Astrocytic upregulation of A 2A R in the mouse CA1.
To determine the impact of astrocytic A2AR upregulation in the mouse hippocampus, we bilaterally injected dedicated adeno-associated viral (AAV) vectors allowing the selective expression of A2AR (AAV-A2A) or GFP (AAV-GFP), taken as a control, in the CA1 astrocytes of 2 months-old C57Bl6/J mice and animals studied 2 months post-injection (Fig. 1A). Immunohistochemistry using antibodies against A2AR or GFP allowed to visualize their hippocampal levels in the different CA1 sublayers i.e. stratum oriens (SO), pyramidal layer (PL) and stratum radiatum (SR) (Fig. 1B; Supplementary Fig. 1). Co-immunostainings using antibodies directed against astrocytic (GFAP, Sox2, S100β), neuronal (NeuN) and microglial (Iba1) markers showed an exclusive expression of A2AR by astrocytes, as expected (Fig. 1C).
Cell-autonomous impact of astrocytic A 2A R upregulation on reactivity and morphological complexity.
Astrocytes can adopt a reactive phenotype which is notably characterized by an increased expression of GFAP, an intermediate filament protein. We found a significant increase of the GFAP+ staining in the CA1 of AAV-A2A mice as compared to the AAV-GFP control (+ 100.2 ± 38.3%; P = 0.0034; Student’s t-test; Figs. 2A-B). Recent studies have also identified the JAK2/STAT3 signaling pathway as a central player in the induction of the astrocytic reactive state in pathological conditions such as Alzheimer's disease and Huntington's disease49. Accordingly, we performed a GFAP/STAT3 co-immunostaining and observed a significant increase of STAT3+ staining within the GFAP+ astrocytes of the CA1 area in the AAV-A2A condition as compared to the AAV-GFP control (+ 68.3 ± 35.9%; P = 0.0024; Student’s t-test; Figs. 2C-D). Moreover, astrocytes continually adapt to their environment and exhibit morphological changes eventually associated with their reactive state50. In order to assess more finely the morphological changes of astrocytes upon A2AR upregulation, a 3D-reconstruction of the latter was performed, following a retro-orbital injection of an AAV-PHP.eB allowing the selective expression of tdTomato in sparse CA1 astrocytes (cytosol and arborization), of both AAV-A2A and AAV-GFP-injected mice (Figs. 2E-F). We thus imaged isolated tdTomato+ astrocytes in the CA1 stratum radiatum sublayer by high-resolution imaging and performed their 3D-reconstruction using Imaris software (Fig. 2G). Our results indicated a significant higher number of intersections, as shown by Sholl analysis, consistent with a higher number of the astrocytic processes in AAV-A2A vs. AAV-GFP mice (Figs. 2H-I). A decreased astrocytic processes length was also observed (Fig. 2J) without change of the overall astrocyte or soma volumes (Fig. 2K-L). Together, these results support that astrocytic A2AR upregulation promotes astrocyte reactivity and complexity.
Aging-like molecular signature induced by astrocytic A 2A R upregulation.
To gain molecular insights on the cell-autonomous impact of astrocytic A2AR upregulation, we performed a RNA sequencing analysis from hippocampal astrocytes of AAV-A2A and AAV-GFP-injected mice. To do so, we used magnetic cell sorting thanks to ACSA-2 (Astrocyte Cell Surface Antigen 239; Supplementary Fig. 2A) magnetic microbeads in order to isolate astrocytes from freshly dissected hippocampi. Following RNA sequencing, and using an available database51, we first validated our enrichment procedure by comparison of the expression of known genes specifically expressed by astrocytes with genes expressed by other cell types (Supplementary Fig. 2B). The first Principal Composant (PC1), which explained 40.24% of the variance, nicely separated our experimental conditions (Fig. 3A). We found 1127 differentially expressed genes (DEG) in the AAV-A2A condition as compared to the AAV-GFP control, 916 being downregulated and 211 upregulated (|Log2 fold-change (FC)|>0.32; Adjusted P-value (Padj) < 0.05; Fig. 3B; list of DEG available on Supplementary Table 3). The top ten of downregulated genes is given on Fig. 3Ci. Functional enrichment analysis provided by DAVID and GSEA analysis indicated that downregulated astrocytic genes in AAV-A2A mice are associated with several metabolic processes such as insulin signaling as well as glucose and glutamate metabolisms. We also observed a significant impact on transcriptional processes (Fig. 3D; Supplementary Fig. 3). Regarding upregulated genes, as expected, we found A2AR (Adora2a). We also observed the upregulation of Timp1 (Log2 FC = 3.19; Padj = 1.59E-07), a gene encoding a metalloproteinase inhibitor, being involved in astrocyte reactivity as well as in neurodegenerative conditions52,53 (Fig. 3Cii). GSEA analysis