Cortical cultures coupled to Micro-Electrode Arrays: A novel approach to perform in vitro excitotoxicity testing

Neurotoxicology and Teratology 34 (2012) 116–127

Contents lists available at SciVerse ScienceDirect

Neurotoxicology and Teratology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / n e u t e r a

Cortical cultures coupled to Micro-Electrode Arrays: A novel approach to perform in vitro excitotoxicity testing Monica Frega a, 1, Valentina Pasquale a, 1, Mariateresa Tedesco b, Manuela Marcoli c, Andrea Contestabile a, Marina Nanni a, Laura Bonzano d, Guido Maura c, Michela Chiappalone a,⁎ a

Department of Neuroscience and Brain Technologies, Italian Institute of Technology, Genova, Italy Neuroengineering and Bio-nano Technology Group (NBT), Department of Biophysical and Electronic Engineering (DIBE), University of Genova, Genova, Italy Department of Experimental Medicine and Center of Excellence for Biomedical Research, University of Genova, Genova, Italy d Department of Neurosciences, Ophthalmology and Genetics and Magnetic Resonance Research Centre on Nervous System Diseases, University of Genova, Genova, Italy b c

a r t i c l e

i n f o

Article history: Received 21 January 2011 Received in revised form 3 July 2011 Accepted 4 August 2011 Available online 11 August 2011 Keywords: In vitro neurotoxicity Spiking Bursting Synchronization Long-term Glutamatergic transmission

a b s t r a c t In vitro neuronal cultures exhibit spontaneous electrophysiological activity that can be modulated by chemical stimulation and can be monitored over time by using Micro-Electrode Arrays (MEAs), devices composed by a glass substrate and metal electrodes. Dissociated networks respond to transmitters, their blockers and many other pharmacological substances, including neurotoxic compounds. In this paper we present results related to the effects, both acute (i.e. 1 hour after the treatment) and chronic (3 days after the treatment), of increasing glutamatergic transmission induced by the application of rising concentrations of glutamate and its agonists (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid — AMPA, N-methyl-Daspartate — NMDA and AMPA together with cyclothiazide — CTZ). Increase of available glutamate was obtained in two ways: 1) by direct application of exogenous glutamate and 2) by inhibiting the clearance of the endogenously released glutamate through DL-threo-β-benzyloxyaspartate (TBOA). Our findings show that fine modulations (i.e. low concentrations of drug) of the excitatory synaptic transmission are reflected in the electrophysiological activation of the network, while intervention leading to excessive direct stimulation of glutamatergic pathways (i.e. medium and high concentrations of drug) results in the abolishment of the electrophysiological activity and eventually cell death. The results obtained by means of the MEA recordings have been compared to the analysis of cell viability to confirm the excitotoxic effect of the applied drug. In conclusion, our study demonstrates that MEA-coupled cortical networks are very sensitive to pharmacological manipulation of the excitatory ionotropic glutamatergic transmission and might provide sensitive endpoints to detect acute and chronic neurotoxic effects of chemicals and drugs for predictive toxicity testing. © 2011 Elsevier Inc. All rights reserved.

1. Introduction Primary neuronal cultures dissociated from different regions of the Central Nervous System (CNS) have been a classical model for in vitro electrophysiological studies of single neuron's neurobiological mechanisms (Dichter and Fischbach, 1977; Sakmann and Neher, 1984). Developed at the beginning of the '80s (Gross et al., 1977; Pine, 1980), the Micro-Electrode Array (MEA) technique nowadays offers a useful experimental approach for in vitro electrophysiological investigations. MEA technology and culture methods, in parallel, have continuously improved during these years. To date, MEAs find several

⁎ Corresponding author at: NeuroTech Group, Department of Neuroscience and Brain Technologies (NBT), Italian Institute of Technology (IIT), Via Morego 30, 16163 Genova, Italy. Tel.: + 39 010 71781 743; fax: + 39 010 720321. E-mail address: [email protected] (M. Chiappalone). 1 Equal contribution. 0892-0362/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.ntt.2011.08.001

applications in many research fields, such as neuroscience, pharmacology, physiology, biophysics and cardiac electrophysiology. In particular, the pioneering works by Gross and co-workers demonstrated the possibility to use dissociated neuronal networks coupled to MEAs as a first prototype of cell-based biosensor (Gross et al., 1977, 1992). This system showed both high sensitivity to neuroactive and neurotoxic compounds and reproducible results (Gramowski et al., 2000). The biocompatibility of the used materials (i.e. titanium nitride for the electrodes and glass for the substrate) and the non invasive nature of the extracellular measurement, make this system a perfect candidate to routinely record and evaluate the dynamics of the network behavior, both in spontaneous condition and under chemical manipulation (Chiappalone et al., 2006; Martinoia et al., 2005), either on short or long time-scales. This experimental system, that constitutes on a mesoscopic level a simplified model of a cortical tissue, allows to explore network properties, while preserving the morphological, molecular and functional properties of the individual neurons in the cortex (Marom and Shahaf, 2002).

M. Frega et al. / Neurotoxicology and Teratology 34 (2012) 116–127

Excitatory neurotransmission in the brain is predominantly mediated by glutamate release via synaptic transmission. Glutamate activates ionotropic (ligand-gated) receptors of both NMDA and AMPA/kainate subtypes (Collingridge and Lester, 1989) and metabotropic (G-protein coupled) receptors. While synaptic glutamatergic transmission onto ionotropic receptors plays a critical role during brain development and in plasticity, memory, learning processes and cognition in adult brain (Antzoulatos and Byrne, 2004; Meldrum, 2000; Riedel et al., 2003), stimulation of ionotropic glutamate receptors may also activate neuro-excitotoxic cascades (Rothman and Olney, 1995, 1987). In this study, we used cortical cultures coupled to MEAs to investigate the short/long term effects of increasing concentrations of ionotropic glutamate's agonists, which could cause excitotoxicity. Cell viability at different time points after glutamate agonist treatment was evaluated by fluorescein diacetate uptake by viable cells and propidium iodine staining of dead cells. In the attempt to get insight on the effects directly dependent on activation of a specific ionotropic glutamate receptor subtype, we selectively activated ligand-gated glutamatergic receptor subtypes by using glutamate agonists (NMDA, AMPA and AMPA together with cyclothiazide — CTZ) or added exogenous glutamate, at different concentrations to modulate the dynamics of cortical networks. Then, in order to rapidly enhance the extracellular concentration of endogenous glutamate, we applied DLthreo-β-benzyloxyaspartate (TBOA), a glutamate uptake inhibitor (Jabaudon et al., 1999; Shimamoto et al., 1998); moreover, to contribute to elucidation of mechanistic issues, the glutamate concentration in the extracellular medium (basal and drug phase) was evaluated by high-performance liquid chromatography (HPLC) analysis, as a measure of the activation of glutamatergic transmission. We show that high concentrations of either glutamate agonists or glutamate uptake inhibitor are responsible of a complete loss of activity in our cortical cultures (due to the excitotoxic effect of glutamate which brings to cell death, as proven by the live/dead assay results), while low concentrations of the same compounds bring network activity to a stable level. These results will be discussed in the view of homeostatic mechanisms such as receptor desensitization and glutamate clearance, responsible of the maintenance of the balance between excitatory and inhibitory activity (Watt et al., 2004). On the other hand, due to high sensitivity to ionotropic glutamate receptor activation and to the ability to detect interference with glutamatergic transmission, suitability of MEA-coupled cortical cultures as a candidate for in vitro neurotoxicity testing would deserve to be evaluated. 2. Materials and methods 2.1. Cell culture Dissociated neuronal cultures were obtained from cortices of embryonic rats. Pregnant rats (Sprague–Dawley derived by Charles River in 1955, IGS) were deeply anesthetized and sacrificed by inhalation of CO2: 18-day embryos (E18) were removed immediately by cesarean section, and killed by decapitation. All experimental procedures and animal care were conducted in conformity with institutional guidelines, in accordance with the European legislation (European Communities Directive of 24 November 1986, 86/609/EEC) and with the NIH Guide for the Care and Use of Laboratory Animals. Culture preparation was performed as previously described (Bologna et al., 2010a). Briefly, the cerebral cortices of 4–5 embryos were dissected out from the brain and dissociated first by enzymatic digestion in trypsin solution 0.125% (20 min at 37 °C) and subsequently by mechanical dissociation with a fine-tipped Pasteur pipette. The resulting tissue was resuspended in Neurobasal medium supplemented with 2% B-27, 1% Glutamax-I, 1% Pen-Strep solution and 10% Fetal Bovine Serum (Invitrogen, Carlsbad, CA), at the final

