Quantal Release of Dopamine and Action Potential Firing Detected in Midbrain Neurons by Multifunctional Diamond-Based Microarrays

Micro-Graphitic Single Crystal Diamond Multi Electrode Arrays (μG-SCD-MEAs) have so far been used as amperometric sensors to detect catecholamines from chromaffin cells and adrenal gland slices. Besides having time resolution and sensitivity that are comparable with carbon fiber electrodes, that represent the gold standard for amperometry, μG-SCD-MEAs also have the advantages of simultaneous multisite detection, high biocompatibility and implementation of amperometric/potentiometric protocols, aimed at monitoring exocytotic events and neuronal excitability. In order to adapt diamond technology to record neuronal activity, the μG-SCD-MEAs in this work have been interfaced with cultured midbrain neurons to detect electrical activity as well as quantal release of dopamine (DA). μG-SCD-MEAs are based on graphitic sensing electrodes that are embedded into the diamond matrix and are fabricated using MeV ion beam lithography. Two geometries have been adopted, with 4 × 4 and 8 × 8 microelectrodes (20 μm × 3.5 μm exposed area, 200 μm spacing). In the amperometric configuration, the 4 × 4 μG-SCD-MEAs resolved quantal exocytosis from midbrain dopaminergic neurons. KCl-stimulated DA release occurred as amperometric spikes of 15 pA amplitude and 0.5 ms half-width, at a mean frequency of 0.4 Hz. When used as potentiometric multiarrays, the 8 × 8 μG-SCD-MEAs detected the spontaneous firing activity of midbrain neurons. Extracellularly recorded action potentials (APs) had mean amplitude of ∼-50 μV and occurred at a mean firing frequency of 0.7 Hz in 67% of neurons, while the remaining fired at 6.8 Hz. Comparable findings were observed using conventional MEAs (0.9 and 6.4 Hz, respectively). To test the reliability of potentiometric recordings with μG-SCD-MEAs, the D2-autoreceptor modulation of firing was investigated by applying levodopa (L-DOPA, 20 μM), and comparing μG-SCD-MEAs, conventional MEAs and current-clamp recordings. In all cases, L-DOPA reduced the spontaneous spiking activity in most neurons by 70%, while the D2-antagonist sulpiride reversed this effect. Cell firing inhibition was generally associated with increased APs amplitude. A minority of neurons was either insensitive to, or potentiated by L-DOPA, suggesting that AP recordings originate from different midbrain neuronal subpopulations and reveal different modulatory pathways. Our data demonstrate, for the first time, that μG-SCD-MEAs are multi-functional biosensors suitable to resolve real-time DA release and AP firing in in vitro neuronal networks.

Micro-Graphitic Single Crystal Diamond Multi Electrode Arrays (µG-SCD-MEAs) have so far been used as amperometric sensors to detect catecholamines from chromaffin cells and adrenal gland slices. Besides having time resolution and sensitivity that are comparable with carbon fiber electrodes, that represent the gold standard for amperometry, µG-SCD-MEAs also have the advantages of simultaneous multisite detection, high biocompatibility and implementation of amperometric/potentiometric protocols, aimed at monitoring exocytotic events and neuronal excitability. In order to adapt diamond technology to record neuronal activity, the µG-SCD-MEAs in this work have been interfaced with cultured midbrain neurons to detect electrical activity as well as quantal release of dopamine (DA). µG-SCD-MEAs are based on graphitic sensing electrodes that are embedded into the diamond matrix and are fabricated using MeV ion beam lithography. Two geometries have been adopted, with 4 × 4 and 8 × 8 microelectrodes (20 µm × 3.5 µm exposed area, 200 µm spacing). In the amperometric configuration, the 4 × 4 µG-SCD-MEAs resolved quantal exocytosis from midbrain dopaminergic neurons. KCl-stimulated DA release occurred as amperometric spikes of 15 pA amplitude and 0.5 ms half-width, at a mean frequency of 0.4 Hz. When used as potentiometric multiarrays, the 8 × 8 µG-SCD-MEAs detected the spontaneous firing activity of midbrain neurons. Extracellularly recorded action potentials (APs) had mean amplitude of ∼−50 µV and occurred at a mean firing frequency of 0.7 Hz in 67% of neurons, while the remaining fired at 6.8 Hz. Comparable findings were observed using conventional MEAs (0.9 and 6.4 Hz, respectively). To test the reliability of potentiometric recordings with µG-SCD-MEAs, the D 2 -autoreceptor modulation of firing was investigated by applying levodopa (L-DOPA, 20 µM), and comparing µG-SCD-MEAs, conventional MEAs and current-clamp recordings. In all cases, L-DOPA reduced

Ion-Induced Damage Simulation
The ion-induced damage density profiles were obtained using a numerical simulation performed with the "Stopping and Range of Ion in Matter" SRIM-2013.00 Monte Carlo code (Ziegler et al., 2010) in "Detailed calculation with full damage cascades" mode, by setting a displacement energy value of 50 eV (Wu and Fahy, 1994). The output of the simulation process (i.e., the number of vacancies created per unit of depth and ion, in #vac #ions −1 µm −1 units) was then multiplied by the ion fluence, assuming the absence of non-linear effects such as damage saturation or formation of defect complexes.

