Synaptic mechanisms of adenosine A2A receptor‐mediated hyperexcitability in the hippocampus

Adenosine inhibits excitatory neurons widely in the brain through adenosine A1 receptor, but activation of adenosine A2A receptor (A2AR) has an opposite effect promoting discharge in neuronal networks. In the hippocampus A2AR expression level is low, and the receptor's effect on identified neuronal circuits is unknown. Using optogenetic afferent stimulation and whole‐cell recording from identified postsynaptic neurons we show that A2AR facilitates excitatory glutamatergic Schaffer collateral synapses to CA1 pyramidal cells, but not to GABAergic inhibitory interneurons. In addition, A2AR enhances GABAergic inhibitory transmission between CA1 area interneurons leading to disinhibition of pyramidal cells. Adenosine A2AR has no direct modulatory effect on GABAergic synapses to pyramidal cells. As a result adenosine A2AR activation alters the synaptic excitation ‐ inhibition balance in the CA1 area resulting in increased pyramidal cell discharge to glutamatergic Schaffer collateral stimulation. In line with this, we show that A2AR promotes synchronous pyramidal cell firing in hyperexcitable conditions where extracellular potassium is elevated or following high‐frequency electrical stimulation. Our results revealed selective synapse‐ and cell type specific adenosine A2AR effects in hippocampal CA1 area. The uncovered mechanisms help our understanding of A2AR's facilitatory effect on cortical network activity. © 2014 The Authors Hippocampus Published by Wiley Periodicals, Inc.


INTRODUCTION
Adenosine is well known for its inhibitory effect on neocortical and hippocampal glutamatergic principal cells via the A 1 receptor (A 1 R) (Dias et al., 2013). In addition, the high affinity adenosine A 2A receptor (A 2A R) is expressed in the brain, and although present at low levels in the neocortex and hippocampus Dixon et al., 1996) its activation in pathological conditions promotes epileptiform activity and facilitates excitotoxic neuronal death (Jones et al., 1998;Etherington and Frenguelli, 2004;Zeraati et al., 2006;El Yacoubi et al., 2009). However, evidence for A 2A R-mediated facilitation of cortical excitatory neuron discharge is largely based on results in epilepsy and neuronal trauma models, and function of A 2A R under physiological conditions in the cortex is less well known. Facilitatory effect of A 2A R on excitatory neurons in healthy brain is well characterized in basal ganglia where it is involved in controlling arousal and motor responses (Rebola et al., 2005a;Ciruela et al., 2006;Shook and Jackson, 2011;Wei et al., 2011;Lazarus et al., 2012). Adenosine A 2A R-mediated modulation of neuronal activity has also been reported in the hippocampus and neocortex where the receptor activation facilitates excitatory input from the CA3 area to CA1 enhancing glutamatergic synapses directly or by altering glutamate transport (Cunha et al., 1994;Rebola et al., 2005c;Dias et al., 2012;Matos et al., 2013). In physiological conditions adenosine A 2A Rs are involved in synaptic longterm plasticity in hippocampal glutamatergic mossy fibers (Rebola et al., 2008;Chamberlain et al., 2013), and a recent study demonstrated that deletion of A 2A R selectively in the hippocampus compromizes contextual memory formation (Wei et al., 2013).
The paucity of apparent adenosine A 2A R expression in the hippocampus hints that the receptor may be localized to specific neuron subpopulations or subtypes of synapses Dixon et al., 1996). Although reported facilitatory effects on glutamatergic transmission between pyramidal cells could explain, at least partly, why A 2A R activation promotes cortical pyramidal cell discharge (Jones et al., 1998;Zeraati et al., 2006;El Yacoubi et al., 2008;El Yacoubi et al., 2009;Moschovos et al., 2012), it is unknown if modulation of GABAergic inhibitory interneurons contributes to A 2A R-mediated effects on hippocampal function. Adenosine A 2A R expression level increases in posttraumatic and epileptic neocortex and hippocampus (Dixon et al., 1996;Rebola et al., 2005b), and this may emphasize a role of the receptor in the activity modulation in pathological conditions. Knowledge of the action of A 2A R on identified hippocampal synaptic circuits is crucial for understanding adenosine function in physiological conditions in the cortex and the therapeutic potential of high affinity adenosine receptors in pathological conditions such as epilepsy.

