Male Wistar rats (adults, 220–250 g, n = 78: 18 controls treated with vehicle, nine controls treated with BBG, nine controls treated with caffeine; 18 stressed treated with vehicle, nine stressed treated with BBG, nine stressed treated with caffeine, six for electrophysiology) were obtained from Charles River (Barcelona, Spain) and were maintained at 23–25°C, with 12 h light / 12 h dark cycle and standard chow and tap water ad libitum. All procedures in this study were conducted following the principles and procedures outlined as “3Rs” in the guidelines of the European Union (2010/63/EU), FELASA, and ARRIVE, and were approved by the Portuguese Ethical Committee (DGAV) and by the Institution’s Ethics’ Committee (ORBEA 238-2019/14102019). Since the behavioral alterations caused by this protocol of restraint stress were so far only validated in male rats, the “3Rs” guidelines imposed the use of only male rats to obtain the first proof-of-concept supporting the existence of any interaction between P_2X7R and A_2AR.

As done previously (Carmo et al., 2014), the blood-brain barrier-permeant and efficacious P_2X7R antagonist brilliant blue G (BBG, 45 mg/kg dissolved in saline; from Sigma–Aldrich, Portugal) or saline were administered intraperitoneally every 48 h at 7 PM, starting 3 days before the protocol of restraint stress, until the sacrifice of the animals. The tested dose of BBG has previously been shown to yield a brain concentration of 200–220 nM (Díaz-Hernández et al., 2012), which is within the effective and selective range of BBG towards central P_2X7R and is without evident side-effects in control rodents (Donnelly-Roberts and Jarvis, 2007).

Caffeine (Sigma, Portugal) was administered through the drinking water as previously reported (Duarte et al., 2009; Cognato et al., 2010) at a dose (0.3 g/L) estimated to correspond to a daily intake of 3–4 cups of coffee by humans (Fredholm et al., 1999), which rodents consume without modification of their water intake (Duarte et al., 2009, 2012; Silva et al., 2013). This yields a concentration of circa 30 μM in the brain parenchyma (Costenla et al., 2010; Silva et al., 2013), which selectively targets adenosine receptors (Lopes et al., 2019) and mimics the neuroprotective impact of A_2AR antagonists, rather than of A₁R (Cunha et al., 2006; Dall’Igna et al., 2007), namely in animal models of stress and depression (Kaster et al., 2015; Machado et al., 2017). Caffeine intake was allowed only overnight (7 PM-7 AM), starting 3 days before the protocol of restraint stress, until the sacrifice of the animals and this repeated exposure to caffeine is expected to afford neuroprotection without major modification of behavioral or physiological parameters in control rodents (Duarte et al., 2009; Yang et al., 2009; Cognato et al., 2010).

The stress model used consisted of a repeated physical restraint of rats, as done previously (Cunha et al., 2006). The rats were individually placed in a room adjacent to their colony in an independent plastic compartment and immobilized in a 25 × 7 cm plastic bottle, with a plastic taper on the outside and a 1 cm hole at one end for breathing. After the termination of each daily restraint stress session, the rats were returned to their home cages. The schedule of sub-chronic restraint stress consisted of a daily 4 h immobilization period (between 10 AM and 4 PM) during 14 consecutive days, the time previously defined to be required to cause stable behavioral modifications for at least 1 week in adult male rats (Cunha et al., 2006). Control age-matched rats were handled as their tested littermates except that they were not isolated or immobilized.

Behavioral tests were carried out from 9 AM until 4 PM on the 15^th until the 18^th day after beginning the restraint stress protocol (Figure 1). As shown in Figure 1, the animals were subject to a tight schedule of behavioral characterization, with a minimal time interval between each test, which could lead to cross-testing interferences. However, the analysis of the performance of control animals in the successive tests did not show evident differences from historic controls where rats of the same age and strain were tested in each different test with wider time gaps between the different tests (Cunha et al., 2006; Cognato et al., 2010; Carmo et al., 2014; Coelho et al., 2014; Matheus et al., 2016). All behavior tests were carried out by two experimenters who were unaware of the phenotypes or drug treatments, in a sound-attenuated room with an eight lux illumination and visual cues on the walls, to which the animals were previously habituated. The apparatuses were cleaned with 20% ethyl alcohol to remove any odors after testing each animal.