Cell Cultures
The methods for the primary culture of mesencephalic dopamine neurons from substantia nigra (SN) was adapted from Pruszak et al. (2009). The ventral mesencephalon area was dissected from embryonic (E15) C57BL6 TH-GFP mice (Sawamoto et al., 2001;Matsushita et al., 2002). TH-GFP mice were kept heterozygous via breeding TH-GFP mice with C57BL/6J mice. All animals were housed under a 12-h light/dark cycle in an environmentally controlled room with food and water ad libitum. All experiments were conducted in accordance with the European Community's Council Directive 2010/63/UE and approved by the Italian Ministry of Health and the Local Organism responsible for animal welfare at the University of Turin (Authorization DGSAF 0011710-P-26/07/2017). HBSS (Hank's balanced salt solution, without CaCl 2 and MgCl 2 ), enriched with 0.18% glucose, 1% BSA, 60% papain (Wortington, Lakewood, NJ, United States), 20% Dnase (Sigma-Aldrich) was stored at 4 • C and used as digestion buffer. Neurons were plated at final densities of 600 cells mm −2 on petri dishes, or 2000 cells mm −2 on conventional MEAs or µG-SCD-MEAs. Cultured neurons were used at 7 DIV for current-clamp experiments and at 14 DIV with MEAs or µG-SCD-MEAs. Petri dishes, as well as MEAs and µG-SCD-MEAs were coated with poly-L-Lysine (0.1 mg ml −1 ) as substrate adhesion. Cells were incubated at 37 • C in a 5% CO 2 atmosphere, with Neurobasal Medium containing 1% pen-strep, 1% ultra-glutamine, 2% B-27 and 2.5% FBSd; pH 7.4.
Potentiometric recordings with MEAs and µG-SCD-MEAs were performed while keeping the cells under a controlled CO 2enriched atmosphere and stable temperature conditions.

Amperometric Recordings
Amperometric recordings were performed by means of µG-SCD-MEAs (4 × 4 channels geometry) and dedicated electronics, which were designed at the Institute of Electron Devices and Circuits (Ulm University). The whole electronic chain was inserted into a Faraday cage to minimize noise. The chip carrier was directly plugged-in to the front-end electronics connected to a data acquisition unit (National Instruments USB-6216). The circuit was grounded by means of a reference Ag/AgCl electrode, which was immersed in the extracellular solution. Amperometry was performed by holding the 16 electrodes at a constant potential of +0.65 V relative to the Ag-AgCl reference electrode.
The acquisition electronics consisted of low-noise transimpedance amplifiers with an input bias current of 1 pA and a gain, set by feedback-resistors, of 100 M . The amplified signals were filtered at 4 kHz with 4 th order Bessel low-pass filters and were subsequently acquired at a sampling rate of 25 kHz per channel. The National Instruments DAQ interface was connected to a computer via a high-speed USB. We used data acquisition control software that was developed in LabView. The noise level was evaluated in spike-free trace segments and then averaged over the 16 electrodes, leading to a mean amplitude of 5.5 ± 0.7 pA, with a mean signal-to-noise ratio (S/N) of ∼3. Spike analysis was performed using "Quanta Analysis" routine (Mosharov and Sulzer, 2005) in Igor pro 5.00 data analysis software by waveMetrics. No change in current output was observed when electrode polarization was lowered below 50 mV.

Potentiometric Recordings Using µG-SCD-MEAs and MEAs
Potentiometric recordings were performed while the cells were kept in their culture medium. Recordings took place inside a dedicated incubator, at a controlled temperature and 5% CO 2 atmosphere. A MCS MEA 1060-Inv-BC amplifier from Multi Channel Systems (Reutlingen, Germany) was used as the readout unit for both with µG-SCD-MEAs (8 × 8 geometry) and conventional MEAs (60MEA200/300iR-Ti).
Data acquisition was controlled using MC_Rack software. The threshold for spike detection was set at −30 µV and the sampling frequency at 10 kHz. Data were then analyzed using Clampfit software (Molecular Devices, Silicon Valley, CA, United States).