Electrophysiology
Whole cell and field potential recording electrodes (5-9 MX) were pulled (P-97, Sutter Instrument Co.) from borosilicate glass capillaries (GC150F-10, Harvard Apparatus). Intracellular solution for experiments in Figure 2 was (in mM): 145 Cs-Methansulfonate, 20 HEPES, 10 CsOH, 8 NaCl, 0.2 CsOH-EGTA, 2 ATP-Mg, 0.3 GTP-Na (295 mOsm, pH 7.2); in Figure 3 (in mM), 145 K-gluconate, 10 KOH, 0.2 KOH-EGTA were used instead; in figs 4, Cs-Methanosulphonate was replaced with CsCl. QX-314 (5 mM) and Neurobiotin (0.2-0.5%, Vector Laboratories) were included in all intracellular filling solutions. Field potential electrodes were filled with saline. Ratio of baseline fEPSP slope values and popspike amplitudes evoked with different intensities were fitted with regression line in each experiment baseline. The fEPSP slope-popspike relation was considered linear when regression fitting index was > 0.8 (0.89 6 0.03, n 5 11, mean 6 s.e.m, Sigma Plot). fEPSP slope values recorded following wash-in of CGS21680 were fitted in the baseline condition regression line. Then, measured popspike amplitude in CGS21680 and popspike estimate given to same fEPSP value in baseline linear slope-popspike relation were compared. This gave D popspike/fEPSP used in Figure 1E. Because lowest intensity often failed to elicit stable popspike in baseline, intensities from 75 ls till 150 ls stimulus duration were used to determine linear relation of fEPSP slope and popspike amplitude in baseline conditions with regression line. The fEPSP values in the presence of agonist, which were potentiated out of the baseline fEPSP slope range, were excluded in analyses because no linear relation between fEPSP slope and popspike could be confirmed.
Data in Figures 1-6 were recorded with a Multiclamp 700B amplifier (Molecular Devices), low-pass filtered (4 kHz), digitized (10 kHz), and acquired by Clampex software (Molecular Devices). Field potential recordings in interface chamber (data for Fig. 7) were performed with an AC preamplifier and AC/ DC amplifiers Neurolog NL104 and NL106 (0.3 Hz high-pass filtering) (Digitimer Ltd.). The signal was digitized by a Power 1401 plus (Cambridge Electronic Design). Additionally, a Humbug 50/60 Hz (Digitimer Ltd.) was used to remove noise locked to the electrical mains supply. Data were stored for offline analysis using Signal5 software (Cambridge Electronic Design) at 10 kHz acquisition rate. In Figure 7 experiments a single-pulse electrical stimuli was delivered (every 20 s), and elicited fEPSPs (100 ms from stimulation) were excluded from spontaneous activity analysis.
Access resistance (<20 MX) was not compensated. Wholecell recordings with >25% change were rejected. Liquidjunction potential was not corrected. Single, paired-pulse and HFS electrical stimuli (50-250 mA) were applied with concentric bipolar electrodes (CBAPC75PL1, FHC) connected to stimulus isolator boxes and triggered via computer. In Figure  FIGURE 1.
Activation of adenosine A 2A receptor facilitates glutamatergic transmission in hippocampal Schaffer collaterals and amplifies CA1 pyramidal cell input-output function. A-C: A selective agonist CGS21680 (30 nM) increases fEPSP slope and population spike amplitude evoked by stimulation of Schaffer collaterals. A: Schematic shows experimental design. Paired-pulse (50 ms interval) electrical stimulation (S) was delivered in the CA1 area. The CA3 area was removed by surgical cut to avoid recurrent excitation. Averaged field potential traces (10) evoked with mid-strength stimulation (100 ls pulse duration) in baseline (bl, black) and following application of GCS21680 (30 nM) (CGS, red). (a) shows prespike volley amplitude (between horizontal dotted lines), (b) fEPSP slope was measured between dotted vertical lines, and (c) popspike amplitude between horizontal lines. Stimulation artifact (S) is truncated. B: Increase of fEPSP slope by CGS21680 (30 nM). fEPSPs were elicited in every experiment with five stimulation intensities gradually increasing stimulus pulse duration from 50 to 150 ls. Open boxes show median (with 25% and 75% quartiles) of baseline-normalized fEPSP slope in 8 experiments following wash-in of CGS21680. Solid boxes show CGS21680 wash-in results in presence of the A 2A R antagonist SCH58261 (100 nM) (n 5 3). Significant difference between open and solid boxes is indicated by asterisk (*P < 0.05, Mann-Whitney test). C: Increase of popspike amplitude by CGS21680 (30 nM) in experiments shown in B. When popspike data are not available in all experiments n is indicated in parenthesis. Asterisks show difference between the open and solid boxes (*P < 0.05, Mann-Whitney test). D, E: GCS21680 increases popspike amplitude -fEPSP slope ratio. D: Relation of popspike amplitude and fEPSP slope in one experiment in baseline (black trace and symbols) and following wash-in of CGS21680 (red). fEPSPs were evoked with various intensities using stimulation pulse duration from 75 to 125 ls. Inset: Averaged (10) field potential responses in baseline (black) and following wash-in of CGS21680 (red). Popspikes appearing in the fEPSP following wash-in of CGS21680 are indicated by arrows. (Data in the plot show first popspike amplitude when more than one popspike is elicited in CGS21680.) E: Effect of CGS21680 on popspike amplitude -fEPSP slope relation in all experiments. In baseline conditions popspike -fEPSP slope relation was determined in each experiment (see Materials and Methods). Plot shows a relation of popspike amplitude associated with similar size fEPSP slope in CGS21680 and baseline. This is indicated as D popspike/fEPSP slope. Open boxes represent median of means of individual experiments (circles). fEPSPs upon 2nd stimulation of paired-pulse generated significantly higher popspikes than similar magnitude fEPSPs in baseline (P < 0.05, Mann-Whitney test). For 1st stimulation pulse response, there was no significant difference between baseline and CGS21680. Solid boxes correspond to control experiments where CGS21680 was applied in the presence of A 2A R blocker SCH58261 (30 nM). Antagonist blocks the agonist-induced increase in D popspike/fEPSP slope (*P < 0.05, Mann-Whitney test). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] 6, stimulation with S2 electrode was suspended after baseline during SCH58261 wash-in and resumed after 10 min. Data were analyzed offline using Clampfit 10.2 software (Molecular Devices) or Spike2 software (Cambridge Electronic Design). Recorded signals were low-pass filtered on-line at 6 kHz and off-line in Figure 7 experiments as reported in results using Spike2 software. Drugs were purchased from Abcam, Ascent Scientific, Sigma-Aldrich, and Tocris Bioscience. Drugs were diluted (1 : 1,000) in ddH 2 O, DMSO or ethanol, and applied via superfusion.

