Modification of astrocytic Cx43 hemichannel activity in animal models of AD: modulation by adenosine A2A receptors

Increasing evidence implicates astrocytic dysfunction in Alzheimer’s disease (AD), a neurodegenerative disorder characterised by progressive cognitive loss. The accumulation of amyloid-β (Aβ) plaques is a histopathological hallmark of AD and associated with increased astrocyte reactivity. In APP/PS1 mice modelling established AD (9 months), we now show an altered astrocytic morphology and enhanced activity of astrocytic hemichannels, mainly composed by connexin 43 (Cx43). Hemichannel activity in hippocampal astrocytes is also increased in two models of early AD: (1) mice with intracerebroventricular (icv) administration of Aβ1-42, and (2) hippocampal slices superfused with Aβ1-42 peptides. In hippocampal gliosomes of APP/PS1 mice, Cx43 levels were increased, whereas mice administered icv with Aβ1-42 only displayed increased Cx43 phosphorylation levels. This suggests that hemichannel activity might be differentially modulated throughout AD progression. Additionally, we tested if adenosine A2A receptor (A2AR) blockade reversed alterations of astrocytic hemichannel activity and found that the pharmacological blockade or genetic silencing (global and astrocytic) of A2AR prevented Aβ-induced hemichannel dysregulation in hippocampal slices, although A2AR genetic silencing increased the activity of astroglial hemichannels in control conditions. In primary cultures of astrocytes, A2AR-related protective effect was shown to occur through a protein kinase C (PKC) pathway. Our results indicate that the dysfunction of hemichannel activity in hippocampal astrocytes is an early event in AD, which is modulated by A2AR. Supplementary Information The online version contains supplementary material available at 10.1007/s00018-023-04983-6.


Introduction
Alzheimer's disease (AD) is a neurodegenerative disorder characterised by a progressive decline of cognitive functions, namely learning and memory, which is linked with an abnormal accumulation of amyloid-β peptides (Aβ).Although the presence of Aβ plaques is a major histopathological hallmark of established AD, it is considered that the extracellular accumulation of soluble Aβ oligomers is a causative agent of neurodegeneration, mainly of synaptic deterioration that is usually associated with the first signs of memory impairments preceding the formation of Aβ deposits that are characteristic of early AD (reviewed in [1]).The hippocampus is particularly affected in early stages of AD, undergoing structural and functional changes typified by alterations of synaptic plasticity that are considered the neurophysiological basis of learning and memory encoding (reviewed in [2]).Astrocytes are glial cells with multiple processes that establish contact with 60-70% of hippocampal synapses and regulate synaptic plasticity, mainly through their capacity to release gliotransmitters, such as ATP and glutamate, and to uptake glutamate and GABA from the synaptic cleft [3][4][5].Accordingly, increasing evidence supports a role of astrocytes in AD onset and progression, as heralded by several morphological and molecular changes of astrocytes in AD mouse models (e.g.APP/PS1) and in human patients [6,7] as well as by the link between Aβ plaques formation and changes in astrocytes morphology and activity [8,9].Astrocytic responses depend on the functions of connexins (Cx), which are proteins forming gap junction channels and hemichannels that allow inter-cellular fluxes of ions and small molecules, such as IP 3 , ATP, glutamate, and energy metabolites (reviewed in [10,11]).Astrocytic hemichannels are mainly composed of hexamers of Cx43 and Cx30, which accumulate in astrocyte processes allowing the release of gliotransmitter and ion fluxes that can regulate hippocampal synaptic transmission and plasticity, and consequently memory [12,13].Astrocytic Cx43 levels are upregulated in animal models of AD, namely in transgenic APP/PS1 and 5xFAD mice [14][15][16].Moreover, there are evidences of dysfunctional astrocytic hemichannels in animal models of AD that do not exhibit alterations on gap junction astrocytic communication [17][18][19].Furthermore, in adult APP/PS1 mice with 8-18 months of age, it was reported that activated microglial cells did not contribute to the elevated connexin immunoreactivity that was concentrated in astroglial processes infiltrating the amyloid plaques [14], implying that the AD-related alteration of connexins hemichannels is intrinsic to astrocytes.
Our group previously showed that cultured astrocytes exposed to Aβ peptides display an enhancement of Cx43 hemichannels and that adenosine A 2A receptors (A 2A R), which are closely associated with Cx43, regulate the levels and activity of hemichannels [20].A 2A R modulate key astrocytic functions that are affected by Aβ exposure, such as glutamate uptake [21], ATP release and hemichannel activity [20], and Ca 2+ signalling [22], and we recently showed that the genetic deletion of astrocytic A 2A R in the hippocampus of adult mice impairs synaptic plasticity and cause memory deficits [23].Moreover, several studies support the idea that A 2A R are a promising target to manage AD, since the pharmacological or genetic blockade of A 2A R confers neuroprotection [24][25][26][27][28].Likewise, as occurs for Cx43, astrocytic A 2A R are increased in the hippocampus of mice exposed intra-cerebro-ventricularly to Aβ [21], mice expressing human amyloid precursor protein [29], 3xTg-AD mice [28] and APP/PS1 mice [26] as well as in AD patients [29].However, it remains to be clarified if the activity of Cx43 hemichannels is modified in early AD and whether A 2A R regulate the activity of hemichannels in hippocampal astrocytes.Thus, in the present study, we resorted to in vivo and ex vivo models of early AD to investigate alterations in astrocytic hemichannel activity in hippocampal slices and tested if A 2A R modulate Cx43 hemichannel activity under physiological and AD-like conditions.