Current Clamp Recordings
Patch-clamp experiments were performed using Pclamp software (Molecular Devices, Silicon Valley, CA, United States). All experiments were performed at a temperature of 22-24 • C. Data analysis was performed using Clampfit software.

Fluorescence Images
Images were acquired using a Zeiss microscope primovert 40x (Carl Zeiss, LLC United States). In the fluorescence configuration, samples were excited with radiation in the visible spectrum at a characteristic wavelength λ ex = 470 nm, and the emission wavelength was λ em = 505 nm, which is typical of GFP staining.

INTRODUCTION
Dopamine (DA) plays fundamental roles in a variety of neurophysiological functions and neurological diseases. Dopaminergic microcircuits are involved in movement, reward, memory and cognition (Waelti et al., 2001), while the degeneration of the nigrostriatal pathway in Parkinson's disease (PD) impairs control and planning of movement, causing tremors and postural instability.
Fluctuations of DA concentration occur on the seconds to subseconds time scale, thus making them suitable for study with carbon fiber electrodes (CFEs) and fast-scan cyclic voltammetry (Hafizi et al., 1990;Kawagoe and Wightman, 1994;Heien and Wightman, 2006;Patel and Rice, 2013). However, in vivo DA electrochemical detection is hampered by electrode fouling, which is caused by accumulation of oxidized products and by interference of ascorbic and uric acid, ultimately limiting electrode sensitivity and selectivity (Suzuki et al., 2007).
Since the early investigation of synaptic dysfunction is a target when attempting to understand the molecular mechanisms that lead to neurodegenerative processes, the development of mutifunctional sensing tools for simultaneous monitoring neurotransmitter release and electrical activity are extremely relevant for addressing key aspects of neurotransmission in the early stages of neurodegenerative diseases (Suzuki et al., 2013;Schirinzi et al., 2016;Castagnola et al., 2018;Ghiglieri et al., 2018;Picconi et al., 2018).
In this regard, conventional multielectrode arrays (MEAs) have been employed to investigate the firing properties in SN pars compacta slices (Berretta et al., 2010), while amperometric detection of DA release from cultured neurons, was initially performed using CFEs (Pothos et al., 1998;Pothos, 2002;Staal et al., 2004;Mosharov and Sulzer, 2005). A range of different amperometric microarrays to detect DA release from PC12 cells (Chen et al., 1994;Lin et al., 2012;Trouillon and Ewing, 2014), striatal slices (Suzuki et al., 2013) and from isolated dopaminergic somas from the pond snail Lymnaea stagnalis have been designed . Dopamine release from striatal slices has been detected by carbon nanotube multi electrode arrays and the same device could successfully detect APs and field postsynaptic potentials from cultured hippocampal neurons and slices (Suzuki et al., 2013). In spite of this, no data concerning the detection of quantal release and electrical activity from the same cultured neurons using the same multiarray prototypes have been reported, to the best of our knowledge. Micro-graphitic single-crystal diamond multielectrode arrays (µG-SCD-MEAs) are a powerful sensor for investigating neurosecretion in living cells (Picollo et al., 2013(Picollo et al., , 2015b. Previous findings have demonstrated their ability to monitor spontaneous and evoked quantal catecholamine release from cultured mouse and bovine adrenal chromaffin cells (Picollo et al., 2016b) as well as from fresh mouse adrenal slices (Picollo et al., 2016a;Carabelli et al., 2017). Besides providing simultaneous recordings from a variety cells, which have been plated and cultured on the planar array for a number of days, µG-SCD-MEAs possess high-time resolution and sensitivity for the detection of amperometric events with different shape, such as the small amplitude, or previously identified "stand-alone-foot" events (Picollo et al., 2016a). Taking advantage of diamond biocompatibility (Bonnauron et al., 2008;Nistor et al., 2015;Piret et al., 2015;Alcaide et al., 2016), we have succeeded in culturing primary midbrain neurons on µG-SCD-MEAs. In the present work, we have provided the first evidence that µG-SCD-MEAs can detect the quantal exocytosis of neuronal synaptic vesicles as well as spontaneous neuronal firing activity.

