Consumption of the methylxanthine caffeine is beneficial in age-related cognitive decline.1 More importantly, longitudinal and retrospective epidemiological studies show that it delays dementia onset and reduces Alzheimer’s disease (AD) risk.1 Consistent with these epidemiological observations, chronic treatment with caffeine in AD animal models improves memory and mitigates accumulation of amyloid peptides and hyperphosphorylated Tau proteins, the two neuropathological hallmarks of this disorder.2, 3

Caffeine is a non-selective adenosine A2A receptor (A2AR) antagonist.4 A2ARs are constitutively activated G-protein-coupled receptors, preferentially expressed in striato-pallidal medium spiny neurons, but are also present, to a lower extent, in cortical and hippocampal neurons.5 Neuronal A2ARs modulate excitability by pre-synaptic control of neurotransmitter release, notably glutamate, as well as by post-synaptic control of excitability.6, 7 A2ARs are also expressed by astrocytes and microglia, controlling their activation state and their ability to uptake glutamate or to release pro-inflammatory factors.6 Blockade of A2ARs has been shown to afford neuroprotection in various neurodegenerative conditions, such as stroke and Parkinson’s disease.8 Interestingly, the protective effects of caffeine achieved in Parkinson’s disease models are mimicked by A2AR pharmacological blockade and gene deletion.9, 10 In sharp contrast, the amount of data regarding the impact of A2AR modulation in AD is still scarce. So far, A2AR blockade was found to be protective against the acute toxicity of amyloid peptides.11 However, the impact of A2AR blockade on the progressive development of AD-related lesions and associated memory impairments has not been investigated yet. Furthermore, even if Tau pathology has a pivotal role in AD,12, 13 the relationships between A2ARs and Tau pathology have been totally overlooked.

In the present study, we explored the outcome of A2AR gene deletion in the THY-Tau22 transgenic mouse model, which progressively develops hippocampal Tau pathology and spatial memory defects.14, 15 Our data demonstrate that A2AR deletion normalizes memory, hippocampal plasticity and neurotransmitter imbalance, but also reduces Tau hyperphosphorylation in these transgenic animals. In addition, oral administration of a selective A2AR antagonist (MSX-3) improves spatial memory and reduces Tau hyperphosphorylation in Tau mice. These findings support the concept of A2ARs as valuable targets against AD progression and pathology.

Materials and methods


A2AR knockout animals and THY-Tau22 Tau transgenic mice have been described elsewhere.14, 15, 16 Parental lines used were both on C57Bl6/J background. Experimental animals were generated by mating THY-Tau22 male mice with A2A+/− female animals. F1 THY-Tau22 A2A+/− males were then crossed with A2A+/− females to generate F2 double-mutant animals, that is, THY-Tau22 A2A−/− (referred as Tau A2A−/−), and related littermate controls, that is, WT A2A+/+, WT A2A−/− and THY-Tau22 A2A+/+ (referred as Tau A2A+/+). Genotyping was realized by PCR analysis of tail DNA as described.14, 16 Mice were housed in a pathogen-free facility, 5 to 6 per cage (Techniplast cages 1284L; Lyon, France), maintained on a 12-h light/12-h dark cycle with ad libitum access to food and water. Body temperature was evaluated before animal sacrifice and showed no statistical differences between groups (not shown). We did not identify any differences in the mouse survival rate, regardless of the genotype. Since no overt gender differences were observed (see for instance Supplementary Figure 1), data from both males and females were analysed as a single group. All experiments were performed at 6–7 months of age, when Tau transgenic mice are known to exhibit significant Tau pathology and memory alterations.17, 18 All protocols were approved by an ethics committee (no. 342012, CEEA).

Treatment of THY-Tau22 mice with MSX-3

MSX-3 is a water-soluble prodrug of the potent and highly selective A2AR antagonist MSX-219 that crosses the blood–brain barrier.20 The drug was given to THY-Tau22 male mice in drinking water at a dose of 0.3 g l−1. Chronic delivery at this dose achieved, in 7-month-old C57Bl6/J mice, plasma and brain concentrations of MSX-2 of about 10–20 nM (not shown), compatible with A2AR blockade (Ki=ca. 8 nM). At this dose, MSX-3 alone was devoid of effects on hippocampal-dependent memory in C57Bl6/J mice (not shown). Animals were randomized according to their body weight into the following three experimental groups: WT/H2O, THY-Tau22/H2O and Tau/MSX-3. The MSX-3 solution was kept in bottles protected from light and changed weekly. Treatment was started at 6 months of age, an age at which Tau pathology and memory impairments are already significant in THY-Tau22 mice (see Figures 1 and 2), and was continued until 8.5–9 months of age. Body weight was measured weekly. MSX-3 consumption was assessed throughout treatment for each experimental cage, allowing to estimate the average consumption per mice. Body temperature was evaluated at completion of the protocol before the Y-maze task.

Figure 1
figure 1

Adenosine A2A receptor (A2AR) deletion prevents behavioural impairments in Tau transgenic mice. (a, b) Effect of A2AR deletion on spatial memory using Y-maze task. (a) During the exposure phase, all groups explored the maze equally, spending a similar amount of time in each available arm (P>0.05, one-way analysis of variance (ANOVA)). (b) During the test phase, the discrimination index, representative of a spatial preference for the ‘novel’ arm, was significantly set above chance (50%) for WT A2A+/+ and WT A2A−/− animals (P<0.05 for WT A2A+/+, P<0.001 for WT A2A−/− vs 50%; Student’s t-test). While hippocampus-dependent spontaneous spatial novelty preference was impaired in Tau A2A+/+ mice (**P<0.01 vs WT A2A+/+, one-way ANOVA), Tau A2A−/− mice exhibited a normalized preference for the ‘novel’ arm (###P<0.001 vs Tau A2A+/+, one-way ANOVA), indicating that A2A deletion prevented memory alterations in Tau transgenic mice. N=7–11 per group. (ce) Effect of A2AR deletion on spatial memory using the Morris water maze. (c) Learning was found to be similar for all genotypes, as indicated by the equivalent path length needed to find the hidden platform (P>0.05, two-way ANOVA). (d) All genotypes exhibited a comparable velocity in the maze, suggesting no motor deficits (P>0.05, one-way ANOVA). (e) Spatial memory was assessed 72 h after the last day of learning. Results represent the percentage of time spent in the target (T) vs non-target (NT) quadrants. Wild-type (WT) animals (both A2A+/+ and A2A−/−) spent significantly more time in the T quadrant, indicative of a preserved spatial memory. While Tau A2A+/+ mice displayed memory deficits as underlined by their lack of preference for the T quadrant, Tau A2A−/− mice behaved as non-transgenic animals, suggesting a rescue of Tau-related memory impairments by A2AR deletion. XXXP<0.001, NT vs T (Student’s t-test); **P<0.01, Tau A2A+/+ vs WT A2A+/+, and ##P<0.01, Tau A2A+/+ vs Tau A2A−/− (one-way ANOVA); N=19–24 per group. Results are expressed as mean±s.e.m.