117

concentration of 36–40 k cells/ml. Cells were afterwards plated onto 60-channel MEAs previously coated with poly-D-lysine and laminin to promote cell adhesion (final density around 1200 cells/mm 2) and maintained with 1 ml of nutrient medium (i.e. serum-free Neurobasal medium supplemented with B27 and Glutamax-I). They were then placed in a humidified incubator having an atmosphere of 5% CO2-95% air at 37 °C. Half of the medium was changed weekly. 2.2. Micro-Electrode Array recordings Microelectrode arrays (Multichannel Systems, MCS, Reutlingen, Germany) consisted of 59 TiN/SiN planar round electrodes (30 μm diameter; 200 μm center-to-center interelectrode distance) arranged in an 8 × 8 square grid excluding corners. One recording electrode was replaced by a bigger ground electrode. The activity of all cultures was recorded by means of the MEA60 System (MCS). After 1200× amplification, signals were sampled at 10 kHz and acquired through the data acquisition card and MC_Rack software (MCS). To reduce thermal stress of the cells during the experiment, MEAs were kept at 37 °C by means of a controlled thermostat (MCS) and covered by PTFE lid (ALA Scientific Instruments, NY, USA) to avoid evaporation and prevent changes in osmolarity. 2.3. Experimental protocols The general protocol adopted for chemical stimulation included 30 minutes of recording in culture solution (Neurobasal (Invitrogen cat.no. 21103049) + 2% B27 (Invitrogen cat.no. 17504044) + 1% Glutamax-I 200 mM (Invitrogen cat.no. 35050038) + 1% PenicillinStreptomycin sol. (Invitrogen cat. no. 15140122)), defined as control condition. All drugs were added to the bath solution at increasing concentrations by directly pipetting in the medium. For each concentration, the electrophysiological activity was recorded for 30 min. Since we noticed that mechanical perturbation due to the pipette injection in the medium could cause a temporary instability of the firing rate, we discarded the first 10 min of each recording phase. According to this choice, the presented data refer to a recording period of 20 min for each experimental phase. All tested networks were measured between the 3rd and the 4th week in vitro. 2.3.1. Glutamate and agonists of the glutamatergic receptors α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and N-methyl-D-aspartate (NMDA) are agonists of ionotropic glutamatergic receptors at the post-synaptic site. We performed experiments by adding to the culture medium progressively increasing concentrations of each drug at intervals of 30 min. More specifically, we tested low (i.e. 0.05–0.2 to 1–5 μM), medium (i.e. 10 – 30 μM) and high (i.e. 50– 100 μM) concentrations of AMPA/NMDA to different sets of cultures (N= 4 cultures for AMPA and N = 5 cultures for NMDA). At the same time, we tested the effects of increasing doses of glutamate (0.05 - 0.2 1 – 5–10 – 30 – 50–100 – 500 μM, N = 4 cultures). We mainly based our choices on the available literature by ours and other groups in the MEA field (Gramowski et al., 2006, 2000; Gross et al., 1992; Martinoia et al., 2005). 2.3.2. Inhibitor of AMPA receptor desensitization Following activation, many ligand-gated ion channels enter in a desensitized state in which the neurotransmitter remains bound but the ion channel is closed (Sun et al., 2002). Both AMPA and NMDA receptors are subject to desensitization, with a different degree also related to the activating agonist. To prevent the desensitization of AMPA receptors, we used cyclothiazide (CTZ) at different concentrations (Trussell et al., 1993). We treated a subset of our cultures with either low (i.e. 1–3 μM) or high (i.e. 10–30 μM) concentrations of CTZ, combined with AMPA at a fixed concentration of 0.2 μM (N = 3 cultures).

118

M. Frega et al. / Neurotoxicology and Teratology 34 (2012) 116–127

2.3.3. Inhibitor of glutamate uptake In order to maintain physiological activity (in vitro as well as in vivo) and avoid glutamatergic excitotoxicity, a specific mechanism such as glutamate clearance from the synaptic cleft is required. DLthreo-β-benzyloxyaspartate (TBOA) is a blocker of glutamate clearance (Shimamoto et al., 1998), a process by which released neurotransmitters are absorbed for subsequent re-utilization by neurons and glia. Since we noticed that the main effects, in terms of change in the electrophysiological activity, were seen at high and low concentrations of the glutamate agonists, TBOA was administered at 100 μM (high dose, N = 3 cultures) or 5 μM (low dose, N = 3 cultures). Aliquots of the bath solutions at different recording phases were collected and rapidly frozen for the determination of the extracellular fraction of endogenous glutamate (see further).

involved automatic precolumn derivatization (Waters 715 ultra wisp; Milford, MA, USA) with o-phthalaldehyde, followed by separation on a C18 reverse-phase chromatography column (Chrompack International, 10 cm × 4.6 mm, 3 μm) and fluorimetric detection. Homoserine was used as an internal standard. The detection limit was 100 fmol/μl. The amount of endogenous glutamate was expressed as pmol/ml. For each culture, the effects of drugs on the endogenous glutamate efflux were measured as percent variation of endogenous glutamate amount in the presence of drugs with respect to the corresponding control value in the absence of drugs (i.e. the glutamate amount in the basal period immediately preceding the drug-exposure period). For endogenous glutamate level, means ± SEM of the numbers of experiments (n) are indicated. Significance of the difference between related pairs of values was analyzed by the Student's t-test; the level of significance was set at p b 0.05.

2.4. Live/dead assay 2.6. Data analysis and statistics Cell viability was evaluated with a modified fluorescein diacetate/propidium iodine staining protocol (Jones and Senft, 1985). In brief, cells were stained for 3 min at room temperature with 15 μg/ml fluorescein diacetate, 5 μg/ml propidium iodine and 3.3 μg/ml Hoechst-33342 (all from Sigma, Milan, Italy) in Tyrode solution (NaCl 140 mM, KCl 4 mM, CaCl2 2 mM, MgCl2 1 mM, glucose 10 mM, Hepes 10 mM; pH 7.4). Cells were then washed in Tyrode solution and immediately imaged by fluorescent microscopy with a Nikon upright microscope (Nikon Corp., Kawasaki, Japan) equipped with a 16× water-immersion objective. The cell-permeable green fluorescent dye fluorescein diacetate (Excitation maximum: 488; Emission maximum: 513 nm) is only retained in viable cells with intact cell membrane and ongoing esterase enzymatic activity while the cellimpermeable red fluorescent dye propidium iodine (Excitation maximum: 536; Emission maximum: 620 nm) can only enter dead cells were cell membrane integrity is compromised. The cellpermeable blue fluorescent nuclear dye Hoechst-33342 (Excitation maximum: 352; Emission maximum: 454 nm) enters both viable and dead cells and is used to visualize the total number of cells. For the assay, we used a total of 18 cultures in three different experimental conditions (i.e. 6 cultures for each condition): 1) control; 2) NMDA treatment; 3) AMPA treatment. The experimental conditions were monitored during two time points: i) 1 hour after the treatment, to observe the acute effects of the drug; ii) 72 hours (three days) after the treatment, to observe the chronic effects of the drug. A total of 3 cultures were used for each condition at one specific time point. The treatment reserved to these cultures was exactly the same adopted for the MEA recordings. Each culture underwent an experiment in which increasing concentrations of the selected drug were added at intervals of 30 min (cf. Section 2.3.1). Once the final concentration of drug was reached (i.e. 500 μM AMPA or NMDA), the medium was changed and fresh medium was added to the culture. The culture was put into the incubator for one hour and then the staining protocol was initiated. The above explained procedure has been performed also in a set of three cultures treated with AMPA + CTZ (at the final concentration of 200 nM AMPA + 30 μM CTZ, cf. Section 2.3.2) to observe only the acute effect of the drug cocktail. At the end of the procedure, we selected 8 random fields for each sample, obtaining a total of 24 images per condition per time point. Fluorescent images were taken with an Orca CCD camera (Hamamatsu Photonics, Hamamatsu City, Japan). Images were then analyzed with ImageJ software (http://rsbweb.nih.gov/ij/) to count the total cell number and the number of live and dead cells. 2.5. Endogenous glutamate determination The amount of endogenous glutamate in the collected aliquots was measured by high-performance liquid chromatography (HPLC), as previously described (Marcoli et al., 2008). The analytical method