Fabrication of µG-SCD-MEAs
The electrochemical sensors that have been used in the present work consist of multi electrode arrays with either 16 or 60 graphitic electrodes that have been embedded into an artificial single-crystal diamond substrate. The two devices combine the properties of diamond, including: (1) biocompatibility, guaranteeing the plating and maintenance of primary cultures for weeks (Tang et al., 1995;Nistor et al., 2015); (2) chemical inertness, which prevents modifications to the employed solutions; and (3) wide optical transparency.
Conventional fabrication schemes cannot be used for the assembly of these sensors due to the extreme chemical/physical characteristics of diamond, meaning that an advanced MeVion-beam-implantation-based process was used (Olivero et al., 2010;Picollo et al., 2015b). This fabrication technique allows the selective phase transition from the diamond to graphite to be promoted by taking advantage of the metastable nature of the substrate. Indeed, if the density of ion beam-induced defects (commonly parametrized in terms of vacancy density) overcomes a critical threshold, the graphitization of the damaged region is obtained upon high-temperature thermal treatment. Moreover, the fact that the defects created by irradiation with MeV ions follow a typical distribution, which is characterized by the so-called "Bragg peak" and the ion end of range ( Figure 1A), means the position of the graphitic electrodes can be modulated along the substrate depth, thus guaranteeing intrinsic electrical passivation due to the presence of the diamond cap layer. Only the electrodes end-points emerge to the surface, thus allowing the interfacing with the front-end electronic in the peripheral region of the sensor and the cells coupling in the central region.
Figures 1B,C shows the sensor and the magnification of the graphitic electrodes, which were arranged in a 4 × 4 matrix, while Figures 1D,E show analogous representations for the device with a 8 × 8 matrix of electrodes (without the four electrodes on the corners). In both cases, active areas were regularly spaced with a ∼200 µm step.

Real-Time Detection of Quantal DA Release From Midbrain Neurons Using µG-SCD-MEAs
DA release from cultured midbrain neurons can either occur at the somato-dendritic or the axon-terminal level (Rice and Patel, 2015). In this work, we have cultured midbrain neurons for 20 DIV on µG-SCD-MEAs and found that quantal exocytotic events can be detected after 10 DIV. Under our experimental conditions, the density of cell plating on the Frontiers in Neuroscience | www.frontiersin.org multiarray (see section "Materials and Methods"), allowed us to reveal amperometric signals from approximately 25-30% of the electrodes. Unstimulated (i.e., spontaneous) release was barely detectable (5% of trials) and occurred at low frequency (0.11 ± 0.07 Hz) in 2 mM CaCl 2 (Figure 2A). Amperometric spikes were characterized by a mean maximum current amplitude (I max ) of 13.2 ± 1.0 pA and a half-time width (t half ) of 0.57 ± 0.03 ms (n = 5). Stimulation with 30 mM KCl (Figures 2A-C) increased the release frequency to 0.40 ± 0.03 Hz. However, spike parameters were unaffected: I max was 18.5 ± 1.1 pA and t half was 0.52 ± 0.01 ms (n = 13 cells, from 4 µG-SCD-MEAs). On the other hand, 200 µM CdCl 2 suppressed Ca 2+ -dependent exocytosis through voltage-gated Ca 2+ channels, as shown in Figure 2A. No events were detected when the recording electrodes were polarized to 0 mV to nullify dopamine detection (bottom trace). Representative recordings of simultaneous acquisition from five different electrodes of the same µG-SCD-MEA are shown in Figure 2B: multiple events, such as the one visible in the first trace, were discarded from the analysis. Some representative spikes that were recorded in the presence of KCl are shown at a more expanded time scale in Figure 2C.

