Reactive Oxygen Species Mediate the Potentiating Effects of ATP on GABAergic Synaptic Transmission in the Immature Hippocampus*

Reactive oxygen species (ROS) constitute important signaling molecules in the central nervous system. They regulate a number of different functions both under physiological conditions and under pathological conditions. Here we tested the hypothesis that in the immature hippocampus ATP, the most diffuse neurotransmitter in the brain, modulates synaptic transmission via ROS. We show that ATP, acting on metabotropic P2Y1 receptors, increased the frequency of GABAA-mediated spontaneous postsynaptic currents (SPSCs) in CA3 principal cells, an effect that was prevented by the antioxidant N-acetyl-cysteine or by catalase, an enzyme that breaks down H2O2. The effect of ATP on SPSCs was mimicked by H2O2 or by the pro-oxidant, Fe2+, which, through the Fentol reaction, catalyzes the conversion of H2O2 into highly reactive hydroxyl radicals. MRS-2179, a P2Y1 receptor antagonist, removed the facilitatory action of Fe2+ on SPSCs, suggesting that endogenous ATP acting on P2Y1 receptors is involved in Fe2+-induced modulation of synaptic transmission. Imaging ROS with the H2O2-sensitive dye DCF revealed that ATP induces generation of peroxide in astrocytes via activation of P2Y1 receptors coupled to intracellular calcium rise. Neither N-acetyl-cysteine nor catalase prevented Ca2+ transients induced by ATP in astrocytes. Since a single hippocampal astrocyte can contact many neurons, ATP-induced ROS signaling may control thousands of synapses. This may be crucial for information processing in the immature brain when GABAergic activity is essential for the proper wiring of the hippocampal network.

Growing evidence suggests that, in physiological conditions, ROS regulate neuronal signaling both in the central and in the peripheral nervous system. Although at the periphery, ROS contribute to the inhibitory effect of ATP on quantal acetylcholine release from motor nerve endings (5), in the central nervous system, ROS produce both enhancement and depression of synaptic transmission (6 -9). In pathological conditions, increased amounts of ROS are involved in apoptosis and neurodegenerative disorders such as Parkinson and Alzheimer disease (3,4,10).
In the nervous system, ROS are produced mainly by microglia (11) and by astrocytes (10,12). In cultured astrocytes, activation of metabotropic P2Y receptors by extracellular ATP induces ROS via membrane-bound NADPH oxidase (10). However, it is unclear whether a similar mechanism also occurs in situ and whether ATP-induced ROS production in astrocytes can reach target neurons and affect synaptic transmission.
In the present study, electrophysiological and imaging techniques were combined to investigate ROS effects on GABAergic transmission in principal cells on acute hippocampal slices from newborn rats. Early in postnatal life, spontaneous activity in the hippocampus is characterized by network-driven membrane potential oscillations, the so-called giant depolarizing potentials (GDPs (13)), and spontaneous ongoing postsynaptic potentials (14). Both types of activity are required for proper wiring of developing circuits (15). During this period, ROS signaling should be more pronounced because of the relative deficiency of superoxide dismutase and glutathione peroxidase (16), known as powerful scavengers of ROS. In our previous studies, we have demonstrated that GDPs and GABA A -mediated spontaneous postsynaptic potentials are modulated by ATP, which is endogenously released during neuronal activity (14). Since ATP may induce ROS production in glial cells (12), it is important to test whether ROS induction is involved in the action of ATP. We found that ATP, via metabotropic P2Y 1 receptor, excites astrocytes and induces the release of diffusible ROS, which in turn facilitate GABA release onto CA3 pyramidal cells in the immature hippocampus.

EXPERIMENTAL PROCEDURES
Slice Preparation-Experiments were performed on hippocampal slices obtained from postnatal day 1 (P1)-P6 Wistar rats as described (17). Briefly, animals were decapitated after being anesthetized with an intraperitoneal injection of urethane (2 g/kg). All experiments were carried out in accordance with the European Community Council Directive of November 24, 1986 (86/609EEC) and were approved by local authority veterinary service. The brain was removed from the skull and placed in ice-cold artificial cerebrospinal fluid (ACSF) containing 130 mM NaCl, 3.5 mM KCl, 1.2 mM NaH 2 PO 4 , 25 mM NaHCO 3 , 1.3 mM MgCl 2 , 2 mM CaCl 2 , and 25 mM glucose and then saturated with 95% O 2 and 5% CO 2 (pH 7.3-7.4). Transverse hippocampal slices (500 m thick) were cut with a vibratome and stored at room temperature in a holding bath containing the same solution as above. After a recovery period of at least 1 h, an individual slice was transferred to the recording chamber, where it was continuously superfused with oxygenated ACSF at a rate of 2-3 ml/min at 33-34°C.