Data analysis was performed off-line by using a custom software developed in MATLAB© (The Mathworks, Natick, MA, USA) named SPYCODE (Bologna et al., 2010b), which collects a series of tools for processing multi-channel neural recordings. Briefly, data were imported in MATLAB from .mcd files (MCS format) and then spike-detected using the PTSD (Precise Timing Spike Detection) algorithm, whose procedure has been previously reported (Maccione et al., 2009). Afterwards, spike trains were analyzed by using a custom burst detection method (Pasquale et al., 2010), whose input parameters were directly estimated from the inter-spike interval distribution of each channel. Once spike and burst detection procedures were performed, several measures describing the electrophysiological patterns could be extracted, such as firing rate (mean number of spikes per second calculated on the active channels of the network), bursting rate (mean number of burst per minute calculated on the active channels of the network), burst duration (mean burst length expressed in ms) and the mean frequency within the burst (MFB, mean number of spikes per second within a burst). By applying a custom algorithm (Pasquale et al., 2010), we were able to identify the number of synchronized bursts involving all the active channels (i.e. network bursts). If not differently reported, the above parameters were normalized for each experiment with regard to the corresponding values of the reference (basal) activity (native condition, for which the initial value for each experiment was 100%) for direct comparability. All the above parameters were extracted solely for the active channels, which are those channels displaying a firing rate above a minimum threshold of 0.0008 spikes/s, i.e. at least 1 spike in 20 min of acquisition. For the agonists' experiments, the changing firing rate as a function of the concentration was fitted by the Hill equation which has long been proposed as a model to analyze nonlinear drug concentration–response relationships (Wagner, 1968): y = ySTART + yEND −ySTART




1 + 10

½ loglC50 − logxHC


ð1Þ

where y is the observed value, ySTART is the highest observed value, yEND is the lowest observed value (usually at the highest used concentration), and HC the Hill coefficient which gives the largest absolute value of the slope of the curve. From this model the IC50 values (half maximal inhibitory concentration) were determined (using the fitting tool from OriginPro, OriginLab Corporation, Northampton, MA, USA) for AMPA, NMDA and GLU and in each experiment. In order to highlight other possible changes on the spike/burst patterns of the network, a cross-correlation algorithm was applied to spike trains (Chiappalone et al., 2006, 2007). This approach is necessary in order to identify synchronization mechanisms (i.e. synchronized firing) induced by specific concentrations of a drug.

M. Frega et al. / Neurotoxicology and Teratology 34 (2012) 116–127

The cross-correlation function (i.e. cross-correlogram) is defined by the incidence of a spike at electrode y after that a spike was fired at electrode x. More specifically, given two spike trains (i.e. x and y) from two electrodes of a MEA, we count the number of spikes in the y train within a time frame around the spikes of the x train of ±T (in the order of tens of milliseconds), using bins of amplitude Δτ (usually set at multiple of the sampling frequency). The correct Cxy(τ) is obtained by means of a normalization procedure, by dividing each element of the array by the square root of the product between the number of peaks in the x and the y train. If the obtained Cxy(τ) shows a distribution that clearly deviated from flat, electrodes x and y are considered correlated. For each cross-correlogram Cxy(τ) we then estimated the coefficient Cpeak according to the definition below: τpeak + kðΔτ = 2Þ

Cpeak =

∑

τ = τpeak −kðΔτ = 2Þ

Cxy ðτÞ

where τpeak indicates the position of the maximum peak of the crosscorrelogram, Δτ is the bin size of the correlogram function and k is the number of bins around the peak. Cpeak represents the value of the

119

cross-correlogram in an area around the maximum detected peak and it is usually evaluated in order to quantify the correlation level among two recording channels (see Suppl. Fig. 1). The statistical distribution of all Cpeak values was computed during each recording phase (e.g. in control condition and under drug treatment) to highlight differences in the overall cross-correlation level due to chemical stimulation. Data within the text are expressed as means ± standard error of the mean (se). Statistical tests were employed to assess the significant difference among different experimental conditions. The normal distribution of experimental data was assessed using the Kolmogorov-Smirnov normality test. According to the distribution of the data, we performed either parametric (e.g. t-test or ANOVA) or non parametric (e.g. Mann–Whitney U test or Kruskal–Wallis multiple comparison) tests and p values b 0.05 were considered significant. Statistical analysis was carried out by using OriginPro (OriginLab Corporation, Northampton, MA, USA) and Sigma Stat (Systat Software Inc., San Jose, CA, USA). More specifically, for the statistics used in the data reported in Figs. 2, 3 and 4, we adopted a ‘One Sample t-Test’, for which we made the following null hypothesis: Mean bN 100 (where 100 is the level of the ‘basal’ condition). Being the alternative hypothesis equal to ‘Mean = 100’, we considered as statistically significant either an

Fig. 1. Cortical neurons plated on MEAs form high-dense active networks during development. A. DIC (Differential Interference Contrast) image of a 27-DIV culture on top of a 60electrode (8 × 8) MEA (10X). B. Sample trace recorded from a single micro-electrode: bursts of high-frequency action potentials are mixed with isolated spikes (top inset: zoom on a single burst, where the detected spikes are marked in black; bottom inset: zoom on a single isolated spike). C, D. Neurofilament staining of a cortical network of 28DIV cultured over a MEA. The recording electrodes of the MEA are clearly visible in C (i.e. the big round circles), while the parallel strips in D are the electrode pathways. The reported images qualitatively show the tangle structural framework built by neurons after 4 weeks of culturing over the MEA device.

120

M. Frega et al. / Neurotoxicology and Teratology 34 (2012) 116–127

Fig. 2. Effects of low (0.05, 0.2, 1, 5 μM), medium (10, 30 μM) and high (50, 100 μM) doses of AMPA on the firing and bursting dynamics of cortical networks. A. AMPA was applied in the culture medium at increasing concentrations and the effects on the network activity are clearly visible comparing the raster plots (30 s of activity acquired from a representative cortical culture of 28 DIV) relative to different experimental phases (30 out of 60 electrodes per experimental phase are depicted: each small vertical bar represents a spike, each line an electrode). B. We found that in the analyzed group of experiments (N = 4), the firing rate significantly increases for very low concentrations of AMPA (i.e. 50 nM, 200 nM and 1 uM), while it significantly decreases starting from AMPA 5 μM till 100 μM. Inset. Fitting of the dose-response curve by means of the Hill equation, starting from the maximum obtained values of the firing rate (i.e. 200 nM and 1 uM). IC50 value is reported in the text. C. The bursting rate follows a profile very similar to that observed for the firing rate, showing a significant increase for low doses of the drug (50 nM, 200 nM and 1 uM). D. Mean frequency of spikes within the bursts (MFB) statistically decreases at AMPA 1 μM and for concentrations higher than 10 μM. E. Burst duration has a significant increase for 1 μM AMPA, but it drops down from 10 μM on. F. The number of network bursts does not change significantly till AMPA 1 μM, and then it starts decreasing. Statistics presented in panels B, C, D, E, F was done by using one-sample t-test, *p b 0.05. G. The correlation peak, calculated on the cross-correlograms between couples of spike trains, starts to statistically decrease at 200 nM (Kruskal–Wallis one way analysis of variance on ranks, Dunn's Method for multiple comparison versus basal group, **p b 0.001).

increase (black asterisk) or a decrease (gray asterisk) of a specific parameter with respect to the test mean.

