Exocytosis of ATP From Astrocytes Modulates Phasic and Tonic Inhibition in the Neocortex

1ATP acts as neurotransmitter mediating excitatory synaptic transmission and synaptic plasticity in the central nervous system,. There is growing evidence that ATP can also play an important role in signal transfer between neuronal and glial circuits and within glial networks –. ATP can regulate growth and development of neural cells and contribute to various pathological processes –. Action of ATP is mediated by ionotropic P2X and metabotropic P2Y receptors abundantly expressed in many types of neurons and glial cells,. By virtue of the high Ca2+-permeability of P2X receptors and the ability of P2Y receptors to stimulate IP3-dependent Ca2+ release from endoplasmic reticulum, purinergic receptors can transmit robust Ca2+-signals and thereby modulate activity and trafficking of excitatory and inhibitory receptors,. In addition to direct actions mediated by P2 purinoreceptors, ATP can initiate secondary neuromodulation via P1 adenosine receptors after rapid degradation by ecto-nucleotidases to adenosine,. Different mechanisms of ATP release have been identified, such as vesicular release from nerve terminals, and several nonvesicular pathways, including concentration gradient-driven diffusion through gap-junction hemichannels, anion channels, and dilated P2X7 receptors,,,.

2A physiological role of ATP release from astrocytes has been suggested by the participation of ATP in the propagation of glial Ca2+-waves – and significant contribution of ATP and adenosine to the astroglia-driven modulation of neuronal activity and sleep homeostasis,,. There is growing evidence, albeit obtained mostly in cell cultures, that the release of gliotransmitters may share common mechanisms of vesicular neurotransmitter release such as a dependence on the proton gradient and vesicular transporters, SNARE proteins, andintracellular Ca2+ elevation,,. Importantly, astroglial-driven release of ATP and modulation of synaptic plasticity in the hippocampus were suppressed in transgenic mice expressing a dominant-negative SNARE (dn-SNARE) domain selectively in astrocytes.

3However, the mechanism of gliotransmitter release from astrocytes has been disputed. Ca2+-dependent exocytosis of glutamate and ATP, mainly from cultured hippocampal astrocytes, has been reported –. By contrast, alteration of astroglial InsP3-mediated Ca2+-signaling did not have a significant effect on glutamatergic synaptic transmission in the hippocampal slices, fuelling the debate on the role for glial exocytosis in more intact tissue. Still, more recent in situ and in vivo data demonstrated an effect of astroglial InsP3-mediated Ca2+-signaling on cholinergic modulation of synaptic plasticity in hippocampus and neocortex –. At the same time, two recent studies reported the possibility of nonvesicular release of astroglial glutamate through the TREK-1 and best1 channels and the lack of immunostaining for vesicular glutamate transporters in brain astrocytes, contrasting with the bulk of evidence for glial exocytosis obtained by variety of different techniques –,. Thus, physiological relevance of Ca2+-dependent exocytosis of gliotransmitters remains controversial.

4In this study, to avoid possible artifacts of cell culture, we investigate release of ATP from acutely isolated cortical astrocytes and astrocytes in the neocortical slices. We provide several lines of evidence for (1) existence of functional vesicular mechanism of Ca2+-dependent gliotransmitter release in neocortical astrocytes, (2) quantal P2X receptor-mediated currents directly activated in neocortical neurons by release of ATP from astrocytes, and (3) glia-driven purinergic modulation of GABAergic transmission that is impaired by astrocytic expression of dn-SNARE or deletion of P2X4 receptors.

5As a first step in demonstrating the presence of quantal Ca2+-dependent release of ATP from astrocytes, we used a sniffer cell approach where ATP release from astrocyte was detected by HEK293 expressing P2X2 receptors (see Materials and Methods for details). We compared release of ATP from astroglia of somatosensory cortex of wild-type mice and from mice conditionally expressing the SNARE domain of VAMP2 selectively in astrocytes (dn-SNARE mice),.

6Acutely isolated cortical astrocytes were separately loaded with Ca2+-indicator Fluo-4 and photosensitive Ca2+-chelator NP-EGTA and then distributed into a recording chamber containing preplated HEK293-P2X2 cells (Figure 1). Whole-cell voltage-clamp recordings were made from a HEK293-P2X2 cell lying in the immediate vicinity to astrocytes (Figure 1A). Identification of astrocytes was confirmed by their functional characterization at the end of experiment, including low input resistance, lack of voltage-gated Na+-conductance, large K+-conductance, large conductance mediated by glutamate transporters, and NMDA receptors lacking Mg2+-block (Table S1). Thus, influence of nonastrocytic cells in the experiments reported below can be ruled out.

7Brief flash of UV-light caused the uncoupling of Ca2+ from NP-EGTA (monitored by Fluo-4 Ca2+-indicator). This was accompanied by an asynchronous burst of phasic currents in the adjacent ATP-sensitive sniffer cell in 17 of 20 experiments (Figure 1B). P2 receptor antagonist PPADS (10 µM) prevented the detection of UV-evoked phasic currents (n = 7 of 7; Figure 1B), confirming that they were mediated by ATP acting via P2X receptor.

8Some phasic currents were observed in the sniffer cells having astrocytes laying on their surface even before triggering the Ca2+-rise in the astrocyte, but at much lower frequency (Figure 1B). These baseline events were not detected in the absence of astrocytes in any of 10 native HEK293-P2X2 cells tested. Application of PPADS (10 µM) reduced the frequency of baseline and UV-evoked events correspondingly by 94%±3% and 97%±3% (Figure 1C), confirming the purinergic nature of all phasic currents. An exocytotic mechanism of ATP release was suggested by the activity-dependent staining of astrocytes with vesicular marker FM1-43 (Figure 1C).

9To further test the role of an elevation of cytosolic Ca2+ concentration as a trigger for the release of ATP from astrocytes, we applied agonists of astroglial metabotropic PAR-1 receptors (see also Figure S1), and ionotropic NMDA receptors,. The PAR-1 receptors were chosen as an ample method of astrocyte activation because of the ability to activate IP3 pathway and their predominant expression in glial but not in neuronal cells,. Similar to UV-uncaging, activation of cytosolic Ca2+-transients in the astrocyte either by application of the PAR-1 agonist TFLLR (10 µM; n = 10) or by application of NMDA (20 µM; n = 7) elicited bursts of spontaneous currents in the adjacent ATP-sensitive sniffer cells (Figure 2A). TFLLR and NMDA did not evoke any activity in the HEK293-P2X2 cells when applied in the absence of astrocytes (n = 5 agonist, unpublished data). The phasic currents in the sniffer cells had amplitudes of 11.2±2.4 pA (n = 34) and rise and decay time of 1.6±0.5 ms and 15.2±3.9 ms correspondingly (Figure 2D), thus resembling parameters of purinergic synaptic currents,.

10As a test of a vesicular mechanism of ATP release, we isolated astrocytes from cortical slices pretreated with blocker of vacuolar-type H-ATPase bafilomycin A1 (1 µM) for 2 h. Treatment with bafilomycin caused a decrease in the amplitudes and frequency of phasic currents initiated by Ca2+-elevation in the astrocytes (Figure 2B). The mean amplitude of phasic purinergic currents activated in the sniffer cell by stimulation of bafilomycin-treated astrocytes was only 3.1±0.4 pA (n = 19). The overall charge transferred by phasic currents activated after stimulation of bafilomycin-treated astrocytes by UV, TFLLR, and NMDA was 9.7%±4.2% (n = 7), 11.8%±5.6% (n = 5), and 12.6%±5.8% (n = 7) of the corresponding control values (Figure 2E).