Electrophysiological Recordings-Electrophysiological experiments were performed from CA3 pyramidal cells using the whole-cell configuration of the patch clamp technique in voltage clamp mode. Patch electrodes were pulled from borosilicate glass capillaries (Hingelberg, Malsfeld, Germany). They had a resistance of 5-7 megaohms when filled with an intracellular solution containing (in mM): 140 KCl, 1 MgCl 2 , 1 NaCl, 1 EGTA, 2 HEPES, and 2 K 2 ATP. The pH was adjusted to 7.3 with KOH. Recordings were made with a patch clamp amplifier (Axopatch 1D; Axon Instruments). The whole-cell capacitance was fully compensated, and the series resistance (10 -20 megaohms) was compensated at 75-80%. The stability of the patch was checked by repetitively monitoring the input and series resistance during the experiment. Substances and drugs used were ATP and DPCPX (purchased from Tocris Cookson; Bristol, UK) and N-acetyl-L-cysteine (NAC), iron sulfate heptahydrate, 6,7-dinitroquinoxaline-2,3-dione (DNQX), catalase, and H 2 O 2 (purchased from Sigma, Milan, Italy). Solutions of NAC and Fe 2ϩ were prepared immediately before use. All drugs were dissolved in artificial cerebrospinal fluid; DNQX was dissolved in Me 2 SO. The final concentration of Me 2 SO in the bathing solution was 0.1%. At this concentration, Me 2 SO alone did not modify the shape or the kinetics of synaptic currents. Drugs were applied in the bath via a three-way tap system, by changing the superfusion solution to one differing only in its content of drug(s). The ratio of flow rate to bath volume ensured complete exchange within 1.5-2 min.
Analysis of Electrophysiological Data-Data were stored on magnetic tape and transferred to a computer after digitization with an A/D converter (Digidata 1200, Axon Instruments, Foster City, CA). Data were sampled at 20 kHz and filtered with a cut-off frequency of 1 kHz. Acquisition and analysis were performed with Clampex 7 (Axon Instruments, Foster City, CA). The analysis of SPSCs was performed with the Mini Analysis Program (Synaptosoft Inc., Decatur, GA). General principles and details on the analysis of synaptic currents can be found elsewhere (18). The analysis of spontaneous synaptic currents was performed on sequences with at least 100 events each. Signals were repeatedly averaged for at least 15 min before infusion of agonists or antagonists to ensure stable baseline conditions. Spontaneous events were detected using amplitude thresholds set as a multiple (4 -5 times) of the standard deviation of the noise. The cumulative amplitude and interevent plots obtained for cells in control and after drug application were compared using the Kolmogorov-Smirnov test. Val-ues are given as mean Ϯ S.E. Significance of differences was assessed by Student's t test or Mann-Whitney test. The significance level is p Ͻ 0.05.
Imaging of Calcium and ROS-Hippocampal slices were incubated for 1 h in continuously oxygenated ACSF containing 30 M of the cell-permeable H 2 O 2 -selective fluorescent dye carboxy-2,7-dichlorodihydrofluorescein diacetate (DCF, Molecular probes, Eugene, OR). After cleavage by cell esterases and oxidation, this probe was monitored using excitation sources and filters appropriate for fluorescein. Cells were imaged with the Bio-Rad Radiance 2100 laser scanning confocal microscope using the 488-nm laser line. Emitted green fluorescence was collected through a 520 long-pass filter. Bio-Rad software was used for image acquisition. Some of the major artifacts when DCF is used are photo-oxidation and photo-damage (19). To prevent these artifacts (see also Ref. 20), we used multiphoton excitation of DCF with Mai-Tai pulsed infrared laser (Spectra Physics). Excitation wavelength was set to 800 nm, and emission was collected through the same long-pass filter as in the single-photon confocal imaging mode. Another artifact that is not often taken into account in long-lasting experiments is leakage of the dye from cells (21). Therefore, we kept the dye constantly present in bath solution at the same concentration as that used during incubation. Prior to Ca 2ϩ imaging experiments, cells were incubated for 1 h in oxygenated ACSF containing 4 M fluo-4AM (Molecular Probes). An additional period of 30 min was allowed for de-esterification of the dye. Under these incubation conditions, astrocytes were loaded preferentially, whereas the dye uptake by neurons was minimal (22,23). Images were acquired every 5 s using the same filters as indicated above. Fluorescence intensity in background-subtracted frame averages was normalized and presented as (⌬F/ F ) ϭ [(intensity-basal)/(basal)]*100%.

