. Average firing rate (AFR; open bars) and coefficient of variation (CV; black dots) of the firing from three selected embryonic stem cell (ES)-derived neurons (A) and from the entire network (B). (C) and (D) as in (A) and (B), but for neurons from hippocampal culture.

 
The variability of the evoked electrical activity can be significantly reduced when several neurons are considered, and evoked APs are averaged or pooled together [1, 15, 82, 84]. When APs recorded from all the active electrodes (i.e., therefore from an ensemble of more than 30 ES-derived neurons) were pooled, the CV of the evoked activity was lower than 0.5 for 60 ms (Fig. 6B). A similar behavior was observed when the firing of individual hippocampal neurons was analyzed (Fig. 6C) and when their evoked electrical activity was pooled together, as shown in Figure 6D.

Functional neuronal networks can process information by encoding important features of the stimulus, such as its intensity [1]. Therefore, coding of stimulus intensity in ES-derived networks was investigated at the level of a single neuron or when APs from a population of cells were pooled. The same analysis was carried out in hippocampal networks (Fig. 7). Because pooling APs in ES-derived neurons exhibited a reproducible response (CV < 0.5), we investigated whether ES-derived networks distinguish the stimulus intensity, as observed in hippocampal networks [1]. Voltage pulses of different intensities were applied to the same row of electrodes. Raster plots of the evoked electrical activity from four distinct electrodes are shown in the four columns of Figure 7A. Similar data obtained from hippocampal networks are shown for comparison in Figure 7D. In both networks, by increasing the intensity of the voltage pulse from 300–600 to 900 mV, the number of APs recorded on each electrode increased, and the evoked APs became more frequent and more reliable. When the evoked electrical activity among several dozens of ES-derived neurons was pooled, it was possible to reliably distinguish the three stimulus intensities. As shown in Figure 7B, the histograms of the evoked electrical activity from about three dozen of neurons for the three voltage stimuli were well separated. Therefore, ES-derived networks were able to distinguish the stimulus intensity as hippocampal networks do (Fig. 7E). To quantify the ability of ES-derived networks to process information, we computed the mutual information IM, calculating the amount of information that could be decoded. As shown in Figure 7C, IM depended on the considered bin width and the size of neuronal pooling. With a bin width of 35 ms, and by pooling the electrical activity from 20 electrodes, approximately 1.3 bits of information were extracted from a theoretical maximum of 1.52 bits.


Figure 7
View larger version (37K):
[in this window]
[in a new window]

  		Figure 7. Information processing in embryonic stem cell (ES)-derived and hippocampal networks. (A): Raster plots of evoked action potentials (APs) recorded from four different electrodes (each column) by voltage pulses of 300, 600, and 900 mV (each row). (B): frequency histograms of all APs recorded in a time window from 0 to 51 ms from the stimulus evoked by the three different voltage stimulations. (C): Dependence of the mutual information on bin width and number of pooled electrodes. (D, E, F): As in (A), (B), and (C), but from hippocampal cultures. In (B) and (E), red corresponds to 300 mV, blue to 600 mV, and green to 900 mV.

 
Similar results were obtained in hippocampal networks, where a maximum of 1.4 bits could be extracted by using a bin width of 25 ms and pooling the activity over 20 electrodes. These results show that ES-derived networks possess some basic computational properties, as we previously demonstrated for hippocampal networks [1].


   DISCUSSION
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
Disclosures
 Acknowledgments
 References
 
In the present article, we provide the first investigation to our knowledge of the parallel processing of ES-derived networks. The study of neuronal networks in vitro is conceivable because it has been shown that functional characteristics of ex vivo neuronal networks are similar to those observed in vivo in terms of connectivity, inhibition/excitation ratio, electrophysiological and electrical stimuli (for a review, see Van Pelt et al. [19]). Moreover, in vitro networks possess information-processing ability with basic computational properties similar to those found in in vivo networks [35]. Recent work in our laboratory using dissociated hippocampal networks has also shown that different decoding strategies can be used to extract relevant features [1], giving the appropriate parameters to investigate ES-derived networks. By culturing ES-derived neurons over an MEA, we show that these networks are functional and exert some computational properties similar to those of hippocampal networks in culture.

