TRPA1 channels promote astrocytic Ca2+ hyperactivity and synaptic dysfunction mediated by oligomeric forms of amyloid-β peptide

Background Excessive synaptic loss is thought to be one of the earliest events in Alzheimer’s disease (AD). However, the key mechanisms that maintain plasticity of synapses during adulthood or initiate synapse dysfunction in AD remain unknown. Recent studies suggest that astrocytes contribute to functional changes observed during synaptic plasticity and play a major role in synaptic dysfunction but astrocytes behavior and involvement in early phases of AD remained largely undefined. Methods We measure astrocytic calcium activity in mouse CA1 hippocampus stratum radiatum in both the global astrocytic population and at a single cell level, focusing in the highly compartmentalized astrocytic arbor. Concurrently, we measure excitatory post-synaptic currents in nearby pyramidal neurons. Results We find that application of soluble Aβ oligomers (Aβo) induced fast and widespread calcium hyperactivity in the astrocytic population and in the microdomains of the astrocyte arbor. We show that astrocyte hyperactivity is independent of neuronal activity and is repaired by transient receptor potential A1 (TRPA1) channels blockade. In return, this TRPA1 channels-dependent hyperactivity influences neighboring CA1 neurons triggering an increase in glutamatergic spontaneous activity. Interestingly, in an AD mouse model (APP/PS1–21 mouse), astrocyte calcium hyperactivity equally takes place at the beginning of Aβ production, depends on TRPA1 channels and is linked to CA1 neurons hyperactivity. Conclusions Our experiments demonstrate that astrocytes contribute to early Aβo toxicity exhibiting a global and local Ca2+ hyperactivity that involves TRPA1 channels and is related to neuronal hyperactivity. Together, our data suggest that astrocyte is a frontline target of Aβo and highlight a novel mechanism for the understanding of early synaptic dysregulation induced by soluble Aβo species. Electronic supplementary material The online version of this article (doi:10.1186/s13024-017-0194-8) contains supplementary material, which is available to authorized users.


Background
Formation of toxic amyloid-β (Aβ) species and their accumulation in plaques are key hallmarks in the pathogenesis of Alzheimer's disease (AD) but it has now been recognized that, far ahead of plaques formation, soluble Aβ oligomers (Aβo) are the pathology-triggering species [1]. More specifically, Aβo manage a progressive loss of synaptic connectivity leading to neurodegeneration. Astrocytes are important safeguards of synaptic function and it becomes increasingly evident that these cells accomplish a more important role in brain function than previously thought. The loss of synapses may reflect functional downfall of astrocytes. These cells possess receptors and signaling machinery for all known neurotransmitters thus sensing neuronal activity [2]. They also actively secrete gliotransmitters such as ATP, glutamate, D-serine hence modulating activity of neuronal receptors [3,4]. Consequently, through their involvement in the tripartite synapse, they both sense and modulate synaptic output [5].
Unlike neurons, astrocytes are electrically non-excitable cells but they are equipped with numerous channels, receptors or exchangers triggering Ca 2+ signals. Thus, astrocytic excitability is based on highly spatiotemporally coordinated fluctuations of intracellular Ca 2+ concentration relying on plasmalemmal and intracellular channels [6]. Important progress has been made in studying astrocytes branches Ca 2+ signaling since they are the primary sites for interactions with neurons [7][8][9]. Direct monitoring of Ca 2+ dynamics in the processes of adult mouse hippocampal astrocytes has revealed intense local and compartmentalized activity that is dissociated from activity in the cell body [8]. However, until recently, it was difficult to specifically explore this astrocytic calcium activity since channels and receptors involved in astrocytic calcium signaling were commonly expressed in neurons. However, among the receptors involved in astrocytic calcium signaling, transient receptor potential A1 (TRPA1) channel seems to be specifically expressed in astrocytes and absent from neurons within hippocampus stratum radiatum [10,11]. The discovery of TRPA1 channel as an important mediator of Ca 2+ entry restricted to astrocytes in the mouse hippocampus provided new opportunity to explore astrocyte signaling in relation to neuron-astrocyte interaction particularly in case of synaptic dysregulation.
In late transgenic AD mouse models, i.e. in phases associated with plaques formation, it has been shown that astrocytic calcium activity dramatically increases, becomes synchronous nearby Aβ plaques and coordinates calcium signals at long distance [12]. Within somatosensory cortex, this astrocyte hyperactivity around plaques is mediated by purinergic signaling [13]. Surprisingly, astrocytes behavior and reactions in early phases of AD remained largely undefined despite their probable involvement in the progressive loss of synaptic connectivity and in the complex and critical cellular phase of AD [14]. We thus chose to study the impact of soluble Aβ oligomers on astrocytic function at the onset of early defensive cellular phase well before astrogliosis or inflammatory processes.
In this work, we monitored astrocytic calcium activity within the CA1 stratum radiatum region of the mouse hippocampus both at the astrocytic population level and at a single cell level, focusing in the astrocytic arbor. We characterized spontaneous Ca 2+ signaling properties at these two related levels and showed that Aβo exposition induced at once a global and a local Ca 2+ hyperactivity. We showed that this hyperactivity was independent of neuronal activity and was totally restored to physiological level by blocking TRPA1 channels. This TRPA1 channels-dependent influence of Aβo on astrocyte activity consequently impacted neighboring CA1 neurons, increasing glutamatergic spontaneous activity. In an AD mouse model, we showed that astrocyte hyperactivity was an early phenomenon setting up at the onset of Aβ production, being also related to neuronal hyperactivity and preceding TRPA1 channel overexpression. Overall, our data provide a novel mechanism for the understanding of early toxicity of soluble Aβo species.
Slices bulk loading with Fluo-4 AM.
Single-astrocyte dye loading with Fluo-4 Coronal 300 μm slices were transferred to a chamber allowing constant perfusion with ACSF at room temperature, bubbled with 95% O 2 and 5% CO 2 on the stage of an upright compound microscope (Eclipse E600 FN, Nikon, Paris, France) equipped with a water immersion 60× objective (NA 1.0) and an infrared-differential interference contrast optics with CCD camera (Optronis VX45, Kehl, Germany). Glass pipettes 8-11 MΩ (Harvard apparatus) were filled with intracellular solution containing (in mM): 105 K-gluconate, 30 KCl, 10 phosphocreatine, 10 HEPES, 4 ATP-Mg, 0.3 GTP-Tris, 0.2 Fluo-4 pentapotassium salt (Life Technologies), adjusted to pH 7.2 with KOH. Signals were amplified by Axopatch 200B, sampled by a Digidata 1440A interface and recorded with pClamp8 software (Molecular Devices, Foster City, USA). Astrocytes were identified based on morphological, localization in the stratum radiatum and negative resting potential (between −70 and −80 mV). Input resistance was calculated by measuring current in response to a 10 mV pulse with 80 ms duration, near the end of the voltage command. Only passive astrocytes showing linear I/V relationship and low input resistance (~50 MΩ) were kept for dye loading. After achieving whole-cell configuration, access resistance was constantly monitored, and astrocytes were excluded from this study when this parameter varied >20% throughout the experiment. To allow sufficient diffusion of the dye and avoid astrocyte dialysis, the time in whole-cell configuration was limited to less than 5 min. Then, the patch pipette was carefully withdrawn to allow the astrocyte to recover. In order to maximize the diffusion of the dye into the astrocytic processes, we waited at least 15 min before calcium imaging [8,18].