RESULTS
The Antioxidant NAC Prevents the Facilitatory Action of ATP on GABAergic Transmission-During the first postnatal week, whole-cell recordings from CA3 pyramidal neurons revealed two types of spontaneous activity: GDPs and SPSCs. Although GDPs are network-driven events generated by the synergistic action of glutamate and GABA (13), SPSCs are almost exclusively driven by GABA released from GABAergic terminals since they were unaffected by DNQX (20 M) and completely blocked by GABA A receptor antagonists bicuculline or picrotoxin (14,24). As shown in Fig. 1, A, B, and E, bath application of ATP (50 M) in the presence of the adenosine receptor antagonist DPCPX (10 M) to prevent the activation of adenosine receptors (25,26) increased the frequency of spontaneous postsynaptic currents (SPSCs) from 4.8 Ϯ 1.0 to 9.4 Ϯ 0.4 Hz (n ϭ 6, p Ͻ 0.01). ATP application did not affect the amplitude of SPSCs (the peak amplitude was 32.8 Ϯ 7.9 and 29.9 Ϯ 6.8 pA before and during ATP application, respectively; n ϭ 6, p Ͼ 0.05; Fig. 1F ). Moreover, the potentiating effect of ATP was prevented by the P2Y 1 -selective antagonist MRS-2179 (10 M, see Fig. 4 of Ref. 14), indicating that it was mediated by metabotropic P2Y 1 receptors.
To test whether the potentiating effect of ATP on GABAergic SPSCs was mediated by ROS, we used the antioxidant NAC, which is a widely used ROS scavenger (5,27). The action of ATP on SPSCs was prevented by NAC ( Fig. 1, C, D, and E). In the presence of NAC (100 M), ATP failed to increase the frequency of SPSC (the frequency was 5.1 Ϯ 1.1 and 5.9 Ϯ 1 Hz before and during ATP, respectively, p Ͼ 0.05, n ϭ 6). NAC per se did not significantly modify the frequency (this was 5.8 Ϯ 1.2 and 5.1 Ϯ 1.1 Hz, before and during NAC, respectively, n ϭ 6; p Ͼ 0.05) or the amplitude of SPSCs (this was 16.7 Ϯ 3.6 and 14.8 Ϯ 3.2 pA before and during superfusion of NAC, respectively; p Ͼ 0.05). Similar results were obtained with NAC (1 mM, data not shown).
Consistent with the data obtained with NAC, catalase, an enzyme known to specifically break down H 2 O 2 (28,29), prevented the potentiating action of ATP on SPSCs (Fig. 2). Thus, in the presence of catalase (1200 units/ml), the frequency or amplitude of SPSCs was not modified by ATP (50 M; the frequency of SPSCs was 6.3 Ϯ 1.3 and 7.5 Ϯ 1.7 Hz in the absence or presence of ATP, respectively; n ϭ 7; p Ͼ 0.05, whereas the amplitude was 21.7 Ϯ 4.8 and 26.2 Ϯ 6.1 pA in the absence or presence of ATP, respectively; n ϭ 7; p Ͼ 0.5). Altogether, these results suggest that the action of ATP on spontaneous GABAergic synaptic currents is mediated by reactive oxygen species.   before and during H 2 O 2 , respectively ( p Ͼ 0.05; Fig. 3D). The effects of H 2 O 2 on SPSCs were prevented by NAC (1 mM; the frequency of SPSCs was 4.2 Ϯ 0.6 and 4.6 Ϯ 0.6 Hz in NAC and in the presence of NAC plus H 2 O 2 , respectively; n ϭ 6; p Ͼ 0.05; Fig. 3, B and C), indicating that they were mediated via ROS (30).