3. Results To study the acute and chronic effects of an increase of glutamate and its agonists, we analyzed the changes exhibited by the electrophysiological activity of cortical networks of primary neurons prepared from E18 rat embryos and plated onto standard MEAs (Fig. 1A). After a couple of weeks in culture, cells started showing a typical electrophysiological pattern which comprised strongly synchronized bursts of action potentials, mixed with desynchronized random firing (Fig. 1B). Neurofilament staining performed on a 28 DIV culture qualitatively shows a good level of connectivity among the cells (Fig. 1C–D).

Fig. 2 shows the main effects on the electrophysiological patterns due to the administration of increasing concentrations of AMPA. Looking at the raster plot of a representative experiment (Fig. 2A, each vertical bar represents a spike), we can notice that low doses of the agonist (50 nM, 200 nM and 1 μM) increase the level of activity. At those concentrations, the electrophysiological activity is characterized by the presence of both spikes and bursts of action potentials. Moreover, the majority of the electrical activity takes place as part of network-wide collective events, known as ‘network bursts’ (i.e. bursts fired at the same time by most of the recording channels) (Pasquale et al., 2010). At 5 μM, only few channels are still active: they are not synchronized and network bursts are missing. At higher concentrations (above 10 μM) the electrophysiological behavior remains similar, indicating that medium and high doses of AMPA abolish most of the spontaneous electrophysiological activity of cortical networks.

M. Frega et al. / Neurotoxicology and Teratology 34 (2012) 116–127

121

Fig. 3. Effects of low (0.05, 0.2, 1, 5 μM), medium (10, 30 μM) and high (50, 100 μM) doses of NMDA on the firing and bursting dynamics of cortical networks. A. NMDA was applied in the culture medium at increasing concentrations and the effects on the network activity are clearly visible comparing the raster plots (200 s of activity acquired from a representative cortical culture of 26 DIV) relative to different experimental phases (15 out of 60 electrodes per experimental phase are depicted: each small vertical bar represents a spike, each line an electrode). B. We found that in the analyzed group of experiments (N = 5), the firing rate significantly increases for low concentrations of NMDA (i.e. 1 μM and 5 μM), while it does not change for lower (50 and 200 nM) and medium concentrations (10, 30, and 50 μM). A significant decrease is observed for 100 μM NMDA. Inset. Fitting of the dose-response curve by means of the Hill equation, starting from the maximum obtained values of the firing rate (i.e. 5 μM). IC50 value is reported in the text. C. The bursting rate does not change till NMDA 30 μM, when it starts significantly decreasing. D. Mean frequency of spikes within bursts starts to statistically decrease at NMDA 5 μM. E. Burst duration has a significant decrease from NMDA 30 μM on. F. The number of network bursts does not change significantly till NMDA 5 μM, and then it decreases from 10 μM on. Statistics presented in panels B, C, D, E, F was done by using the one-sample t-test, *p b 0.05. G. The correlation peak, calculated on the cross-correlograms between couples of spike trains, starts to statistically decrease at 1 μM (Kruskal–Wallis one way analysis of variance on ranks, Dunn's Method for multiple comparison versus basal group, **p b 0.001).

To quantify our results, we derived a set of dose-response curves based on different spike/burst parameters. Low doses of AMPA induce an increase of spike and burst rates, which reach a maximum at the concentration of 200 nM - 1 μM (Fig. 2B, C). The IC50 calculated for AMPA is equal to 2.8 ± 0.9 μM [N = 4 exp] (Fig. 2B, inset). The bursting pattern tends to remain stable in terms of both MFB (frequency within bursts) and burst duration, apart from a drop at 1 μM for MFB (Fig. 2D) and a peak at 1 μM for burst duration (Fig. 2E). These results indicate that, for medium concentrations, the few bursting channels display longer bursts characterized by high frequency spikes. We also quantified the changes in number of generated network bursts, whose number remains statistically stable and then drops down at AMPA 5 μM (Fig. 2F). The correlation level (Fig. 2G, measured as the height of the correlogram peak computed from the spike trains of each pair of electrodes, see Materials and methods) does not significantly change at 50 nM, while it starts to significantly decrease at higher doses of AMPA (from 200 nM to 5 μM), falling down to zero from 10 μM on. Upon a complete medium change, we noticed a total depression of the activity, with no recovery during an acquisition time of 60 min. After 1 day in the incubator we again measured our cultures, noting a very slow assessment of the firing rate to low levels (firing

rate = 0.27 ± 0.2 spikes/s, equal to 6.27 ± 4.47% with respect to the native state). An even slighter increase was observed 3 days after the AMPA treatment (firing rate = 1.30 ± 0.82 spikes/s, equal to 30.27 ± 17.94% with respect to the native state), indicating that the long term effect of the drug consists in a reduction (not a total abolishment) of the initial activity, but with a complete loss of synchronized patterns (bursts and network bursts). NMDA causes similar effects to those due to AMPA administration (Fig. 3A). The electrophysiological pattern does not change for very low doses of NMDA (50 and 200 nM), while it appears different at 1 and 5 μM, when the network is characterized by less bursts and an increased level of firing. The activity drops down at 10 μM NMDA, when few spiking channels are still active but the synchronized patterns totally disappear. The change in firing rate profile indicates a clear-cut increase of the activity at 1–5 μM NMDA and a fast decrease for higher doses of the drug (Fig. 3B). The obtained IC50 is equal to 33.4 ± 15.7 μM [N = 5 exp] (Fig. 3B, inset). The bursting rate does not change statistically till NMDA 30 μM (Fig. 3C). MFB tends to decrease starting from 10–30 μM NMDA (Fig. 3D), while burst duration does not change significantly for low-medium doses (Fig. 3E), meaning that bursts have lengths similar to the native state with lower intra-burst frequency. Synchronized activity, measured in terms of change in the

122

M. Frega et al. / Neurotoxicology and Teratology 34 (2012) 116–127

Fig. 4. Effects of low (0.05, 0.2, 1, 5 μM), medium (10, 30 μM) and high (50, 100, 500 μM) doses of glutamate on the firing and bursting dynamics of cortical networks. A. Glutamate was applied in the culture media at increasing concentrations and the effects on the network activity are clearly visible comparing the raster plots (200 s of activity acquired from a representative cortical culture of 25 DIV) relative to different experimental phases. (15 out of 60 electrodes per experimental phase are depicted: each small vertical bar represents a spike, each line an electrode). B. We found that in the analyzed group of experiments (N = 4), the firing rate significantly increases at glutamate 5 μM, while it does not change for the other doses. A significant decrease is observed only at 500 μM. Inset. Fitting of the dose-response curve by means of the Hill equation, starting from the maximum obtained values of the firing rate (i.e. 5 μM). IC50 value is reported in the text. C. The bursting rate statistically increases only at glutamate 5 μM, dropping down at 500 μM. D. Mean frequency of spikes intra bursts starts to statistically decrease at glutamate 100 μM. E. Burst duration has a significant decrease for glutamate 50 nM, 1 μM and 5 μM, then it falls down at 500 μM. F. The number of network bursts significantly decreases for glutamate 50 and 200 nM, then it does not change with respect to the basal condition for low and medium doses (1-5-10-3050 μM), but it drops down at 100 and 500 μM. Statistics presented in panels B, C, D, E, F was done by using the one-sample t-test, *p b 0.05. G. The correlation peak, calculated on the cross-correlograms between couples of spike trains, statistically increases for glutamate 50 nM, 200 nM, 1 μM, 5 μM and 10 μM, then it significantly drops down starting from 50 μM (Kruskal–Wallis one way analysis of variance on ranks, Dunn's Method for multiple comparison versus basal group, **p b 0.001).