Statistics
All data presented were tested for normal distribution (Shapiro-Wilk test, Sigma Plot), and when passed t-test or single way ANOVA and Tukey's post hoc test was used to confirm significance, and data were shown as mean 6 sem. Otherwise Mann-Whitney was used instead and data shown as median and quartiles.

Stereotaxic Injections
An adeno-associated virus serotype 2 or 5 construct (AAV2/ 5:ChR2-eYFP) was stereotaxically injected into dorsal hippocampus of heterozygous PV-Cre, CCK-Cre, and CaMKII-Cre mice (CA1-CA3 area) via 33-gauge needle attached to a Microlitre Syringe (Hamilton). Craniotomy was made for mice anesthetized with 2-4% isoflurane. In each hemisphere, 800 nL of virus suspension was delivered at 80 nL/min by a Micro Syringe Pump Controller (World Precision Instruments). Following suturing of the wound, mice were allowed to recover for 14-21 days after injections.

Optogenetics
ChR2 was activated by a fixed-spot 20 or 80 mm diameter laser light spot (pulse 3 ms, max. 100 mW, Rapp OptoElectronics) via the microscope objective (diameter measured under objective). All experiments with 20 Hz 5-pulse stimulation were performed in the presence of blockers for high-frequency stimulation-elicited long-term plasticity. Paired-pulse ratios are presented as 2nd versus 1st IPSC amplitude. Compound IPSC and EPSC charge was measured in 500 ms window from current onset.

Cell Visualization, Anatomical Analysis, and Immunohistochemistry
Processes and analyses are described in Oren et al. (2009). Briefly, slices were fixed overnight at 4 C, washed in 0.1 M Adenosine A 2A R facilitates glutamatergic synapses to pyramidal cells, but not to two major feed-forward GABAergic inhibitory interneuron populations expressing either PV or CCK. Electrical stimulation of Schaffer collaterals in the presence of GABA receptors blockers (PiTX, 100 mM) and CGP55845, 1 mM). Timing of bath-applied A 2A R agonist and antagonist is indicated by horizontal bars. A,B: Facilitation of EPSCs by CGS21680 in identified pyramidal cells. A: Illustration of one recorded, neurobiotin-filled and visualized pyramidal cell (soma and dendrites red; axon blue). (s.r., stratum radiatum, s.p., stratum pyramidale, s.o., stratum oriens). B1: Bath-applied A 2A R agonist CGS21680 (30 nM) facilitates glutamatergic EPSC amplitude (mean 6 sem, baseline-normalized, ***P < 0.001, t-test). Insets; experimental design. Averaged EPSCs (10) from one cell in baseline (bl) and following CGS21680 application (at 15-20 min time point). B2: Adenosine A 2A R antagonist SCH58261 (100 nM) has no effect on EPSC amplitude in the experimental conditions (t-test). Plot (mean 6 sem) and averaged EPSCs as in B1. C-F: The A 2A R agonist fails to modulate EPSCs in interneurons. C: PV1 interneurons were identified by Cre-dependent fluorophore (tdTomato, tdTom) expression. Confocal images showing tdTom (above) and immunoreaction for PV (below, visualized with Alexa-488) in the CA1 area in a fixed slice. D: EPSCs in PV1 cells were not altered by CGS21680 (mean 6 sem). Insets: experimental design and averaged EPSCs (10) from one postsynaptic PV1 cell. E: Postsynaptic CCK1 interneurons were identified by positive immunoreaction for pro-CCK in post hoc analysis. Confocal images from one postsynaptic neurobiotinfilled (nb, Alexa-488) and pro-CCK1 (Cy5) interneuron. F: Adenosine A 2A R agonist CGS21680 does not change EPSCs in CCK1 GABAergic interneurons (mean 6 sem). Insets: experimental design and averaged EPSCs (10)   A 2A R selectively facilitates GABAergic synapses between feed-forward interneurons, but has no direct effect on GABAergic inhibitory synapses to pyramidal cells. A-C: A 2A R agonist facilitates IPSCs elicited from GABAergic PV1 cells to various inhibitory interneurons, but not to pyramidal cells. A: Optogenetic stimulation of GABAergic synapses from PV1 interneurons. ChR2 is expressed in Cre-dependent manner. Confocal images show eYFP-ChR2 (above) in PV1 cells (below, Cy5). Arrows point to positive somata (fixed slice). B: Plot shows that CGS21680 (30 nM) facilitates IPSC amplitude in postsynaptic interneurons (open symbols; mean 6 sem of baseline-normalized IPSCs, ***P < 0.001, t-test). Facilitation by CGS21680 is blocked in the presence of A 2A R antagonist (SCH58261, 100 nM; gray symbols, t-test). Insets: Schematic of experimental design. Averaged IPSCs (10) in baseline and after 15 min in CGS21680. C: CGS21680 