11In support of exocytotic mechanism of ATP release, astrocytes obtained from dn-SNARE mice elicited a diminished burst of purinergic currents in sniffer cells regardless of the method used to elevate cytosolic Ca2+-level (Figure 2C,E). We also observed the SNARE-complex-dependent release of ATP from the isolated hippocampal astrocytes (Figure S2).

12The amplitude histograms of the phasic P2X-mediated currents activated by elevation of astrocytic Ca2+ exhibited prominent second peak (Figure 2D; see also Figure S3) at amplitude twice that of the primary peak. Fitting of amplitude distribution with simple multiquantal binomial model (shown in Figure 2D as dotted line) gave a quantal size of 7.9±0.13 pA and release probability of 0.28±0.04. Similarly, fitting of the distributions of P2X-currents evoked in sniffer cell after application of NMDA and TFLLR gave quantal size of 8.14±0.15 pA and 8.05±0.11 pA, respectively.

13It should be noted that astrocyte-driven purinergic currents observed in our experiments had much faster kinetics (10–25 ms) than nonvesicular release of gliotransmitters from hippocampal astrocytes, which was mediated by TREK-1 potassium channels and best1 chloride channels. To verify the lack of contribution of nonvesicular mechanisms to the quantal purinergic phasic currents in the sniffer cells, we activated astrocytes by TFLLR in presence of TREK-1 channels inhibitor fluoxetine and large conductance chloride channels inhibitors DIDS and NPPB. Application of fluoxetine (100 µM), DIDS (300 µM), and NPPB (100 µM) did not have marked effect on the astrocyte-evoked phasic currents in the HEK293 cells in any of five experiments for each inhibitor (Figure S4).

14Combined, the above results strongly suggest that activity-dependent release of ATP from cortical astrocytes occurs mainly via quantal exocytotic mechanism, dependent on SNARE protein complex.

15Vesicular mechanism of ATP release from neocortical astrocytes was also supported by immunostaining of living isolated astrocytes with antibodies to vesicular nucleotide transporter VNUT1 and various vesicular, neuronal, and glial marker proteins (Figures S5 and S6). Although immunostaining of live cells has certain limitations (see Text S1) and should be interpreted with great caution, our data suggest the good co-localization of VNUT1 and synaptic vesicle (SV) markers (Figure S5A), which is in agreement with previous reports of vesicular location of VNUT1 and presence of synaptic-like vesicles in astrocytes,. We observed weaker correlation between VNUT1 and lysosomal markers cathepsin D and LAMP3 (Figure S5B,C), which goes in line with previous observation of astroglial ATP release by the lysosome exocytosis.

16However, lysosomal exocytosis from astrocytes exhibited much slower kinetics than the purinergic currents measured in the sniffer cells (Figures 1 and 2). This argues against a major contribution of this mechanism to the present observations. As the kinetics of sniffer cell responses are more consistent with millisecond time-scale of SV exocytosis from astrocytes, we suggest that the astrocyte-driven purinergic currents observed in our experiments could be triggered by exocytosis of ATP from synaptic-like vesicles. We also observed an immunoreactivity for vesicular glutamate transporter (VGLUT1) in the fraction of cortical astrocytes (Figure S5D,E; Figure S6A), which goes in line with data reported previously for hippocampal and cortical astrocytes,. Of course, further investigation of exocytosis of glutamate from neocortical astrocytes is required, which is beyond the scope of this article.

17Previously we have shown that cortical pyramidal neurons express ionotropic P2X purinoreceptor, which can be activated by synaptic release of ATP. Hence, it might be plausible to detect the glia-driven contribution to purinergic current in neurons. We recorded whole-cell currents in the pyramidal neurons of neocortical layer 2/3 of brain slice at membrane potential of −80 mV in the presence of glutamate receptor antagonists CNQX (50 µM) and D-APV (30 µM) and irreversible blocker of GABA receptors picrotoxin (100 µM). Like our previous results, we observed residual nonglutamatergic excitatory spontaneous synaptic currents (Figure 3A,B); neither amplitude nor frequency of residual currents were affected by further increase in concentrations of glutamate and GABA receptors antagonists (n = 10 cells tested, unpublished data).

18The amplitude of inward spontaneous excitatory currents (sEPSCs) was reduced by specific antagonists of P2X receptors PPADS (10 µM) and NF-279 (3 µM) correspondingly by 45%±1 3% (n = 7) and 56%±19% (n = 16); the sEPSC frequency was reduced by PPADS and NF-279 by 65%±22% and 69%±27%, respectively (Figures 3 and S7). At concentrations used, both PPADS and NF279 are selective for P2X receptors –. Based on these data as well as our previous work,, the spontaneous inward currents observed in cortical neurons in the presence of glutamatergic and GABAergic antagonists can be confidently attributed to the ATP receptors.

19The partial inhibitory action of PPADS and NF-279 on nonglutamatergic sEPSCs could be explained by participation of homomeric P2X4 receptors, which are insensitive to these antagonists,,. Since P2X4 subunit-containing receptors are abundantly expressed in the brain and could potentially contribute to neuronal purinergic signaling,–, we used previously characterized P2X4 receptor knockout mice (P2X4 KO). In the P2X4 KO mice, application of 10 µM PPADS decreased the amplitude and frequency of nonglutamatergic sEPSCs (Figure 3C) by 74%±10% and 97%±6% correspondingly (n = 12); difference from the wild-type mice was statistically significant with p = 0.05 and 0.01. Taking into account that significant attenuation of sEPSCs can put them below the detection threshold, nonglutamatergic sEPSCs can be confidently attributed to neuronal P2X receptors.

20We triggered the release of gliotransmitters from the astrocytes by rapid application of an agonist of PAR-1 receptor TFLLR to neocortical slices. As in hippocampus,, TFLLR (10 µM) triggered cytosolic Ca2+ rise predominantly in astrocytes (Figure S1). Application of TFLLR caused a dramatic increase in the frequency of ATP-mediated spontaneous currents (Figure 3A,D). TFLLR also elevated astrocytic Ca2+ in dn-SNARE mice, but the burst of purinergic spontaneous currents was not detected (Figure 3B,E). In the P2X4 knockout mice, the average increase in the purinergic sEPSCs frequency reached only 28%±15% (n = 12), which was significantly lower (p<0.005) than 72%±21% increase observed in wild-type mice (Figure 3C,F). Combined, these results demonstrate that activation of astrocytes can evoke synaptic-like purinergic currents in neurons. In addition to effects of knocking out and inhibiting P2X receptors, the purinergic nature of glia-driven spontaneous currents was corroborated by inhibitory action of ATP-hydrolyzing enzyme apyrase (Figure S8). The apyrase application significantly decreased the mean amplitude and frequency of sEPSCs and abolished the TFLLR-induced burst. The decrease in the sEPSCs frequency is most likely related to the reduction of their amplitude below threshold of detection.