number of network bursts, totally disappears at 10 μM NMDA (Fig. 3F), as also demonstrated by the correlation value calculated on spike trains (Fig. 3G). After a double wash-out from NMDA treatment, we completely lost the few spiking channels. No recovery, either in terms of spikes or bursts, was observed after 1 and 3 days (firing rate = 0 spikes/s). To complete our analysis on the effects of glutamate agonists on cortical cultures, we made experiments employing exogenous glutamate, to be added to the culture medium. Looking at the raster plots reported in Fig. 4A, it is worth noting that glutamate administration slightly changes the native electrophysiological pattern. Fluctuations in the level of the firing rate are visible (see also Fig. 4B), but the overall activity does not qualitatively change, apart from a statistical increase of the activity at glutamate 5 μM and a drop at 500 μM. The fitted dose– response curve produces an IC50 equal to 137.8 ± 40.3 μM [N= 4 exp] (Fig. 4B, inset). A clear-cut change can be noticed at medium concentrations (from 30 to 50 μM), when ‘superburst’ patterns (i.e. groups of network bursts) can be observed (Wagenaar et al., 2006). Also

the bursting rate presents a statistical increase at the dose of 5 μM (Fig. 4C). From medium doses the MFB tends to decrease, reaching the minimum (around zero) for very high concentrations (500 μM), thus indicating the total absence of bursts (Fig. 4D). After a slight decrease for low doses, the burst duration stays at the initial levels till 50–100 μM (Fig. 4E), showing a profile similar to that obtained for network burst number (Fig. 4F). The correlation level (Fig. 4G) significantly increases with respect to the native state (i.e. from 50nM to 10 μM) till Glu 30 μM (no statistical difference), then it significantly decreases at higher doses (50 and 100 μM), dropping to zero at 500 μM. Also in the case of glutamate treatment, after a double wash-out of the culturing medium, we observed a complete loss of activity, which showed no recovery after either 1 or 3 days (firing rate = 0 spikes/s). To test the effect of a prolonged activation of the AMPA receptor, we decided to perform a set of experiments by employing cyclothiazide (CTZ). We combined increasing concentrations of CTZ with a constant concentration of AMPA. We used a very low dose (i.e. 200 nM) of the AMPA agonist since we knew that it caused a slight increase in the firing

M. Frega et al. / Neurotoxicology and Teratology 34 (2012) 116–127

activity (see Fig. 2) and, under this condition, the effect of CTZ administration could be better highlighted. For a qualitative judgment, let us consider the raster plots presented in Fig. 5A. The effect of low doses of CTZ is striking: either at 1 or 3 μM, there is an evident increase in the bursting rate with respect to the basal condition. The effect drastically changes when the concentration of CTZ rises to 10 μM, during which there is a progressive loss of activity without the presence of bursting, up to 30 μM. In this last condition the network is silent, only isolated spikes can be found on few channels, completely desynchro-

123

nized, indicating that the collective activity is lost. The effect of the low doses of CTZ is also proven by the bar graphs of Fig 5B, which demonstrates a significant increase of firing (and bursting) rate and a significant decrease of burst duration. The reported results refer to a set of three cultures treated with AMPA 200 nM combined with CTZ at either 1 or 3 μM. Looking at the array-wide firing rate (i.e. the profile of the firing rate evaluated on all the channels at the same time) of Fig. 5C, CTZ at low concentrations (1 or 3 μM, combined with 200 nM AMPA) induces

Fig. 5. Increasing concentrations of CTZ cause a complete abolishment of electrophysiological activity. A. Raster plot of 30 electrodes (out of 60) coming from a set of MEAs treated with different doses of CTZ added to a constant concentration of AMPA (0.2 μM). More specifically: basal condition, AMPA 0.2 μM + CTZ 1 μM, AMPA 0.2 μM + CTZ 3 μM, AMPA 0.2 μM + CTZ 10 μM, AMPA 0.2 μM + CTZ 30 μM. Note that at low doses of CTZ (1, 3 μM) there is a clear-cut increase in bursting activity, while at high doses there is a progressive (10 μM) or abrupt (30 μM) depression. B. Bursting rate during basal condition and under administration of low doses of CTZ. There is a statistically significant increase in the bursting rate (Mann–Whitney rank sum test, *p b 0.001) and the percentage of increase is 251.30 ± 6.41% (mean ± SEM). Under low doses CTZ administration, there is a statistically significant decrease in the burst duration (Mann–Whitney rank sum test, *p b 0.001) and the percentage of change is 62.73 ± 5.47% (mean ± SEM). C. Array wide firing rate during time course of a representative experiment of AMPA 0.2 μM + CTZ 1 μM (inset, zoom of the highlighted areas — black: spontaneous activity; gray: AMPA 0.2 μM + CTZ 1 μM treatment). Low doses of CTZ (1 or 3 μM) combined with low doses of AMPA (0.2 μM) induce transient periods of high-frequency short bursts. Bin size = 100 ms. D. Array wide firing rate during time course of a representative experiment of AMPA 0.2 μM + CTZ 30 μM (inset, zoom of the highlighted areas – black: spontaneous activity; gray: AMPA 0.2 μM + CTZ 30 μM treatment). High doses of CTZ completely abolish the electrophysiological activity. Bin size = 100 ms.

124

M. Frega et al. / Neurotoxicology and Teratology 34 (2012) 116–127

transient periods of closely packed short bursts (gray profile). In Fig 5D, we can see that CTZ at high concentrations (10–30 μM, combined with 200 nM AMPA) completely abolishes the spontaneous electrophysiological activity, obtaining a result similar to that obtained with high doses of either AMPA or NMDA. To check whether the depression induced by the increase in the level of glutamate's agonists is a temporary or an irreversible process, we tested the viability of our cells by using a modified fluorescein diacetate/propidium iodine staining protocol (cf. Materials and methods). The results of this procedure are reported in Suppl. Fig. 2. To further confirm the effect of high doses of glutamate or its agonists on the network activity, we decided to employ an inhibitor of the glutamate uptake, named TBOA (see methods). High doses of TBOA (100 μM) suppress the electrophysiological activity in a progressive way (Fig. 6A, B), as confirmed by the analysis of the firing rate which significantly decreases after the treatment with the drug (Fig. 6C). HPLC analysis reveals that the addition of 100 μM TBOA also increased the extracellular endogenous glutamate concentration of a percentage equal to 80% (Fig. 7D). Low doses of TBOA (5 μM) do not affect the spontaneous firing and bursting dynamics (Fig. 7A, B) of cortical neurons, but burst duration is significantly increased (Fig. 7C). Also synchronization is not affected (Fig. 7D), indicating that burst elongation is the only consequence of glutamate increase (due to the block of its uptake). From a quantitative point of view, HPLC analysis reveals that TBOA at 5 μM is responsible of the increase of extracellular glutamate, but in a concentration-dependent manner. At 5 μM (Fig. 7D), in fact, TBOA evokes an increase of the extracellular endogenous glutamate by about 30% instead of 80%, as obtained for higher concentrations. 4. Discussion Dysregulation of glutamatergic transmission, involving increased extracellular glutamate levels and subsequent excessive stimulation of NMDA or AMPA/kainate receptors, is responsible for excitotoxic