Animals and surgeries
Animals were kept with food and water available ad libitum under a controlled environment (23 ± 2 °C, 50 ± 10% relative humidity and 12 h light/dark cycle).Mice were handled following the "3R" principles, and all experiments were approved by the Ethical Committee of the Center for Neuroscience and Cell Biology of the University of Coimbra (ORBEA_300_2021/24092021) and certified by Direção Geral de Alimentação e Veterinária (DGAV; Portuguese National Authority for Animal Health and Well-Being, 0421/000/000/2021).

Morris water maze
The Morris water maze test was performed as previously described [32] in order to evaluate hippocampal-dependent spatial learning and memory.The test was done using a Page 3 of 19 340 105 cm diameter circular pool filled with water opacified with non-toxic white paint.A platform was submerged 1 cm underneath water surface.The acquisition training phase consisted of four swimming trials of 60 s per day with a 20 min inter-trial interval.In each trial, mice were placed in the pool from a different drop location and given 60 s to find the location of the hidden platform.To complete the trial, the animals must remain 10 s on the platform.This stage was completed when the average of escape latency from the four drop points reached 20 s in WT mice.It was then followed by a retention/probe trial in which the platform was removed and mice were placed on the pool from a random drop location and allowed to swim freely for 60 s.The number of crossings of the hidden platform location was recorded with Any-Maze version 4.99 tracking software (Stoelting, Wood Dale, USA, RRID:SCR_014289).The strategies used by mice to find the platform location in the retention trial were analysed offline and classified as hippocampus-dependent (allocentric) or hippocampus-independent (egocentric), as previously described [33].

Ethidium bromide uptake assay and immunolabelling of astrocytes
The activity of hemichannels was carried out with an ethidium bromide (EtBr) assay as previously described [35,36].Briefly, after the pharmacological treatments, hippocampal slices were incubated with EtBr (20 μM; Sigma-Aldrich) in aCSF solution in an oxygenated chamber for 5 min at room temperature (RT) and rinsed with gassed aCSF for 15 min to stop EtBr uptake.Hippocampal slices were then fixed by immersion overnight at 4 °C in a 4% paraformaldehyde (PFA) solution in PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 and 1.9 mM KH 2 PO 4 , pH 7.4) and then rinsed with a glycine solution (0.5 M) for 5 min to block unreacted aldehydes that produce background signals.This step was followed by several washes with PBS.Slices were protected from light throughout the entire procedure and were further processed for GFAP immunolabelling, as previously described [23].Briefly, hippocampal slices were permeabilized for 30 min with 0.1% Triton X-100 in PBS and then with blocking solution (10% horse serum, 0.1% Triton X-100 in PBS) for 2 h.Slices were then incubated for 2 days at 4 °C with the primary antibody, rabbit anti-GFAP (1:1000, Millipore Cat# AB5804, RRID:AB_2109645) diluted in blocking solution.After washing, slices were incubated first with a blocking solution for 30 min and then with a donkey antirabbit 488 (1:1000, Thermo Fisher Scientific Cat# A-21206, RRID:AB_2535792) secondary antibody for 4 h at RT.In the case of hippocampal slices obtained from GFAP-CRE-A 2A R mice, we also used as a primary antibody goat anti-GFP (1:500, Abcam Cat# ab6673, RRID:AB_305643) and anti-goat 488 (1:1000, Molecular Probes Cat# A-11055, RRID:AB_2534102) and an anti-rabbit 647 (1:1000, Molecular Probes Cat# A-21245, RRID:AB_141775) as secondary antibodies.Hippocampal slices were then washed three times with PBS during 10 min, incubated with Hoe-chst33258 (1:2000, Sigma Aldrich) for 30 min at RT to stain the nuclei, washed again three times during 10 min with PBS and mounted onto Ibidi μ-slide 8 well treated using Fluoromount Aqueous Mounting Medium (Sigma-Aldrich).Hippocampal slices were visualised with a LSM 710 confocal inverted microscope (Zeiss, RRID:SCR_018063) with a 40× objective (Plan-Apochromat 40×/1.4Oil DIC M27 objective) and Z-stacks were acquired using Black Zen software (RRID:SCR_018163).
Analysis of EtBr fluorescence signal in astrocytes involved the processing of the images obtained by confocal microscopy through deconvolution using Huygens software (RRID:SCR_014237) to reduce out-of-focus information, thus increasing spatial resolution through a classic maximum likelihood estimation (CMLE) algorithm.Then, a normalisation operation was conducted in the GFAP channel in order to improve visualisation of GFAP + cells with Imaris software (RRID:SCR_007370), which was used in the subsequent processing stages and analysis.The next step was a nuclei segmentation through creation of surfaces in the nuclei channel.Astrocyte nuclei were manually selected according to the following criteria: (i) cells positive for GFAP which comprehend a single nucleus enwrapped by a GFAP immuno-labelled structure, (ii) nucleus not truncated and (iii) nucleus exhibiting EtBr fluorescence above background.An additional criterion was considered when analysing hippocampal slices from GFAP-CRE-A 2A R mice: only GFAP-positive cells with GFP immuno-labelling were evaluated.EtBr uptake was quantified as a ratio between EtBr fluorescence intensity (in arbitrary units, au) and nucleus volume (µm 3 ) for each astrocyte (i.e.GFAP + cell); we then calculated, for each experimental condition, the mean of EtBr uptake normalised (ratio) to the corresponding control condition.It should be mentioned that all GFAP + cells take up EtBr and that we checked for differences in the volume of astrocyte nuclei amongst the different experimental conditions analysed and no statistically significant differences were found (data not shown).