Detection of Spontaneous AP Firing From Cultured Midbrain Neurons
After assessing the sensitivity of µG-SCD-MEAs to reveal quantal DA release in dopaminergic neurons, we tested if these sensors were able to measure the electrical activity of cultured midbrain neurons. For this purpose, µG-SCD-MEAs were patterned with a higher electrode number (8 × 8 array) than to those used for amperometry (4 × 4 array). Recordings were performed in parallel using µG-SCD-MEAs and conventional MEAs [Multi Channel Systems (MCS)], for a more rigorous interpretation of acquired data. As has already been observed in cultured hippocampal neurons (Gavello et al., 2012Allio et al., 2015) and other brain regions (Martinoia et al., 2005), mesencephalic DA neurons start generating spontaneous APs after 7 DIV (Henderson et al., 2016), while network functionality was wellresolved at 14 DIV.
Representative recordings of spontaneously firing midbrain neurons, measured using µG-SCD-MEAs and conventional MEAs, are shown in Figure 3A. This spontaneous spiking activity occurred under physiological conditions (2 mM Ca 2+ ) and was suppressed by blocking the firing during the exogenous application of 300 nM TTX (data not shown). Unlike amperometric spikes, which exibit monopolar waveforms, single APs ( Figure 3B) were characterized by a fast downward deflection (negative peak), which corresponds to the AP rising phase, followed by an upward deflection (positive antipeak), which is associated to the AP repolarising phase (Fromherz, 1999;Marcantoni et al., 2007). The mean amplitude of the negative peaks recorded by µG-SCD-MEAs (n = 10) was −50.2 ± 3.6 µV, with S/N of ∼4, while the mean signal amplitude was equal to −54.0 ± 4.7 µV, with S/N of ∼5, for conventional MEAs (n = 10). The amplitude of the positive antipeak, when detectable, was approximately 30% of the negative peak amplitude. For instance, for the channels indicated by the asterisks in Figure 3A, the mean positive antipeak amplitudes was 25.4 ± 0.4 and 26.0 ± 0.4 µV (with µG-SCD-MEAs and MEAs, respectively), while the mean peak amplitude was −60.1 ± 0.6 and −67.1 ± 0.5 µV (with µG-SCD-MEAs and MEAs, respectively). Since the positive antipeak was not detectable in all neurons, we limited our analysis to the negative peak, in good agreement with our previous observations (Vandael et al., 2010).
Firing frequencies were comprised within 0.1 and 15 Hz (Figure 3C), which is consistent with the presence of distinct neuronal populations within the network (Berretta et al., 2010;Ramayya et al., 2014). Most neurons (67%) were spontaneously active, fired with a basal frequency which not exceeded 4 Hz and had a mean firing frequency of 0.66 ± 0.14 or 0.90 ± 0.10 Hz, when measured with µG-SCD-MEAs and conventional MEAs (p > 0.1), respectively. The remaining neurons had a much higher basal firing frequency, ranging between 4 and 11 Hz and characterized by mean values of 6.8 ± 1.4 and 6.4 ± 0.5 Hz, when measured with µG-SCD-MEAs and MEAs (p > 0.1), respectively. This heterogeneity of responses can, most likely, be ascribed to the presence of distinct neuronal populations such as DA neurons from SN, GABAergic and DA neurons from the nearby ventral tegmental area (Berretta et al., 2010;Cucchiaroni et al., 2011). Optical images of GFP-TH + neurons that were cultured on µG-SCD-MEAs are provided in Figure 3D.
Despite the above-mentioned heterogeneity, these recordings are the first experimental evidence that µG-SCD-MEAs are suitable for potentiometric recordings from primary cultures of brain neurons.