PowerPoint slide

Figure 2
figure 2

Adenosine A2A receptor (A2AR) deletion reduces hippocampal Tau phosphorylation and impacts Tau kinases in THY-Tau22 mice. (a) Representative two-dimensional profile of total human Tau in Tau A2A+/+ (top) vs Tau A2A−/− (bottom) mice, showing a reduced amount of Tau acidic species in Tau A2A−/− animals (arrow; Total Tau C-ter antibody). (b) Quantification of Tau phosphorylation on Thr212/Ser214, Ser262, Thr181, Ser404 and Ser214 epitopes, as well as dephosphorylated Tau (Tau1) in Tau A2A+/+ and Tau A2A−/− mice (#P<0.05 vs Tau A2A−/−, N=6–7 per group, Student’s t-test). Results are expressed as a percentage of Tau A2A+/+±s.e.m. (c) Representative western blots of sarkosyl-soluble (S) and -insoluble (I) hippocampal Tau species of Tau A2A+/+ (top) and Tau A2A−/− (bottom) mice, showing no overt difference between the two genotypes (Total Tau C-ter antibody). T shows the total amounts of Tau in non-fractionated samples. Quantification of insoluble Tau species is represented in the graph below (N=5 per group). (d) Quantification of the Tau kinase changes (GSK3β, AMPK, Erk, P38, JNK and cdk5 and its regulator p35/p25), as well as of the catalytic subunit of its main phosphatase PP2A (#P<0.05 and ##P<0.01 vs Tau A2A−/−, N=6–7 per group, Student’s t-test). Results are expressed as a percentage of Tau A2A+/+±s.e.m.

PowerPoint slide

Behavioural evaluation

Exploratory test. Open-field exploration was examined using a 50 cm × 50 cm arena. Each mouse was placed in the arena for 10 min. Movements were recorded using EthoVision video tracking equipment and software (Noldus Bv, Wageningen, The Netherlands). Total path length was measured as an index of mouse locomotion.

Rotarod test. Motor coordination and equilibrium were tested on an accelerating rotarod (MED Associates Inc., St Albans, VT, USA). Mice were placed on the rod with their head directed against the direction of the rotation so that the animal had to progress forward to maintain its balance. Mice were first trained on a constant speed (4 r.p.m., 2 min) before starting with four test trials (intertrial interval, 10 min). During these trials, mice had to balance on the rotating rod that accelerated from 4 to 40 r.p.m. within 5 min. The latency until the mice fell from the rotating rod was recorded, up to a maximum of 5 min.

Anxiety test. The elevated plus maze was used to investigate anxiety-related behaviour. The apparatus consisted of a plus-shaped maze with two closed and two open arms (5 cm wide). Mice were placed at the centre of the maze with their face in the direction of a closed arm, and were allowed to explore freely for 10 min. Five infrared beam detectors, connected to a PC activity logger, recorded the behaviour of the mice. Percentage of time spent in the open arms was recorded.

Y-maze test. Short-term spatial memory was assessed in a spontaneous novelty-based spatial preference Y-maze test as described.21 Each arm of the Y-maze was 22 cm long, 6.4 cm wide, with 15-cm-high opaque walls. Different extramaze cues were placed on the surrounding walls. Sawdust was placed in the maze during the experiments and mixed between each phase. Allocation of arms was counterbalanced within each group. During the exposure phase, mice were placed at the end of the ‘start’ arm and were allowed to explore the ‘start’ arm and the ‘other’ arm for 5 min (beginning from the time the mouse first left the start arm). Access to the third arm of the maze (‘novel’ arm) was blocked by an opaque door. The mouse was then removed from the maze and returned to its home cage for 2 min. In the test phase the mouse was placed again in the ‘start’ arm of the maze, the door of the ‘novel’ arm was removed and the mouse was allowed to explore the maze for 1 min (from the time the mouse first left the start arm). The amount of time the mouse spent in each arm of the maze was recorded during both exposure and test phases using EthovisionXT (Noldus Information Technology, Wageningen, The Netherlands). For the exposure phase, we calculated the percentage of time spent in the ‘other’ arm vs the ‘start’ arm. For the test phase, a discrimination index [novel arm/(novel+other arm)] × 100 was calculated.

Morris water maze. Spatial memory abilities were evaluated in the standard hidden platform (PF) acquisition and retention version of the water maze, as we previously described.3 A 100-cm circular pool was filled with water, opacified with non-toxic white paint (Viewpoint, Lyon, France) and kept at 21 °C. A 10-cm round PF was hidden 1 cm beneath the surface of the water at a fixed position. Four positions around the edge of the tank were arbitrarily designated 1, 2, 3 and 4, thus dividing the tank into four quadrants (clockwise): target (hidden PF contained), adjacent 1, opposite and adjacent 2.

During the learning procedure, mice were tested during the light phase between 08:00 and 18:00 hours. Each mouse was given four swimming trials per day (20-min intertrial interval) for five consecutive days. The start position (1, 2, 3 or 4) was pseudo-randomized across trials. A trial consisted of placing the mouse into the water facing the outer edge of the pool in one of the virtual quadrants and allowing it to escape to the hidden PF. A trial terminated when the animal reached the PF, where it was allowed to remain for 15 s. If the animal failed to find the target before 120 s, it was manually guided to the PF, where it was allowed to stay for 15 s. After completion of a trial, mice were removed from the pool and placed back to their home cages beneath heat lamps in order to reduce the loss of temperature. Time required to locate the hidden escape PF (escape latency) and distance travelled to the hidden escape PF (path length) was recorded using the Ethovision XT tracking system (Noldus, Wageningen, The Netherlands). Swimming speed (that is, velocity, as a measure of possible motor defects that could interfere with their ability to perform in this task) was also measured. Seventy-two hours following the acquisition phase, a probe trial was conducted. During this probe trial (60 s), the PF was removed and the search pattern of the mice was tracked again. Proportion of time spent in the target quadrant vs averaged non-target quadrants was determined.