Locomotion and exploratory behavior were monitored using an open-field arena made of dark gray PVC measuring 100 × 100 cm² (divided by white lines into 25 squares of 20 × 20 cm²) and was surrounded by 40-cm high walls. Each rat was placed in the center of the open field and the following variables were recorded for 10 min: number of peripheral squares (adjacent to the walls) crossed (peripheral locomotion), number of central squares (away from the walls) crossed (central locomotion) and total locomotion (peripheral locomotion plus central locomotion).

Anxiety was further assessed using the elevated plus-maze, which consisted of four arms of the same size (40 cm × 5 cm) arranged in the form of a cross and raised 50 cm above the floor. Two opposed arms were surrounded by 30 cm high opaque black Plexiglas walls, except for the entrance (closed arms) while the other two had no walls (open arms). Each animal was placed on the central square of the maze facing an enclosed arm and was allowed to explore the maze for 5 min. The number of entries and the time spent in both open and closed arms were recorded, considering an entry only when the whole body and four paws were inside an arm.

The depressive-like behavior was evaluated in the forced swimming test, where rats were placed in individual glass cylinders (40 cm in height and 17 cm in diameter) containing water (water depth was 30 cm, kept at 25 ± 1°C) to measure the total duration of immobility, climbing, and swimming during a 10-min session. A rat was regarded as immobile when floating motionless or making only those movements necessary to keep its head above the water. The climbing behavior was defined as upward-directed movements of the forepaws usually along the side of the swimming chamber and the swimming behavior is defined as movement (usually horizontal) throughout the swimming chamber; diving and face shaking behaviors were not considered.

Anhedonic-like behavior was evaluated with the sucrose preference test, where rats were first single-housed in a cage with two bottles and free access to food. After 4 h of habituation, one bottle was randomly switched to contain 1.2% sucrose solution and the total consumption of water and sucrose solution was measured at the end of a 16 h test period (12 h dark phase plus 4 h light phase). Sucrose preference was calculated as the ratio of sucrose vs. total intake.

Spatial memory was evaluated using a 2-trials Y-maze paradigm (Dellu et al., 1997). The test was carried out in a Plexiglas apparatus with equal three arms (10 cm wide, 35 cm long, and walls of 25 cm height) in a Y-shape, separated by equal angles. The test consists of two sessions of 5 min duration separated by a 2-h inter-trial interval. During the first session, the rat was placed at the end of one arm and allowed to explore the two available arms since the third arm (the novel arm) was blocked by a guillotine door. During the second session, the “‘novel”’ arm was opened and the rat was again placed in the start arm and allowed to explore the three arms. Memory performance was evaluated by measuring the time spent exploring the “novel” arm compared to the exploration of the other two arms. An entry into an arm was defined as the placement of all four paws into the arm.

Hippocampal-dependent memory was also evaluated using the object displacement test, where rats were exposed to two identical objects in the same open field apparatus in which they were habituated and were allowed to explore for 5 min the objects fixed in opposite corners 10 cm away from walls and 70 cm apart from each other. In the test trial, carried out 2-h after, rats were again placed for 5 min in the open field arena, except that one of the objects was moved to a novel position. Memory performance was quantified with an object displacement index defined as the ratio between the time exploring the object in the novel location over the total time exploring both objects. Exploration of an object is defined as directing the nose to the object at a distance equal to or less than 2 cm from the object and/or touching it with the nose; rearing on to the object was not considered exploratory behavior.

The sequence of the tests is indicated in Figure 1.

After completion of the battery of behavior analysis, rats were sacrificed by decapitation under deep anesthesia upon exposure to a halothane-saturated atmosphere. One hippocampus or part of the prefrontal cortex of each rat was used to extract total RNA with a MagNA Lyser Instrument and a MagNA Pure Compact RNA Isolation kit (Roche, Portugal), according to the manufacturer’s instructions. The integrity, quantity, and purity of the RNA yields were checked by electrophoresis and spectrophotometry. Reverse transcription for first-strand cDNA synthesis from each sample was performed using a random hexamer primer with the Transcriptor First Strand cDNA Synthesis kit (Roche), according to the manufacturer’s instructions. The resulting cDNAs were used as templates for real-time PCR, which was carried out on the LightCycler instrument (Roche) using the FastStart DNA Master SYBR Green I kit (Roche). The mRNA expression of the marker of microglia “activation” Iba1 (ionized calcium-binding adaptor molecule 1), of the pro-inflammatory cytokines interleukin-1β (IL1β) and tumor necrosis factor α (TNFα) and of P_2X7R, was calculated relative to GADPH (glyceraldehyde 3-phosphate dehydrogenase) mRNA expression, using the following primers (from Tib MolBiol, Germany): Iba1 (forward: 5′-TGC GCA AGA GAT CTG CCA TC-3′; reverse: 5′-ACC AGT TGG CTT CTG GTG TT-3′); IL1β (forward: 5′-ATG AGA GCA TCC AGC TTC AAA TC-3′; reverse: 5′-CAC ACT AGC AGG TCG TCA TCA TC-3′); TNFα (forward: 5′-CGA GAT GTG GAA CTG GCA GA-3′; reverse: 5′-CTA CGG GCT TGT CAC TCG A-3′); P2rx7 (forward: 5′-CTG CCT CCC GTC TCA ACT AC-3′; reverse: 5′-GCC TCT CTG GAT AGC ACG AT-3′); GAPDH (forward: 5′-CCC TTC ATT GAC CTC AAC TAC-3′; reverse: 5′-CTT CTC CAT GGT GGT GAA GAC-3′). Quantification was carried out based on standard curves run simultaneously with the test samples generated by conventional PCR amplification, as previously described (Costenla et al., 2011; Rebola et al., 2011). The purity and specificity of the resulting PCR products were assessed by melting curve analysis and electrophoresis. Control reactions were performed to verify that no amplification occurred without cDNA.