21In the wild-type mice, purinergic sEPSCs showed bimodal amplitude distributions (Figure 3G–I; see also Figure S9) with peaks at 3.1±0.9 pA and 5.7±1.6 pA (n = 14); decay time distributions had peaks at 9.1±0.9 ms and 15.3±1.8 ms (Figure 3G). TFLLR selectively increased the probability of detection of smaller and slower sEPSCs in all 14 neurons tested. In contrast, recordings made from cortical neurons of dn-SNARE mice did not show two peaks in the distributions of amplitude or decay time; the amplitude and decay time were not altered after TFLLR application (Figure 4H). In the P2X4 KO mice, the amplitude distribution of purinergic sEPSCs showed peaks as 2.5±0.7 pA and 4.2±1.1 pA; decay time distribution had peaks at 9.0±0.9 and 15.7±1.7 in control. Activation of astrocytes caused just a moderate increase in the proportion of smaller and slower sEPSCs in the P2X4 KO neurons (Figure 3I).

22Elimination of the distinct population of smaller and slower currents by astrocytic dn-SNARE expression strongly suggests that this population of purinergic currents was elicited by exocytosis of ATP from astrocytes. The vesicular origin of slower purinergic sEPSCs was also supported by elimination of these events by treatment of the cortical slices with 1 µM bafilomycin A1 (Table S2). The slower and faster purinergic currents recorded in cortical neurons exhibited different quantal behavior (Figure S7). The slower purinergic currents evoked by application of TFLLR in the presence of TTX (Figure S3A) exhibited multiquantal amplitude distribution, whereas faster currents exhibited monoquantal distribution typical for miniature synaptic currents (Figure S7B,C).

23Thus, detailed analysis of purinergic sEPSCs in the pyramidal cortical neurons revealed two distinct populations of events, which differ by their amplitude and kinetics. Based on their insensitivity to astrocytic dn-SNARE expression, larger and faster sEPSCs most likely have a neuronal origin. In contrast, sEPSCs of smaller amplitude and slower kinetics can be attributed to the vesicular release of ATP from astrocytes. In the following sections, we provide further experimental support of this notion.

24We sought to obtain a parallel line of evidence for the vesicular ATP release from cortical astrocytes via an alternative approach: we measured ATP concentration in neocortical slice using microelectrode biosensors (see Materials and Methods), a technique that has been applied previously for evaluation of ATP release in several brain areas,. Selective activation of astrocytic PAR-1 receptor by TFLLR induced a robust increase in the extracellular ATP concentration in the cortical tissues of wild-type mice; this increase was impaired in the dn-SNARE mice and was blocked by bafilomycin, confirming its astroglial origin and vesicular nature (Figure S10A). The increase in the “tonic” concentration of extracellular ATP after activation of astrocytes in the wild-type mice reached 1.1±0.4 µM (Figure S10B) and was inhibited by 84%±5% (n = 7) after incubation with bafilomycin. The TFLLR-evoked elevation of ATP concentration in the dn-SNARE mice was decreased by 56%±12% (n = 12) as compared to wild-type. These results support the significant contribution of vesicular mechanism to the activity-dependent release of ATP from cortical astrocytes. Taking into account that bafilomycin can inhibit only re-charging of ATP-storing vesicles and not all the cortical astrocytes express dn-SNARE protein, one could not expect the full inhibition of vesicular ATP release in these experiments. Thus, the incomplete inhibition of TFLLR-evoked ATP transients in the dn-SNARE mice and in the bafilomycin-treated neocortical slices of wild-type mice could hardly be attributed to a large contribution of nonvesicular release.

25We have shown previously that stimulation of intracortical afferents is able to significantly elevate cytosolic Ca2+-level in the cortical astrocytes acting via ionotropic and metabotropic receptors to glutamate and ATP,,. We asked whether an episode of high-frequency stimulation (HFS) could similarly trigger release of ATP from astrocytes in situ. As before, we monitored spontaneous purinergic currents in the pyramidal cortical neurons at a membrane potential of −80 mV in the presence of CNQX and picrotoxin (Figure 4). The short HFS train triggered more than a 2-fold elevation of the frequency of purinergic sEPSCs in the pyramidal neurons of wild-type mice (Figure 4A,C). Such an elevation did not occur in the dn-SNARE mice, where HFS train caused only a modest transient increase in the sEPSC frequency (Figure 4B,D).

26The HFS-induced changes in the amplitude and kinetics of phasic purinergic currents had a complex pattern in the wild-type mice (Figure 4C,E). During the first 30 s after HFS train, the average amplitude of events increased to 12.4±4.2 pA (n = 6) as compared to 8.6±2.4 pA in the baseline conditions and their decay time was slightly larger (11.2±2.3 ms, n = 6) than in control (9.9±2.7 ms). The sEPSCs appeared in the neurons of wild-type mice 1–3 min after HFS train had lower amplitudes (6.8±1.6 pA, n = 6) and much larger decay times (13.4±3.6 ms) than in control. Analysis of the amplitude and decay time distributions (Figure 4E) revealed a significant increase in the number of smaller, slower sEPSCs after HFS train that formed the distinct fraction of purinergic spontaneous events. In addition to smaller and slower currents, the number of fast sEPSCs with fast decay times (9.2±2.5 ms) and large amplitudes (19.6±2.7 pA) was observed during the first 30 s immediately after stimulation (Figure 5E, purple lines). The amplitudes of these large currents corresponded to the double of unitary amplitude of fast purinergic currents; this explains a short-lived increase in the average amplitude immediately after HFS train.

27The existence of two functionally distinct populations of purinergic events in the wild-type mice was corroborated by correlation between amplitude and decay time of sEPSCs (Figure 4G). The slower currents (decay time of 15.4±2.2 ms) had smaller amplitudes (5.5±1.3 pA, n = 6), but the faster currents (decay time of 9.2±1.3 ms) had higher amplitude (9.9±2.4 pA, n = 6). The amplitudes of slower currents closely agree with quantal amplitude of TFLLR-evoked slow purinergic currents recorded at −80 mV, whereas quantal amplitude of fast currents is close to the unitary size of TFLLR-insensitive fast purinergic sEPSCs (Figure S6B). The train of HFS significantly increased the number of slower spontaneous currents with smaller amplitude (Figure 4G).

28In contrast to wild-type mice, the only effect produced by HFS train on purinergic sEPSCs in the dn-SNARE mice was the transient increase in the number of double-quantal fast currents (Figure 4D,F), which led to the brief increase in the average amplitude. The expression of dn-SNARE in astrocytes caused a selective loss of the smaller and slower sEPSCs (Figure 4H). These data strongly support the different origins of fast and slow purinergic sEPSCs, from neuronal terminals and astrocytes correspondingly.

29To verify that slower purinergic sEPSCs originated from astrocytic ATP release directly, we tested the effect of diadenosine triphosphate (AP3A) and UTP, which have been shown presviously to strongly inhibit transport of ATP into astrocytic vesicles. Since these substance are not specific VNUT antagonists and can have an action on purinergic receptors, we applied them intracellulary to minimize side effects. In order to increase the impact of single-cell perfusion, we chose neuron-astrocyte pairs lying in a close proximity (Figure 5A). A similar strategy has been previously used to test the effects of perfusion of Ca2+-chelators into astrocytes. The feasibility of this approach is based on the high probability of synapses of a single neuron falling within functional island enwraped and controlled by single nearby astrocyte. When astrocytes of wild-type mice were perfused only with fluorescent dye, two consecutive HFS episodes caused the burst of slow purinergic sEPSCs in the neighboring neurons (Figure 5B) of the magnitude similar to previous experiments with intact astrocytes (Figure 4). Prolonged intracellular perfusion of astrocyte with 1 mM of AP3A or UTP significantly attenuated the frequency of purinergic sEPSCs (Figure 5B–E) in 10 of the 12 experiments. The effect was more prominent after the second HFS episode. This was most likely related to the depletion of the releasable pool of ATP in astrocytes. Analysis of amplitude and decay time distributions showed that inhibitors of vesicular nucleotide transporters selectively affected the fraction of slower sEPSCs (Figure 5C), significantly decreasing their amplitude and frequency (Figure 5D). Although slower purinergic sEPSCs were not abolished completely, this might be explained by incomplete perfusion of distal astrocytic processes or release from other astrocytes. These results strongly suggest that smaller and slower purinergic sEPSCs originated directly from vesicular release of ATP from neighboring astrocytes.