Aβ immunohistochemistry and thioflavin-S staining
APP/PS1 and WT mice were transcardially perfused with ice-cold PBS followed by 4% PFA in PBS.Brains were removed, post-fixed for 24 h in PFA, dehydrated in 30% sucrose solution for 72 h at 4 °C and cryopreserved at −80 °C.Coronal brain sections (30 μm) were obtained using a cryostat (CryoStar NX50, ThermoScientific, RRIDD:SCR_022732) for Aβ immunolabelling and thioflavin-S staining following a previously described protocol with minor modifications [37].First, antigen retrieval was performed through the incubation of air-dried sections in citrate buffer (10 mM sodium citrate tribasic dehydrate, 0.05% Tween 20, pH 6.0) at 95 °C for 15 min.Then, sections were rinsed three times with PBS for 5 min, permeabilized with a 0.2% Tween 20 solution in PBS for 15 min and non-specific binding blocked by incubation of 1% bovine serum albumin (BSA) 0.05% Tween 20 in PBS for 30 min.Sections were incubated overnight in an airtight humidity chamber with the mouse anti-β-amyloid (βA) primary antibody (1:250, Covance Cat# SIG-39320, RRID:AB_662798) in blocking solution.Sections were washed again three times in PBS for 5 min prior to incubation for 2 h with the secondary anti-mouse Alexa594 antibody (1:1000, Molecular Probes Cat# A-21203, RRID:AB_141633).Following three rinses with PBS for 5 min, sections were stained with 1% thioflavin-S solution for 10 min, differentiated in 70% ethanol, rinsed thrice with water, mounted with DAKO mounting medium and cover-slipped.Finally, hippocampal sections were visualised by fluorescence microscopy (Zeiss, Axio Imager Z2 microscope, RRID:SCR_018856) and the pictures were captured using the AxioVision Imaging System (RRID:SCR_002677 version 4.8).

3D reconstruction of astrocytes
The morphology of astrocytes was studied as previously described [23,38,39] by tridimensional reconstruction of astrocytic structures using an open access tool, Simple Neurite Tracer (SNT, RRID:SCR_016566) plugin available in Fiji-ImageJ software.The tridimensional reconstruction of astrocytic processes within the stratum radiatum of the CA1 subregion of the dorsal hippocampus was carried out using Z-stack images as reported above.Astrocytes were selected for the tridimensional reconstruction of arbors of astrocytic processes according with the following criteria: (i) a single nucleus enwrapped by a GFAP-immunolabelled structure, (ii) the main structure of the astrocyte did not present truncated processes and (iii) reconstruction was carried out in the first five astrocytes fulfilling the previously mentioned criteria in each animal of both genotypes.The morphometric analysis of astrocytic arbor complexity was performed by quantifying the number of processes, their total length and the number of intersections with concentric spheres starting at the centre of the soma with intervals of 4 μm (Sholl analysis).