fails to modulate IPSCs from PV1 GABAergic synapses to identified pyramidal cells (mean 6 sem, t-test). Insets: Schematic shows experimental design. Averaged IPSCs (10) in baseline and after 15 min in CGS21680. D: The CGS21680-induced IPSC facilitation in interneurons is associated with reduced paired-pulse ratio (PPR). A plot shows baseline-normalized IPSC amplitude (for 1st IPSC) and PPR (2nd vs. 1st IPSC amplitude) following wash-in of CGS21680 Circles, mean in individual experiments; tri-angles mean 6 sem of the means (***P < 0.001, t-test). Averaged IPSCs (10) shown on the top, scale 50 ms. Traces are scaled by 1 st IPSC amplitude and dotted line indicates 2 nd IPSC peak in baseline. E: Facilitation of IPSCs by CGS21680 in interneurons involves protein kinase A (PKA). Histogram shows baselinenormalized IPSC amplitude following CGS21680 application in control (mean 6 sem, n 5 6), and in the presence of a PKA inhibitor H-89 (1 mM, n 5 5) (**P < 0.01, t-test). IPSCs were elicited by electrical stimulation of GABAergic fibers (glutamate receptors blocked with NBQX 25 mM and DL-APV, 100 mM). F: IPSC facilitation by CGS21680 occurs in various different postsynaptic interneuron types. Illustration of a basket cell (above; collapsed zstack epifluorescence image from one 60 mm-thick section, soma and dendrites red, axon blue) with positive axonal immunoreaction for CB1R (below; confocal images of CB1R at Cy3 and a neurobiotin-filled axon in Alexa488, pointed by arrows). s.r. and s.p 5 stratum radiatum and pyramidale, scale 20 mm. Histogram shows baseline-normalized IPSC in CGS21680 in all recorded interneurons (n 5 12). Analyses revealed four putatively PV1 cells (two O-LM cells and two CB1R-basket cells) and six putative CCK1 cells immunopositive for axonal CB1R. Two interneurons remained unidentified. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] (Vector Laboratories) and examined with an epifluorescent microscope (DM5000 B, Leica Microsystems) using appropriate filter sets (L5 or Y3) and a CCD camera (ORCA-ER, Hamamatsu). Pyramidal cells were identified by mushroom spines on dendrites, basket cells and oriens-lacunosum molecular (O-LM) cells by their axon arborisation inside stratum pyramidale or lacunosum moleculare, respectively (Oren et al., Nissen et al., 2010). Digital micrographs were constructed from z-stack images recorded with epifluorescence microscope, collapsed, and analyzed with Image-J software (Somogyi et al., 2012).
Free-floating sections were washed in 50 mM TBS-Tx, blocked in 20% normal horse serum (NHS, Vector Laboratories) in TBS-Tx, and incubated in primary antibodies at 4 C for 48 h. Fluorochrome-conjugated secondary antibodies were applied overnight at 4 C. After another wash-in TBS-Tx, sections were mounted in Vectashield under coverslips. Immunoreactivity was evaluated at 403 magnification with 23 zoom using confocal laser-scanning microscopy (LSM710, Carl Zeiss) with Zen2008 software. Details of primary and secondary antibodies are reported in Nissen et al. (2010).

Adenosine A 2A R Facilitates Glutamatergic Schaffer Collateral Synapses and Amplifies CA1 Pyramidal Cell Input-Output Transformation
We studied effect of A 2A R activation on hippocampal Schaffer collateral synapses in the CA1 area using paired-pulse microelectrode stimulation (50 ms interval, delivered every 15 s) and field potential recording in mouse hippocampal slices. Wash-in of selective A 2A R agonist CGS21680 (30 nM) after a baseline (at least 10 min) enhanced stimulus-evoked field EPSP (fEPSP) slope and increased population spike (popspike) amplitude (P < 0.05), but did not alter prespike volley (Mann-Whitney test) (Figs. 1A-C). Stimulus-evoked fEPSP and popspike details are shown in Figures 1B,C. Baselinenormalized prespike volley in CGS21680 was 1.02 6 0.02 for 1st stimulation pulse and 1.01 6 0.03 for 2nd (n 5 11) (Sebastiao and Ribeiro, 1992). Facilitatory effects of CGS21680 on fEPSP slope and popspike amplitude were fully blocked in Effect of CGS21680 on field potential responses was studied in each experiment with five stimulation intensities. In all experiments stimulation intensity was adjusted to generate a popspike with mid-range intensity in baseline conditions (popspike amplitude 0.25 6 0.06 mV for 1st pulse, and 0.59 6 0.17 mV for 2nd pulse, n 5 11, mean 6 sem. This corresponded to fEPSP slope of 0.32 6 0.06 mV/ms and 0.57 6 0.11 mV/ms, respectively).