The binding assays were performed as previously described (Cunha et al., 2006), using the second hippocampus and the rest of the prefrontal cortex from each rat. After purifying whole membranes by centrifugation-based fractionation (Rebola et al., 2005), the membranes were resuspended in Tris-Mg solution (containing 50 mM Tris and 10 mM MgCl₂, pH 7.4) with 4 U/ml of adenosine deaminase (to remove endogenous adenosine). Binding with 2 nM of ³H-SCH58261 (specific activity of 77 Ci/mmol; prepared by GE Healthcare and offered by E.Ongini, Schering-Plough, Italy), a supramaximal concentration of this selective A_2AR ligand (Lopes et al., 2004), was performed for 1 h at room temperature with 286–343 (hippocampus) or 54–71 μg of protein (prefrontal cortex), with constant swirling. The binding reactions were stopped by the addition of 4 ml of ice-cold Tris-Mg solution and filtration through Whatman GF/C filters (GE Healthcare). The radioactivity was measured with 2 ml of scintillation liquid (AquaSafe 500 Plus, Zinsser Analytic). The specific binding was expressed as fmol/mg protein and was estimated by subtraction of the non-specific binding, which was measured in the presence of 12 μM of xanthine amine congener (XAC; Sigma), an antagonist of adenosine receptors. All binding assays were performed in duplicate.

A murine microglial cell line, N9 (a kind gift from Professor Claudia Verderio, CNR Institute of Neuroscience, Milan, Italy), was grown as previously described (Gomes et al., 2013) in an RPMI medium supplemented with 30 mM glucose (Sigma), 100 U/ml penicillin and 100 μg/ml streptomycin (GIBCO, Invitrogen, Portugal) and maintained at 37°C in an incubator with a humidified atmosphere with 5% CO₂, until reaching confluence. N9 cells were then detached using 0.05% trypsin (T3924, Sigma) for 5 min, resuspended in RPMI after washing and centrifugation, and counted using a hemocytometer with trypan blue. N9 cells were then seeded in a 48-multiwell at a density of 0.02 × 10⁶ cells and remained in culture for 48 h. Then, cells were incubated for 45 min with Fluo-4-AM (4 μM; Life Technologies) dissolved in recording buffer (132 mM NaCl, 4 mM KCl, 1.4 mM MgCl₂, 6 mM glucose, 10 mM HEPES, 1.8 mM CaCl₂; pH 7.4) with 0.05% bovine serum albumin to facilitate probe entry into the cells, as previously described (Simões et al., 2012). The cells were then washed and left in a recording buffer for 15 min to allow complete Fluo-4 AM de-esterification. In some experimental conditions, the following modifiers of the evoked signals were added to the recording buffer during the de-esterification and kept until the end of the experiment: 1 μM JNJ47965567 (2-(phenylthio)-N-[[tetrahydro-4-(4-phenyl-1-piperazinyl)-2H-pyran-4-yl]methyl-3-pyridinecarboxamide, a selective P_2X7R antagonist from Tocris), 50 nM SCH58261 (2-(2-furanyl)-7-(2-phenylethyl)-7H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-amine, a selective A_2AR antagonist from Tocris) or 100 nM CGS21680 (4-[2-[[6-amino-9-(N-ethyl-β-D-ribofuranuronamidosyl)-9H-purin-2-yl]amino]ethyl]benzenepropanoic acid, a selective A_2AR agonist from Tocris).