30In summary, our data provide compelling evidence that quantal release of astrocytic ATP activates a distinct population of purinergic currents in cortical pyramidal neurons.

31ATP release from astrocytes can activate neuronal P2X and P2Y receptors. The following increase in cytosolic Ca2+-signals may trigger a variety of intracellular cascades implicated in the modulation of synaptic strength,,. In particular, phosphorylation of postsynaptic GABAA receptors might provide an endogenous pathway for Ca2+-dependent regulation of synaptic strength,.

32To test this hypothesis, we recorded inhibitory synaptic currents (IPSCs) in neocortical pyramidal neurons at a membrane potential of −40 mV (Figure 6) in the presence of glutamate receptor antagonists CNQX (50 µM) and D-APV (30 µM). Under these conditions, we observed inward purinergic currents simultaneously with outward Cl−-currents mediated by GABAA receptors (Figure 6). Similar to our previous results, outward IPSCs were completely inhibited by bicuculline in all 12 cells tested (unpublished data). The burst of ATP-mediated currents, induced in cortical neurons by activation of astrocytic Ca2+ signaling via PAR1 receptors (as shown in Figure 3A), was accompanied by significant decrease in the amplitude of GABA-mediated synaptic currents (Figure 6A,B). Amplitudes of evoked IPSCs and spontaneous mIPSCs recorded in the wild-type mice were reduced after application of 10 µM TFLLR by 44.3%±6.1% (n = 7) and 39.4%±6.3% (n = 12) correspondingly. In the dn-SNARE mice, the inactivation of GABA receptor-mediated synaptic currents was greatly diminished (Figure 6A,B). Application of TFLLR reduced the amplitude of evoked and spontaneous IPSCs in the cortical neurons of dn-SNARE mice just by 6.4%±8.7% (n = 8) and 4.3%±6.6% (n = 8), respectively. The difference in the action of TFLLR on the IPSCs in the wild-type and dn-SNARE was statistically significant with p<0.005 for both evoked and spontaneous currents. These results confirm the importance of astroglial exocytosis for the observed inactivation of GABA receptors. It also indicates the lack of unspecific action of PAR-1 agonist on inhibitory synaptic transmission.

33The effect of TFLLR on inhibitory synaptic currents was mimicked by application of nonhydrolysable ATP analog ATP-γS (10 µM) and was considerably reduced by P2 receptor antagonist PPADS (10 µM). In the P2X4 KO mice, activation of astrocytes decreased the amplitude of evoked and spontaneous IPSCs (Figure 6A,B) only by 14.3%±8.2% and 15.7%±9.5% (n = 12); the difference between P2X4 KO and wild-type mice was significant with p<0.005. These data strongly support the participation of neuronal ATP receptors in the astrocyte-driven modulation of IPSCs.

34We found out that exocytosis of gliotransmitters also caused long-term homeostatic modulation of inhibitory synaptic transmission. We observed a marked difference in the amplitude distribution of the baseline (control) amplitude of postsynaptic inhibitory currents in the wild-type and dn-SNARE mice (Figure 6C). The amplitude of mIPSCs in the cortical neurons before the application of TFLLR was much higher in the dn-SNARE than in the WT mice. The average baseline amplitude of mIPSCs was 23.5±8.3 pA in the dn-SNARE mice (n = 8) and 14.9±6.9 in the WT mice. Application of TFLLR caused the leftward shift of amplitude distribution of mIPSCs only in the WT mice; this shift was significantly reduced by PPADS (Figure 6C). The considerable difference in the baseline amplitude of miniature inhibitory currents in the WT and dn-SNARE mice provides the first evidence that vesicular release of gliotransmitters may be involved in the long-term homeostatic regulation of inhibitory transmission in the neocortex.

35In order to elucidate the role of post- and presynaptic mechanisms in the modulation of inhibitory transmission, we evaluated the changes in the mean quantal content (Figure 6A), changes in the frequency of spontaneous mIPSCs (Figure 6B), and paired-pulse ratio (PPR) of evoked IPSCs (Figure 7). Neither the mean quantal content of IPSCs (Figure 6A) nor the mIPSCs frequency (Figure 6B) exhibited marked changes. However, we observed a moderate change in the PPR of IPSCs in the neocortical pyramidal neurons of wild-type mice (Figure 7A). The IPSCs evoked with a 50 ms interval in the control showed paired-pulse depression with mean PPR of 0.79±0.11 (n = 12), and application of TFLLR increased PPR by 0.16±0.07 (n = 8). The application of TFLLR did not cause a significant change in PPR in the dn-SNARE mice (Figure 7B,D), indicating the involvement of glial exocytosis in the mechanism. The effect of TFLLR was significantly reduced by PPADS (Figure 7A,D) and reproduced by application of ATP-γS both in the wild-type and dn-SNARE mice (Figure 7A,B,D). Hence, the presynaptic increase in the PPR of IPSCs was most likely mediated by ATP acting via P2 purinoreceptors. This result is consistent with previous reports of presynaptic facilitation of GABAergic synaptic transmission by ATP and P2 receptors,,.

36There was no significant difference in the effects of TFLLR and ATP-γS between wild-type and P2X4 KO mice (Figure 7C,D). So the P2X4 receptors hardly make the large contribution in the increase of PPR, contrasting with their prominent role in the astrocyte-induced decrease of IPCS amplitude (Figure 6). One could speculate that presynaptic facilitation of IPSCs by glia-driven ATP can be mediated by other subtypes of P2X receptors and P2Y receptors, whose role in the presynaptic modulation in the brain was widely reported,.

37More importantly, the above data clearly demonstrate that the large decrease in the amplitude of evoked and spontaneous IPSCs cannot be attributed to the presynaptic modulation of GABA release, and astrocyte-induced down-regulation of inhibitory transmission operates via postsynaptic mechanism.

38The postsynaptic mechanism of IPSCs inactivation was corroborated by experiments where the Ca2+-dependent phosphorylation of GABA receptors was impaired by intracellular agents. First, we found that activation of P2X receptors in pyramidal neocortical neurons caused marked reduction of GABA-activated currents via Ca2+- and protein kinase C–dependent mechanism (Figure S11). Second, application of TFLLR to the cortical slices of wild-type mice did not cause marked reduction in the mIPSCs recorded in neurons perfused with intracellular Ca2+-chelator EGTA (10 µM) or protein kinase C antagonist GF109203X (30 nM). The mIPSC amplitude was reduced by just 2.5%±5.9% (n = 6) in the presence of intracellular EGTA and by 11.2%±5.5% (n = 4) in the presence of the protein kinase C antagonist (Figure S12). These results agree with previous reports on Ca2+- and PKC-dependent down-regulation of GABA receptors.