When an extracellular voltage pulse was applied to the MEA electrodes, electrical activity was evoked and propagated for several millimeters in the culture with a velocity of approxiamtely 400 mm/second, similar to that observed in hippocampal cultures. The firing pattern of neurons in response to the stimulus was mainly composed by a first, reliable AP followed by less reliable APs, again comparable to hippocampal [1] and cortical [81, 85] neuronal networks in vitro. The variability of firing in ES-derived neurons was reduced when the APs were pooled over a population of neurons, and the high reproducibility of the pooled response allowed distinguishing, at the level of a single trial, stimuli varying in intensity. The ability of ES-derived neurons to process information was illustrated by measuring the mutual information between the evoked response and the stimulus intensity. This analysis indicates that a population of ES-derived neurons can reliably discriminate three stimulus intensities (corresponding to 1.52 bits of information) as observed in naturally matured neurons [1] and that ES-derived neurons can process information as mature neurons do.

Despite the striking similarity observed in the evoked activity, we observed some differences in the spontaneous activity. Although bursts of synchronous firing were present in both networks, interburst activity was more prominent in the ES-derived networks (Fig. 3), reminiscent of immature hippocampal networks (data not shown). This could reflect a difference in the ongoing maturational process characterized by different numbers of gap junctions, differential voltage-gated channel and ligand receptor expression and/or composition, difference in the rectifying function of the chloride ions [86–91], or percentage of glia in the cultures. One or more of these factors could also explain the difference in the duration of firing observed in the patch-clamp experiments between the two networks (Fig. 2C, 2D).

Our data also demonstrate that purely ES-derived neurons undergo functional maturation at a rate comparable to that observed after coculturing ES cells on top of hippocampal slice [53]. The appearance of neuronal postmitotic markers occurred much earlier than did the electrophysiological maturation of neurons. Moreover, the composition of the culture changed in terms of presence of glial cells during the observation period, indicating that a prolonged follow-up of the culture for therapeutic purposes is required.

The finding that ES-derived neurons form functional networks in vitro opens up a variety of new perspectives for future applications. ES cells are a potent source for the generation of various neuronal cell types, and by changing the composition of the culture, it is possible to promote the differentiation in specific neuronal types [38, 43–46]. Therefore, in the near future, it will be feasible to engineer neuronal networks with different neuronal types and to investigate the properties of the resulting networks. Similarly, because ES cells are amenable to genetic modification [92, 93], it will be possible to obtain neuronal types with modified properties in which selected ion channels or synaptic receptors are over- or underexpressed. This technology will allow the construction of neuronal networks with entirely new computational properties. The same technology can provide valid models of neuronal networks linked to a variety of neurodegenerative diseases and can verify the action of new ES-derived neurons on these model networks, providing a new perspective for therapeutic recovery.

The availability of unlimited number of ES-derived neurons paired with an intrinsic ability of these cells to form neuronal networks suggests the potential of this system to establishing large-scale, self-assembling networks. Such complex networks could provide a first key step towards ES cell-based neurocomputing [1, 33].


   DISCLOSURES
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 Disclosures
Acknowledgments
 References
 
The authors indicate no potential conflicts of interest.


   ACKNOWLEDGMENTS
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 Disclosures
 Acknowledgments
References
 
This work was supported by the EU Grant 12788 NEURO, and by a FIRB grant from the Italian Ministers. We thank Manuela Schipizza Lough for carefully reading the manuscript. P.B. is currently affiliated with the University of Cambridge, Department of Physiology, Cambridge, U.K.