After de-esterification, cytosolic Ca²⁺-dependent fluorescence was recorded using a VICTOR³ Multiplate reader (Perkin Elmer) with Wallac 1420 software, using an exciting wavelength of 485 nm and recording the emission wavelength at 530 nm, close to the ideal wavelength to monitor Fluo-4 fluorescence (494/506 nm). The baseline fluorescence was recorded at 0.2 Hz. [Ca²⁺]_i transients were triggered by the application of different stimuli, either 100 μM BzATP [2′(3′)-O-(4-benzoylbenzoyl)adenosine 5′-triphosphate, a selective P_2X7R agonist from Sigma], 100 nM CGS21680 or 100 μM glutamate (to mimic excitotoxic conditions, from Sigma) and fluorescence was recorded for 5 min at 0.6 Hz (Janks et al., 2018). When glutamate was used as a trigger of Ca²⁺ transients, experiments were performed with the recording buffer without MgCl₂ and with 133.4 mM NaCl. After recording the stimulus-induced [Ca²⁺]_i transient response, cells were exposed to ionomycin (10 μM, Tocris) to induce a steep increase of extracellular Ca²⁺ influx and consequently a maximum fluorescence response.

The fluorescence data were background-corrected by subtracting the mean fluorescence value of N9 cells that were not incubated with Fluo-4-AM. Intracellular calcium concentration was estimated for each time point using the formula: [Ca²⁺] = K_d × (F − F_min)/(F_{max -} F), in which K_d is the dissociation constant of Fluo-4 (345 nM), F is the fluorescence recorded at each time point, F_max is the maximal fluorescence, obtained upon ionomycin application, and F_min is the minimal fluorescence. The magnitude of [Ca²⁺]_i transients evoked by each stimulus (Δ[Ca²⁺]_i) was obtained subtracting the mean of basal levels from the maximum value after stimulus application.

Rats were decapitated after anesthesia and the brain was quickly removed and placed in ice-cold, oxygenated (95% O₂, 5% CO₂) artificial cerebrospinal fluid (ACSF; in mM: 124.0 NaCl, 4.4 KCl, 1.0 Na₂HPO₄, 25.0 NaHCO₃, 2.0 CaCl₂, 1.2 MgCl₂, 10.0 glucose). Using a McIlwain tissue chopper (Brinkmann Instruments, NY, USA), slices (400 μm-thick) from the dorsal hippocampus were cut transverse to its long axis and placed in a holding chamber with oxygenated ACSF. Slices were allowed to recover for at least 1 h before being transferred to a submerged recording chamber and superfused at 3 mL/min with oxygenated ACSF kept at 30.5°C.

Extracellular field excitatory post-synaptic potential (fEPSP) were recorded as previously described (Costenla et al., 2011) with the stimulating bipolar concentric electrode placed in the proximal CA1 stratum radiatum for stimulation of the Schaffer collaterals and the recording electrode, filled with 4 M NaCl (2–5 MΩ resistance), placed in the CA1 stratum radiatum targeting the distal dendrites of pyramidal neurons. Stimulation was delivered every 20 s with rectangular pulses of 0.1 ms duration using either a Grass S44 or a Grass S48 square pulse stimulator (Grass Technologies, RI, USA). After amplification (ISO-80, World Precision Instruments, Hertfordshire, UK), the recordings were digitized (BNC-2110, National Instruments, Newbury, UK), averaged in groups of three, and analyzed using the WinLTP version 2.10 software (WinLTP Limited, Bristol, UK; Anderson and Collingridge, 2007). The intensity of stimulation was chosen between 30–50% of maximal fEPSP response, determined based on input/output curves in which the fEPSP slope was plotted vs. stimulus intensity. Alterations of basal synaptic transmission were quantified as the percentage change of the average value of the fEPSP slope taken from 15–20 min after beginning exposure to the tested drug applied through the superfusion medium, relative to the average value of the fEPSP slope during the 5 min that preceded the application of each modifying drug. Long-term potentiation (LTP) was induced by a high-frequency stimulation (HFS) train (100 Hz for 1 s). LTP was quantified as the percentage change of the average fEPSP slope taken between 55 and 60 min after LTP induction relative to the average slope of the fEPSP measured during the 10 min that preceded LTP induction. The effect of drugs on LTP was assessed by comparing LTP magnitude in the absence and presence of the drug in experiments carried out in different slices from the same animal.