39In addition to the fast IPSCs (“phasic inhibition”), central neurons also receive a diffusional inhibitory signal mediated by the extrasynaptic GABA receptors continuously activated by small concentrations of GABA present in the extrasynaptic space (“tonic inhibition”),. Extrasynaptic GABA receptors, responsible for tonic inhibition, have been reported to undergo Ca2+-dependent phosphorylation,. Thus, one could expect the impact of astrocyte-driven ATP not only on phasic but also on tonic inhibitory signaling in the neocortical pyramidal neurons. To verify this hypothesis, we used a conventional experimental paradigm where tonic inhibition is assessed by the change in the whole-cell holding current under action of GABA receptor blockers, in our case 50 µM bicuculline. In the wild-type mice (Figure 8A, top), layer 2/3 neurons showed the tonic current of 39.9±8.3 pA (n = 20) at membrane potential of −80 mV. The tonic current in the pyramidal neurons dn-SNARE mice was almost two times higher than in the wild-type (Figure 8B, top), reaching 76.9±15.1 pA (n = 10). A significant difference in the amplitude of tonic current between wild-type and dn-SNARE mice suggests that exocytosis of gliotransmitters from cortical astrocytes can modulate the activity of extrasynaptic GABA receptors in the adjacent neurons. Consistent with this notion, activation of gliotransmitter release by TFLLR caused a marked upward shift in the holding current in the pyramidal neurons of wild-type mice (Figure 8A, middle), and the rest of holding current was efficiently diminished by bicuculline. Application of TFLLR did not have a notable effect on tonic current in the dn-SNARE mice (Figure 8B, middle). The amplitude of tonic current recorded under action of TFLLR in the wild-type, dn-SNARE, and P2X4 KO mice was correspondingly 12.6±6.8 pA (n = 14), 72.2±9.1 pA (n = 7), and 43.4±7.2 pA (n = 6). The relative decrease in the amplitude of tonic current caused by TFLLR was, respectively, 68%±14%, 6%±8%, and 25%±11%. The down-regulation of tonic current by PAR-1 agonist was considerably attenuated by antagonist of ATP receptors (Figure 8A, bottom). The effects of TFLLR and PPADS in the wild-type mice as well as a difference between mice strains were statistically significant as indicated in Figure 7C. It has been recently shown that Ca-dependent modulation of astrocytic GABA GAT3 transporters in the hippocampus can alter an extracellular GABA level elevating the tonic current and decreasing the mIPSCs due to GABA receptor desensitization. One should note that this pathway did not contribute significantly to PAR-1 agonist-induced modulation of inhibitory currents in neocortical neurons, since we observed a decrease in both phasic (Figure 6) and tonic currents (Figure 7). This notion was corroborated by our finding that blocking of the astroglial GAT3 GABA transporter increased the tonic current and decreased the amplitude of mIPSCs both in the wild-type and dn-SNARE mice (Figure S13). These data also show that dn-SNARE expression does significantly alter the activity of GABA transporters in cortical astrocytes.

40Taken together, our experiments in brain slices demonstrate that exocytosis of ATP from cortical astrocytes in situ can activate the ATP receptors in the adjacent pyramidal neurons, leading to down-regulation of synaptic and extrasynaptic GABA receptors.

41Detection of ATP released from acutely isolated cortical astrocytes using sniffer cells demonstrated the SNARE protein-dependent exocytosis of ATP from cortical (Figures 1 and 2) and hippocampal (Figure S2) astrocytes. Our experiments in situ show that release of ATP from astrocytes can be sensed by neighboring neurons where glia-derived ATP can activate neuronal purinoreceptors. Quantal behavior of astrocyte-induced phasic purinergic currents (Figures 3, 4, and S6) and their elimination in the bafilomycin-treated cortical slices (Table S2) confirm that they originated from the vesicular release of ATP. Combined with previous observation of astroglial purinergic modulation of neuronal activity in the hippocampus, cortex, and basal forebrain,, these data strongly support the universal character of a vesicular mechanism of ATP release as a gliotransmitter.

42In the sniffer cell experiment we used P2X2 receptors as a detector for ATP. They have a moderate affinity for ATP (EC50 around 10 µM) that allowed us to avoid saturation and resolve the quantal behavior of astrocyte-evoked purinergic currents with a mean quantal size of about 8 pA (Figure 2D). To activate such response, the concentration of ATP released from astrocytes should have reached at least the low micromolar range. Detailed calculations, performed using the approach applied for extrasynaptic release and diffusion of glutamate,,, show that such concentration can be reached after release of about 1,000 molecules from a single glial synaptic-like vesicle, and the size of the “active spot” due to diffusion of ATP can reach as far as 1–2 µM (Figures S14A and S15). This estimation is in line with our data obtained using ATP microelectrode biosensors (Figure S10) as well as with previous evidence that the extracellular ATP level in the brain can reach 1–100 µM depending on the physiological and pathological context,,.

43Similar calculations, as above, argue against involvement of lysosomal release in generation of purinergic currents observed in our experiments. Indeed, the typical lysosome has a diameter of about 300 nm and can contain more than 1,000,000 molecules of gliotransmitter —that is, 1,000 times more than in the synaptic-like vesicle. Even though lysosome undergoes kiss-and-run exocytosis releasing only 10% of its content, one could expect the peak of a sniffer cell response to reach at least 200–500 pA, with most receptors on its surface saturated with agonist. So the observed quantal size of sniffer cell response (about 8 pA) is in much better agreement with release of ATP from synaptic-like vesicles.

44The low micromolar level of ATP concentration can be sufficient for activation of P2X purinoreceptors abundantly expressed in the central neurons,,. The data of electron microscopy and single molecular imaging,, show both synaptic (mostly at the periphery of synaptic density) and extrasynaptic location of P2X receptors. Interestingly, P2X2 and P2X4 receptors are generally located peri-synaptically (i.e., at the edges of postsynaptic density), which makes them accessible by ATP released from both nerve terminals and extrasynaptic glial release sites. This is consistent with our observation of strong sensitivity of astrocyte-driven sIPSCs to deletion of P2X4 receptors and inhibition of P2X2 receptors (Figures 3, 4, and S7). The distance to the source of ATP would affect the kinetics and amplitude of input mediated by peri-synaptic P2X receptors. Due to diffusion and rapid conversion to ADP, the transient of ATP concentration reaching P2X receptor after vesicular release from the distal (astroglial) site will have less magnitude but will decay longer than ATP transient after release from a close intrasynaptic site (as illustrated in Figure S15). This would lead to smaller quantal amplitude and slower kinetics of purinergic current of the glial origin as compared to neuronally activated synaptic currents. So the diversity in location of the ATP source could be the most plausible explanation for the difference in the amplitude and kinetics of purinergic sEPSCs of neuronal and glial origin observed in the neocortical neurons (Figures 3 and 4). An alternative explanation might be that EPSCs of smaller amplitude and slow kinetics were generated at neuronal synapses at a much longer electrotonic distance. In this case, however, the attenuation factor would increase gradually with distance, leading to smooth single-peaked distributions of EPSC amplitude and decay time due to the presence of large “intermediate” EPSCs. This contrasts with our observation of two peaks in the amplitude and decay time histograms (Figures 3–5). These arguments are supported by results of computer simulation using the model of neurotransmitter spillover, adapted for release of ATP from glial and synaptic sites (Figures S14 and S15). Importantly, the key experimental evidence of the astroglial origin of purinergic currents of slower kinetics and smaller quantal size has been provided by their selective inhibition by disruption of vesicular ATP transport in astrocytes (Figure 5).