   REFERENCES
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 Disclosures
 Acknowledgments
 References
 

  1. Bonifazi P, Ruaro ME, Torre V. Statistical properties of information processing in neuronal networks. Eur J Neurosci 2005;22:2953–2964.[CrossRef][Medline]

  2. Hopfield JJ. Neural networks and physical systems with emergent collective computational ability. Proc Nat Acad Sci U S A 1982;79:2554–2558.[Abstract/Free Full Text]

  3. Rumelhart DE, McClelland JL. Explorations in Parallel Distributed Processing. Cambridge, MA: MIT Press,1998;.

  4. DeCharms RC, Merzenich MM. Primary cortical representation of sounds by the coordination of action-potential timing. Nature 1996;381:610–613.[CrossRef][Medline]

  5. Georgopoulos AP, Schwartz AB, Kettner RE. Neuronal population coding of movement direction. Science 1986;233:1416–1419.[Abstract/Free Full Text]

  6. Gray CM, Konig P, Engel AK et al. Oscillatory responses in cat visual cortex exhibit inter-columnar synchronization which reflects global stimulus properties. Nature 1989;338:334–337.[CrossRef][Medline]

  7. Nicolelis MA, Ghazanfar AA, Stambaugh CR et al. Simultaneous encoding of tactile information by three primate cortical areas. Nat Neurosci 1998;1:621–630.[CrossRef][Medline]

  8. Thorpe S, Delorme A, Van Rullen R. Spike-based strategies for rapid processing. Neural Network 2001;14:715–725.[CrossRef]

  9. Nicolelis MA, Ghazanfar AA, Faggin BM et al. Reconstructing the engram: Simultaneous, multisite, many single neuron recordings. Neuron 1997;18:529–537.[CrossRef][Medline]

  10. Kerr JND, Greenberg D, Helmchen F. Imaging input and output of neocortical networks in vivo. Proc Natl Acad Sci U S A 2005;102:14063–14068.[Abstract/Free Full Text]

  11. Stosiek C, Garaschuk O, Holthoff K et al. In vivo two-photon calcium imaging of neuronal networks. Proc Natl Acad Sci U S A 2003;100:7319–7324.[Abstract/Free Full Text]

  12. Singer W, Gray CM. Visual feature integration and the temporal correlation hypothesis. Ann Rev Neurosci 1995;18:555–586.[CrossRef][Medline]

  13. O'Keefe J, Recce MLP. hase relationship between hippocampal place units and the EEG theta rhythm. Hippocampus 1993;3:317–330.[CrossRef][Medline]

  14. Hopfield JJ. Pattern recognition computation using action potential timing for stimulus representation. Nature 1995;376:33–36.[CrossRef][Medline]

  15. Johansson RS, Birznieks I. First spikes in ensembles of human tactile afferents code complex spatial fingertip events. Nat Neurosci 2004;7:170–177.[CrossRef][Medline]

  16. Van Pelt J, Vajda I, Wolters PS et al. Dynamics and plasticity in developing neuronal networks in vitro. Prog Brain Res 2005;147:173–188.[Medline]

  17. Jimbo Y, Tateno T, Robinson H. Simultaneous induction of pathway-specific potentiation and depression in networks of cortical neurons. Biophys J 1999;76:670–678.[Abstract/Free Full Text]

  18. Welsh DK, Logothetis DE, Meister M et al. Individual neurons dissociated from rat suprachiasmatic nucleus express independently phased circadian firing rhythm. Neuron 1995;14:697–706.[CrossRef][Medline]

  19. Van Pelt J, Wolters PS, Corner MA et al. Long-term characterization of firing dynamics of spontaneous burst in cultured neural networks. IEEE Trans Biomed Eng 2004;51:2051–2062.[CrossRef][Medline]

  20. Marom S, Shahaf G. Development, learning and memory in large random networks of cortical neurons: Lessons beyond anatomy. Q Rev Biophys 2002;35:63–87.[CrossRef][Medline]

  21. Potter SM. Distributed processing in cultured neuronal networks. Prog Brain Res 2000;130:49–62.

  22. Neale EA, Oertel WH, Bowers LM et al. Glutamate decarboxylase immunoreactivity and gamma-[3H] aminobutyric acid accumulation within the same neurons in dissociated cell cultures of cerebral cortex. J Neurosci 1983;3:376–382.[Abstract]

  23. Huettner JE, Baughman RW. Primary culture of identified neurons from the visual cortex of postnatal rats. J Neurosci 1986;6:3044–3060.[Abstract]