neuronal injury, one of the major pathological factors leading to neuronal death. Excitotoxic mechanism activation represents a common way for neuron damage shared by neuroinflammatory and neurodegenerative conditions such as ischemia, epilepsy, amyotrophic lateral sclerosis, Huntington's and Parkinson's disease (Chapman, 2000; Chase and Oh, 2000; Danbolt, 2001; Maragakis and Rothstein, 2001; Tzschentke, 2002). In this study we investigated the effects of increasing glutamatergic transmission induced by pharmacological manipulation acting on the main ionotropic glutamatergic receptors (i.e. AMPA and NMDA receptors) and on inhibition of glutamate clearance through TBOA. To perform our experiments, we used dissociated cortical cultures between 20 and 30 DIV plated onto standard MEAs, a well-suited approach for quantitative multi-parametric analyses of physiologically functional neuronal networks (Johnstone et al., 2010). In addition, when we inhibited the glutamate uptake we also measured actual extracellular level of endogenous glutamate. To the best of our knowledge, this is the first time that such a thorough study, i.e. comparative evaluation of the effects of exogenously added glutamate receptor agonists and of increasing the level of endogenously released glutamate by blocking the glutamate transporters on cortical cultures over MEAs, has been presented in the literature. AMPA and NMDA are prototypic selective agonists whose effects reflect activation of the ionotropic glutamate receptor of the AMPA and NMDA type, respectively (Traynelis et al., 2010). The effects of TBOA, a potent non transportable broad spectrum inhibitor of the excitatory aminoacid transporters that did not show any significant effects on either the ionotropic or metabotropic glutamate receptors, can be exclusively ascribed to inhibition of glutamate transporters (Shimamoto et al., 1998). We found that medium to high micromolar concentrations of AMPA as well as NMDA (from 10 up to 100 μM) produced an overstimulation of glutamate receptors that is immediately followed by loss of electrophysiological activity, possibly reflecting excitotoxic effects, as proven by the live/dead assay results (cf. Suppl. Fig. 2). When very low concentrations of AMPA (50 or

Fig. 6. High doses of TBOA progressively decrease the level of electrophysiological activity. A. Raster plot of 30 out of 60 recording channels in basal (top) and TBOA 100 μM (bottom) condition in a culture of 27 DIV. B. Array wide firing rate of a representative experiment (DIV 21, time bin = 30s). C. Significant decrease of the firing rate under the TBOA 100 μM administration (Mann–Whitney rank sum test, *p b 0.001, N = 3 cultures). D. Effects of TBOA on glutamate release: endogenous glutamate released in basal conditions (white bars) and in the presence of TBOA 100 μM (gray bars). The extracellular glutamate is expressed as pmoles/ml. Data are means ± SEM (bars) of 5 experiments. *p b 0.05 (Student's t-test between related pairs of values) when compared to the control basal level.

M. Frega et al. / Neurotoxicology and Teratology 34 (2012) 116–127

125

Fig. 7. Low doses of TBOA maintain stable the level of activity producing a significant increase in the burst duration. A. Raster plot of 30 out 60 recording channels in basal (top) and TBOA 100 μM (bottom) condition in a representative culture of 20 DIV. B. Array wide firing rate the same experiment presented in A (time bin = 30 s). C. Firing rate and bursting rate are not affected by the administration of 5 μM TBOA, while burst duration significantly increases (167.56 ± 5.64%, mean ± SEM) under chemical stimulation with TBOA (Mann– Whitney rank sum test, *p b 0.001, n = 3 cultures). D. TBOA does not affect the synchronization level of the treated cultures. E. Effects of TBOA on glutamate release: endogenous glutamate released in basal conditions (white bars) and in the presence of TBOA 5 μM (gray bars). The extracellular glutamate is expressed as pmoles/ml. Data are means ± SEM (bars) of 4 experiments. *p b 0.05 (Student's t-test between related pairs of values) when compared to the control basal level.

200 nM) were applied to our cultures we noted a significant increase of activity with respect to basal condition, indicating a high sensitivity of our experimental model to glutamate receptor agonists (Martinoia et al., 2005). Low and medium concentrations of AMPA (5, 10 and 30 μM) already showed a decrease of the level of firing, indicating an initial fatigue of the excitatory system, also proven by a significant increase of the burst duration (up to 120%, see Fig. 2E). In the case of NMDA administration, very low concentrations (50 and 200 nM) did not affect the behavior of our cultures, while 5 μM brought to a slight increase of the firing rate (Fig. 3). No significant changes were observed for the burst-related parameters. The range of concentrations for which we observed an effect at the NMDA receptor might be different from those seen for the AMPA receptor because of the different affinity of the activating agonist (Traynelis et al., 2010; Zorumski et al., 1996). It is reported in literature that NR2B-containing NMDA receptors were expressed in mature cerebrocortical cultures (Sobczyk and Svoboda, 2007; Thomas et al., 2006): a possible explanation for the observed high sensitivity to agonists could be due to the activation of extrasynaptic NMDA receptors, predominantly NR2B-containing (Liu et al., 2004) which are particularly sensitive to glutamate and are known to be prominent players in synaptic depression and contribute to excitotoxicity in diverse animal models of neuronal injury (Chen et al., 2008; Léveillé et al., 2008; Picconi et al., 2006; Tu et al., 2010; von Engelhardt et al., 2007).

When glutamate was directly administered to the network, we noticed a significant increase of the firing for low concentrations (5 μM, as in case of NMDA), notwithstanding we observed a complete abolishment of the activity for very high concentrations (500 μM), with no recovery upon an immediate medium change. The same behavior was observed also 3 days after the treatment, as well as for NMDA treatment. Differently, AMPA treatment showed a slight recovery after three days (not observed in acute experiments) that could be related to the ability of AMPA receptor to desensitize (see further) and not only cell death. Differently, the loss of activity in case of NMDA or glutamate application was not induced by a transient depression of activity but by a significant cell death, as proven by the performed live/dead assay (cf. Suppl. Fig. 2). These results indicate that the high concentrations used for our in vitro experiments have an irreversible neuro-toxic effect, as observed in other experimental models (Garthwaite and Garthwaite, 1986). AMPA receptors open readily upon glutamate exposure, but desensitize quickly (Trussell and Fischbach, 1989) and present low affinity (Patneau et al., 1993). The desensitization of this receptor can be modulated by adding CTZ (Trussell et al., 1993). CTZ at different concentrations and in addition to AMPA agonist (200 nM, see Results) exhibits a high non-stationary (i.e. non-stable) behavior showing a wide range of electrical patterns related to the rapid increase of glutamate transmission. Besides of a postsynaptic allosteric modulation of AMPA