Endogenously Released ATP Affects Synaptic Release of GABA from GABAergic Interneurons-It is known that, in the presence of Fe 2ϩ , H 2 O 2 is readily converted via the Fenton reaction into the hydroxyl radical (OH ⅐ ), which is the strongest oxidant among other ROS (31). In this set of experiments, we tested whether Fe 2ϩ can enhance synaptic activity via endogenous ROS. Fe 2ϩ in a relatively low concentration (10 M) increased the frequency of SPSCs from 6.0 Ϯ 0.7 to 8.7 Ϯ 1.0 Hz (n ϭ 7; p Ͻ 0.05; Fig. 4, B and C). Iron did not change the amplitude of SPSCs (the peak amplitude of synaptic currents was 35.8 Ϯ 9.5 and 32.9 Ϯ 9.5 pA before and during application of Fe 2ϩ , respectively; p Ͼ 0.05 (Fig. 4C)). The potentiating effect of iron on SPSCs was prevented by NAC (1 mM), indicating that it was mediated by ROS (Fig. 4, A and C). Thus, the frequency of SPSCs was 7.9 Ϯ 0.9 and 8.6 Ϯ 1.0 Hz in the presence of NAC and NAC plus iron, respectively (n ϭ 6; p Ͼ 0.05), whereas the amplitude was 13.4 Ϯ 3.6 and 13.8 Ϯ 3.8 in the presence of NAC and NAC plus iron, respectively (n ϭ 6; p Ͼ 0.05).
To test whether Fe 2ϩ -induced enhancement of synaptic activity was mediated by endogenously released ATP, we tried to antagonize the effect of Fe 2ϩ with the selective P2Y 1 receptor blocker MRS-2179. Indeed, MRS-2179 at the concentration of 10 M, which per se did not significantly modify the frequency of spontaneous synaptic events (14), strongly antagonized the facilitatory action of Fe 2ϩ on spontaneous GABA A -mediated synaptic currents (Fig. 4, B and C). The frequency of SPSCs was 9.5 Ϯ 0.8 and 6.9 Ϯ 0.9 Hz in the presence of Fe 2ϩ and Fe 2ϩ plus MRS-2179, respectively, whereas the amplitude was 45.5 Ϯ 16.2 and 39.5 Ϯ 11.1 pA in the presence of Fe 2ϩ and Fe 2ϩ plus MRS-2179, respectively. These data indicate that endogenous ATP acting on P2Y 1 receptors is involved in iron-induced modulation of synaptic transmission.
Ca 2ϩ Imaging in Astrocytes and the Action of NAC-The increase in intracellular calcium in astrocytes has been shown to be an important step in the generation of ROS via NADPH oxidase (10). Previous experiments from acute hippocampal slices obtained from adult rats have revealed calcium transients in astrocytes following activation of P2Y 1 receptors (25, 32). To test whether, in immature hippocam-  pal slices, ATP was able to induce calcium transients in astrocytes, slices were loaded with the Ca 2ϩ -sensitive dye fluo-4AM. Consistent with previous studies (25,32), application of ATP (100 M) elicited large Ca 2ϩ transients in astrocytes ( Fig. 5A; for astrocyte identification, see the section below). ATP-induced calcium transients were blocked by MRS-2179 (50 M ; Fig. 5, A and D), indicating that they were mediated by P2Y 1 receptor subtypes.
Interestingly, the application of NAC at the same concentration used to prevent the effects of ATP on SPSCs (1 mM) did not modify ATP-induced Ca 2ϩ responses (Fig. 5, B and D). Similarly, catalase (2000 units/ml) failed to modify ATP-induced Ca 2ϩ transients in astrocytes (Fig. 5, C and D). These data indicate that the blocking effect of NAC or catalase on GABAergic synaptic transmission occurs downstream of P2Y 1 receptormediated Ca 2ϩ rise.
ROS Imaging-The above mentioned experiments clearly indicated that ATP is able to induce calcium transient in astrocytes via P2Y 1 receptors. To directly test whether ATP triggers ROS generation in astrocytes, we performed imaging of ROS in brain slices using the fluorescent dye DCF (5,10,21).