Preparation of gliosomes
Gliosomes were obtained from hippocampal tissue through a discontinuous Percoll gradient as previously described [40,41].Briefly, hippocampal tissue was homogenised in ice-cold isolation sucrose solution (0.25 M sucrose, 10 mM HEPES, pH 7.4 at 4 °C) using a glass-Teflon tissue grinder.Nuclei and debris were removed by centrifugation (1000g, 5 min at 4 °C) and the supernatant was carefully placed on top of the discontinuous gradient composed by 23, 10, 6 and 2% v/v of Percoll in a sucrose solution (0.32 M sucrose, 1 mM EDTA, pH 7.4 at 4 °C), which was stratified by centrifugation (31,000g for 5 min at 4 °C), turning off the centrifuge brake for the last 2000g to avoid a sudden stop.Gliosomes were collected between the 2% and 6% v/v Percoll layers, whereas synaptosomes (purified synapses) were collected in the interface between the 23% and 10% of Percoll layers.Each fraction was washed with isotonic physiological solution 140 mM NaCl, 5 mM KCl, 5 mM NaHCO 3 , 1.2 mM NaH 2 PO 4 , 1 mM MgCl 2 , 10 mM glucose and 10 mM HEPES, pH 7.4 at 4 °C) and further centrifuged (30,000g for 20 min at 4 °C).Pellets were washed again in isotonic physiological solution and centrifuged (22,000g for 20 min at 4 °C), the supernatant was discarded and the pellets solubilised in RIPA lysis buffer supplemented with 1 mM dithiothreitol (Sigma Aldrich), 1 mM phenylmethylsulfonyl fluoride (PMSF, Sigma Aldrich), 0.001% of a protease inhibitor cocktail (CLAP; Sigma Aldrich) and phosphatases inhibitor cocktail phosphoSTOP (Roche).The comparative analysis of gliosomes and synaptosomes by Western blot allowed confirming the enrichment of our gliosomal preparation in proteins found in perisynaptic astrocytic processes, such as Cx43 and glutamate transporters (GLT-1 and GLAST), as can be observed in Figure S1, the gliosomes preparation of adult wild-type mice had an enrichment not only in GFAP levels, but also of Cx43, GLT-1 and GLAST, as compared with synaptosomal preparations that were enriched in synaptophysin, a widely used synaptic marker.

Astrocytic cultures
Primary cultures of astrocytes were prepared following a previously used protocol [20].Briefly, mixed glial cultures were obtained from the cerebral cortex of 2-5 days Wistar rats, grown in astrocyte medium [DMEM, supplemented with 10% foetal bovine serum, penicillin (100 U/mL), streptomycin (100 μg/ mL), HEPES (6 g/L), and sodium bicarbonate (0.84 g/L)] and maintained at 37 °C in a humified 5% CO 2 incubator for 10-15 days until reaching confluency.Microglia cells were removed by mechanical shaking of the mixed cultures in an orbital shaker for 4 h at 200 rpm and astrocytes detached by a mild trypsinization procedure, reseeded at a density of 5 × 10 4 cells/coverslip, and maintained in culture for 2-3 days prior to EtBr uptake.
Following the exposure to these pharmacological agents, the activity of hemichannels was assessed through EtBr uptake as previously described [20].Briefly, astrocytes were exposed for 10 min to EtBr (5 μM)

Statistical analysis
Data are presented as mean ± SEM of the indicated number (n) of animals (in the case of behaviour and Western blot analysis) or hippocampal slices from different animals (for EtBr uptake quantification).In the case of data from tri-dimensional reconstructions of astrocytic processes, n refers to the number of cells reconstructed.Comparisons between experimental groups were performed with two-way ANOVA followed by Tukey's or Sidak's multiple comparisons test.Comparison between two experimental conditions was performed using either a paired or unpaired Student's t test.To identify differences between strategies used in the Morris water maze test, a table of contingency was built and a Fisher's exact test of independence was performed.Differences in astrocytic morphometric analysis between WT and APP/PS1 mice (Sholl analysis) were assessed by multiple Student's t test using the Holm-Sidak method for correction of multiple comparisons.Statistical significance was set for p values < 0.05.All statistical tests were carried out using Graphpad Prism software (version 8.0.1, RRID:SCR_002798).
Next, we checked for the presence of Aβ accumulation and deposition in the hippocampus of APP/PS1 mice. Figure 1D illustrates the accumulation of Aβ in the hippocampus of APP/PS1 mice, and thioflavin-S staining further supports the existence of an accumulation of these peptides, suggesting the presence of amyloid plaques.By contrast, in the hippocampus of WT mice no thioflavin-S staining was observed.Additionally, we investigated whether APP/PS1 mice displayed alterations in astrocytic arbor complexity through tri-dimensional reconstructions of astrocytes in the CA1 hippocampal region.Astrocytes of APP/PS1 mice displayed a significant increase in the total length of processes (APP/PS1: 1651.15 ± 43.18 µm vs WT: 1171.43 ± 54.27 µm, n = 15 astrocytes from three mice, p < 0.0001, t 28 = 6.917,Fig. 1F) and also in the number of processes (APP/PS1: 98.07 ± 3.30 vs WT: 61.47 ± 2.30, n = 15 astrocytes from three mice, p < 0.0001, t 28 = 9.098, Fig. 1G) relatively to hippocampal astrocytes from WT mice.This suggests an increment in morphologic complexity of hippocampal astrocytes in APP/PS1 compared to WT mice.These results were supported by Sholl analysis which unveiled a significant augmentation of the astrocytic arbor complexity of APP/PS1 compared to WT mice at 8 µm and at 16-24 µm from the central soma (8 µm: p = 0.0312, 16 µm: p = 0.0023, 20 µm: p = 0.0035 and 24 µm: p = 0.0033, Fig. 1H).