Data are presented as the mean ± SEM of n experiments (i.e., n independent rats or cell cultures). The comparison of control and stressed rats and the effect of drugs was analyzed using a two-tailed unpaired Student’s t-test. When testing the impact of a drug on the effects of stress, the data were first analyzed with a two-way ANOVA followed by a Newman–Keuls post hoc test. The comparison between the effect of multiple drugs was carried out using a Dunnett’s test. All tests were performed using Prism 6.0 software (GraphPad, San Diego, CA, USA) considering significance at a 95% confidence interval.

The P_2X7R-prefering antagonist Brillant Blue G (BBG, 45 mg/kg) was devoid of effects in control rats but attenuated or prevented the behavioral and neurochemical alterations caused by repeated stress (Figures 4, 5). Thus, BBG prevented the stress-induced decrease of the number of crossings in the central area of the open field (effect of stress F_(1,32) = 22.13, p < 0.001; effect of BBG F_(1,32) = 20.41, p < 0.001; interaction F_(1,32) = 17.66, p < 0.001; two-way ANOVA; Figure 4B), the stress-induced decrease of the number of entries in the open arms of the elevated plus maze (effect of stress F_(1,32) = 17.96, p < 0.001; effect of BBG F_(1,32) = 6.248, p = 0.018; interaction F_(1,32) = 23.89, p < 0.001; two-way ANOVA; Figure 4C), the stress-induced decrease of the time spent in the open arms of the elevated plus maze (effect of stress F_(1,32) = 23.28, p < 0.001; effect of BBG F_(1,32) = 4.187, p = 0.044; interaction F_(1,32) = 25.61, p < 0.001; two-way ANOVA), the stress-induced decrease of sucrose preference (effect of stress F_(1,32) = 8.737, p = 0.006; effect of BBG F_(1,32) = 4.753, p = 0.037; interaction F_(1,32) = 6.044, p = 0.019; two-way ANOVA; Figure 4D), the stress-induced increase of immobility in the forced swimming test (effect of stress F_(1,32) = 39.91, p < 0.001; effect of BBG F_(1,32) = 181.4, p < 0.001; interaction F_(1,32) = 13.02, p = 0.001; two-way ANOVA; Figure 4E), the stress-induced decrease of the time climbing the wall in the forced swimming test (effect of stress F_(1,32) = 16.35, p < 0.001; effect of BBG F_(1,32) = 36.11, p < 0.001; interaction F_(1,32) = 12.87, p = 0.001; two-way ANOVA; Figure 4F), the stress-induced decrease of the time spent in the novel arm of the Y-maze (effect of stress F_(1,32) = 9.243, p = 0.005; effect of BBG F_(1,32) = 6.434, p = 0.016; interaction F_(1,32) = 3.596, p = 0.067; two-way ANOVA; Figure 4G), and the stress-induced decrease of the relative time exploring the displaced object (t = 1.928, p = 0.072 between displaced and non-displaced object in stressed rats treated with vehicle and t = 6.246, p < 0.001 between displaced and non-displaced object in stressed rats treated with BBG, unpaired Student’s t-test; Figure 4H).

BBG also attenuated the stress-induced increase in the expression of the marker of “activated” microglia Iba1 in the hippocampus (effect of stress F_(1,20) = 30.51, p < 0.001; effect of BBG F_(1,20) = 5.295, p = 0.032; interaction F_(1,20) = 16.96, p = 0.001; two-way ANOVA; Figure 5A) and prefrontal cortex (effect of stress F_(1,20) = 30.52, p < 0.001; effect of BBG F_(1,20) = 9.150, p = 0.007; interaction F_(1,20) = 7.524, p = 0.012; two-way ANOVA; Figure 5A), as well as in the levels of mRNA of both IL1β in the hippocampus (effect of stress F_(1,20) = 38.13, p < 0.001; effect of BBG F_(1,20) = 23.05, p < 0.001; interaction F_(1,20) = 23.15, p < 0.001; two-way ANOVA; Figure 5B) and prefrontal cortex (effect of stress F_(1,20) = 24.39, p < 0.001; effect of BBG F_(1,20) = 6.641, p = 0.018; interaction F_(1,20) = 7.505, p = 0.013; two-way ANOVA; Figure 5B) and of TNFα in the hippocampus (effect of stress F_(1,20) = 59.05, p < 0.001; effect of BBG F_(1,20) = 19.99, p < 0.001; interaction F_(1,20) = 24.96, p < 0.001; two-way ANOVA; Figure 5C) and prefrontal cortex (effect of stress F_(1,20) = 152.8, p < 0.001; effect of BBG F_(1,20) = 108.7, p < 0.001; interaction F_(1,20) = 85.38, p < 0.001; two-way ANOVA; Figure 5C).