Calcium imaging
Bulk or single-astrocyte loaded slices were placed in a constantly perfused chamber on the stage of an upright compound microscope (Eclipse E600 FN, Nikon, Paris, France) equipped with a water immersion 40× (NA 0.8) or 60× (NA 1.0) objective and a confocal head (confocal C1 head, Nikon, Paris, France). Excitation was achieved with light at 488 nm and emission was filtered with a 515 ± 15 nm filter. Images were acquired with EZ-C1 software (Nikon, Paris, France) at 1.2 s intervals in a single confocal plane over a period of 5 min.

Bulk loading calcium imaging data analysis
Prior to analysis, raw images were stabilized (when needed if slight x-y drift occurred during recordings, z drifts were excluded) using ImageJ plugin Template Matching. CalSignal software was used to measured intracellular Ca 2+ activity, analyzing the fluorescence signal F within each region of interest (ROI) corresponding to the cell body area of each astrocyte [19]. F 0 was calculated for each ROI on the recording period. Based on the ΔF/F 0 ratios, significant fluorescence variations were detected and a Ca 2+ event was defined as a significant and continuous signal increase larger than a fixed threshold followed by a significant and continuous signal decrease larger than the same threshold. Thus, cells were defined as active when fluorescence increased ≥2 standard deviations relative to baseline fluorescence. After peak detection, each Ca 2+ transients were visually checked by the operator. The theoretical Poisson distribution was calculated by the method of least squares approximating λ until it most closely fits the observed frequency distribution.

Single astrocyte loading data analysis
Ca 2+ transients were measured in two-dimensional images, in individual subregions matching the shape of the astrocyte structure. Manually selected ROIs (~1 μm 2 ) were placed along astrocytic processes lying in the focal plane [8] and a ROI was also selected in the soma if accessible. Prior to analysis, raw images were stabilized (when needed if slight x-y drift occurred during recordings, z drifts were excluded) using ImageJ plugins Template Matching and filtered with 3D Hybrid Median Filter [7]. CalSignal software was used to measure intracellular Ca 2+ activity, analyzing the fluorescence signal F within each ROI. As described above, significant changes in fluorescence were detected on the basis of the calculated ΔF/F 0 ratios. Each Ca 2+ transients within ROI were visually checked by the operator and reported in a raster plot in order to discriminate focal activities from expanded ones (as described by [8]). At the end of each recording, z-stacks (0.5 μm steps) were performed to obtain tri-dimensional projections of astrocyte territories revealed by Fluo-4 loading. Images were then filtered with 3D Hybrid Median Filter plugin in ImageJ.