supported that upregulated genes were associated with inflammatory processes (NFκB, TNFɑ and Il1) as well as oxidative stress and DNA repair pathways (Fig. 3E; Supplementary Fig. 3), all previously linked to cellular processes associated with aging54,55. One of the hallmarks of aging, cellular senescence, has been particularly linked with astrocytes in AD56,57. Interestingly, it has been recently demonstrated that astrocytic senescence associates with a reduction of cdk6 through an impairment of the activity of YAP (Yes1 Associated Transcriptional Regulation)58. Accordingly, our data highlighted significant reductions in the expressions of cdk6 (Log2 FC=-3.03; Padj = 8.85E-31; Fig. 3Ci) and yap1, coding YAP (Log2 FC=-1.12; Padj = 6.93E-05). In order to validate impairment of YAP, instrumental to astrocyte senescence58, we performed co-immunostainings against YAP and GFAP (Fig. 3F). Our data showed a significant reduction in the percentage of YAP+ cells among GFAP+ cells in AAV-A2A vs. AAV-GFP mice (-25.0 ± 11.2%; P = 0.002; Student’s t-test) as well as a reduction of the nuclear YAP staining of GFAP+ astrocytes (-45.0 ± 11.5%; P = 0.0183; Student’s t-test) in the CA1 of AAV-A2A vs. AAV-GFP mice (Fig. 3G). Further, we also found that the volume of astrocyte nuclei was significantly enlarged in YAP+ GFAP+ of A2AR expressing astrocytes as compared to GFP controls (+ 22.8 ± 8.0%; P = 0.0344; Student’s t-test; 6–8 nuclei per animals; N = 4–5/group; not shown), another hallmark of senescence59,60. Together, these data suggested that the upregulation of A2AR in hippocampal astrocytes is sufficient to elicit an aged and senescent-like phenotype. In this sense, we compared our transcriptomic data with the signature of astrocytes isolated from aged mice61 but also to a list of “astrocytic senescence-related genes” extracted from the literature (Fig. 3H; Supplementary Table 4; Supplementary Table 5). We found an overlap with 33 genes (out of 358) varying similarly in the AAV-A2A animals when compared the “aged” signature. Same applied regarding 13 genes (out of 62) belonging the “astrocytic senescence-related genes” signature (Fig. 3I).
Astrocytic upregulation of A 2A R favors neuronal excitability.
Astrocytes regulate glutamate neurotransmission and synaptic plasticity62. Interestingly, the above-mentioned transcriptomic data supported that astrocytic upregulation of A2AR significantly impacts genes involved in glutamate release, in line with previous data reporting that A2AR tightly controls (inhibits) glutamate reuptake by astrocytes5,6. This led us to address the effect of A2AR upregulation on neuro-astroglial communication and particularly markers of neuronal excitability. We first prepared hippocampal synaptosomes from AAV-GFP and AAV-A2A mice to evaluate the phosphorylation of glutamatergic receptor subunits important for the synaptic trafficking and conductance of NMDA and AMPA receptors, focusing on Y1472 of GluN2B and S831 of GluA163,64. While pY1472 GluN2B/GluN2B ratio did not significantly change in AAV-A2A mice, we observed a significant increase (+ 161.4 ± 94.7%; P = 0.003; Student’s t-test) of the synaptosomal pS831 GluA1/GluA1 ratio as compared to AAV-GFP animals (Fig. 4A). Then, we used a DREADD chemogenetic tool to determine whether A2AR upregulation in astrocytes was prone to affect neuronal activation. We jointly expressed the activatory Gq-coupled DREADD receptor (hM3Dq; with an mCherry reporter) in CA1 neurons with A2AR or GFP in CA1 astrocytes (Fig. 4B). A co-immunostaining against A2AR and RFP, in order to amplify and visualize the hM3Dq-mCherry signal, confirmed the astrocytic expression of A2AR (green; Fig. 4C) and the neuronal expression of the hM3Dq (red; Fig. 4C). Two months post-AAV injections, we intraperitoneally injected the exogeneous and synthetic ligand CNO of hM3Dq, or saline as control, to selectively activate hippocampal neurons. Animals were studied 90 min later (Fig. 4B). We evaluated the mRNA expression of immediate early genes (IEGs) as an indicator of neuronal activation by CNO injection. As expected, our data demonstrated a significant hippocampal increase of Dusp1, JunB and c-fos levels in CNO-treated (vs. saline) animals (Figs. 4D-F). Notably, the magnitude of IEG activation was significantly larger in AAV-A2A mice as compared to AAV-GFP animals in CNO-treated condition (Figs. 4D-F). C-Fos immunohistochemistry confirmed qPCR data, as shown on Figs. 4G-H (controls are shown in supplementary Fig. 4). Altogether these data demonstrated that astrocytic changes elicited by A2AR upregulation led to an alteration of neuro-astroglial communication towards an exacerbated neuronal response upon DREADD-mediated activation.