  24. Nakanishi K, Kukita F. Intracellular [Cl(-)] modulates synchronous electrical activity in rat neocortical neurons in culture by way of GABAergic inputs. Brain Res 2000;863:192–204.[CrossRef][Medline]

  25. Morin FO, Takamura Y, Tamiya E. Investigating neuronal activity with planar microelectrode arrays: Achievements and new perspectives. J Biosci Bioeng 2005;100:131–143.[CrossRef][Medline]

  26. Jimbo Y, Kawana A. Electrical stimulation and recording from cultured neurons using a planar microelectrode array. Bioelectrochem Bioenerg 1992;29:193–204.[CrossRef]

  27. Robinson HP, Kawahara M, Jimbo Y et al. Periodic synchronized bursting and intracellular calcium transients elicited by low magnesium in cultured cortical neurons. J Neurophysiol 1993;70:1606–1616.[Abstract/Free Full Text]

  28. Robinson HP, Torimitsu K, Jimbo Y et al. Periodic bursting of cultured cortical neurons in low magnesium—cellular and network mechanisms. Jpn J Physiol 1993; (Suppl 43)l25–130.

  29. Maeda E, Robinson HP, Kawana A. The mechanisms of generation and propagation of synchronized bursting in developing networks of cortical neurons. J Neurosci 1995;15:6834–6845.[Abstract/Free Full Text]

  30. Maeda E, Kuroda Y, Robinson HP et al. Modification of parallel activity elicited by propagating bursts in developing networks of rat cortical neurones. Eur J Neurosci 1998;10:488–496.[CrossRef][Medline]

  31. Kamioka H, Maeda E, Jimbo Y et al. Spontaneous periodic synchronized bursting during formation of mature patterns of connections in cortical cultures. Neurosci Lett 1996;206:109–112.[CrossRef][Medline]

  32. Shahaf G, Marom S. Learning in networks of cortical neurons. J Neurosci 2001;21:8782–8788.[Abstract/Free Full Text]

  33. Ruaro ME, Bonifazi P, Torre V. Toward the neurocomputer image processing and learning with a neuronal culture. IEEE Trans Biomed Eng 2005;52:371–383.[CrossRef][Medline]

  34. Bain G, Kitchens D, Yao M et al. Embryonic stem cells express neuronal properties in vitro. Dev Biol 1995;168:342–357.[CrossRef][Medline]

  35. Fraichard A, Chassande O, Bilbaut G et al. In vitro differentiation of embryonic stem cells into glial cells and functional neurons. J Cell Sci 1995;108Pt 10:3181–3188.[Abstract]

  36. Kawasaki H, Mizuseki K, Nishikawa S et al. Induction of midbrain dopaminergic neurons from ES cells by stromal cell-derived inducing activity. Neuron 2000;28:31–40.[CrossRef][Medline]

  37. Keller GM. In vitro differentiation of embryonic stem cells. Curr Opin Cell Biol 1995;7:862–869.[CrossRef][Medline]

  38. Lee SH, Lumelsky N, Studer L et al. Efficient generation of midbrain and hindbrain neurons from mouse embryonic stem cells. Nat Biotechnol 2000;18:675–679.[CrossRef][Medline]

  39. Li M, Pevny L, Lovell-Badge R et al. Generation of purified neural precursors from embryonic stem cells by lineage selection. Curr Biol 1998;8:971–974.[CrossRef][Medline]

  40. Ying QL, Stavridis M, Griffiths D et al. Conversion of embryonic stem cells into neuroectodermal precursors in adherent monoculture. Nat Biotechnol 2003;21:183–186.[CrossRef][Medline]

  41. Nakayama T, Momoki-Soga T, Yamaguchi K et al. Efficient production of neural stem cells and neurons from embryonic stem cells. Neuroreport 2004;15:487–491.[CrossRef][Medline]

  42. Okabe S, Forsberg-Nilsson K, Spiro AC et al. Development of neuronal precursor cells and functional postmitotic neurons from embryonic stem cells in vitro. Mech Dev 1996;59:89–102.[CrossRef][Medline]