In the CA3 stratum radiatum region of newborn rat hippocampal slices, bath application of ATP (100 M) produced a fast and reversible increase in the DCF fluorescent signal of identified astrocytes but not in adjacent interneurons or pyramidal cells (Fig. 6, A, C, and D). Astrocytes were identified on the basis of the lack of action potential generation in response to depolarizing current pulses (Fig. 6B) and on the basis of their morphology (33). To calibrate the ATP-induced increase in the DCF fluorescence, H 2 O 2 (100 M) was applied at the end of the experiments. As shown in Fig. 6D, H 2 O 2 induced an increase in DCF fluorescence similar to that produced by ATP. These data confirm that in immature hippocampal slices, astrocytes are the primary source of ATP-induced free oxygen radicals.

DISCUSSION
The main finding of the present study is that in the immature hippocampus, ATP, a widely distributed transmitter and neuromodulator, controls GABA release onto CA3 principal cells via induction of ROS. Imaging experiments with the selective ROS-sensitive dye DCF have clearly demonstrated that ROS are produced by glial cells following activation of P2Y 1 receptors by  ATP. Thus, purinergic modulation of synaptic transmission in the brain involves cross-talk between glia and neurons mediated by ROS acting as diffusible messengers.
ROS Are Involved in ATP-induced Modulation of GABAergic Transmission-During the first postnatal week, when associative commissural fibers are still poorly developed (34,35), CA3 pyramidal cells receive a strong GABAergic innervation from the mossy fibers and GABAergic interneurons (36). Therefore, during this period, spontaneous synaptic activity recorded from CA3 principal cells is mediated mainly by GABA acting on GABA A receptors, as demonstrated by the observation that in the presence of GABA A receptor antagonists, spontaneous glutamatergic events are merely detectable (24). In previous work, we have demonstrated that either endogenously released or exogenously applied ATP, via the activation of metabotropic P2Y 1 receptors, exerts a powerful excitatory action on spontaneous GABA A -mediated synaptic events (14). The present data provide evidence that ATP-induced increase in GABA release is mediated by ROS because: (i) the effect of ATP was prevented by NAC, a precursor of the most efficient endogenous antioxidant glutathione (27); (ii) the potentiating effect of ATP was abolished by catalase, an enzyme that specifically breaks down H 2 O 2 (28,29); (iii) the effect of ATP on GABA release was mimicked by low concentrations of H 2 O 2 , a mild oxidant that, unlike other ROS, is relatively stable and capable of diffusing at relatively long distances (4); and (iv) a similar facilitatory effect was produced by the pro-oxidant Fe 2ϩ , which, via the Fenton reaction, catalyzes the transformation of H 2 O 2 into highly reactive hydroxyl radicals (31).
In addition, imaging experiments with a selective ROS-sensitive dye have provided direct evidence that ROS production is triggered by ATP via activation of P2Y 1 receptors. However, the present experiments do not allow excluding the possibility that other radicals are involved in ATP action on synaptic transmission. It has been recently demonstrated that ATP can induce in astrocytes the production of nitric oxide (37), which may interact with ROS (38). Thus, nitric oxide, either alone or in cooperation with superoxide (to produce highly aggressive ONOO Ϫ (28, 39)), can control GABAergic signaling in the hippocampus. Altogether, our data reveal that, in the immature hippocampus, ATP controls the local network through ROS production.
Endogenous ATP Contributes to ROS Generation-ATP is the most ubiquitous molecule stored at high concentrations (mM range) in synaptic vesicles from which it can be released in an activity-dependent manner (40 -42). We have previously shown that, in the immature hippocampus, synchronized network activity such as GDPs enhances the level of endogenous ATP, which in turn regulates GDP induction (14). Endogenous ATP can also be released in a Ca 2ϩ -dependent way from astrocytes (Fig. 7), which abundantly express purinergic receptors (25,32,43,44). Accumulation of endogenous ATP in the extracellular space may lead to ROS production by astrocytes and facilitation of GABA release from interneurons. Consistent with this view, Fe 2ϩ -induced facilitation of synaptic transmission was prevented when the action of endogenous ATP on P2Y 1 receptors was reduced by MRS-2371, a specific P2Y 1 antagonist (Fig. 7). It should be mentioned that P2Y 1 receptor can also be activated by ADP, the breakdown product of ATP (25,32,45,46). Therefore, we cannot exclude the possibility that endogenous ADP acting on P2Y 1 receptors contributes to ROS generation by astrocytes. In turn, the contribution of ADP would be dependent on the activity of ectoATPases, which control the rate of ATP breakdown (14,45).