45Although the possibility of nonvesicular gliotransmitter release from astrocytes has been previously reported in several brain regions,,,, our results do not show a strong contribution of nonvesicular pathways to the release of ATP in the neocortex. On a contrary, both sniffer cell and biosensor data suggest that vesicular mechanism brings major contribution to activity-dependent ATP release. One should note that incomplete inhibition of ATP release in the biosensor experiments could be explained by incomplete inhibitory action of bafilomycin treatment and dn-SNARE expression on vesicular release rather than notable contribution of nonvesicular mechanisms. It is possible that vesicular and nonvesicular ATP release operates in the neocortex in the different time scale, similarly to spinal cord astrocytes, where the fast initial exocytosis of ATP can be followed by slow secondary release of ATP through the pannexin and connexin hemichannels. So it is plausible that vesicular and nonvesicular mechanisms of ATP release coexists in the neocortex, but nonvesicular release was not activated in our experimental context.

46It is traditionally considered that physiological astroglial Ca2+ signaling is driven mainly by InsP3-mediated release from intracellular stores. Calcium signals arising from activation of metabotropic receptors are believed to be primarily responsible for control of exocytotic release of gliotransmitters,,. Although the role of InsP3-induced Ca2+ signaling in astroglial physiology was questioned, more recent data strongly support the importance of this pathway for glia-derived modulation of synaptic transmission in the hippocampus, and neocortex,. We have shown previously that ionotropic NMDA and P2X receptors can mediate a significant fraction of the synaptically driven Ca2+ rise in cortical astrocytes,,. Our present data suggest that astrocytic NMDA receptors can trigger the release of ATP, acting in parallel to the metabotropic receptors (Figure 2). The capability of ionotropic receptors to control of exocytosis of gliotransmitters (Figure 2) might account for a lesser than expected impact of modulation of astroglial InsP3 signaling on synaptic transmission and plasticity.

47Our results shown in Figures 1–4 provide a strong evidence of ATP release by exocytosis and support previous observations of presence of vesicular ATP transporters and synaptic-like vesicles in astrocytes,,. Observation of vesicular location of VNUT1 (Figures S5 and S6), even with inherent limitations of immunostaining procedure (see Text S1), agrees with previous reports,, and with our functional data as well. Co-localization of VNUT1 and synaptic vesicle markers (Figure S5), rather small quantal size and fast kinetics of glia-driven purinergic currents (Figures 1–4), suggest that release of ATP from synaptic-like vesicles can make a significant contribution into purinergic gliotransmission in cortical astrocytes.

48Until recently, the action of ATP as gliotransmitter was associated mainly with presynaptic adenosine receptors,,. It was also shown that astrocyte-derived ATP could operate through P2X7 receptors to enhance presynaptic release of glutamate in hypothalamic neurons. Although we observed a moderate presynaptic modulation of inhibitory transmission (Figure 7), our main finding is that glia-driven ATP can directly activate excitatory currents in neurons acting via postsynaptic P2X receptors (Figures 3–5). We showed that activation of neuronal purinoreceptors by astrocyte-driven ATP caused a dramatic change in the inhibitory transmission in neocortex inhibiting postsynaptic and extrasynaptic GABAA receptors (Figures 6 and 7). ATP receptors can act through the Ca2+-dependent phosphorylation by protein kinase C (Figures S11 and S12); the latter mechanism is intrinsic for GABA receptors,.

49The physiological relevance of exocytosis of gliotransmitters was strongly supported by our finding of significant increase in the baseline phasic and tonic inhibitory signaling in the cortical neurons of dn-SNARE mice (Figures 6 and 7). This result suggests that release of gliotransmitters, presumably ATP, can be involved in long-term homeostatic regulation of neuronal GABA receptors by a mechanism that has yet to be investigated. Our evidence of down-regulation of inhibitory synaptic transmission in the neocortex goes in line with observations that enhanced Ca2+ signaling in cortical astrocytes contributed to neuronal excitotoxicity and epilepsy,.

50Our results outline a novel mechanism of action of ATP as a gliotransmitter: in addition to catabolism of ATP to ADP and adenosine and modulation of synaptic transmission via presynaptic purinoreceptors, ATP can enhance neuronal excitability by down-regulating the phasic and tonic inhibition acting via postsynaptic P2 receptors. Our present (Figures 6–8) and previous results show that modulation of postsynaptic receptors activated by Ca2+ influx via purinergic P2X receptors can provide an efficient mechanism of regulation of signaling within tripartite synapse. Data on a large contribution of P2X4 receptors to glia-driven modulation of neuronal signaling (Figures 6–8) go in line with a previous observation of a facilitatory role for astroglial release of ATP and P2X4 receptors in the long-term potentiation of synaptic transmission in the hippocampus.

51Our findings give further insight into the role of P2X receptors in the CNS. There is growing consensus that, despite a clear evidence of participation of P2X receptors in the excitatory synaptic transmission in several brain areas, they bring notable contribution to slow neuromodulation rather than fast excitation. The above results suggest that P2X receptors can also mediate glia-to-neuron signals, which can be activated in the millisecond time scale and have more long-lasting consequences for neuronal excitability.

52Hippocampal astrocytes have been recently reported to decrease the amplitude of mIPSCs via Ca2+-dependent modulation of GABA transport and increase the frequency of mIPSCs via P2Y receptors of inhibitory interneurons, presumably activated by slow release of ATP from astrocytes through connexin hemichannels.

53Our data suggest that presynaptic ATP receptors can enhance GABA release in the neocortex (Figure 7). Down-regulation of both phasic and tonic postsynaptic GABA receptors by astrocyte-driven ATP (Figures 7 and 8) can act downstream of these pathways and significantly affect their impact on inhibitory transmission. Interplay between post- and presynaptic pathways of glial modulation could have diverse effects on neuronal excitability in different physiological contexts. Apparently, the down-regulation of inhibitory transmission provided by postsynaptic P2X receptors prevails in the neocortical pyramidal neurons, at least in our experimental conditions (Figures 6 and 7).

54It becomes evident now that even release of one gliotransmitter, ATP (followed by formation of ADP and adenosine), can activate a variety of pre- and postsynaptic regulatory cascades that can affect synaptic efficacy in opposite ways. Combined with diversity of vesicular and nonvesicular pathways of ATP release and the possibility of release of other gliotransmitters, such as glutamate and D-serine,,, this may confer more complex behavior to tripartite synapse than previously assumed,.

55In summary, our data suggest that Ca2+ elevation in cortical astrocytes, which might occur in response to signals from neurons and/or propagation of glial Ca2+ waves, can trigger exocytosis of ATP from synaptic-like vesicles and activate neuronal P2X receptors that are located at the periphery of synapse. Ca2+ signaling via neuronal ATP receptors can cause phosphorylation-dependent inhibition of postsynaptic GABA receptors acting downstream of astroglial modulation of presynaptic GABA release and GABA uptake. Our results strongly support physiological importance of exocytosis of gliotransmitters, in particular ATP, in communication between astrocytes and neurons and modulation of synaptic efficacy.

56Experiments were performed on astrocytes and neurons in the somato-sensory cortex of dn-SNARE transgenic mice, and their wild-type (WT) littermates; in some experiments, the P2X4 receptor knockout mice were used. Genotypes of animals were verified by PCR from ear samples. Initial experiments to investigate release of ATP from astrocytes were also performed in transgenic mice expressing enhanced green fluorescent protein (EGFP) under the control of the glial fibrillary acidic protein (GFAP) promoter,,. Data obtained in the experiments on GFAP–EGFP mice (n = 5–6 for each type of experiments) did not differ significantly from data obtained in the WT mice of the same age. For clarity, all data referred to here as wild-type are reported solely for wild-type littermates of dn-SNARE mice; usage of GFAP–EGFP mice was explicitly stated where appropriate.