Biochemical analyses

Mice used for biochemical evaluations and mRNA studies were quickly killed by decapitation, as anaesthesia promotes Tau hyperphosphorylation.22 Brains were removed and hippocampi were dissected out using a coronal acrylic slicer (Delta Microscopies, Mauressac, France) at 4 °C and stored at −80 °C until use.

For all biochemical experiments, tissue was homogenized in 200 μl Tris buffer (pH 7.4) containing 10% sucrose and protease inhibitors (Complete; Roche Diagnostics, Meylan, France), sonicated and kept at 80 °C until use. Protein amounts were evaluated using the BCA assay (Pierce, Rockford, IL, USA).

Bi-dimensional electrophoresis experiments were performed as previously described.23 Briefly, lysates were precipitated with methanol/chloroform. Fifteen micrograms of proteins were dissolved in 2D buffer (7 M urea, 2 M thiourea, 4% CHAPS and 0.6% pharmalytes). Lysates were loaded on immobilized pH gradient strip 3–11 ReadyStrip (Amersham GE, Velizy-Villacoublay, France) and isoelectrofocused with the Protean IEF cell (Amersham GE) according to the manufacturer’s instructions. The strips were layered onto a 4–12% Bis–Tris poly-acrylamide gel. Membranes were incubated with anti-total Tau antibody (Cter).

For sarkosyl-soluble/insoluble protein preparations, hippocampi were homogenized by sonication in a lysis buffer containing 10 mM Tris-HCl, pH 7.4, 0.32 M sucrose, 800 mM NaCl and 1 mM EGTA with protease inhibitors (Complete w/o EDTA, Roche Diagnostics), and centrifuged at 12 000 g for 10 min at 4 °C.24 The supernatant incubated 1 h in 1% sarkosyl (N-laurylsarkosine sodium salt, Sigma, Saint-Quentin-Fallavier, France) at room temperature was then centrifuged at 100 000 g for 1 h at 4 °C, thus forming the supernatant and pellet containing sarkosyl-soluble and -insoluble Tau species, respectively. Sarkosyl-insoluble proteins were directly resuspended in LDS 2 × and sarkosyl-soluble samples were mixed with LDS 2 ×, supplemented with reducing agents (Invitrogen, Saint Aubin, France). Sarkosyl-soluble and -insoluble samples were loaded onto NuPage Novex (Invitrogen) gels at a ratio of 1:2 (v:v).

For western blots, proteins were diluted with LDS 2 × supplemented with reducing agents (Invitrogen) and then separated on NuPage Novex gels (Invitrogen). Proteins were transferred to nitrocellulose membranes, which were then blocked (5% non-fat dry milk or 5% bovine serum albumin in TNT:Tris-HCl 15 mM, pH 8, NaCl 140 mM, 0.05% Tween) and incubated with primary and secondary antibodies. Signals were visualized using chemiluminescence kits (ECLTM, Amersham Velizy-Villacoublay, Villacoublay, France) and a LAS3000 imaging system (Fujifilm, Tokyo, Japan). Total proteins were quantified vs glycéraldéhyde-3-phosphate déshydrogénase. Phosphorylated proteins were quantified vs total counterpart. Quantifications were performed using ImageJ software. All primary antibodies used are described in Table 1.

Table 1 Antibodies used in this study

mRNA extraction and quantitative real-time reverse transcription (RT)-PCR analysis

Total RNA was extracted from hippocampi and purified using the RNeasy Lipid Tissue Mini Kit (Qiagen, Courtaboeuf, France). One microgram of total RNA was reverse-transcribed using the Applied Biosystems (Saint-Aubin, France) High-Capacity cDNA reverse transcription kit. Quantitative real-time reverse transcriptase-PCR analysis was performed on an Applied Biosystems Prism 7900 System using Power SYBR Green PCR Master Mix. The thermal cycler conditions were as follows: hold for 10 min at 95 °C, followed by 45 cycles of a two-step PCR consisting of a 95-°C step for 15 s followed by a 60-°C step for 25 s. Sequences of the primer used are given in Table 2. Cyclophilin A was used as internal control. Amplifications were carried out in triplicate and the relative expression of target genes was determined by the ΔΔCT method.

Table 2 Primer sequences used in this study

ELISA analysis

Hippocampal tissues were homogenized in RIPA buffer (Tris HCl 25 mM, NaCl 150 mM, NP40 1%, sodium deoxycholate 1%, SDS 0.1%, pH 7.6). The resulting lysates were sonicated and incubated for 1 h at 4 °C under agitation before centrifugation (12 000 g, 4 °C, 15 min). Supernatants were collected and protein amounts evaluated using the BCA assay (Pierce). CCL3, CCL4 and CCL5 levels were quantified using commercially available ELISA assays (R&D Systems, Abingdon, UK) after loading 300 μg of proteins in 50 μl volume.


GFAP immunohistochemistry was performed as previously described21 using 40 μm floating brain sections from paraformaldehyde-perfused brains of the different experimental groups. Quantification of the GFAP staining intensity was performed using Image J Software (Scion Image, Bethesda, MD, USA).

Electrophysiological recordings

Animals were killed by cervical dislocation, decapitated, the brain rapidly removed and the hippocampi dissected in ice-cold Krebs solution composed of (mM): NaCl 124; KCl 3; NaH2PO4 1.25; NaHCO3 26; MgSO4 1; CaCl2 2; and glucose 10 (previously gassed with 95% O2 and 5% CO2, pH 7.4). Slices (400 μm thick) were obtained with a McIlwain tissue chopper, left to recover for at least 1 h in Krebs solution, and field excitatory postsynaptic potentials were recorded as previously described in the CA1 stratum radiatum.25 Long-term depression protocol (LTD, 3 trains with 10 min interval of 2 Hz, 1,200 pulses) was adapted from Van der Jeug et al.,26 recorded at 32 °C with a constant flow of 3 ml min−1.