  43. Barberi T, Klivenyi P, Calingasan NY et al. Neural subtype specification of fertilization and nuclear transfer embryonic stem cells and application in Parkinsonian mice. Nat Biotechnol 2003;21:1200–1207.[CrossRef][Medline]

  44. Morizane A, Takahashi J, Takagim Y. Optimal conditions for in vivo induction of dopaminergic neurons from embryonic stem cells through stromal cell-derived inducing activity. J Neurosci Res 2002;69:934–993.[CrossRef][Medline]

  45. Westmoreland JJ, Hancock CR, Condie BG. Neuronal development of embryonic stem cells: A model of GABAergic neuron differentiation. Biochem Biophys Res Commun 2001;284:674–680.[CrossRef][Medline]

  46. Park C-H, Minn Y-K, Lee J-Y et al. In vitro and in vivo analyses of human embryonic stem cell-derived dopamine neurons. J Neurochem 2005;92:1265–1276.[CrossRef][Medline]

  47. Bibel M, Richter J, Schrenk K et al. Differentiation of mouse embryonic stem cells into a defined neuronal lineage. Nat Neurosci 2004;7:1003–1009.[CrossRef][Medline]

  48. Finley MF, Kulkarni N, Huettner JE. Synapse formation and establishment of neuronal polarity by P19 embryonic carcinoma cells and embryonic stem cells. J Neurosci 1996;16:1056–1065.[Abstract/Free Full Text]

  49. Hancock CR, Wetherington JP, Lambert NA et al. Neuronal differentiation of cryopreserved neural progenitor cells derived from mouse embryonic stem cells. Biochem Biophys Res Commun 2000;271:418–421.[CrossRef][Medline]

  50. Jüngling K, Nägler K, Pfrieger FW et al. Purification of embryonic stem cell-derived neurons by immunoisolation. FASEB J 2003;7:2100–2102.

  51. Strübing C, Ahnert-Hilger G, Shan J et al. Differentiation of pluripotent embryonic stem cells into neuronal lineage in vitro gives rise to mature inhibitory and excitatory neurons. Mech Dev 1995;53:275–287.[CrossRef][Medline]

  52. Miles GB, Yohn DC, Wichterle H et al. Functional properties of motoneurons derived from mouse embryonic stem cells. J Neurosci 2004;24:7848–7858.[Abstract/Free Full Text]

  53. Benninger F, Beck H, Wernig M et al. Functional integration of embryonic stem cell-derived neurons in hippocampal slice cultures. J Neurosci Neurosci 2003;23:7075–7083.

  54. Ruschenschmidt C, Koch PG, Brustle O et al. Functional properties of ES cell-derived neurons engrafted into the hippocampus of adult normal and chronically epileptic rats. Epilepsia 2005;46 (suppl 5):174–178.[CrossRef][Medline]

  55. Wernig M, Benninger F, Schmandt T et al. Functional integration of embryonic stem cell-derived neurons in vivo. J Neurosci 2004;24:5258–5268.[Abstract/Free Full Text]

  56. Baizabal JM, Furlan-Magaril M, Santa-Olalla J et al. Neural stem cells in development and regenerative medicine. Arch Med Res 2003;34:572–588.[CrossRef][Medline]

  57. Kim J-H, Auerbach JM, Rodriguez-Gomez JA et al. Dopamine neurons derived from embryonic stem cells function in an animal model of Parkinson's disease. Nature 2002;418:50–56.[CrossRef][Medline]

  58. Perrier AL, Studer L. Making and repairing the mammalian brain—in vitro production of dopaminergic neurons. Semin Cell Dev Biol 2003;14:181–189.[CrossRef][Medline]

  59. Sánchez-Pernaute R, Studer L, Bankiewicz KS et al. In vitro generation and transplantation of precursor-derived human dopamine neurons. J Neurosci Res 2001;65.

  60. Tabar V, Panagiotakos G, Greenberg ED et al. Migration and differentiation of neural precursors derived from human embryonic stem cells in the rat brain Nat Biotechnol. 2005;23:601–606.

  61. Brüstle O, Spiro AC, Karram K et al. In vitro-generated neural precursors participate in mammalian brain development. Proc Natl Acad Sci U S A 1997;94:14809–14814.