The treatment with BBG also attenuated the stress-induced up-regulation of P_2X7R in the hippocampus (effect of stress F_(1,20) = 21.31, p < 0.001; effect of BBG F_(1,20) = 8.316, p = 0.009; interaction F_(1,20) = 8.222, p = 0.009; two-way ANOVA; Figure 5D) and prefrontal cortex (effect of stress F_(1,20) = 18.12, p < 0.001; effect of BBG F_(1,20) = 5.305, p = 0.032; interaction F_(1,20) = 10.52, p < 0.001; two-way ANOVA; Figure 5D). Remarkably, BBG also attenuated the stress-induced up-regulation of A_2AR in the hippocampus (effect of stress F_(1,20) = 10.85, p = 0.004; effect of BBG F_(1,20) = 13.01, p = 0.002; interaction F_(1,20) = 3.766, p = 0.067; two-way ANOVA; Figure 5E) and prefrontal cortex (effect of stress F_(1,20) = 12.21, p = 0.002; effect of BBG F_(1,20) = 11.84, p = 0.003; interaction F_(1,20) = 16.96, p = 0.001; two-way ANOVA; Figure 5E).

The non-selective adenosine receptor antagonist, caffeine (0.3 g/L, p.o.), which affords neuroprotection through the antagonism of A_2AR (e.g., Dall’Igna et al., 2007; Cognato et al., 2010; Kaster et al., 2015), was devoid of effects in control rats but attenuated or prevented the behavioral and neurochemical alterations caused by repeated stress (Figures 6, 7). Thus, caffeine prevented the stress-induced decrease of the number of crossing in the central area of the open field (effect of stress F_(1,32) = 8.160, p = 0.007; effect of caffeine F_(1,32) = 8.459, p = 0.007; interaction F_(1,32) = 8.160, p = 0.007; two-way ANOVA; Figure 6B), the stress-induced decrease of the number of entries in the open arms of the elevated plus maze (effect of stress F_(1,32) = 23.91, p < 0.001; effect of caffeine F_(1,32) = 1.463, p = 0.235; interaction F_(1,32) = 16.27, p < 0.001; two-way ANOVA; Figure 6C), the stress-induced decrease of sucrose preference (effect of stress F_1,64 = 15.96, p < 0.001; effect of caffeine F_{3, 64} = 6.544, p = 0.001; interaction F_{3, 64} = 3.828, p = 0.014; two-way ANOVA; Figure 6D), the stress-induced increase of immobility in the forced swimming test (effect of stress F_(1,32) = 29.31, p < 0.001; effect of caffeine F_(1,32) = 10.13, p = 0.003; interaction F_(1,32) = 13.58, p = 0.001; two-way ANOVA; Figure 6E), the stress-induced decrease of the time climbing the wall in the forced swimming test (effect of stress F_(1,32) = 16.45, p < 0.001; effect of caffeine F_(1,32) = 8.564, p = 0.006; interaction F_(1,32) = 7.247, p = 0.001; two-way ANOVA; Figure 6F), the stress-induced decrease of the time spent in the novel arm of the Y-maze (effect of stress F_(1,32) = 5.879, p = 0.021; effect of caffeine F_(1,32) = 9.671, p = 0.004; interaction F_(1,32) = 6.851, p = 0.013; two-way ANOVA; Figure 6G), and the stress-induced decrease of the relative time exploring the displaced object (t = 1.492, p = 0.161 between displaced and non-displaced object in stress rats treated with vehicle and t = 8.637, p < 0.001 between displaced and non-displaced object in stress rats treated with caffeine, unpaired Student’s t-test; Figure 6H).