Microdialysis and neurotransmitter determinations

For microdialysis studies, mice were anaesthetized with urethane (1.62 g kg−1, i.p.) and placed in a stereotaxic frame (David Kopf, Tujunga, CA, USA) with the body temperature maintained close to 37.5 °C using a heated under-blanket (Harvard Instruments, Les Ulis, France). The skull was exposed and, after drilling an appropriate hole, a homemade microdialysis probe was implanted in a randomized manner in the right or left dorsal hippocampus (coordinates relative to the bregma: AP −1.8 mm, L ±1.5 mm, V −2.4 mm below the brain surface) according to the atlas of Franklin & Paxinos. Concentric microdialysis probes were constructed from regenerated cellulose dialysis tubing (MWCO 6000 Da, 225 mm o.d., 1 mm active dialysis length) and fused-silica capillary tubing, the body of the probe being made of a 3-cm 26-G stainless steel tube. The probes were perfused at a rate of 1 μl min−1 with artificial cerebrospinal fluid (149 mM NaCl, 2.80 mM KCl, 1.2 mM MgCl2, 1.2 mM CaCl2, 2.78 mM phosphate buffer, pH 7.4). At least 2 h was allowed to elapse after microdialysis probe implantation before collection of basal samples (fractions 1, 2 and 3). Samples were collected every 5 min. Samples were stored and kept at −80 °C before analyses. At the end of the experiment, the mice were decapitated; the implanted hemisphere was removed and the placement of the cannula was verified on the frozen hemisphere.27

To evaluate gamma-amino butyric acid (GABA) and glutamate levels in dialysates, 5 μl of sample and 5 μl of standard solutions were derivatized at room temperature by adding 2 μl of a mixture (1:2:1 v/v/v) of (i) internal standard (10–4 mol l−1 cysteic acid in 0.117 mol l−1 perchloric acid), (ii) a borate/NaCN solution (100:20 v/v mixture of 500 mmol l−1 borate buffer, pH 8.7, and 87 mmol l−1 NaCN in water) and (iii) a 2.925-mmol l−1 solution of naphthalene-2,3-dicarboxaldehyde in acetonitrile/water (50:50 v/v). The samples were then analysed for amino-acid content using an automatic capillary electrophoresis P/ACE™ MDQ system (Beckman, Brea, CA, USA) equipped with a ZETALIF laser-induced fluorescence detector (Picometrics, Labege, France). Excitation was performed using a diode laser (Oxxius, Lannion, France) at a wavelength of 410 nm, the emission wavelength being 490 nm. Separations were carried out on a 63 cm × 50 μm i.d. fused-silica capillary (Composite Metal Services, Worcester, England) with an effective length of 52 cm. Each day, before the analyses, the capillary was sequentially flushed with 0.25 mol l−1 NaOH (15 min), ultra-pure water (15 min) and running buffer (75 mmol l−1 sodium borate, pH 9.20±0.02, containing 10 mmol l−1 HP-β-CD and 70 mmol l−1 SDS) (5 min). The separation conditions were an applied voltage of 25 kV, hydrodynamic sample injection (10 s at 0.6 psi) and a temperature between 41 and 43 °C. The capillary was sequentially flushed for 30 s each with 0.25 mol l−1 NaOH, ultra-pure water and running buffer between analyses. Electropherograms were acquired at 15 Hz using Karat 32 software.28


Results are expressed as means±s.e.m. Differences between mean values were determined using the Student’s t-test, two-way analysis of variance (ANOVA) or one-way ANOVA, followed by a post-hoc Fisher’s LSD test using Graphpad Prism Software. P values <0.05 were considered significant.


A2AR deletion prevents behavioural impairments in Tau transgenic mice

We evaluated the impact of A2AR deletion on behavioural impairments of THY-Tau22 mice at the age of 7 months. First, we determined the effects of A2AR deletion on spatial memory using Y-maze task. During the exposure phase, all groups explored the maze equally, spending a similar amount of time in the two available arms (P>0.05; Figure 1a). During the test phase, we found that the discrimination index, reflecting the preference of mice for the novel arm vs the familiar arm, was significantly above chance (that is, >50%) for WT A2A+/+ (P=0.03) and WT A2A−/− (P<0.0001) animals, both groups exhibiting a similar performance (P=0.6, one-way ANOVA; Figure 1b). Importantly, while Tau A2A+/+ mice exhibited an impaired preference for the new arm (P=0.24 when compared to chance, that is, 50%; P=0.0023 in Tau A2A+/+ vs WT A2A+/+, one-way ANOVA), the discrimination index of Tau A2A−/− animals was found to be significantly higher than that of Tau A2A+/+ mice (P=0.0003 in Tau A2A−/− vs Tau A2A+/+, one-way ANOVA; P=0.0015 when compared to chance, that is, 50%; Figure 1b), indicative of an improved spatial memory.