We discovered that following wash-in of CGS21680 (30 nM), popspike amplitude-fEPSP slope ratio also changed. In the presence of CGS21680, fEPSPs were associated with higher amplitude popspikes than during baseline (Fig. 1D). We used linear regression to fit fEPSP slope and popspike amplitude values (evoked with various stimulus intensities) in baseline conditions in each experiment (see Material and Methods). Following wash-in of CGS21680 (30 nM), fEPSPs upon 2nd stimulation of the paired-pulse generated significantly higher amplitude popspikes than similar magnitude fEPSPs during baseline (P < 0.05, Mann-Whitney test) (Fig. 1E). Popspike amplitude-fEPSP slope relation details are shown in Figures  1D and E. The results show that A 2A R facilitates glutamatergic synapses in the hippocampus, and in addition increases CA1 pyramidal cells' output in response to Schaffer collateral excitation.
Because pyramidal cells in the CA1 area can express low levels of CCK and Cre, light-evoked ChR2 currents could mask synaptic EPSCs in these experiments (Geibel et al., 2014). We therefore washed in glutamate receptor blockers NBQX (25 mM) and DL-APV (100 mM) at the end to measure ChR2contribution to light stimulation-evoked excitatory currents. In all tested cells glutamatergic current was predominant (78 6 8% of total charge, n 5 7 cells) showing that the facilitatory effect of A 2A R agonist on excitatory currents is caused by increased glutamatergic EPSCs.
The results show that A 2A R activation modulates Schaffer collateral-driven synaptic input from CA3 area to CA1 pyramidal cells in two ways; facilitating monosynaptic glutamatergic excitation and suppressing network-driven disynaptic GABAergic inhibition simultaneously. These changes can at least partially explain our above findings on A

Adenosine A 2A R Facilitates Glutamatergic Schaffer Collateral Synapses Selectively to Pyramidal Cells
We repeated Schaffer collateral electrical stimulation experiments (see Fig. 1) and recorded intracellularly from either postsynaptic CA1 pyramidal cells or interneurons. Bath-applied adenosine A 2A R agonist CGS21680 (30 nM) facilitated glutamatergic EPSC amplitude to 1.30 6 0.04 from baseline (10-15 min following application, P < 0.001, n 5 9, t-test) in synapses onto identified CA1 pyramidal cells (see Material and Methods). Wash-in of A 2A R antagonist SCH58261 (100 nM) after baseline failed to change EPSCs, and baseline-normalized EPSC amplitude in SCH58261 was 0.94 6 0.04 (n 5 6, t-test) indicating that A 2A Rs are not activated by endogenous adenosine under the experimental conditions (Figs. 3A,B). Next, we studied EPSCs in two major interneuron populations involved in feed-forward inhibition in area CA1; GABAergic cells expressing either parvalbumin (PV1) or cholecystokinin (CCK1) with axonal cannabinoid receptor Type 1 (CB1R) (Katona et al., 1999;Glickfeld and Scanziani, 2006;Nissen et al., 2010;Armstrong and Soltesz, 2012). EPSCs in PV1 and CCK1 interneurons were not altered by A 2A R agonist (ttest), and baseline-normalized EPSC amplitudes in CGS21680 (30 nM) were 1.05 6 0.05 (n 5 8) and 1.04 6 0.02 (n 5 7) accordingly (Figs. 3C-F). Thus, activation of A 2A R facilitates excitatory Schaffer collateral synapses in target-specific manner. Mean 6 sem of EPSCs during baseline was 79.6 6 8.1 pA in pyramidal cells (n 5 9) and 77.8 6 15.8 pA in the interneurons (n 5 15). GABA receptors were blocked with PiTX (100 mM) and CGP55845 (1 mM), and cells were filled with neurobiotin for post hoc anatomical and immunohistochemical studies (Figs. 3A,E). Modulation of spontaneous epileptiform pyramidal cell discharge by A 2A R antagonist and agonist in hyperexcitable conditions with elevated extracellular potassium. A-C: Adenosine A 2A R blocker SCH58261 (100 nM) suppresses spontaneous epileptiform discharges in hippocampal slices exposed to elevated (8-9 mM) extracellular potassium. Spontaneous interictal-like synchronous bursting activity was recorded with field potential electrode in CA3 area. A: A sample trace from one experiment showing inhibition of spontaneous epileptiform burst activity by SCH58261 (unfiltered trace). Timing for washin of A 2A R antagonist SCH58261 (100 nM) is indicated by horizontal bar. Histogram below shows occurrence of spontaneous epileptiform bursts in 2 min bins. For burst occurrence analysis data were band-pass filtered (1-100 Hz) to avoid detection of occasional single unitary extracellular spikes. B: Epileptiform population bursts are characterized by 1-100 Hz band-pass filtered (BP 1-100 Hz) field potential deflection associated with extracellular spikes (high-pass filtered at 60 Hz, HP 60 Hz). An unfiltered epoch shown on top with filtering below as indicated. C: Plot shows suppression in occurrence of spontaneous epileptiform events by SCH58261 in the three of three experiments.