D 2 -Autoreceptor Induced Inhibition of Repetitive Firing in Current-Clamp Recordings
The firing of nigral dopaminergic neurons is down-regulated by DA release through a D 2 -autoreceptor mediated pathway (Aghajanian and Bunney, 1977;Mercuri et al., 1990). Since this down-regulatory pathway has been observed in midbrain slices and we were recording from primary cultured midbrain dissociated neurons (Lacey et al., 1987;Guatteo et al., 2013), we aimed to identify this inhibitory down-regulation under our experimental conditions first (Bigornia et al., 1990). Preliminary experiments were performed in whole-cell currentclamp configuration, by applying the DA precursor levodopa (L-DOPA) (20 mM). Recordings were selectively performed on 7 DIV dopaminergic neurons that were identified by means of GFP staining (Figure 3D). Although the responses to applied L-DOPA varied, it caused a 70 ± 4% reduction of the firing frequency in 80% of cases (n = 20 cells, from 1.36 ± 0.02 to 0.41 ± 0.11 Hz; Figures 4A,F). Maximum inhibition occurred within 2-5 min of L-DOPA perfusion, and was reversed some minutes after the application of the D 2 antagonist sulpiride (10 µM). The repetitive firing frequency measured in the presence of the D 2 antagonist recovered to 1.2 ± 0.2 Hz, thus confirming the autocrine inhibition that is induced by released DA (Guatteo et al., 2013). It is worth mentioning that the reduced firing frequency was associated, in 70% of the cases, to a membrane hyperpolarization of −7.8 ± 1.1 mV and by a sharp increase in AP peak amplitude (from 27 ± 3 to 35.6 ± 1.6 mV; n = 14, p < 0.05; Figures 4C,D). All this was most likely induced by the DAmediated activation of a G-protein-coupled potassium channel (GIRK) (Lacey et al., 1987). Both effects were reversed after perfusion with sulpiride.
In the remaining 30% of neurons, the nearly threefold reduction of firing frequency occurred without causing either the significant hyperpolarization of the membrane potential, or alterations in the AP waveform (Figures 4F,G). For this subset of neurons, in some cases sulpiride restored the control firing frequency, even though the recovery was not always complete. This variability reveals the probable existence of distinct modulatory pathways that may originate from different midbrain neuron subpopulations (Dragicevic et al., 2015;Duda et al., 2016).
Phase plane plot analysis was performed in order to gain further insights into the APs properties and their modulation by L-DOPA (Vandael et al., 2012;Marcantoni et al., 2014). By plotting the time derivative of voltage versus voltage (dV/dt), parameters such as the AP threshold can be easily inferred from the voltage value at which dV/dt suddenly increases. The phaseplane plots in Figure 4B are referred to the same APs that are indicated by the asterisks in Figure 4A. From the plot we found that: (i) the maximum derivative (dV/dt max ), which is associated with the maximum current density through voltagegated Na v channels, was drastically enhanced by L-DOPA (from 67 to 129 mV ms −1 , p < 0.05, Figure 4D), suggesting a sustained recruitment of Na v channels (Guarina et al., 2018); (ii) the AP hyperpolarization peak was significantly augmented by L-DOPA, from from −37.2 ± 1.3 to −40 ± 2 mV (p < 0.05, Figure 4D); (iii) the AP threshold, measured from the phase-plane plot when an abrupt change in dV/dt was observed (at 4.5 ± 1.2 mV ms −1 for control and 6.4 ± 0.9 mV ms −1 for L-DOPA-treated neurons), decreased from −25.3 ± 1.8 to −31.9 ± 1.8 mV (p < 0.05), respectively. This again confirms a potentiated recruitment of Na v channels during L-DOPA treatment.
In order to compare the AP waveform recorded intracellularly with those recorded extracellularly, the negative first derivative of AP traces shown in Figure 4C is reported in Figure 4E. They correspond to the AP shape recorded extracellularly by the MEAs (Fromherz et al., 1991) and identified as a biphasic AP waveform in which, similarly to that of Figure 3B, a large negative peak and a small positive antipeak component can be distinguished.
Finally, a range of different effects on neuronal activity were detected in the neurons that were not inhibited by L-DOPA (20% of neurons). L-DOPA accelerated repetitive firing by 80 ± 20% (17% of neurons), while it was ineffective in the remaining ones (3% of neurons).

Heterogeneity of L-DOPA Induced Responses in Cultured Midbrain Neurons Observed Through µG-SCD-MEAs and Conventional MEAs
Potentiometric recordings using µG-SCD-MEAs were performed to simultaneously detect spikes arising from different neuronal populations and to investigate their responses to the applied drugs. With respect to patch-clamp experiments, performed on isolated and young neurons (7 DIV), these trials were designed to provide a rapid screening of the effects of L-DOPA on mature networks (14 DIV). After the firing properties under control conditions were monitored for a couple of minutes, the addition of L-DOPA to the culture medium revealed three different responses, confirming the existence of heterogeneous firing, as measured in SN slices (Berretta et al., 2010). In most cases (70% of neurons), the firing activity was significantly reduced by L-DOPA and the inhibitory effect required some minutes for completion (Mercuri et al., 1990). As shown in a representative recording using µG-SCD-MEAs, the firing frequency was reduced by 80% after 2-3 min, and the extracellular AP peak increased from −75 ± 1 to −87 ± 3 µV, while sulpiride reversed both effects, suggesting that D 2 autoreceptors are involved (Figures 5A,C). On average L-DOPA decreased the spontaneous spiking activity from 1.1 to 0.3 Hz and increased the negative peak amplitude by 14% (n = 5 µG-SCD-MEAs, p < 0.05), suggesting a prominent recruitment of Na v channels following L-DOPA hyperpolarization. In order to validate these experimental findings, we repeated the same experiments using conventional MEAs. Once again, the majority of neurons (64%), responded to L-DOPA by reducing the mean firing frequency, on average from 1.5 ± 0.7 to 0.29 ± 0.09 Hz (p < 0.05, n = 10 MEAs; Figure 5B), while sulpiride restored the basal frequency to 1.1 ± 0.3 Hz (Figure 5C). In this subset of neurons, firing frequency reduction was also associated to a 20% increase in the negative peak amplitude, confirming the prominent role that D 2 -autoreceptors play in L-DOPA induced inhibition.
Nevertheless, a relevant fraction of neurons in the mature networks (30 and 36%, respectively for µG-SCD-MEAs and MEAs) also displayed a significant increase (up to sixfold) in spontaneous frequency and a 30% reduction in the negative peak amplitude following exposure to L-DOPA. This is in good agreement with the heterogeneity of responses that we observed in dissociated neurons under current-clamp conditions. In the example shown in Figure 6A for µG-SCD-MEAs, the negative peak amplitude decreased from −44.1 ± 1.2 to −34.2 ± 1.1 µV, while the firing frequency increased from 0.5 to 2.9 Hz (Yasumoto et al., 2004). Similarly, the potentiation of firing activity by L-DOPA occurred with a mean threefold increase in firing frequency when using conventional MEAs, and was usually accompanied by a 28% decrease in the negative peak amplitude (Figure 6B).
L-DOPA reduced the spiking activity without affecting the AP shape in the remaining cases. Examples of this modulation are shown in Figure 6C. The unaltered amplitude of the negative AP peak, that have been revealed by µG-SCD-MEAs and conventional MEAs (Figure 6D), confirms the findings obtained under patch-clamp conditions, in which 30% of neurons displayed a reduced firing frequency without alterations of the AP rising phase.