Astrocytic Cx43 hemichannel activity is increased in the hippocampus of APP/PS1 mice
The activity of astrocytic hemichannels was assessed in hippocampal slices using the ethidium bromide (EtBr) uptake assay.Data showed a significant effect of the genotype in EtBr uptake by hippocampal astrocytes (p = 0.0179, F 1,8 = 8.809) and further analysis with a post hoc multiple comparisons test unveiled a significant increase of EtBr uptake in hippocampal astrocytes of APP/PS1 compared with WT mice (1.00 ± 0.036 for WT vs 1.69 ± 0.087 for APP/PS1, n = 5, p = 0.0004, Fig. 2B).To investigate the contribution of Cx43 to the altered hemichannel activity in APP/PS1 mice, Cx43 hemichannels were selectively blocked with Gap19 [43].Two-way ANOVA showed a significant effect of Gap19 (p = 0.0014, F 1,7 = 26.13)and an interaction between Gap19 and genotype (p = 0.0015, F 1,7 = 25.35).Sidak's multiple comparisons tests disclosed a significant effect of Gap19 on EtBr uptake by hippocampal astrocytes of APP/PS1 mice (1.685 ± 0.087 for CTR vs 1.066 ± 0.163 for Gap19, n = 5, p = 0.0003).Thus, Cx43 hemichannels were major contributors to the EtBr uptake and their blockade rescued the dysfunctional hemichannel activity, restoring astrocytic EtBr uptake to levels similar to WT (Fig. 2B).
Given the relevant role of Cx43 hemichannels in the altered EtBr uptake in APP/PS1 mice, we further evaluated putative alterations of Cx43 levels in gliosomes (membranes Fig. 2 APP/PS1 mice showed an increase of Cx43 hemichannel activity in hippocampal astrocytes and an increase of Cx43 levels.A Representative images of EtBr fluorescence signal (red) in astrocytes in hippocampal slices of WT and of APP/PS1 mice.Astrocytes were immuno-labelled with anti-GFAP (green) and nuclei were stained with Hoechst 33,258 (blue).Scale bars: 20 μm.B EtBr uptake was significantly higher in hippocampal astrocytes of APP/PS1 mice than in WT mice.The selective blockade of Cx43 hemichannels significantly reduced EtBr uptake in astrocytes of APP/PS1 mice, indicating the involvement of these channels in EtBr uptake.Hippocampal slices from APP/PS1 mice were treated with the Cx43 hemichannel inhibitor, Gap19 (250 µM, 90 min), prior to EtBr uptake assay.***p < 0.001, two-way ANOVA followed by post hoc Sidak's mul-tiple comparisons test.The activity of hemichannels was assessed through the fluorescence intensity of EtBr signal, per astrocyte (GFAP + cells) nucleus volume (µm 3 ) from WT or APP/PS1 mice and expressed as a ratio of control condition.Data are mean ± SEM of five independent experiments.C Levels of total Cx43, but not of phospho-Cx43 at Ser 368 (D), were increased in gliosomes (membranes from astrocytic processes) obtained from the hippocampus of APP/PS1 relatively to WT mice.The ratio between immunoreactivities for total Cx43 or for phospho-Cx43 at Ser 368 to α-tubulin levels were expressed as a percentage of values obtained in WT mice.Data are mean ± SEM of 3-4 independent experiments.*p < 0.05, unpaired Student's t test.Representative immunoblots for Cx43, phospho-Cx43 and α-tubulin are shown below the average bar graphs of astrocytic processes) obtained from the hippocampus of APP/PS1 compared to WT mice.Data showed a substantial enhancement of Cx43 levels in gliosomes from APP/PS1 mice when compared with WT mice (100.00 ± 4.68% for WT vs. 251.33± 55.02% for APP/PS1 mice, n = 4, p = 0.0337, t 6 = 2.741, Fig. 2D).Since the activity of hemichannels is modulated by phosphorylation, we also evaluated alterations of Cx43 phosphorylation at Ser368 and no alterations were observed between gliosomes of WT and APP/PS1 mice (100.00 ± 43.85% for WT vs. 87.95± 32.30% for APP/PS1, n = 4, p = 0.8323, t 6 = 0.2212, Fig. 2E).The ratio between pCx43 and total Cx43 was 1.00 ± 0.12 for WT mice and of 0.64 ± 0.12 for APP/PS1 mice, supporting an increase in total Cx43 levels.