Caffeine also attenuated the stress-induced increase in the expression of the marker of “activated” microglia Iba1 in the hippocampus (effect of stress F_(1,20) = 63.06, p < 0.001; effect of caffeine F_(1,20) = 19.07, p < 0.001; interaction F_(1,20) = 18.35, p < 0.001; two-way ANOVA; Figure 7A) and prefrontal cortex (effect of stress F_(1,20) = 57.53, p < 0.001; effect of caffeine F_(1,20) = 10.02, p = 0.005; interaction F_(1,20) = 12.92, p = 0.002; two-way ANOVA; Figure 7A), as well as in the levels of mRNA of both IL1β in the hippocampus (effect of stress F_(1,20) = 19.2, p < 0.001; effect of caffeine F_(1,20) = 13.58, p = 0.001; interaction F_(1,20) = 12.34, p = 0.002; two-way ANOVA; Figure 7B) and prefrontal cortex (effect of stress F_(1,20) = 22.15, p < 0.001; effect of caffeine F_(1,20) = 9.351, p = 0.006; interaction F_(1,20) = 30.24, p < 0.001; two-way ANOVA; Figure 7B) and of TNFα in the hippocampus (effect of stress F_(1,20) = 82.54, p < 0.001; effect of caffeine F_(1,20) = 57.24, p < 0.001; interaction F_(1,20) = 31.66, p < 0.001; two-way ANOVA; Figure 7C) and prefrontal cortex (effect of stress F_(1,20) = 106.0, p < 0.001; effect of caffeine F_(1,20) = 85.53, p < 0.001; interaction F_(1,20) = 92.56, p < 0.001; two-way ANOVA; Figure 7C).

The treatment with caffeine also attenuated the stress-induced up-regulation of P_2X7R in the hippocampus (effect of stress F_(1,20) = 35.32, p < 0.001; effect of caffeine F_(1,20) = 15.30, p = 0.001; interaction F_(1,20) = 8.046, p = 0.010; two-way ANOVA; Figure 7D) and prefrontal cortex (effect of stress F_(1,20) = 5.011, p = 0.048; effect of caffeine F_(1,20) = 0.569, p = 0.094; interaction F_(1,20) = 51.4, p < 0.001; two-way ANOVA; Figure 7D), as well as the stress-induced up-regulation of A_2AR in the hippocampus (effect of stress F_(1,20) = 4.282, p = 0.045; effect of caffeine F_(1,20) = 5.256, p = 0.033; interaction F_(1,20) = 4.369, p = 0.050; two-way ANOVA; Figure 7E) and prefrontal cortex (effect of stress F_(1,20) = 10.98, p = 0.004; effect of caffeine F_(1,20) = 9.302, p = 0.006; interaction F_(1,20) = 8.317, p = 0.009; two-way ANOVA; Figure 7E).

Since we observed crosstalk between BBG and caffeine upon restraint stress, whereby BBG controlled the up-regulation of A_2AR and caffeine controlled the upregulation of P_2X7R expression, and the stress-induced behavioral modifications were accompanied by a parallel control of markers of microglia “activation” and neuroinflammation, we next used a microglial N9 cell line to directly investigate a putative crosstalk between P_2X7R and A_2AR, since both receptors are present and functional in this microglia cell model (e.g., Ferrari et al., 1996; Gomes et al., 2013).

The P_2X7R-preferring agonist BzATP (100 μM) evoked an elevation of intracellular free Ca²⁺ levels (Δ[Ca²⁺]_i) of 94.8 ± 14.5 nM (n = 16), which was inhibited (−76.51 ± 20.02%, n = 10–16, F_{(3, 44)} = 13.21, p = 0.029) in the presence of the selective P_2X7R antagonist, JNJ4796556 (1 μM), added 15 min before BzATP (Figures 8A,B). The selective A_2AR agonist CGS21680 (100 nM) also evoked a Δ[Ca²⁺]_i of 79.3 ± 11.9 nM (n = 7), which was inhibited (−63.0 ± 14.0%, n = 6, F_{(2, 18)} = 8.67, p = 0.003) in the presence of the selective A_2AR antagonist, SCH58261 (50 nM), added 15 min before CGS21680 (Figures 8C,D). Notably, the Δ[Ca²⁺]_i evoked by BzATP (100 μM) was inhibited (−69.5 ± 15.7%, n = 10–16, F_{(3, 44)} = 12.13, p = 0.048) by SCH58261 (50 nM) and potentiated (+80.3 ± 19.5%, n = 12–16, F_{(3, 44)} = 12.13, p = 0.012) by CGS21680 (100 nM; Figure 8B), whereas the Δ[Ca²⁺]_i triggered by CGS21680 (100 nM) was inhibited (−54.3 ± 14.7%, n = 8, F_{(2, 18)} = 8.67, p = 0.005) by JNJ47965567 (1 μM; Figure 8D), indicating a crosstalk between P_2X7R and A_2AR in the control of Δ[Ca²⁺]_i responses in microglial N9 cells. This P_2X7R-A_2AR crosstalk is further reinforced by the observation that neither JNJ47965567 (1 μM) nor SCH58216 (50 nM) affected basal [Ca²⁺]_i levels (control, no drugs: 284.2 ± 30.2 nM, n = 10; 1 μM JNJ47965567: 224.2 ± 32.0 nM, n = 9; F_{3, 34} = 1.92, p = 0.488 vs. control; 50 nM SCH58261: 219.1 ± 17.6 nM, n = 8, F_{3, 34} = 1.92, p = 0.449 vs. control), indicating a lack of tonic P_2X7R- or A_2AR-mediated control of Δ[Ca²⁺]_i that could hinder the interpretation of the cross-inhibition between both purinergic receptor systems.