In addition, we evaluated the impact of A2AR deletion on the development of memory impairments of THY-Tau22 mice using Morris water maze. As shown in Figure 1c, following path length calculations, there was no significant difference between groups during the learning phase using two-way ANOVA analysis (P>0.05). Also, neither Tau nor A2A genotypes influenced animal velocity (P>0.05; Figure 1d). Seventy-two hours following acquisition, a probe trial was performed. Regardless of A2AR expression, WT animals exhibited a significant preference for the target quadrant (Figure 1e; WT A2A+/+: P<0.0001; WT A2A−/−: P<0.0001 vs non-target quadrants), with a significantly greater proportion of time spent in the former than expected by chance (that is, >25%; WT A2A+/+: 36.1±2.3%, P<0.0001; WT A2A−/−: 35.1±1.9%, P<0.0001). Therefore, in line with the Y-maze evaluation, A2AR deletion itself did not impact spatial memory in WT animals. As expected, Tau A2A+/+ mice exhibited major memory impairments with no preference for the target quadrant (Figure 1e; P>0.05 vs non-target quadrants; P=0.17 when compared to chance, that is, 25%), as well as a significant reduction of the percentage of time spent in the latter as compared to WT animals (WT A2A+/+: 36.1±2.3% vs Tau A2A+/+: 29.1±2.9%, P=0.009; one-way ANOVA). In sharp contrast, A2AR deletion prevented memory impairments in THY-Tau22 animals. Indeed, Tau A2A−/− mice manifested a significant preference for the target quadrant (Figure 1e, P<0.0001 vs non-target quadrants) and spent a significantly greater proportion of their time in the former than expected by chance (37.6±2.9%, P=0.0003). Furthermore, one-way ANOVA analysis indicated that Tau A2A−/− mice spent a higher percentage of time in the target quadrant than Tau A2A+/+ animals (37.6±2.9% vs 29.1±2.9%; P=0.002; Figure 1e), while they exhibited a score similar to WT A2A+/+ (37.6±2.9% vs 36.1±2.3%; P=0.6) and WT A2A−/− groups (37.6±2.9% vs 35.1±1.9%; P=0.36). In THY-Tau22 mice, similar beneficial effects of A2AR deletion were found regardless of mice gender (Supplementary Figure 1). Altogether, these behavioural evaluations indicated that A2AR deletion prevents spatial memory alterations in THY-Tau22 mice.

Of note, we also checked motor performances using open-field and rotarod tasks. We did not find significant differences in path length and latency to fall between experimental groups (Supplementary Figures 2A and B). We also analysed the impact of A2AR deletion on anxiety using the elevated plus maze task. THY-Tau22 mice spent significantly more time in the open arms as compared to controls (WT: 28.7±1.6% vs Tau A2A+/+: 36.7±3.7%; P=0.04, one-way ANOVA; Supplementary Figure 2C), whereas A2AR knockout animals exhibited the opposite phenotype (WT: 28.7±1.6% vs A2A−/−: 19.9±2.1%, P=0.04, one-way ANOVA). In the THY-Tau22 background, A2AR deletion normalized the anxiety profile. Tau A2A−/− mice spent significantly less time in the open arms than Tau A2A+/+ animals (Tau A2A−/−: 28.9±2.8%, P=0.04 vs Tau A2A+/+; one-way ANOVA) and exhibited a similar score to that of WT animals (P=0.04 vs WT; P=0.95, one-way ANOVA; Supplementary Figure 2C). A2AR deletion thus appears to normalize anxiety impairments in THY-Tau22 mice.

A2AR deletion reduces hippocampal Tau phosphorylation in THY-Tau22 mice

Spatio-temporal progression of Tau pathology in the AD brains correlates with the progression of cognitive symptoms.12, 13 Accordingly, progressive development of hippocampal Tau pathology parallels memory impairments in THY-Tau22 mice.15 Therefore, we investigated whether the memory improvement seen in Tau A2A−/− mice were related to any changes in Tau protein. In a first attempt, given the important number of phosphorylation sites on Tau (>80),29 we performed a two-dimensional (2D) gel electrophoresis analysis to evaluate global charge changes of human Tau protein in Tau A2A+/+ vs Tau A2A−/− animals. We observed a reduction of Tau isovariants at the acidic pH range in Tau A2A−/− mice as compared to Tau A2A+/+ animals, suggesting a reduced Tau phosphorylation (arrow; Figure 2a).

A moderate reduction of phosphorylation was observed by sodium dodecyl sulphate-polyacrylamide gel electrophoresis at the level of several Tau phospho-epitopes. As compared to Tau A2A+/+ animals, Thr181 and Ser404 epitopes showed a trend for reduced phosphorylation in Tau A2A−/− animals (Thr181: −24.9±6.1%, P=0.09; Ser404: −27.4±8.8%, P=0.09 vs Tau A2A+/+, Student’s t-test) and phosphorylation on Ser214 was found to be significantly reduced (−32.9±5.1% of Tau A2A+/+, P=0.04, Student’s t-test; Figure 2b and Supplementary Figure 3a). Conversely, phosphorylation on Thr212/Ser214 (AT100) and Ser262 epitopes as well as Tau1 staining, reflecting dephosphorylated Tau at Ser195/198/199/202, remained unchanged in Tau A2A−/− vs Tau A2A+/+ animals (P=0.26, P=0.35 and P=0.56, respectively, Student’s t-test; Figure 2b and Supplementary Figure 3a). Noteworthy, A2AR deletion did not influence transgene expression as human Tau mRNA levels remained unchanged in Tau A2A−/− vs Tau A2A+/+ animals (not shown). Tau protein levels and proteolytic fragments, known to favour aggregation,30 were also unaffected in both groups (not shown), suggesting that A2AR deletion does not affect the stability or the degradation of human Tau proteins. Finally, to study the impact of the A2AR knockout on Tau aggregation, biochemical fractionation was performed and sarkosyl-soluble/insoluble fractions analysed. The amount of sarkosyl-insoluble Tau remained similar in Tau A2A+/+ and Tau A2A−/− mice (P>0.05; Figure 2c).

A2AR deletion and kinase/phosphatase balance in THY-Tau22 mice

Tau phosphorylation is under the tight control of a kinase/phosphatase balance. More than 30 different kinases are able to phosphorylate Tau.29 We thus evaluated the impact of A2AR deletion on activation of major Tau kinases, namely GSK3β, AMPK, p42/44 MAPK (Erk), p38 MAPK, JNK and Cdk5. As shown in Figure 2d and Supplementary Figure 3b, we observed that in THY-Tau22 mice A2AR deletion led to a significant increase of GSK3β phosphorylation at the Ser9 inactivating epitope (+40.8±10.0% of Tau A2A+/+ mice; P=0.009, Student’s t-test), as well as to a mild but significant reduction of Cdk5 immunoreactivity (−18.4±2.9% of Tau A2A+/+ mice; P=0.04, Student’s t-test) without any change in the level of its pathological activator p25. Conversely, phosphorylation of AMPK, Erk, p38 and JNK remained unchanged in Tau A2A−/− mice. In addition, we addressed expression, demethylation and Y307 phosphorylation of the catalytic subunit of PP2A (PP2Ac), the main Tau phosphatase. Results showed that PP2Ac remained unaffected by A2AR knockout in THY-Tau22 animals (Figure 2d and Supplementary Figure 3b). The reduced Tau phosphorylation seen in Tau A2A−/− mice thus appears to be likely associated to changes in GSK3β and cdk5 kinases.