Discussion
The present study re-enforces the relevance of astrocytic alterations at the onset of memory deficits and in experimental models of established AD, by showing that the activity of astrocytic hemichannels composed by Cx43 is modified in animal models both of early stages as well as of established AD.Furthermore, the observed different alterations of total Cx43 levels and of phosphorylated Cx43 at Ser368 in different models mimicking different AD stages from in vitro to in vivo suggest that the enhancement in astroglial hemichannel activity might be due to different mechanisms throughout the evolution of AD.
The analysis of astrocyte morphology in the hippocampus of APP/PS1 mice when memory dysfunction became evident, clearly showed global alterations of astrocyte morphology, typified by an increase in arbor complexity and an enhancement in the number and total length of astrocytic processes.This finding is aligned with previous observations that an increase in astrocyte reactivity is linked to the accumulation of Aβ associated with AD, as assessed by the analysis of GFAP immunoreactivity in human samples and animal models of AD [6,27,49,50], namely in APP/PS1 mice [7,14,51,52].Given the active role of astrocytes in synaptic communication and memory processes [38,53,54], these alterations of astrocytes emerge as potential contributors to the deficits in hippocampal-dependent learning and memory observed in the performance of APP/PS1 mice in the Morris water maze test.
Our present study also revealed that the increased reactivity of astrocytes in the AD brain was associated with an increase of astroglial hemichannel activity.In the present study, although we did not use scrambled peptides, we pinpointed the specific contribution of Cx43 to the measured astroglial hemichannel activity by resorting to Gap19, which was previously shown to selectively target Cx43 hemichannels without affecting gap junctions [43].Accordingly, we observed increased total levels of Cx43 in APP/PS1 mice that display scarce amyloid plaques.This is in agreement with previous studies also reporting an increase of Cx43 protein and mRNA levels in the brain of both patients [49,55] and animal models [14,18,56] with established AD.This increased density of total Cx43 was originally associated with sites of Aβ deposits [14,49], but was later recognised to be widespread in the afflicted brain regions of AD [18].It should be mentioned that although we observed an increase in astrocytic process length, number and complexity in APP/PS1 mice, the increase in total Cx43 levels does not seem to be a direct consequence of the increased number of astrocytic processes, because we quantified the levels of total and phosphorylated (at Ser 368) Cx43 in a similar amount of astrocytic processes membranes (gliosomes) of wild-type and APP/PS1 mice.
Interestingly, we now observed that the increased activity of hemichannels in hippocampal astrocytes is an early event of AD pathogenesis, which is maintained in advanced stages of the disease.In early AD modelled by Aβ-icv exposure, there was an increase in Cx43 phosphorylation at Ser368 without alterations of the total levels of Cx43, whereas in hippocampal gliosomes of 9-month-old APP/PS1 mice already displaying sparse Aβ plaques, there was an increase in Cx43 levels without alterations in Cx43 phosphorylation at Ser368.This is in agreement with the ability of soluble forms of Aβ to control Cx43 levels at cell membrane [57], bolstering gliotransmitters release from cultured astrocytes, namely ATP and glutamate [20,56,58], which might promote neuronal injury [58].These findings further indicate that the onset of AD seems to be related with alterations in Cx43 phosphorylation, whereas advanced AD stages involve alterations in Cx43 levels, which is consistent with the progressive increase in Cx43 levels observed in APP/PS1 and 5xFAD mice [59] as well as in AD patients [55].Although we have not directly explored the impact of increased Cx43-HC activity as a trigger of memory dysfunction, this increased of Cx43 might be a contributing factor for AD-related dysfunction of synaptic plasticity and memory since the genetic silencing of Cx43 affords a neuroprotection against Aβ-induced modifications [55] and improves memory deficits in APP/PS1 mice by increasing synaptic function without affecting amyloid plaque formation or the inflammatory response [18,59].
We further explored the mechanism involved in the regulation of hemichannel activity by phosphorylation which remains largely unknown in astrocytes.The observed enhancement in Cx43 phosphorylation in Ser368 residue, which is mainly mediated by PKC [60,61], in hippocampal astrocytes of mice icv administrated with Aβ peptides, is in line with alterations previously reported by us in cultured astrocytes exposed to Aβ peptides [20].These data suggest an Aβ-induced increase in PKC activity, which was already reported in human astrocytes exposed to Aβ [62] as well as in cortical astrocytes of 5xFAD mice [47].Thus, we postulate as a working hypothesis that the increased PKC activation and the subsequent enhancement of Cx43 phosphorylation at Ser368 might be responsible for the increased hemichannel activity under AD conditions in hippocampal astrocytes.Interestingly, a recent study showed that an inflammatory stimulus (Thy-1) caused astrocytic Cx43 phosphorylation at Ser373 through a PI3K/AKT signalling pathway, being this phosphorylation related with the increased opening of hemichannels and ATP release [63].Some studies performed in endothelial and cardiac cells also showed a link between PKC-mediated phosphorylation of Cx43 at Ser368 residue leading to altered hemichannel activity [64] and of gap junctional intercellular communication [48,64,65]; however, little is known about the control of Cx43 phosphorylation at Ser368 in astrocytes [66].