Astrocytic A 2A R upregulation alters microglial phenotype.
Besides the link with neurons and considering that A2AR upregulation activates a pro-inflammatory signature in astrocytes, we further aimed at determining to which extent such astrocyte-autonomous changes may impact microglial cells. We first performed an immunohistochemistry directed against Iba1, a calcium binding protein increased following microglial activation (Fig. 5A). Our results showed a significant increase of the Iba1+ staining in AAV-A2A animals without change in the density of Iba1+ cells (Figs. 5B-C). Like astrocytes, microglial phenotype exhibits a significant heterogeneity and complexity depending on cellular environment65. To capture potential changes of microglial morphology, we performed a 3D-analysis from Iba1 immunofluorescence. The Sholl analysis revealed a slight but significant reduction of the microglial complexity in AAV-A2A animals as compared to AAV-GFP controls (P < 0.001; Two-Way ANOVA; Figs. 5D-E). In addition, we evaluated the level of CD68, a lysosomal marker associated with microglial phagocytosis, essential for eliminating pathogens and misconformed proteins. To do so, co-immunostaining against both Iba1 and CD68, was performed and, after 3D analysis, we observed a decreased level of the CD68+ staining in the Iba1+ cells, suggestive of a reduced microglia phagocytosis in the AAV-A2A animals as compared to the AAV-GFP control (-41.6 ± 32.4%; P = 0.046; Student’s t-test; Figs. 5F-G). Together these data suggested that the upregulation of A2AR in astrocytes is sufficient to affect microglial phenotype.
Astrocytic A 2A R upregulation in CA1 impairs short-term spatial memory and spatial learning.
Finally, to determine whether cell-autonomous and non-cell-autonomous modifications induced by the astrocytic upregulation of A2AR in the hippocampus were prone to alter memory performances, we performed two hippocampus-dependent spatial memory tests following an Elevated-Plus Maze evaluation. Elevated-Plus Maze revealed no difference in either the locomotor activity or anxiety-like behavior (percentage of the time spent in the open arms) in animals overexpressing astrocytic A2AR (Supplementary Figs. 5A-C). Short-term spatial memory performance was assessed using the Y-maze test (Fig. 6A). We first observed no difference regarding the distance moved and the velocity (Supplementary Figs. 5D-E). A discrimination index, corresponding to the animal preference for the new arm versus the familiar arm during the retention phase, was calculated and showed a preference > 50% i.e. vs. chance for both groups (P < 0.0001; One-Sample Student’s t-test), indicating that AAV-GFP and AAV-A2A mice both had a significant preference for the novel arm (Fig. 6B). However, the discrimination index was significantly lower in AAV-A2A animals as compared to AAV-GFP controls, indicating reduced short-term spatial memory (P = 0.03; Fig. 6B). We also evaluated long-term spatial memory performances using the Barnes maze task (Fig. 6C). During the 4 days of the learning, the AAV-A2A group exhibited significantly higher distance (P < 0.001; Two-Way ANOVA), primary latency (P < 0.001; Two-Way ANOVA) and primary errors (P < 0.001; Two-Way ANOVA) to find the goal box as compared to AAV-GFP animals, suggesting altered spatial learning abilities (Figs. 6D-F). During the retention stage, 24 h after the learning, the animals did not show any difference regarding the distance moved or the velocity (Supplementary Figs. 5F-G). The percentage of time spent in the target quadrant was significantly higher than 25% i.e. than chance (P = 0.021 in AAV-GFP animals; P < 0.0001 in AAV-A2A animals; Student’s t-test) and similar for both groups, indicating intact spatial memory for both groups (Fig. 6G). Overall, these data highlighted that the multicellular alterations induced by the hippocampal upregulation of A2AR in astrocytes ultimately lead to impairments of the short-term spatial memory and learning abilities in mice.