A2AR deletion modulates hippocampal neuro-inflammation in THY-Tau22 mice

Tau transgenic models have been previously shown to exhibit hippocampal neuro-inflammation, progressing along the development of hippocampal Tau pathology and behavioural alterations.3, 31 Of interest, neuro-inflammation favours Tau hyperphosphorylation32 and memory dysfunctions33 and A2AR blockade has been suggested to mitigate neuroinflammatory processes.6, 34 These observations prompted us to investigate the effects of A2AR deletion on the neuro-inflammatory markers previously found to be upregulated in Tau transgenic mice.3 As expected, we found that mRNA expression of microglial (CD68) and astroglial (GFAP) markers as well as of several pro-inflammatory cytokines (CCl3, CCl4, CCl5 and TNFα) were significantly upregulated in the hippocampus of Tau A2A+/+ animals, with most of them significantly reduced in Tau A2A−/− mice (CD68: −18.8±6.9%, P=0.03; GFAP: −38.1±4.6%, P=0.0004; TNFα: −34.3±7.5%, P=0.0004; CCl3: −33.0±11.5%, P=0.02; and CCl5: −29.3±10.1%, P=0.02 vs Tau A2A+/+, one-way ANOVA; Supplementary Figure 4B). Using immunohistochemistry and ELISA, we confirmed that A2AR deletion normalized GFAP expression and CCl3 levels in Tau mice (Supplementary Figures 4B and C). In WT animals, A2AR deletion itself had no significant effect on the levels of all markers studied.

A2AR deletion modulates hippocampal levels of glutamate and GABA

A2ARs regulate neuronal pre-synaptic release as well as astrocyte-based reuptake of glutamate and GABA.6, 35, 36 THY-Tau22 mice exhibit higher hippocampal glutamate levels compared to WT animals, regardless of the A2AR genotype (P=0.021; two-way ANOVA; Figure 3a). Surprisingly, hippocampal GABA levels were found to be significantly upregulated in the hippocampus of Tau A2A−/− animals, reaching 206±16% of Tau A2A+/+ (P=0.003 using one-way ANOVA; Figure 3b). Subsequently, glutamate/GABA ratio was found to be increased in Tau A2A+/+ as compared to WT animals (WT A2A+/+: 7.3±1.2 vs Tau A2A+/+:13.9±2.8, P=0.05, one-way ANOVA) while normalized to control levels in Tau A2A−/− mice (8.4±2.6, P=0.75 vs WT A2A+/+). Hippocampal mRNA levels for the GABA transporters (GATs) 1, 2 and 3 were significantly increased in Tau A2A+/+ mice, whereas both GAT-2 and GAT-3 expression returned to control levels in Tau A2A−/− animals (Supplementary Figure 5).

Figure 3
figure 3

Adenosine A2A receptor (A2AR) deletion modulates hippocampal levels of glutamate and gamma-amino butyric acid (GABA). Glutamate (a) and GABA (b) concentrations in microdialysates recorded from the dorsal hippocampus of anaesthetized mice. The histograms represent the mean±s.e.m. of three consecutive 5-min microdialysates following a 2-h post-implantation equilibration period (*P<0.05 using two-way analysis of variance (ANOVA); ##P<0.01 in Tau A2A−/− vs Tau A2A+/+, N=6–9 per group, one-way ANOVA).

PowerPoint slide

A2AR deletion rescues LTD deficits in THY-Tau22 transgenic mice

We previously demonstrated that impaired hippocampal-dependent learning and memory in THY-Tau22 transgenic mice was associated with attenuated long-term depression.26 Slices from WT animals developed a robust and similar LTD regardless of A2AR expression (WT A2A−/− vs WT A2A+/+, P>0.05; two-way ANOVA; Figure 4a). As expected, Tau A2A+/+ mice exhibited an impaired LTD maintenance as compared to WT A2A+/+ animals (P<0.001 vs WT A2A+/+, two-way ANOVA; Figure 4a and b), with a significant reduction, reaching 31.6±3.5% of controls (P=0.008 vs WT A2A+/+, one-way ANOVA; Figure 4b). In contrast, LTD remained unaffected in Tau A2A−/− mice (P<0.001 vs Tau A2A+/+, two-way ANOVA; Figure 4a), being similar to WT animals (P=0.82, Figure 4b).

Figure 4
figure 4

Adenosine A2A receptor (A2AR) deletion rescues long-term depression (LTD) deficits in THY-Tau22 transgenic mice. (a) Plots of the field excitatory postsynaptic potential (fEPSP) slope variation over time after induction of LTD in hippocampal slices. While Tau A2A+/+ mice displayed an impaired LTD as compared to WT (A2A+/+ or A2A−/−) animals, slices from Tau A2A−/− animals exhibited LTD similar to control animals, suggesting that A2A deletion rescues LTD in Tau mice. (b) Quantification of the percentage of variation of the fEPSP slope during the LTD maintenance phase, 60 min after induction (**P<0.01 in Tau A2A+/+ vs WT A2A+/+; ##P<0.01 in Tau A2A−/− vs Tau A2A+/+; N=4–7 per group, one-way analysis of variance).

PowerPoint slide

Beneficial effects of A2AR-based therapy using the selective antagonist MSX-3 in Tau transgenic mice

We explored the effect of MSX-3, a water-soluble prodrug of MSX-2, which is a potent and highly selective A2AR antagonist.19 THY-Tau22 mice received MSX-3 through drinking water (0.3 g l−1) starting at 6 until 8.5–9 months of age. At completion of the experiment, body weight gain was not different between water- and MSX3-treated Tau animals (Tau/H2O: 120.1±1.4% of their initial body weight; Tau/MSX-3: 122.9±1.1% of their initial body weight; P=0.13 vs water consumers). Body temperature was also similar in all experimental groups (not shown). Mice consumed on average 4.7±0.3 ml of the MSX-3 solution every day, corresponding to an average daily intake of 1.4 mg of the antagonist.