A second major advance provided by the present study is the concept that adenosine A 2A receptors (A 2A R) modulate the impact of Aβ in the activity of astrocytic Cx43 hemichannels.Although, we had previously shown that A 2A R control Cx43 hemichannels in cultured astrocytes exposed to Aβ peptides [20], we now extend our studies to a more complex system, namely mouse hippocampal slices modelling early or advanced stages of AD, in which A 2A R were genetically silenced in astrocytes and/or in neurons.Our group and others showed that A 2A R blockade ameliorates synaptic dysfunction and loss in addition to memory deficits in mice icv administered with Aβ 1-42 [24,31] and in APP/PS1 mice [25,26,67].This A 2A R-mediated control of AD-related dysfunction involves an overfunction of neuronal A 2A R [25,31,67], in accordance with the predominant localization of A 2A R in excitatory synapses in the limbic cortex [68,69].However, A 2A R are also present in astrocytes and the selective manipulation of astrocytic A 2A R has an impact on memory function [23,29,70].Furthermore, A 2A R modulate alterations triggered by Aβ in primary cultures of astrocytes, namely decreased glutamate uptake [21], altered Ca 2+ dynamics [22], and increased hemichannel activity and subsequently ATP release [20], a danger signal in brain disease conditions [71].We now extend these findings by showing that the Aβ-induced alterations in hemichannel activity in hippocampal slices were abrogated by the pharmacological blockade and genetic silencing of A 2A R in astrocytes, whereas they were unaffected in hippocampal astrocytes of mice with a selective genetic silencing of A 2A R in neurons (FbA 2A RKO); this further re-enforces the conclusion that astrocytic A 2A R are responsible for the regulation of Aβ-induced alterations in astrocytic hemichannel activity in the hippocampus.Interestingly, A 2A R also have a relevant role in non-pathological conditions since global, neuronal and astrocytic A 2A R genetic silencing significantly impacted the activity of astrocytic hemichannels in the hippocampus.Indeed, mice lacking astrocytic A 2A R displayed an increased hemichannel activity, which is likely a consequence of an increase of total Cx43 and phospho-Cx43 levels (see also [72]).Thus, astrocytic A 2A R emerge as key modulators of hemichannel activity, which is consistent with the reported physical association between A 2A R and Cx43 in primary cultures of astrocytes [20].This provides a molecular rationale to understand that the removal of A 2A R from the complex with Cx43 can contribute to dysregulate hemichannel activity, which may underlie the observation that the selective silencing of astrocytic A 2A R results in impairments of synaptic plasticity and deficits in hippocampal-dependent reference memory [23].Interestingly, we also observed that the genetic silencing of neuronal A 2A R (using FbA 2A RKO mice) triggered an enhancement of the activity of hippocampal astrocytic hemichannels relatively to WT littermates, despite no alterations in Cx43 or phospho-Cx43 levels were detected, suggesting that alterations in neuronal function due to A 2A R deletion can influence astrocytic activity.This idea is in agreement with a study reporting that, under physiological conditions, the activity of Cx43 hemichannels in astrocytes is promoted by neuronal activity, which, in turn, modulates neuronal network function via a purinergic pathway in the olfactory bulb [73].Altogether our data indicate that astrocytic A 2A R are crucial to modulate Cx43 hemichannel activity in hippocampal astrocytes under AD-like conditions.Furthermore, our data also provides a tentative rationale for a riddle regarding the impact of astrocytic A 2A R on memory performance, whereby the genetic elimination of astrocytic A 2A R is detrimental for memory performance in naïve animals [23,70] but beneficial in AD mouse models [29].This is in agreement with the observed increase of hemichannel activity upon deletion of astrocytic A 2A R in naïve animals and with the lack of additional effects of Aβ on hemichannel activity in hippocampal slices of mice with a genetic deletion of astrocytic A 2A R.
This opposite role of A 2A R in naïve animals and in conditions of early AD is likely due to the upregulation of astrocytic A 2A R [21].Previous studies have shown that the upregulation of A 2A R is coupled with an alteration of the transducing signalling system operated by A 2A R [74].In fact, whereas A 2A R are canonically coupled to the activation of adenylate cyclase and generation of cAMP [46], these pleiotropic receptors seem to be mostly coupled to PKC when they are upregulated in disease conditions associated with increased glutamatergic signalling [74], such as in early AD (see e.g.[75][76][77]).And, in agreement with the impact of PKC on the activity of astrocytic hemichannels, we observed that A 2A R blockade was not able to prevent Aβ-induced alterations of hemichannel activity when PKC was previously activated, which supports the contention that the recruitment of the PKC pathway is a downstream event to A 2A R activation under conditions of Aβ exposure.This is in agreement with several previous reports indicating the involvement of PKC in the signalling of glial A 2A R [74,78,79], as also occurs in neurons [80] and other cell types [81][82][83] under stressful conditions.
In conclusion, the present study reinforces the role of astrocytic dysfunction in AD pathogenesis.In transgenic APP/PS1 mice displaying diffuse Aβ plaques, deficits in hippocampal-dependent memory were accompanied by alterations in astrocytic morphology and by an enhancement of astrocytic Cx43 hemichannel activity in the hippocampus.Similarly, models of early AD also showed an increase in the activity of Cx43 hemichannels involving an increase of Cx43 phosphorylation at Ser368.This increased activity of astrocytic Cx43 hemichannel activity in the hippocampus was regulated by astrocytic A 2A R, as their genetic silencing increased both channels activity and Cx43 and phospho-Cx43 levels.Additionally, astrocytic A 2A R also modulate the impact of Aβ on Cx43 hemichannel activity through a PKC-dependent mechanism (Fig. 8).Overall, these findings re-enforce the contention that astrocytic A 2A R modulate astrocyte-to-neuron communication, which is altered in ADlike conditions.This may contribute to the benefits afforded by A 2A R antagonists in early AD.