DISCUSSION
We have provided the first evidence that µG-SCD-MEAs allow long-term neuronal cultures to be performed and that they can function as sensing devices for recording quantal exocytosis and spontaneous AP firing.

Amperometric Detection of Quantal Dopamine Release
Amperometric recordings have proved that µG-SCD-MEAs are suitable for the real time detection of exocytosis from neuronal networks. Under physiological conditions (2 mM Ca 2+ ), µG-SCD-MEAs can resolve spontaneous secretory events as amperometric spikes of <20 pA I max and mean quantity of charge Q of 0.01 pC, which can most likely be associated with the tonic discharge activity of the network (Sulzer et al., 2016). These exocytotic events are significantly smaller than those of released catecholamines from large dense-core vesicles of adrenal chromaffin cells, whose I max is in the order of 10s of pA, and Q is > 1.5 pC, as has already been reported using the same µG-SCD-MEAs (Picollo et al., 2016b).
When using KCl as a secretagogue to increase the probability of release from DA neurons, a mean release of 3.7 × 10 4 DA molecules/spike was found. Comparable neurotransmitter content values (∼3 × 10 4 DA molecules) were estimated from axonal DA vesicles using CFEs (Sulzer et al., 2016). In other CFE experiments on postnatally derived midbrain neurons, performed at a 100 kHz sampling rate to discriminate between single-spike and flickering events, DA release values were estimated to be around 1 × 10 4 and 2.4 × 10 4 DA molecules, respectively for single-spike and flickering events (Staal et al., 2004). It is worth noting that estimates of quantal size can be affected by the different experimental configurations used, such as detection being performed from the cell apex using CFEs or from the cell bottom using µG-SCD-MEAs (Amatore et al., 2007).