57Mice (8–12 wk and 9 mo old) were anaesthetized by halothane and then decapitated, in accordance with UK legislation. Brains were removed rapidly after decapitation and placed into ice-cold physiological saline containing (mM) NaCl 130, KCl 3, CaCl2 0.5, MgCl2 2.5, NaH2PO4 1, NaHCO3 25, glucose 15, pH of 7.4 gassed with 95% O2 to 5% CO2. Transverse slices (280–300 µm) were cut at 4°C and then placed in physiological saline containing (mM) NaCl 130, KCl 3, CaCl2 2.5, MgCl2 1, NaH2PO4 1, NaHCO3 22, glucose 15, pH of 7.4, and kept for 1–4 h prior to cell isolation and recording.

58Astrocytes were acutely isolated using the modified “vibrating ball” technique,. The glass ball (200 µm diameter) was moved slowly some 10–50 µm above the slice surface, while vibrating at 100 Hz (lateral displacements 20–30 µm). Isolation protocol was adjusted to provide a high yield of astroglial cells. This technique preserves the function of membrane proteins and therefore is devoid of many artifacts of enzymatic cell isolation and culturing procedures. In particular, vibro-dissociated astrocytes retain many morphological features (e.g., GFAP–EGFP fluorescence, size, proximal processes) and functional properties (e.g., high potassium conductance, glutamate transporters, Ca2+ signaling) while being completely isolated from neuronal somata and nerve terminals. The composition of external solution for all isolated cell experiments was (mM) 135 NaCl, 2.7 KCl, 2.5 CaCl2, 1 MgCl2, 10 HEPES, 1 NaH2PO4, 15 glucose, pH adjusted with NaOH to 7.3. Astrocytes were identified by their morphology under DIC observation, EGFP fluorescence (astrocytes from dn-SNARE and GFAP-EGFP mice), and functional properties as described previously (see also Table S1),.

59In the experiments with dn-SNARE mice, administration of Dox, has been removed 4 wk prior to electrophysiological studies. According to our observations, about 70% of astrocytes in the layer 2/3 of somatosensory cortex in situ and freshly isolated cortical astrocytes exhibited EGFP fluorescence. The astrocytic identity of EGFP reporter and dn-SNARE-transgene-expressing cells has previously been confirmed for the dn-SNARE line,. So one can expect that at least 60%–65% of astrocytes in the somatosensory cortex of dn-SNARE mice express the dn-SNARE domain of synaptobrevin II. In the experiments with freshly isolated astrocytes, only the fluorescent cells have been selected to ensure the impairment of SNARE complex function. In the experiments in the somatosensory cortex of dn-SNARE mice in situ, electrophysiological recordings have been performed in the areas with higher density of EGFP-expressing astrocytes to maximize the putative impact of the loss of SNARE function on neighboring neurons. To increase the probability of a neuron lying within a functional island of synapses controlled by dn-SNARE astrocytes, we recorded from the neurons located in the close proximity to at least two fluorescent cells, as illustrated in Figure S16.

60After isolation from a brain slice, cortical astrocytes were incubated with 5 µM Ca2+ indicator Fluo4-AM (or Rhod-2 AM) and 10 µM of photoliable Ca2+ chelator o-nitrophenyl-EGTA-AM (NP-EGTA) for 30 min, re-suspended in a small volume (200–300 µl) of fresh extracellular medium, and placed over cultured HEK293 cells expressing GFP-tagged P2X2 receptors to ATP (HEK293-P2X2, gift from Prof. R. Evans, University of Leicester, UK). To evaluate release of ATP, the transmembrane currents were recorded in the HEK293-P2X2 cells that had an astrocyte lying on their surface; simultaneously, elevation in cytosolic Ca2+ concentration was induced in the astrocytes by UV uncaging.

61It has to be noted that contrary to the experiments with astrocytes in culture, the spatial density of acutely isolated cells in our experiments was rather low, as shown previously. So we could easily select HEK cells contacting astrocytes with no other cell lying in the immediate vicinity (e.g., as shown in Figures 1 and S2). We performed recordings from the HEK293 cell–astrocyte couples, which were separated from any other cells by at least 15–20 microns of free space to ensure that HEK293-P2X2 cells were activated by ATP released only from contacting astrocyte.

62Flash photolysis and fluorescent imaging were performed with the aid of a IX51 inverted microscope and epifluorescent illumination via UPLSAPO60XW/NA1.2 objective (Olympus, Tokyo, Japan). To monitor the intracellular Ca2+ level, astrocytes were constantly illuminated at 480±10 nm using OptoLED light source (Cairn Research, Faversham, UK), and fluorescence was measured at 535±25 nm. Astrocytes from dn-SNARE and GFAP-EGFP mice were loaded with Ca2+ indicator Rhod-2 and illuminated at 530±10 nm; fluorescence was measured at 590±30 nm. The fluorescent images were recorded using Retiga 2000R enhanced CCD camera (QImaging, Canada); exposure time was 35 ms at 2X2 binning. Elevation in the intracellular Ca2+ level was evaluated by a ΔF/F0 ratio averaged over the whole cell image after background subtraction.

63For uncaging of intracellular Ca2+, cells were illuminated by a brief pulse (200 ms) of UV light (365±10 nm) emitted by high-power LED NCSU033AT (Nichia, Tokyo, Japan), peak power 500 mW, and estimated power at an objective >150 mW. Illumination was delivered via OptoLED dual-port epifluorescence condenser (Cairn Research, UK). In addition to photolysis, Ca2+ rise was induced in the astrocytes by fast application of agonists of PAR-1 receptors (10 µM TFLLR) or NMDA receptors (20 µM NMDA); these agonists did not cause any response in HEK293-P2X2 cells when applied without astrocytes placed over HEK cells.

64Whole-cell voltage clamp recordings from HEK293-P2X2 cells, neurons, and astrocytes were made with patch pipettes (4–5 MΩ for neurons and HEK293-P2X2 cells and 6–8 MΩ for astrocytes) filled with intracellular solution (in mM): 110 KCl, 10 NaCl, 10 HEPES, 5 MgATP, pH 7.35. Intracellular solution for recording in neurons and astrocytes contained 0.2 mM EGTA, and the solution for HEK293-P2X2 cells contained 10 mM EGTA and 1 mM CaCl2. In some experiments (simultaneous recording of GABA-mediated and ATP-mediated synaptic currents), KCl was replaced by KGluconate. Currents were monitored using an AxoPatch200B patch-clamp amplifier (Axon Instruments, USA) filtered at 2 kHz and digitized at 4 kHz. Experiments were controlled by the PCI-6229 data acquisition board (NI, USA) and WinFluor software (Strathclyde University, UK); data were analyzed by self-designed software. Liquid junction potentials were compensated with the patch-clamp amplifier. Series and input resistances were, respectively, 5–7 MΩ and 500–1100 MΩ in the HEK cells and neurons and 8–12 MΩ and 50–150 MΩ; both series and input resistance varied by less than 20% in the cells accepted for analysis. For activation of synaptic inputs, axons originating from layer IV–VI neurons were stimulated with a bipolar coaxial electrode (WPI, USA) placed in the layer V close to the layer IV border, approximately opposite the site of recording; the stimulus duration was 300 µs. For the recording of EPSCs and IPSCs, the stimulus magnitude was set 3–4 times higher than minimal stimulus adjusted to activate the single-axon response in the layer II pyramidal neurons as described in. In order to trigger synaptically driven astroglial Ca2+ transients, the single episode of theta-burst stimulation (HFS) was delivered; the HFS episode consisted of 5 pulses of 100 Hz stimulation, repeated 2 times with 200 ms interval (total 10 pulses per episode).