Fig. 1
Fig. 1 APP/PS1 mice displayed deficits in hippocampal-dependent memory, an accumulation of Aβ and of Aβ plaques and an increased astrocytic arbor complexity in the hippocampus.A In the Morris water maze test, APP/PS1 mice displayed a higher latency to find the hidden platform location in the acquisition learning curve than WT mice.****p < 0.0001, two-way ANOVA followed by a post hoc Sidak's multiple comparisons test.B APP/PS1 mice showed an impairment in spatial memory since the number of crossings of the platform location was significantly lower than WT mice in probe trial of the Morris water maze test.****p < 0.0001, unpaired Student's t test.C Analysis of the search strategy patterns unveiled that WT mice preferentially used strategies implicating the hippocampus, whereas APP/PS1 mice mostly used strategies not dependent on the hippocampus to discover the hidden platform location in the probe trial of the Morris water maze.****p < 0.0001, Fisher's exact test of

Fig. 3
Fig. 3 Hemichannel activity was also enhanced in astrocytes of models mimicking early AD.The activity of hemichannels was augmented in astrocytes of hippocampal slices obtained from A WT mice and incubated with Aβ 1-42 (50 nM, 60 min) and B from WT mice administrated icv with Aβ 1-42 (icv-Aβ 1-42 ), compared to control (or icv-vehicle administration) WT mice, being the EtBr uptake assay performed after 15 days of Aβ 1-42 -or vehicle-icv administration.The activity of hemichannels was evaluated by the mean fluorescence intensity of EtBr signal per astrocyte nucleus volume and expressed as a ratio of control for each experimental condition.Data are mean ± SEM of 4-5 independent experiments.*p < 0.05, paired for (A) or unpaired for (B) Student's t test

Fig. 4
Fig. 4 Aβ 1-42 -icv administration increased Cx43 phosphorylation at Ser368 in hippocampal gliosomes of mice mimicking early AD stages.Although no alterations were observed in A Cx43 levels, an enhancement in B phospho-Cx43 levels (at Ser368) was observed in gliosomes obtained from the hippocampus of mice icv administrated with Aβ 1-42 in comparison with icv-vehicle (VEH)-injected control mice.Data are ratios between Cx43 or phospho-Cx43 and α-tubulin (loading control protein) immunoreactivities expressed as percentage of the control condition (icv-VEH).Data are mean ± SEM of 4-5 independent experiments.**p < 0.01, unpaired Student's t test.Representative immunoblots for Cx43, phospho-Cx43 and α-tubulin are shown

Fig. 5
Fig. 5 Astrocytic A 2A R genetic silencing increased the activity of hemichannels and the levels of Cx43 and of phospho-Cx43 at Ser368 in hippocampal astrocytes.A Deletion of A 2A R selectively in astrocytes led to the enhancement of EtBr uptake in hippocampal astrocytes.Hemichannel activity was evaluated by the mean fluorescence intensity of EtBr signal per astrocyte nucleus volume expressed as a ratio of control condition (GFAP-CTR) in hippocampal slices from GFAP-CTR and GFAP-CRE-A 2A R mice.The levels of B Cx43 and

Fig. 6 A
Fig. 6 A 2A R regulated alterations in hemichannel activity caused by an acute challenge with Aβ 1-42 .A The enhancement in astrocytic hemichannel activity driven by the acute exposure of hippocampal slices to Aβ 1-42 was prevented by A 2A R selective blockade.Hippocampal slices were pre-incubated with SCH58261 (SCH, 50 nM) for 30 min prior to challenge with Aβ 1-42 (50 nM, 60 min), before the EtBr uptake assay.B In GbA 2A RKO and C GFAP-CRE-A 2A R mice, Aβ 1-42 failed to increase EtBr uptake in hippocampal astrocytes, whereas D in FbA 2A RKO mice, Aβ 1-42 augmented astrocytic EtBr

Fig. 7
Fig. 7 The increased hemichannel activity in cultured astrocytes exposed to Aβ 1-42 was regulated by PKC.A PKC inhibition with GF109203X (GF) prevented the effect of Aβ 1-42 on EtBr uptake in astrocytes.Cultured astrocytes were exposed to GF (5 µM) for 30 min prior to challenge with Aβ 1-42 (1 µM, 24 h).Data are mean ± SEM of 3-4 independent experiments.*p < 0.05, two-way ANOVA post hoc Tukey's multiple comparisons test.B On the other hand, PKC activation with PMA (phorbol 12-myristate 13-acetate, 10 ng/mL) mimicked the effect of Aβ 1-42 on EtBr uptake in astrocytes.In the presence of PMA, the selective A 2A R antagonist SCH58261 (SCH) did not blunt the Aβ 1-42 -induced increase in EtBr uptake by astrocytes.The selective A 2A R antagonist SCH58261 (SCH) was no longer able