Occurrence of events is shown in 2 min bins. Horizontal bar indicates wash-in of the antagonist. Inset plot shows baselinenormalized effect of the antagonist on burst occurrence (indicated with same symbols as in main plot). Inhibitory effect of SCH58261 was highly significant (**P < 0.01, t-test, at 20-30 min after drug application). D-F: Wash-in of A 2A R agonist CGS21680 (30 nM) is associated with increased spontaneous occurrence of epileptiform bursts. D: Traces from one experiment illustrate spontaneous burst activity in baseline and following agonist application (20-30 min wash-in). E: Illustration of one burst event from same experiment. Unfiltered (top) and filtered (band-pass 1-100 Hz and high-pass 60 Hz) traces of the same event are illustrated as indicated. F: Plot shows effect of A 2A R agonist (CGS21680, 30 nM) on occurrence of spontaneous epileptiform bursts in four experiments (2 min bin). Wash-in of the antagonist is indicated by horizontal bar. G: Baselinenormalized burst occurrence in the presence of agonist in the four experiments above (indicated with same symbols). Burst occurrence is variably modulated, but significantly increased in pool of four experiments (*P < 0.05, t-test, at 20-30 min time point following agonist application.

Adenosine A 2A R Enhances GABAergic Inhibition in the CA1 Area Selectively Between Interneurons
The results above do not explain why feed-forward IPSCs were strongly suppressed by A 2A R activation in experiments shown in Figure 2. To explore this, we investigated whether GABAergic synapses from interneurons to pyramidal cells are modulated by A 2A R agonist, or if GABAergic synapses between interneurons are altered. We utilized Cre-dependent ChR2 expression to optogenetically activate GABAergic synapses from either PV-or CCK-expressing CA1 interneurons. Slices were prepared from heterozygous PV-Cre (Fig. 4) and BAC-CCK-Cre tg/1 mice (Fig. 5) transduced with AAV:ChR2-eYFP (see Materials and Methods) (Geibel et al., 2014). We first stimulated ChR2-expressing PV1 GABAergic interneuron axons with paired-pulse laser light pulses (3 ms, 50 ms interval) in the CA1 area, and found that wash-in of the agonist CGS21680 (30 nM) increased IPSC amplitude in postsynaptic interneurons to 1.35 6 0.04 of baseline (P < 0.001, n 5 12, ttest) (Figs. 4A,B). The facilitation was significant in 11 of 12 anatomically verified interneurons, and was fully blocked when studied in the presence of the A 2A R antagonist SCH58261 (100 nM) (n 5 5, t-test) (Fig. 4B). However, CGS21680 (30 nM) failed to directly modulate GABAergic synapses from PV1 cells to postsynaptic pyramidal cells (t-test) (Fig. 4C). Baseline-normalized IPSC amplitude in postsynaptic pyramidal cells was 0.93 6 0.04 in the presence of CGS21680 (30 nM) (n 5 12).
The results on IPSCs in postsynaptic pyramidal cells and interneurons show that A 2A R-mediated modulation of inhibitory synapses from PV1 GABAergic fibers depends on the postsynaptic cell type. Postsynaptic neurons were filled with neurobiotin during recording for post hoc analysis of the cells (see Materials and Methods). This confirmed that A 2A Rmediated facilitation of IPSCs occurs in various postsynaptic interneuron types including oriens-lacunosum moleculare (O-LM) cells (n 5 2), and basket cells with negative (n 5 2) or positive (n 5 6) axonal immunoreaction for CB1R (Fig. 4F) (Glickfeld and Scanziani, 2006;Lawrence et al., 2006;Klaus-berger and Somogyi, 2008). Two interneurons, of which one showed IPSC facilitation by A 2A R, remained unidentified (Fig. 4F).

Endogenous Adenosine Promotes Synchronous Pyramidal Cell Discharge Via A 2A Rs in Hippocampal Slices
We next studied whether endogenous adenosine released by high-frequency electrical stimulation is sufficient to modulate hippocampal pyramidal cell discharge through adenosine A 2A R (Chamberlain et al., 2013). We utilized experimental design used above in Figure 1 to electrically stimulate Schaffer collaterals with paired pulses (50 ms interval), while recording field potential in the CA1 area. In addition, we applied highfrequency stimulation (HFS, 50 Hz 100 pulse) with a second stimulation electrode positioned in the vicinity of recording electrode aiming to elicit local release of adenosine ( Fig. 6A) (Chamberlain et al., 2013). Schaffer collaterals were stimulated every 5 s and HFS delivered with second electrode every 2 min. To uncover adenosine A 2A R-mediated modulation the experiments were performed in continuous presence of blockers for CB1R (AM-251 2 mM), GABA B receptor (CGP55485, 1 mM), adenosine A 1 R (DPCPX, 200 nM) as well as with DL-APV (100 mM). We analyzed same fEPSP parameters as in Figure 1 and found that HFS was followed by significant increase of pop-spike amplitude in Schaffer collateral -mediated field potential response. Popspike were elicited by 2nd stimulation pulse of the paired-pulse and they were significantly increased from baseline up to 40 s following the HFS. Importantly, the facilitation was blocked after wash-in of SCH58261 (100 nM) (P < 0.001, ANOVA, Tukey's HSD test, Fig. 6B). Although HFS transiently also modulated fEPSP slope in the experiments, application of the A 2A R blocker failed to cause any change in the effect on slope. Neither did HFS or SCH58261 affect prespike volley (ANOVA, Tukey's HSD test, data not shown). The HFS and A 2A R antagonists effects on popspike are shown in detail in Figure 6.