Detection of Spontaneous Firing From Cultured Midbrain Neurons
Besides detecting the quantal release of DA, µG-SCD-MEAs can be exploited to measure the electrical activity of cultured midbrain neurons. The 8 × 8 µG-SCD-MEAs and the conventional MEAs can reveal APs only after 7 DIV in approximately 30% of the electrodes, due to the delayed maturation of the network after dissociation. However, the network activity becomes detectable in most (i.e., 70%) of the electrodes at later stage of maturation (14 DIV). Spontaneous firing at this stage of maturation exhibits great variability of responses, even within the same µG-SCD-MEA or conventional MEA. Neuronal firing frequencies were scattered throughout a range of frequencies varying from 0.1 to 15 Hz ( Figure 3C) for both devices, which is in reasonably good agreement with previous reports of low frequency activity in isolated SN DA neurons (Bean, 2007;Margolis et al., 2010). A broad distribution of firing frequencies has also been observed in SN slices positioned onto conventional MEAs: the majority of neurons (∼94%) exhibit low firing frequencies (i.e., 1-3 Hz), whereas the remaining ones fire at higher frequencies (5-10 Hz) (Berretta et al., 2010). Different firing patterns have also been described for in vivo recordings, where SN DA neurons display both slow single-spike activity (1-10 Hz), and higher frequency discharges (∼13-20 Hz) (Dragicevic et al., 2015;Hage and Khaliq, 2015). A final consideration concerns the low firing rate (<1 Hz) that was recorded in the majority of neurons, and the hardly detectable DA release that occurred at 0.1 Hz. Both processes suggest the existence of a tonic discharge activity at rest, partially tuned by D 2 -autoreceptors (Al-Hasani et al., 2011).
Using µG-SCD-MEAs for Pharmacological Studies: The L-DOPA-Induced Down-Modulation of Spontaneous Firing The neuronal discharge of SN DA neurons is inhibited by D 2autoreceptors-mediated GIRK activation and is prevented by the D 2 -antagonist sulpiride (Mercuri et al., 1990;Dragicevic et al., 2014). To assay the sensitivity of the µG-SCD-MEAs, we tested this inhibitory pathway in current-clamped TH-GFP neurons, as well as in mature midbrain DA neurons, cultured for 2 weeks on µG-SCD-MEAs (or conventional MEAs). This autocrine inhibition is induced by adding L-DOPA, which is converted to DA and then released from dopaminergic neurons.
As shown, L-DOPA caused a range of effects (Figure 4). In most current-clamped neurons, the response caused a firing frequency reduction together with a slow membrane hyperpolarization and an increased AP amplitude that was reverted by sulpiride. The increased AP amplitude and the rapid increase in the AP rising phase was revealed as an enhanced dV/dt peak amplitude in the phase-plane plot analysis of AP recordings. The same occurrence was detected using both the µG-SCD-MEAs and conventional MEAs (Figure 5). Both MEAs revealed the reduced firing frequency and the increased AP rising phase, which were reverted by sulpiride. It is worth noting that, in the case of MEAs, the increased AP rising phase in currentclamp recordings is converted to an increased peak amplitude in the extracellular APs.
From a physiological point of view, both measurements are in excellent agreement and suggest sustained recruitment of Na V channels due to the increased cell hyperpolarization induced by GIRK K + channel activation. Sustained cell hyperpolarizations increase the rate of Na V channels recruitment from steady-state inactivation (Vandael et al., 2015;Guarina et al., 2018), while the recrutiment of different Na v channel isoform that characterized by a lower threshold of activation cannot be excluded.
Concerning the opposing effect that was observed in a minority of neurons, in which L-DOPA increased the spiking activity (Figure 6), variable responses have also been described in Substantia Nigra pars compacta (SNc) neurons, using MEA recordings from midbrain slices (Berretta et al., 2010). In that case, neurons that fired at high rates (>5 Hz) were insensitive to DA, while low-firing neurons were either highly or weakly inhibited by DA. Furthermore, a fraction of lowrate spiking neurons were insensitive to DA, or excited by DA, and a minority of neurons were potentiated by L-DOPA. Under our experimental conditions, where midbrain neurons were cultured for weeks on the microarray, signals detection may occur from different DA subpopulations (Lammel et al., 2008;Liss and Roeper, 2010), either from non-DA neurons or from DA neurons of the nearby ventral tegmental area. Indeed, the excitatory effects of L-DOPA on nigral dopaminergic neurons have been previously described, and were featured as an "early" and a "late" phase of excitation (Guatteo et al., 2013).
Finally, regarding the fraction of neurons that were inhibited by L-DOPA and did not undergo relevant membrane potential hyperpolarization, several pathways may be responsible for this modulation, based on the involvement of K + channels other than GIRK (Yang et al., 2013), or D 1 -mediated signaling cascades (Surmeier et al., 2007).

CONCLUSION
Our data demonstrate that µG-SCD-MEAs are highly reliable as multi-functional sensing multiarrays for long-term recordings of neuronal activity under variable pharmacological conditions. With respect to conventional approaches, the real-time measurements of quantal exocytosis and neuronal firing makes µG-SCD-MEA a promising biosensor for in vitro investigation of neuronal circuit properties as well as a valid tool for studying mistuned neurotransmission in neurodegenerative disorders.

AUTHOR CONTRIBUTIONS
GT performed the experiments and analyzed acquired data. FP fabricated the sensors, performed the experiments, and manuscript preparation. AB fabricated the sensors. BP contributed to critically editing the manuscript. SDM contributed to critically editing the manuscript. AP made hardware and software of the 16-channel setup and revised the manuscript. PO contributed to the design for the diamond biosensor and to the preparation of the manuscript. AM contributed to experimental design and manuscript preparation. PC contributed to critically editing the manuscript. EC contributed to the interpretation of AP recordings and helped with a critical revision of the manuscript. VC contributed to planning the experimental design, manuscript writing, and overall revision.

FUNDING
This work was supported by the following projects: Project 2015FNWP34 (from Italian MIUR) to PC and VC and CSTO165284 (from Compagnia di San Paolo) to VC. DIACELL project (from National Institute of Nuclear Physics) to FP. MiRaDS project (from CRT Foundation), "Finanziamento expost di progetti di ricerca di Ateneo" (from CSP Foundation), "Departments of Excellence" (L. 232/2016) project (from Italian MIUR) to PO. Ion beam irradiation was performed at the AN2000 accelerator of the Legnaro National Laboratories of the Italian Institute of Nuclear Physics (INFN) within the "Dia.Fab." beamtime.