126

M. Frega et al. / Neurotoxicology and Teratology 34 (2012) 116–127

receptor, CTZ was also found to exert a powerful but opposite effect on GABAA receptors, indicating that CTZ interacts with both glutamate and GABAA receptors (Deng and Chen, 2003). Therefore in our dense networks (see Material and methods), at an advanced stage of maturation, CTZ is able to raise the level of activity only for very low doses (1–3 μM). Although the low effective concentrations make it unlikely, we cannot exclude that CTZ inhibition of GABAA responses (see Deng and Chen, 2003) might have contributed to the overall increase of neuronal activity. The activity of the network started to decrease at 10 μM CTZ and completely disappeared at higher concentrations (see Fig. 5). Discriminating the effects of high-concentration CTZ (i.e. dependent on inhibition of AMPA receptor desensitization) from excitotoxicity is not likely attainable. In fact, desensitization of AMPA receptors most likely contributes to dampen the effects of excessive AMPA receptor activation (including depolarization, subsequent opening of NMDA receptors, and excitotoxicity). Therefore, in the presence of high CTZ, that might completely inhibit the (protective) AMPA receptor desensitization, excitotoxicity is well expected. As a matter of fact, AMPA alone caused less cell damage than AMPA plus CTZ (cf. Suppl. Fig. 2), confirming the protective effect of AMPA receptor desensitization. Looking at the effect of the glutamate uptake inhibitor TBOA, we found that high doses bring to an accumulation of extracellular glutamate (Figs. 6D–7D), whose effect is exactly the same observed for high concentrations of the glutamate agonists AMPA and NMDA. At variance with the effect of exogenously added glutamate, synchronization was not affected. In fact, inhibiting endogenous glutamate uptake does not necessarily mirror the effects of adding extracellular glutamate: TBOA effects are expected to depend on activation of synaptic glutamate receptors whose activation is prevented by the uptake processes that clear the endogenously released glutamate. On the contrary, the exogenous glutamate effects are expected to depend on activation of all, synaptic and extrasynaptic, glutamate receptors that are accessible to the compound. At the dose of 100 μM we observed a large increase of the extracellular glutamate accompanied by a complete loss of electrophysiological activity (Fig. 6), while at 5 μM (only slightly increasing the extracellular glutamate level) we witness a no change in the activity frequency but a significant increase of the burst duration (Fig. 7), as also reported in the literature for different experimental models (Campbell and Hablitz, 2005; Milh et al., 2007). As a matter of fact, the increase of the extracellular glutamate concentration we measured after 100 μM TBOA was well below 30–100 μM glutamate, which appeared sufficient to depress neuronal activity. Nevertheless, the depressive action of TBOA is most likely due to inhibition of glutamate uptake rather than to non-specific effects, as the compound behaved as a selective blocker of glutamate transporters and did not show any significant effects on either the ionotropic or metabotropic glutamate receptors at concentration of 100 μM (Jabaudon et al., 1999; Shimamoto et al., 1998). Furthermore, extracellular glutamate level we measured in the medium might not sufficiently reflect the rapid increase (due to inhibition of glutamate uptake) of extracellular glutamate at synaptic cleft onto ionotropic glutamate receptors during network activation (see also Jabaudon et al., 1999). In conclusion, evidence obtained by manipulating glutamatergic transmission through various approaches (i.e., by direct activation of ionotropic glutamate receptors or by increasing the glutamate levels in the receptor biophase) altogether indicates that in MEA-coupled cortical cultures glutamatergic transmission needs to be controlled in a narrow range to sustain the network activity. The finding that only fine modulation of the excitatory synaptic transmission was reflected in physiological network activation (i.e. slight modulation of the parameters related to the electrophysiological patterns — cf. Materials and methods) is consistent with the capability of the in vitro neuronal network to reflect the complex collective behavior of in vivo assemblies. Accordingly, intervention leading to excessive direct stimulation of glutamatergic pathways or altering time course and/or levels of glutamate in the receptor biophase results in disruption of the network

activity (i.e. loss of spiking and bursting activity) and eventually cell death, as proven by the performed staining procedures (cf. Suppl. Fig. 2). Therefore, MEA-coupled cortical networks could be used as a very sensitive system of the physiological or pathological state of the excitatory glutamatergic transmission and the activation of ionotropic glutamate receptors, primary responsible for excitotoxicity. Moreover, MEA-coupled neuronal cultures might provide sensitive endpoints to detect early neurotoxic effects of chemicals and drugs and could be used for predictive toxicity testing (Bal-Price et al., 2010; Hogberg et al., in press; Johnstone et al., 2010). Furthermore, combining the study of network firing/bursting activity (neuronal specific electrophysiological endpoints) with the study of the release of glutamate and of activation of specific ionotropic glutamate receptor subtypes (neuronal specific neurotransmitter endpoints) would allow an integrated mechanisticbased approach to neurotoxicology, useful for a better understanding of neurotoxicity mechanisms. On the other hand, in cultured cortical networks, where indirect effects are operative, we cannot exclude that activation of neuromodulatory metabotropic glutamate receptors might contribute to the final effects of exogenously added glutamate, or of endogenously accumulated glutamate by TBOA. This is indeed a field worth to be accurately investigated in the future. In fact, evidence that metabotropic glutamate receptors do not “mediate”, but rather “modulate” excitatory synaptic transmission (Niswender and Conn, 2010) make them potential targets for neuroprotective drugs able to modulate glutamatergic ionotropic excitotoxicity (Bruno et al., 2001; Byrnes et al., 2009). Accurate characterization of the effects of metabotropic glutamate receptor activation/blockade on network activity might extend potential usefulness of MEA-coupled cortical cultures to the assessment of neurotoxicological endpoints for evaluation of both excitotoxicity and neuroprotection. Supplementary materials related to this article can be found online at doi:10.1016/j.ntt.2011.08.001. Conflict of interest statement All authors disclose that there are no financial and personal relationships with other people or organizations that could inappropriately influence (bias) their work. Acknowledgments The authors wish to thank Prof. Sergio Martinoia for the revision of the whole manuscript and his valuable comments. The authors are grateful to Chiara Venerelli who performed part of the presented experiments during her Bachelor thesis. References Antzoulatos EG, Byrne JH. Learning insights transmitted by glutamate. Trends Neurosci 2004;27:555–60. Bal-Price AK, Hogberg HT, Buzanska L, Coecke S. Relevance of in vitro neurotoxicity testing for regulatory requirements: challenges to be considered. Neurotoxicol Teratol 2010;32:36–41. Bologna LL, Nieus T, Tedesco M, Chiappalone M, Benfenati F, Martinoia S. Lowfrequency stimulation enhances burst activity in cortical cultures during development. Neuroscience 2010a;165:692–704. Bologna LL, Pasquale V, Garofalo M, Gandolfo M, Baljon PL, Maccione A, et al. Investigating neuronal activity by SPYCODE multi-channel data analyzer. Neural Netw 2010b;23:685–97. Bruno V, Battaglia G, Copani A, D'Onofrio M, Di Iorio P, De Blasi A, et al. Metabotropic glutamate receptor subtypes as targets for neuroprotective drugs. J Cereb Blood Flow Metab 2001;21:1013–33. Byrnes KR, Loane DJ, Faden AI. Metabotropic glutamate receptors as targets for multipotential treatment of neurological disorders. Neurotherapeutics 2009;6: 94–107. Campbell S, Hablitz JJ. Modification of epileptic discharges in neocortical neurons following glutamate uptake inhibition. Epilepsia 2005;46:129–33. Chapman AG. Glutamate and epilepsy. J Nutr 2000;130:1043S–5S. Chase TN, Oh JD. Striatal dopamine- and glutamate mediated dysregulation in experimental parkinsonism. Trends Neurosci 2000;23:S86–91.