Finally, we investigated whether A 2A R activation by endogenous adenosine modulates spontaneous epileptiform discharge of hippocampal pyramidal cells in hyperexcitable conditions. Spontaneous inter-ictal like pyramidal cell population bursts were generated exposing slices to elevated (8-9 mM) extracellular potassium ([K o ]) in perfusion solution (Korn et al., 1987;Sagratella et al., 1987). Field potential was recorded in the CA3 area in an interface chamber. Following stable baseline (at least 10 min), either A 2A R blocker SCH58261 (100 nM) or agonist CGS21680 (30 nM) was washed in. Epileptiform activity was quantified analyzing the occurrence of spontaneous inter-ictal like events characterized by a low frequency content field potential deflection associated with a barrage of extracellular spikes. Recordings were band-pass (1-100 Hz) filtered offline to uncover low-frequency deflections and analyze event occurrence (Figs. 7A,B). Amplitude threshold was set to 0.25 mV, and event detection was visually verified. Parallel high-pass filtering (>60 Hz) of recordings uncovered extracellular spikes associated with the events. Occurrence of inter-ictal like events in baseline conditions was 32.7 6 11.7 events/min, ranging from 6.3 to 97.4 events/min (n 5 7). The adenosine A 2A R blocker SCH58261 strongly inhibited the occurrence spontaneous population bursts to 36 6 9% (P < 0.01, n 5 3, t-test) of baseline in 20-30 min following drug application. The activity-suppressing effect of antagonist persisted and in 40-50 min from drug application the burst occurrence dropped to 16 6 5% of baseline level (P < 0.001, n 5 3, t-test) (Fig. 7C). Adenosine A 2A R agonist CGS21680 (100 nM) increased spontaneous epileptiform burst occurrence (Figs. 7D-G) from baseline to 140 6 16% (P < 0.05, n 5 4, t-test) in 20-30 min following drug application. Increase of burst occurrence was significant in three of four experiments, but varied in magnitude (Figs. 7F,G). Samples of band-pass and high-pass -filtered events are illustrated in Figures 7B,E. Modulation of spontaneous activity with A 2A R drugs suggests the receptors are tonically activated in slices with elevated [K o ], possibly because of increased ambient adenosine levels (Marichich and Nasello, 1973;Etherington and Frenguelli, 2004;Dias et al., 2013).

DISCUSSION
Adenosine has a well-established role as an endogenous neuronal inhibitor in the brain. Adenosine's suppressive effect on excitatory glutamatergic transmission via A 1 R is well characterized, but its effect via other adenosine receptor types is not as well known (Dunwiddie and Masino, 2001;Sebastiao and Ribeiro, 2009). In the hippocampus and neocortex the highaffinity A 2A R is expressed in low quantities (Dixon et al., 1996), but elevated levels of extracellular adenosine activate these receptors to facilitate neuronal discharge (Etherington and Frenguelli, 2004;Zeraati et al., 2006;El Yacoubi et al., 2008;El Yacoubi et al., 2009). It has been proposed that excitatory effects of adenosine in the cortex may mainly occur in pathological conditions, because A 2A R expression levels increase in those circumstances in parallel with desensitization and down-regulation of A 1 R (Rebola et al., 2005b;D'Alimonte et al., 2009;Hamil et al., 2012;Moschovos et al., 2012). In addition evidence for A 2A R-mediated modulation of activity in the hippocampus in physiological conditions is emerging (Cunha and Ribeiro, 2000;Rebola et al., 2005a;Rebola et al., 2008;Dias et al., 2012;Chamberlain et al., 2013;Dias et al., 2013;Wei et al., 2013), but A 2A R effect on identified neuronal circuits in this area is still poorly understood.
We identified here two sites of synaptic modulation by which A 2A R acts to shift the balance between synaptic excitation and inhibition in mouse hippocampus to facilitate principal cell discharge. Adenosine A 2A R activation directly enhances excitatory glutamatergic Schaffer collateral synapses to CA1 pyramidal cells, and simultaneously suppresses feed-forward GABAergic inhibition to the same neurons. This at least partially explains the facilitatory effects of A 2A R agonist on Schaffer collateral field potential responses in the CA1 area with increased fEPSP slope and popspike amplitude (Sebastiao and Ribeiro, 1992) (also shown here in Figs. 1-6). Our results also demonstrate that adenosine A 2A R is unlikely to modulate glutamatergic Schaffer collateral axon excitability, for example through axonal receptors , because the agonist did not have effect on extracellular prespike volley. Together our findings provide a simple mechanistic explanation how A 2A R activity increases excitability in the hippocampal CA3-CA1 circuitry modulating identified excitatory and inhibitory synapses. Although modulatory effects of A 2A R are not restricted to synapses, but in addition can include alterations in intrinsic properties of neurons (Rebola et al., 2011) as well as glial glutamate transport (Matos et al., 2013), the synaptic modulatory action can at least partly explain proconvulsive effect of A 2A R reported previously (Jones et al., 1998;Zeraati et al., 2006;El Yacoubi et al., 2008;El Yacoubi et al., 2009) and also demonstrated here.