Spatial memory was evaluated using the Y-maze task. During the exposure phase, all groups spent a similar amount of time in the two available arms (P>0.05; Figure 5a). During the test phase, the discrimination index was significantly impaired in Tau mice treated with water (P=0.0022 in Tau/H2O vs WT/H2O, one-way ANOVA; Figure 5c). Tau/MSX-3 animals exhibited an index higher than that of Tau/H2O mice (P=0.05 in Tau/MSX-3 vs Tau/H2O, one-way ANOVA; Figure 5c) and similar to control animals (Tau/MSX-3 vs WT/H2O, one-way ANOVA; P=0.36), indicative of an improved spatial memory.

Figure 5
figure 5

The specific A2A receptor antagonist MSX-3 improves spatial memory and reduces Tau hyperphosphorylation in THY-Tau22 mice. MSX-3 treatment started at 6 months, an age at which Tau pathology and memory impairments are already significant (see Figures 1 and 2) until 8.5–9 months. (a–c) Effect of MSX-3 on the spatial memory of THY-Tau22 mice using Y-maze task. During the exposure phase (a), all groups explored the maze equally, spending a similar amount of time in each available arm (P>0.05, one-way analysis of variance (ANOVA)). All genotypes exhibited a comparable velocity in the maze (b; P>0.05, one-way ANOVA). During the test phase (c), while hippocampus-dependent spontaneous spatial novelty preference was impaired in Tau/H2O mice (*P<0.05 vs WT, one-way ANOVA), Tau mice treated with MSX-3 exhibited a normalized preference for the ‘novel’ arm (#P<0.001 vs Tau/H2O, one-way ANOVA; N=9–15 per group). (d) Representative two-dimensional profile of total human Tau protein in Tau/H2O (top) vs Tau/MSX-3 (bottom) mice, showing a reduced amount of Tau acidic species in Tau animals treated with the adenosine A2A receptor antagonist (arrow; Total Tau C-ter antibody). (e) Tau phosphorylation on Thr212/Ser214, Ser262, Thr181, Ser404 and Ser214 epitopes, as well as dephosphorylated Tau (Tau1) in Tau/H2O and Tau/MSX-3 mice (*P<0.05 and **P<0.01 vs Tau, N=6–8 per group, Student’s t-test). Results are expressed as a percentage of Tau/H2O ±s.e.m. (f) Representative western blots of sarkosyl-soluble (S) and -insoluble (I) hippocampal Tau species of Tau/H2O (top) and Tau/MSX-3 (bottom) mice, showing no overt difference between the two groups (Total Tau C-ter antibody).

PowerPoint slide

We then investigated whether these memory improvements were related to changes of Tau protein. Using two-dimensional (2D) gel electrophoresis analysis, we observed a reduction of Tau isovariants at the acidic pH range in Tau/MSX-3 mice as compared to Tau/H2O animals, suggesting a reduced Tau phosphorylation (arrow; Figure 5d). Accordingly, we observed a reduction of Tau phosphorylation at several Tau phospho-epitopes using sodium dodecyl sulphate-polyacrylamide gel electrophoresis. As compared to Tau/H2O animals, we found a significantly reduced phosphorylation at Ser262, Ser404 and Ser214 epitopes in Tau/MSX-3 vs Tau/H2O animals (Figure 5e). Phosphorylation at Thr212/Ser214 (AT100) and Thr181 epitopes as well as Tau1 staining remained unchanged. MSX-3 treatment did not affect total Tau protein levels (Figure 5e) nor Tau proteolytic fragments (not shown). Finally, the MSX-3 effect on Tau aggregation was evaluated using biochemical fractionation. The amount of sarkosyl-insoluble Tau remained similar in Tau/H2O and Tau/MSX-3 animals (Figure 5f).


The present study provides the first evidence that A2AR deletion is sufficient to prevent memory defects, hippocampal plasticity impairments as well as Tau hyperphosphorylation, in a mouse model of AD. Accordingly, our data also demonstrate that treatment with a specific A2AR antagonist, starting from a symptomatic stage, improves spatial memory and reduces Tau hyperphosphorylation in THY-Tau22 mice.

The role of A2ARs in AD is ill-defined. Previous studies demonstrated a significant A2AR upsurge in the brain of AD patients, visible at early Braak stages.37 Consistent with the view that such A2AR dysfunction is detrimental towards hippocampal function,25 epidemiological and experimental studies have shown that the consumption of caffeine, a non-selective A2AR antagonist, reduces AD risk and mitigates AD lesions.1, 2, 3 However, the specific relationship between A2ARs and AD has been overlooked. While recent findings suggested that genetic and pharmacological A2AR blockade mitigates acute Aβ synaptotoxicity in mice,11 the impact of A2AR modulation on the pathophysiological progression of AD hallmarks, including specific lesions, remained totally unknown so far.

In the present study, we demonstrate that A2AR deletion significantly prevents the development of spatial memory deficits in THY-Tau22 transgenic mice, using Y-maze and Morris water maze tasks. A2AR deletion itself does not lead to any learning and memory changes in WT animals. This is consistent with previous studies showing that, unlike working memory, global or forebrain-specific A2AR deletion spares spatial reference memory in mice.38, 39 Overall, our results are in agreement with previous data showing that A2A receptor blockade is pro-cognitive in various conditions leading to memory decline, such as ageing,40 chronic stress,25 β-amyloid acute toxicity11 or traumatic brain injury.41 Importantly, MSX-3 treatment rescued hippocampal-dependent memory in Tau mice even after onset of the pathology. This supports the idea that A2A receptor blockade is efficient not only to prevent but also to revert deficits in our model of AD-like Tau pathology. Accordingly, pharmacological A2A receptor blockade was shown to revert memory deficits in a chronic stress model induced by maternal separation.25

Tau hyperphosphorylation has been associated to cognitive impairments in several dementing disorders29 as well as in various experimental models.17, 42 Obviously, Tau-related memory impairments are dissociated from Tau aggregation.3, 24, 43 Our present results demonstrate that A2AR deletion or blockade by MSX-3 promotes a moderate but significant reduction of Tau hyperphosphorylation as notably seen using 2D electrophoresis. This occurs while Tau aggregation remains unchanged, as supported by unmodified AT100 immunoreactivity and sarkosyl-insoluble Tau. This suggests that the observed memory improvements following A2AR deletion or pharmacological blockade are likely related, at least in part, to changes in Tau phosphorylation status. Interestingly, the effects of A2AR deletion on Tau pathology are distinct from those observed following chronic caffeine intake in the same mouse model.3 Indeed, Tau epitopes found changed in the present study (essentially pSer214) are different from those found modulated by caffeine (AT100, Tau1). Moreover, in contrast to caffeine, A2AR deletion does not influence the amount of Tau proteolytic fragments.3 Taken together, these data support that regulation of Tau by caffeine is likely resulting from a broader range of action than only inhibition of A2AR activity.