M. Frega et al. / Neurotoxicology and Teratology 34 (2012) 116–127 Chen M, Lu TJ, Chen XJ, Zhou Y, Chen Q, Feng XY, et al. Differential roles of NMDA receptor subtypes in ischemic neuronal cell death and ischemic tolerance. Stroke 2008;39:3042–8. Chiappalone M, Bove M, Vato A, Tedesco M, Martinoia S. Dissociated cortical networks show spontaneously correlated activity patterns during in vitro development. Brain Res 2006;1093:41–53. Chiappalone M, Vato A, Berdondini L, Koudelka M, Martinoia S. Network dynamics and synchronous activity in cultured cortical neurons. Int J Neural Sys 2007;17:87–103. Collingridge GL, Lester R. Excitatory amino acid receptors in the vertebrate central nervous system. Pharmacol Rev 1989;41:143–210. Danbolt NC. Glutamate uptake. Prog Neurobiol 2001;65:1–105. Deng L, Chen G. Cyclothiazide potently inhibits gamma-aminobutyric acid type A receptors in addition to enhancing glutamate responses. PNAS 2003;100:13025–9. Dichter M, Fischbach G. The action potential of chick dorsal root ganglion neurons manteined in cell culture. J Physiol 1977;276:281–98. Garthwaite G, Garthwaite J. Neurotoxicity of excitatory amino acid receptor agonists in rat cerebellar slices: dependence on calcium concentration. Neurosci Lett 1986;66: 193–8. Gramowski A, Schiffmann D, Gross GW. Quantification of acute neurotoxic effects of trimethyltin using neuronal networks cultured on microelectrode arrays. Neurotoxicology 2000;21:331–42. Gramowski A, Jugelt K, Stuwe S, Schulze R, McGregor GP, Wartenberg-Demand A, et al. Functional screening of traditional antidepressants with primary cortical neuronal networks grown on multielectrode neurochips. Eur J Neurosci 2006;24:455–65. Gross GW, Reiske E, Kreutzberg GW, Mayer A. A new fixed-array multimicroelectrode system designed for long-term recording of extracellular single unit activity in vitro. Neurosci Lett 1977;6:101–5. Gross GW, Rhoades BK, Jordan RJ. Neuronal networks for biochemical sensing. Sens Actuators B Chem 1992;6:1–8. Hogberg HT, Sobanski T, Novellino A, Whelan M, Weiss DG, Bal-Price AK, Application of micro-electrode arrays (MEAs) as an emerging technology for developmental neurotoxicity: evaluation of domoic acid-induced effects in primary cultures of rat cortical neurons. Neurotoxicology 2011;32:158–68. Jabaudon D, Shimamoto K, Yasuda-Kamatani Y, Scanziani M, Gähwiler BH, Gerber U. Inhibition of uptake unmasks rapid extracellular turnover of glutamate of nonvesicular origin. PNAS 1999;96:8733–8. Johnstone AFM, Gross GW, Weiss DG, Schroder O, Gramoswki A, Shafer TJ. Microelectrode arrays: a physiologically based neurotoxicity testing platform for the 21st century. Neurotoxicology 2010;31:331–50. Jones KH, Senft JA. An improved method to determine cell viability by simultaneous staining with fluorescein diacetate-propidium iodide. J Histochem Cytochem 1985;33:77–9. Léveillé F, El gaamouch F, Gouix E, Lecocq M, Lobner D, Nicole O, et al. Neuronal viability is controlled by a functional relation between synaptic and extrasynaptic NMDA receptors. FASEB J 2008;22:4258–71. Liu L, Wong TP, Pozza MF, Lingenhoehl K, Wang Y, Sheng M, et al. Role of NMDA receptor subtypes in governing the direction of hippocampal synaptic plasticity. Science 2004;304. Maccione A, Gandolfo M, Massobrio P, Novellino A, Martinoia S, Chiappalone M. A novel algorithm for precise identification of spikes in extracellularly recorded neuronal signals. J Neurosci Methods 2009;177:241–9. Maragakis NJ, Rothstein JD. Glutamate transporters in neurologic disease. Arch Neurol 2001;58:365–70. Marcoli M, Cervetto C, Paluzzi P, Guarnieri S, Alloisio S, Thellung S, et al. P2X7 presynaptic receptors in adult rat cerebrocortical nerve terminals: a role in ATPinduced glutamate release. J Neurochem 2008;105:2330–42. 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. Martinoia S, Bonzano L, Chiappalone M, Tedesco M, Marcoli M, Maura G. In-vitro cortical neuronal networks as a new high-sensitive system for biosensing application. Biosens Bioelectron 2005;20:2071–8.

127

Meldrum BS. Glutamate as a neurotransmitter in the brain: review of physiology and pathology. J Nutr 2000;130:1007S–15S. Milh M, Becq H, Villeneuve N, Ben-Ari Y, Aniksztejn L. Inhibition of glutamate transporters results in a ‘suppression burst’ pattern and partial seizures in the newborn rat. Epilepsia 2007;48:169–74. Niswender CM, Conn PJ. Metabotropic glutamate receptors: physiology, pharmacology, and disease. Annu Rev Pharmacol Toxicol 2010;50:295–322. Pasquale V, Chiappalone M, Martinoia S. A self-adapting approach for the detection of bursts and network bursts in cortical cultures in vitro. J Comput Neurosci 2010;29: 213–29. Patneau DK, Vyklicky LJ, Mayer ML. Hippocampal neurons exhibit cyclothiazidesensitive rapidly desensitizing responses to kainate. J Neurosci 1993;13:3496–509. Picconi B, Tortiglione A, Barone I, Centonze D, Gardoni F, Gubellini P, et al. NR2B subunit exerts a critical role in postischemic synaptic plasticity. Stroke 2006;37:1895–901. Pine J. Recording action potentials from cultured neurons with extracellular microcircuit electrodes. J Neurosci Methods 1980;2:19–31. Riedel G, Platt B, Micheau J. Glutamate receptor function in learning and memory. Behav Brain Res 2003;140:1–47. Rothman SM, Olney JW. Excitotoxity and the NMDA receptor. Trends Neurosci 1987;10: 299–302. Rothman SM, Olney JW. Excitotoxicity and the NMDA receptor-still lethal after eight years. Trends Neurosci 1995;18:57–8. Sakmann B, Neher E. Patch clamp techniques for studying ionic channels in excitable membranes. Annu Rev Physiol 1984;46:455–72. Shimamoto K, Lebrun B, Yasuda-Kamatani Y, Sakaitani M, Shigeri Y, Yumoto N, et al. DLthreo-β-benzyloxyaspartate, a potent blocker of excitatory amino acid transporters. Mol Pharmacol 1998;53:195–201. Sobczyk A, Svoboda K. Activity-dependent plasticity of the NMDA-receptor fractional Ca2+ current. Neuron 2007;53:17–24. Sun Y, Olson R, Horning M, Armstrong N, Mayer ML, Gouaux E. Mechanisms of glutamate receptor desensitization. Nature 2002;417:245–53. Thomas CG, Miller AJ, Westbrook GL. Synaptic and extrasynaptic NMDA receptor NR2 subunits in cultured hippocampal neurons. J Neurophysiol 2006;95:1727–34. Traynelis SF, Wollmuth LP, McBain CJ, Menniti FS, Vance KM, Ogden KK, et al. Glutamate receptor ion channels: structure, regulation, and function. Pharmacol Rev 2010;62: 405–96. Trussell LO, Fischbach G. Glutamate receptor desensitization and its role in synaptic transmission. Neuron 1989;3:209–18. Trussell LO, Zhang S, Raman IM. Desensitization of AMPA receptors upon multiquantal neurotransmitter release. Neuron 1993;10:1185–96. Tu W, Xu X, Peng L, Zhong X, Zhang W, Soundarapandian MM, et al. DAPK1 interaction with NMDA receptor NR2B subunits mediates brain damage in stroke. Cell 2010;140:222–34. Tzschentke TM. Glutamatergic mechanisms in different disease states: overview and therapeutical implications — an introduction. Amino Acids 2002;23:147–52. von Engelhardt J, Coserea I, Pawlak V, Fuchs EC, Köhr G, Seeburg PH, et al. Excitotoxicity in vitro by NR2A- and NR2B-containing NMDA receptors. Neuropharmacology 2007;53:10–7. Wagenaar DA, Pine J, Potter SM. An extremely rich repertoire of bursting patterns during the development of cortical cultures. BMC Neurosci 2006;7. Wagner JG. Kinetics of pharmacologic response. I. Proposed relationships between response and drug concentration in the intact animal and man. J Theor Biol 1968;20:173–201. Watt AJ, Sjöström PJ, Häusser M, Nelson SB, Turrigiano GG. A proportional but slower NMDA potentiation follows AMPA potentiation in LTP. Nat Neurosci 2004;7: 518–24. Zorumski CF, Mennerick S, Que J. Modulation of excitatory synaptic transmission by low concentrations of glutamate in cultured rat hippocampal neurons. J Physiol 1996;494:465–77.