Facilitation of epileptiform activity through low A 2A R expression level in the hippocampus (Dixon et al., 1996) can be explained by synergistic action of the synaptic modulatory actions shown here. Increased Schaffer collateral excitation of pyramidal cells, but not feed-forward interneurons increases CA1 pyramidal firing to glutamatergic input from the CA3 area (Pouille and Scanziani, 2001;Lamsa et al., 2005;Xu et al., 2006;Pavlov et al., 2011;Lovett-Barron et al., 2012). We studied two major subpopulations of CA1 area GABAergic interneurons, either expressing PV or CCK, which both contribute to CA3-CA1 feed-forward inhibition controlling CA1 area pyramidal cell firing and their input-output transformation (Cobb et al., 1995;Buhl et al., 1996;Glickfeld and Scanziani, 2006;Klausberger and Somogyi, 2008;Lovett-Barron et al., 2012). Inhibitory transmission through these interneurons to CA1 pyramidal cells was not enhanced by A 2A R. Instead A 2A R activation suppressed feed-forward GABAergic inhibition in pyramidal cells through a mechanism, which is likely to include disinhibition. Facilitation of inhibitory synapses between CA1 interneurons has been demonstrated to effectively suppress network activity-driven GABAergic inhibition in the CA1 area pyramidal cells (Chamberland and Topolnik, 2012;Lovett-Barron et al., 2012). This promotes synaptically-driven pyramidal cell discharge and increases their input-output transformation (Toth et al., 1997;Letzkus et al., 2011;Lovett-Barron et al., 2012;Xu et al., 2013). We report that A 2A Rmediated facilitation of IPSCs was present in various postsynaptic CA1 area interneuron types, including O-LM cells specialized to inhibit distal dendrites of pyramidal cells, and basket cells that directly control pyramidal cell action potential firing via perisomatic inhibitory synapses (Zhang and McBain, 1995;Glickfeld and Scanziani, 2006;Klausberger and Somogyi, 2008). Through modulation of the GABAergic circuits A 2A Rs can control co-ordinated rhythmic neuronal activities in the hippocampus (Cobb et al., 1995;Klausberger et al., 2005;Wulff et al., 2009). Interestingly, the A 2A R-mediated facilitation of GABAergic efferents was specific to PV-expressing interneurons, and was not detected in CCK1 GABAergic interneuron fibers (Armstrong and Soltesz, 2012).
Importantly, we showed that A 2A R-mediated facilitation of CA1 pyramidal cell activity also occurs through endogenous adenosine. High-frequency electrical stimulation experiment demonstrated that CA1 area pyramidal cell input-output transformation to Schaffer collateral stimulation is similarly facilitated via endogenous and agonist-induced A 2A R activity. Although high-frequency stimulation-evoked A 2A R activation failed to significantly change synaptic Schaffer collateral responses in the experiments, this can be explained by higher sensitivity of the network-driven input-output function than a monosynaptic pathway to synaptic modulations (Lovett-Barron et al., 2012).
Our results on spontaneous activity modulation by A 2A R antagonist and agonist in hyperexcitable conditions confirm the previously reported findings that A 2A R controls spontaneous epileptiform pyramidal cell discharge in the hippocampus (Sebastiao and Ribeiro, 2009). In addition, the results indicate that in slices with elevated extracellular potassium adenosine A 2A Rs are tonically active promoting synchronous discharge in the hippocampus. This was evidenced by robust effect with A 2A R antagonist suppressing the spontaneous interictal like events in the CA3 area. Variability and occasionally a lack of A 2A R agonist effect to promote synchronous discharge in these conditions could also be explained by vigorous tonic A 2A R activity in baseline conditions (Dias et al., 2013). Given that ambient adenosine levels elevate in epileptic tissue and A 2A R expression increases whereas A 1 R levels go down, A 2A R blockers might provide an effective supplementary treatment in specific forms of epilepsy (Sebastiao and Ribeiro, 2009;Gomes et al., 2011). Adenosine's therapeutic effect via A 1 R might benefit from inhibition of A 2A Rs. A seizure promoting role of A 2A R in humans has recently been highlighted (Shinohara et al., 2013), and adenosine A 2A R antagonists have already entered clinical trials and are safe to use with relatively mild side effects (Lopes et al., 2011;Shook and Jackson, 2011;Muller, 2013). Our findings here identify specific synaptic targets for A 2A R-modulation. This helps to understand how these receptors are involved in generation of aberrant hippocampal activity and can point out specific therapeutic targets in cortical microcircuits.