We next explored if there was a control by P_2X7R and by A_2AR and a crosstalk between both receptors in the control of Δ[Ca²⁺]_i evoked by glutamate to mimic a condition of excitotoxicity-induced “activation” of microglia (reviewed in Zhang et al., 2020), irrespective of the receptors involved. Glutamate (100 mM) triggered a Δ[Ca²⁺]_i of 93.3 ± 13.4 nM (n = 18), which was inhibited either by 1 μM JNJ47965567 (−66.4 ± 13.3%, n = 10–18, F_4,53 = 13.56, p = 0.002) or by 50 nM SCH58261 (−42.67 ± 8.41%, n = 13–18, F_4,53 = 13.56, p = 0.050), each added 15 min before BzATP (Figure 8E). Notably, glutamate-induced Δ[Ca²⁺]_i was 25.1 ± 4.1 nM (n = 7) in the simultaneous presence of JNJ47965567 (1 μM) and SCH58261 (50 nM) indicating an inhibition of −73.1 ± 15.8% (Figure 8E), which was similar to that caused by JNJ47965567 alone (t = 0.997, p = 0.334).

Since we and others have collected evidence for a role of synaptic dysfunction underlying stress-associated behavioral alterations (Duman and Aghajanian, 2012; Kaster et al., 2015) and suggestions of P_2X7R-mediated synaptic dysfunction add-up to the well-established ability of A_2AR to control synaptic function (reviewed in Cunha, 2016), we next investigated if P_2X7R and A_2AR might interact in the control of synaptic plasticity in excitatory synapses of the dorsal hippocampus.

We first tested the effect of P_2X7R agonist BzATP on basal synaptic transmission. BzATP (30 μM) decreased hippocampal synaptic transmission by 54.75 ± 3.96% (n = 4); this effect recovered fully upon washout of BzATP and repeated administrations of 30 μM BzATP caused a similar depression of synaptic transmission (p > 0.05). This allowed exploring the pharmacology of BzATP (30 μM)-induced decreased hippocampal synaptic transmission: this effect was unaffected in the presence of 1 μM BBG (−48.98 ± 4.96%, n = 4, t = 1.245, p = 0.260 vs. the effect of BzATP alone) and was fully prevented in the presence of the adenosine A₁ receptor antagonist, DPCPX (50 nM; 2.97 ± 17.91% alteration of fEPSP slope, n = 4; t = 3.319, p = 0.016 vs. the effect of BzATP alone; Figure 9A). This shows the inexistence of a P_2X7R-mediated effect (lack of effect of BBG) and indicates that BzATP is rapidly converted by ectonucleotidases (Cunha et al., 1998) into an adenosine analog to indirectly alter hippocampal synaptic transmission through inhibitory A₁ adenosine receptors (prevention by DPCPX), as previously proposed (Kukley et al., 2004). This precludes the use of BzATP to search for P_2X7R-mediated effects in hippocampal slices. Instead, we tested the impact of P_2X7R antagonists on high-frequency induced LTP in Schaffer collaterals-CA1 pyramidal cell synapses. LTP magnitude was not significantly altered by either 1 μM BBG (n = 8, t = 0.493, p = 0.630 vs. LTP magnitude in control conditions, i.e., in the absence of tested drugs) or 1 μM JNJ47965567 (n = 6; t = 0.754, p = 0.468 vs. control LTP magnitude; Figures 9B,C). This does not support a role of P_2X7R in the control of synaptic plasticity.

We next investigated if P_2X7R might instead control the known ability of A_2AR to control hippocampal synaptic plasticity (e.g., Costenla et al., 2011; Lopes et al., 2019). SCH58261 (50 nM) decreased LTP magnitude by −48.10 ± 10.77% (n = 5; t = 2.440, p = 0.029 vs. control LTP magnitude; Figure 9D) and a non-significantly different inhibition of −49.11 ± 7.21% (n = 5; t = 0.049, p = 0.962 vs. LTP magnitude in the SCH58261 alone) was observed in the simultaneous presence of BBG (1 μM) and SCH58261 (50 nM; Figure 9D).