Our experiments also reveal facilitation of GABA release by endogenous H 2 O 2 , which is more evident in the presence of Fe 2ϩ . It is worth noting that in newborns, there is an excess of free iron (16), which is a cofactor for the generation of highly reactive hydroxyl radicals.
Sources of ROS in the Hippocampus-What are the cellular sources of ROS? Possible candidates are astrocytes and interneurons since both of these cells express metabotropic P2Y 1 receptors (25,32). However, the present work based on DCF imaging from brain slices combined with electrophysiological recordings has clearly identified astrocytes as the main cellular source of ROS. In the present study, we did not analyze the molecular mechanism of ROS generation. Nevertheless, recent studies on cultured astrocytes have shown that ATP-induced ROS are generated mainly by membrane NADPH oxidase (10).
Physiological Implication of ROS in the Immature Hippocampus-The present data demonstrate that the hippocampus of newborn rats is very sensitive to ROS since H 2 O 2 at low concentrations was effective in enhancing the frequency of spontaneous GABA A -mediated synaptic events. It is known that in particular situations such as reperfusion after ischemia, endogenous H 2 O 2 can reach in certain brain regions a local concentration up to 150 M (47). Such high concentrations of H 2 O 2 may have a variety of adverse effects or even kill the neurons (48). The high sensitivity of newborns to ROS may result not only from the relative deficiency of superoxide dismutase and glutathione peroxidase in the immature brain, known as FIGURE 7. Scheme for the modulatory effect of ATP on SPSCs in neonatal hippocampus. ATP acts on astrocytes, which express P2Y 1 receptors, to increase the level of intracellular Ca 2ϩ . Ca 2ϩ rise stimulates the release of SO 2Ϫ via activation of NADPH oxidase. In the presence of superoxide dismutase, SO 2Ϫ is transformed in H 2 O 2 . H 2 O 2 as a diffusible messenger might affect neighboring GABAergic terminals either directly or via transformation into highly reactive hydroxyl radicals (OH ⅐ ). The transformation of H 2 O 2 into OH ⅐ is accelerated in the presence of divalent iron via the Fenton reaction. A single astrocyte may control ϳ140,000 synapses, providing a widespread control of synaptic transmission. The action of ATP could be blocked either at the astrocyte level by specific P2Y 1 receptor antagonists or by catalase or by NAC, a potent ROS scavenger. Consistent with this, NAC also blocks the facilitatory action Fe 2ϩ or exogenous H 2 O 2 .
the main scavengers of ROS, but also from the excess of free iron (16). Consistent with this, in the present work, even low micromolar concentrations of Fe 2ϩ were able to modulate the action of endogenous ATP, promoting facilitation of GABA release into pyramidal cells. As already mentioned, at early developmental stages, GABAergic transmission provides the major excitatory drive to pyramidal cells (14,36,49). Fig. 7 shows that ATP released from astrocytes induces peroxide via activation of P2Y 1 receptors coupled to intracellular calcium rise. In the presence of Fe 2ϩ , H 2 O 2 is converted in hydroxyl radicals, which in turn facilitate GABA release from mossy fibers and GABAergic interneurons onto principal cells. It is worth noting that ROS act mainly presynaptically without modifying postsynaptic responsiveness. At the neuromuscular junction, ROS were found to affect acetylcholine release, probably interfering with the vesicle fusion protein SNAP25 (50). However, this protein is not expressed in hippocampal GABAergic synapses (51), suggesting another target for modulation of GABAergic synaptic currents by ROS. Considering that a single astrocyte may control ϳ140,000 synapses (33) and that neighboring astrocytes communicate by means of intercellular Ca 2ϩ waves (43), the release of ROS from astrocytes may result in a widespread and synchronized synaptic modulation.
In the immature hippocampus, synaptic GABAergic activity is required for proper wiring of developing circuits (15). This implies that any change in pro-oxidant levels may interfere with ATP-induced ROS production and synaptic transmission, thus contributing to information processing at the network level.