Tau phosphorylation is under the control of phosphatases, essentially PP2A, and a number of distinct kinases (>30).29 Here, we found that the deletion of A2AR did not affect the expression, methylation and phosphorylation of PP2A in Tau mice. Rather, our data show that A2AR deletion increases the levels of pSer9 GSK3β phosphorylation, which inversely correlates with its activity, as well as slightly reduces Cdk5 expression in Tau transgenic mice. Notably, these two kinases target Tau epitopes that were modified by A2AR deletion in THY-Tau22 mice.29

We cannot rule out that glial-based phenomena underlie some of the beneficial effects resulting from A2AR blockade. It is known that activation of microglia and pro-inflammatory cytokines promotes Tau hyperphosphorylation32, 44 through kinase activation.45 A2ARs are expressed by astroglial and microglial cells, modulating both their activation as well as their ability to release pro-inflammatory factors or to reuptake glutamate.6, 35, 46 According to this, A2A receptor blockade has been shown to reduce neuro-inflammation in different pathological situations.34, 47, 48 While further studies would be needed to identify the precise underlying mechanisms, our data suggest that A2AR deletion contributes to the reported changes of Tau hyperphosphorylation presumably through a reduction of hippocampal neuroinflammation in THY-Tau22 transgenic mice. Additionally, astrocytes play a key role in regulating adenosine homeostasis and hippocampal plasticity.49, 50 Adenosine metabolic clearance is highly regulated by adenosine kinase, a phosphotransferase predominantly expressed by the astrocytes of the adult brain.51 Interestingly, astrocytic adenosine kinase upsurge following astrogliosis has been associated with cognitive impairments including hippocampal-dependent memory.52, 53 The fact that A2ARs play a role in astrocyte activation6 and that Tau A2A−/− mice exhibit reduced astrogliosis further supports the hypothesis of an astrocyte contribution to the A2AR-mediated THY-Tau22 deficits.

In addition to pathological markers, we evaluated the impact of A2AR deletion on neurochemical and synaptic hippocampal indexes. A2ARs have been shown to modulate neurotransmitter dynamics by regulating both synaptic release and glial uptake.6, 35, 36 Regardless of A2AR genotype, we found that THY-Tau22 mice exhibited a limited but significant increase in hippocampal glutamate levels. These observations are in accordance with the somewhat increased vulnerability of these Tau mice when challenged with pentylenetetrazol—a GABAA receptor antagonist—as well as the abnormal synaptic response of THY-Tau22 slices to bicucullin, another GABAA receptor antagonist (unpublished data). A2ARs regulate presynaptic glutamate release as well as glutamate re-uptake by astrocytes.6, 35 Also, A2A receptor blockade in astrocytes is able to counteract β-amyloid toxicity by increasing glutamate uptake.54 However, we could not observe any changes in glutamate levels following A2A receptor deletion. Rather, we observed that A2AR deletion led to a significant increase in hippocampal GABA levels in Tau mice. While the differential impact of A2AR deletion on GABA levels in WT and Tau mice remains puzzling, these observations are in line with data reporting regulation of neuronal and glial GABA uptake by A2ARs.36, 55 Accordingly, we observed a significant decrease of GAT-2 and GAT-3 expression levels in the hippocampus of Tau A2A−/− vs Tau A2A+/+ animals. Overall, changes in GABA levels resulted in a normalization of the glutamate/GABA ratio. Previous studies suggested that normalization of the latter could be associated to neuroprotection as well as to memory improvements in various neuropathological models.56, 57, 58

Finally, behavioural improvements observed in Tau A2A−/− animals were found to be associated with a rescue of hippocampal long-term depression (LTD). As we previously reported in older mice, behavioural deficits seen in THY-Tau22 mice are associated with an impaired LTD,26 in the absence of neuronal death and synaptic shrinkage.17 The combined observation of LTD normalization and memory improvement in Tau A2A−/− mice argues for an instrumental role of LTD defect towards cognitive dysfunction in this mouse model. This is in accordance with previous reports linking hippocampal LTD to memory abilities.59, 60 Interestingly, similarly to what has been observed using a specific antagonist,61 A2AR knockout itself did not impact LTD, in our experimental conditions, in line with the preserved spatial memory of WT A2A−/− animals. The mechanisms underlying LTD normalization in Tau A2A−/− mice remain to be determined. Despite moderate changes, reduced Tau hyperphosphorylation and neuro-inflammation might contribute to the rescue of LTD deficit.62 Another explanation may rely on GSK3β activity changes. A recent systematic evaluation of the role of kinases in hippocampal LTD revealed that among 58 kinases investigated only GSK3β inhibition was associated with LTD impairments.63 At first sight, these data are at odds with our observations. However, we demonstrated that, conversely to WT animals, GSK3β inhibition leads to LTD normalization in slices from THY-Tau22 mice (Ahmed et al.64). Reduced GSK3β activation, besides impacting Tau phosphorylation, could then play a direct role towards the synaptic improvement seen in Tau A2A−/− animals. Finally, the ability of BDNF to impair LTD has recently been demonstrated.61 The BDNF upsurge described in the hippocampus of THY-Tau22 mice14 could thus play a role in the LTD impairments seen in this strain. Interestingly, BDNF’s effects on LTD were found to be dependent on A2AR activation.61 It remains thus possible that A2AR deletion in THY-Tau22 mice could gate these BDNF-dependent effects, resulting in LTD normalization. This will deserve further studies.

In conclusion, we have shown for the first time that both A2AR gene deletion and pharmacological blockade are effective in a model of AD-like Tau pathology. Therefore, the present findings highlight A2AR targeting as a promising therapeutic strategy in AD and Tauopathies.