65To monitor the cytoplasmic-free Ca2+concentraton [Ca2+]i, cortical astrocytes were loaded by 40 min incubation with Fura-2AM. For fura-2 excitation, cells were alternately illuminated at wavelengths of 340±5 nm and 380±5 nm using the OptoScan monochromator (Cairn, Faversham, UK). Fluorescent images were recorded using Olympus BX51 microscope equipped with UMPLFL20x/NA0.95 objective and 2× intermediate magnification and Andor iXon885 EMCCD camera; exposure time was 35 ms at 2×2 binning; experiments were controlled by WinFluor software. The [Ca2+]i levels were expressed as F340/F380 ratio averaged over the whole-cell image.

66To investigate the Ca2+ signaling activated by PAR-1 receptor agonist, cortical neurons and astrocytes of wild-type and dn-SNARE mice were loaded with 50 µM Fluo-4. The whole-cell voltage-clamp recordings were used to confirm the identification of neurons and astrocytes and verify the lack of changes in the basic electrophysiological properties of the dn-SNARE astrocytes. The Fluo-4 fluorescence signal was excited at 488±10 nm and measured at 530±20 nm; the fluorescent images were recorded and analyzed as described above.

67In parts of the experiments (Figure 1, Figure 5, Figure S5, Figure S16), two-photon imaging of neurons and astrocytes was performed using Zeiss LSM-7MP multiphoton microscope coupled with the SpectraPhysics MaiTai pulsing laser; experiments were controlled by ZEN LSM software (Carl Zeiss, Germany). Images were further analyzed offline using ZEN LSM (Carl Zeiss) and ImageJ (NIH) software.

68For investigation of vesicular dynamics (Figure 1C), acutely isolated cortical astrocytes were loaded with FM1-43 fluorescent dye (2.5 µM), and FM1-43 fluorescence was excited at 820 nm and observed at 560±20 nm.

69For immunolabeling of vesicular transporters, secretory organelles, and astroglial markers, acutely isolated astrocytes were incubated with 0.1 µg/ml of antibodies following antibodies: rabbit polyclonal anti-VNUT1 (T-12), goat polyclonal anti-cathepsin D (N-19, Santa Cruz Biotechnology), mouse monoclonal anti-VGLUT1 (McKA1), mouse monoclonal anti-PSD95 (6G6-1C9), mouse monoclonal anti-GLT-1 (10B7, Abcam), mouse monoclonal anti-NeuN (A60), rabbit monoclonal anti-S100b (EP1576Y, Millipore), goat monoclonal anti-SV2A, mouse monoclonal anti-NG2, and rabbit polyclonal anti-GFAP (Sigma). Prior to cell loading, antibodies were conjugated to the green fluorescent dye DyLight488 (VNUT1, VGLUT1, PSD-95, NeuN, NG2, and GFAP) or red fluorescent dye DyLight594 (VNUT1, SV2A, cathepsin-D, LAMP3, GLT-1, S100β) using Lighting-Link antibody conjugation system (Innova Bioscience, Cambridge, UK) according to the manufacturer's protocol. Antibodies to VNUT1 (recognizing intraluminal epitopes), VGLUT1 (recognizing cytoplasmic epitope), and cathepsin-D (intraluminal epitope) were applied to living astrocytes directly; other antibodies were conjugated with BioPORTER protein delivery reagent (Genlantis, San Diego, CA) 10 min prior to incubation. Immediately after isolation from the neocortical slice, living astrocytes were pre-incubated with 2% of normal bovine serum (Sigma) in the extracellular saline for 30 min to block unspecific binding sites. After them, cells were gently washed two times with clean extracellular saline for 5 min and then incubated with antibodies for 60 min at room temperature. After incubation, cells were washed with laminar flow of extracellular solution in the microscope recording chamber for 30 min prior to image recording.

70Fluorescence was excited at 820 nm, GFAP-EGFP and DyLight488 signal was observed at 520±10 nm, and DyLight594 signal was observed at 590±20 nm. The photomultiplier gain of the two-photon microscope was adjusted to avoid saturation in both channels but in the same time did not differ more than 10% between red and green channel. Colocalization analysis of images was carried out using ImageJ software and methods described in.

71The concentration of ATP within cortical slices (Figure S10) was measured using microelectrode biosensors obtained from Sarissa Biomedical Ltd (Coventry, UK). A detailed description of the properties of ATP biosensors and recording procedure has been published previously in,,. Briefly, biosensors consisted of ATP metabolizing enzymes immobilized within a matrix on thin (25–50 µM) Pt/Ir wire. This allowed insertion of the sensors into the cortical slice and minimized the influence of a layer of dead surface tissue. The concentration of ATP has been calculated from the difference in the signals of two sensors: a screened ATP sensor and screened null-sensor, possessing the matrix but no enzymes. This allowed us to compensate for release of any nonspecific electro-active interferents. Biosensors show a linear response to increasing concentration of ATP and have a rise time of less than 10 s. Biosensors were calibrated with known concentrations (10 µM) of ATP before the slice was present in the perfusion chamber and after the slice had been removed. This allowed compensation of any reduction in sensitivity during the experiment. Biosensor signals were acquired at 1 kHz with a 1400 CED interface and analyzed using Spike 6.1 software (Cambridge Electronics Design, Cambridge, UK).

72All data are presented as mean ± SD, and the statistical significance of difference between data groups was tested by two-population t test, unless indicated otherwise. The spontaneous transmembrane currents recorded in HEK293-P2X2 cells and neurons were analyzed off-line using methods described previously,. Briefly, phasic transmembrane currents of an amplitude higher than 2 SD of baseline noise were selected for the initial detection of spontaneous events. Then every single spontaneous event was analyzed within the 140 ms time window, and its amplitude and kinetics were determined by fitting the model curve with single exponential rise and decay phases. As a rule, mean square error of fit amounted to 5%–20% of peak amplitude depending on the background noise. Whenever error of fit exceeded 25%, spontaneous currents were discarded from further analyses.

73The amplitude distributions of spontaneous and evoked currents were analyzed with the aid of probability density functions and likelihood maximization techniques; all histograms shown were calculated as probability density functions. The amplitude distributions were fitted with either multiquantal binomial model or bimodal function consisting of two Gaussians with variable peak location, width, and amplitude. The decay time distributions were fitted with bimodal functions. Parameters of models were fit using likelihood maximization routine. For each particular distribution, the fit with quantal or bimodal model was accepted only when it has a confidence level α≤0.05. To monitor and analyze the time course of changes in the amplitude and frequency of spontaneous currents, the amplitude and frequency were averaged over the 1 min time window.

74For the basic analysis of the time course of quantal parameters of the evoked IPSCs, the mean quantal content was evaluated as reciprocal square of coefficient of variation, and the quantal size was calculated as ratio of mean amplitude to the mean quantal content.

75All animal work has been carried out in accordance with UK legislation and "3R" strategy. Research has not involved non-human primates.