Electrophysiological recordings
Whole-cell recordings were made from the somata of visually identified CA1 pyramidal neurons. Patch pipettes (4-6 MΩ) were filled with an internal solution containing (in mM): 105 K-gluconate, 30 KCl, 10 phosphocreatine, 10 HEPES, 4 ATP-Mg, 0.3 GTP-Tris, 0.3 EGTA, adjusted to pH 7.2 with KOH. Spontaneous excitatory post-synaptic currents (sEPSCs) were collected at a membrane holding potential of −65 mV which is close to the reverse potential of GABA. All recordings were done at room temperature (22-24°C) and only a single neuron was studied per slice. sEPSCs and their kinetics were analyzed in 5-min epochs within the time frame of the recordings. Each epoch was compared to the initial 5-min recording and sEPSCs frequencies were normalized to this initial value. Access resistance was constantly monitored and recordings were excluded from this study when this parameter varied >20% throughout the experiment. Recordings were analyzed using the Clampfit module of the pClamp8 software (Molecular Devices, Foster City, USA) with a threshold at −20 pA to exclude miniature EPSCs.
When appropriate, the Aβ monomer is purified on a C18 column (SPE-Chromabond-HRX C18 ec, 200 μl, 5 mg, Macherey-Nagel, France). The column was equilibrated with 0.1% trifluoroacetic acid (TFA) in water. Immediately after dilution in DMSO, the Aβ sample was loaded and the column was washed three times with 0.1% TFA. Then, a gradient of acetonitrile from 30 to 60% was applied (Additional file 1). Fractions (0.1 ml) were collected. The elution profile was determined by measuring the absorbance at 275 nm. The peak fraction was collected and the concentration of peptide was determined by absorbance at 275 nm using ɛ 275 nm = 1400 M −1 cm −1 . The peptide is then stored at −80°C.
When appropriate, amyloid-β deposits were stained using Thioflavine S [21]. Sections were re-hydrated in TBS buffer (0.1 M Tris Base, 0.15 M NaCl), incubated in filtered 1% aqueous Thioflavine S (Sigma, France) for 8 min at room temperature, in the dark and washed several times in TBS buffer.

Image acquisition
Sections were examined with a Zeiss LSM 710 confocal laser scanning microscope with a Plan Apochromat 20× objective (NA 0.8) or an oil immersion Plan Neofluor 40× objective (NA 1.3) and translating platform with motorized crossed roller stages. When appropriate, mosaics were acquired for each channel separately with "Zen" software, in a 12-bit format, using the tile scan function. For TRPA1 and GFAP co-staining, sections were also acquired with a Zeiss Airyscan module with an oil immersion Plan Apochromat 63× objective (NA 1.46) to improve lateral resolution (~140 nm) and signal-tonoise ratios. For illustration, images were merged with ImageJ software.

Immunoblotting
Dissected hippocampi from 1-month old APP/PS1-21 mice were homogenized in cold buffer containing 0.32 M sucrose and 10 mM HEPES, pH 7.4. Samples were maintained at 4°C during all steps of the experiments. Homogenates were cleared at 1000 x g for 10 min to remove nuclei and large debris. Samples in loading buffer were boiled for 10 min and equal amounts of proteins (20 μg, quantified by micro-BCA assay (Pierce) in duplicate extracts) were resolved on a 4-20% gradient Bis-Tris polyacrylamide precast stain free gels (Bio-Rad) in denaturing conditions. Proteins were transferred to a polyvinylidene difluoride membrane (Millipore) for 30 min at 4°C. Membranes were blocked with 3% dry milk in Tris-Buffered Saline (TBS: 10 mM Tris, 150 mM NaCl, pH 7.4) containing 0.1% Tween for 1 h at room temperature. Membranes were probed with anti-TRPA1 antibody (Novus, USA; 1:2000) and anti-GFAP antibody (Dako, USA, rabbit polyclonal; 1:100000) diluted in 3% dry milk in 0.1% Tween TBS overnight at 4°C. Membranes were washed in 0.2% Tween TBS and probed with HRP-conjugated anti-rabbit IgG (Fab') (Interchim, France; 1:40,000) antibody for 45 min at room temperature. After washes, specific proteins were visualized with an enhanced chemiluminescence ECL Detection System (Bio-Rad) and the chemidoc system (Bio-Rad). Chemiluminescence signals were normalized to protein loading signals acquired using Stain-free pre-cast gels (Bio-Rad).

Statistical analysis
Data were analyzed using R (the R Project for Statistical Computing) [22]. Comparisons between two groups were conducted with the two-tailed Mann-Whitney test. Kruskal-Wallis test followed by Pairwise comparison using Wilcoxon rank sum test was used when needed for multiple comparisons. Proportions of hyperactive/active astrocyte and focal/expanded activities were compared with χ 2 -test. Data were expressed as mean ± SEM accompanied by distribution of experimental points. Graphic significance levels were *, p < 0.05; **, p < 0.01 and ***, p < 0.001.

Results
Ca 2+ activity in the astrocytic population and individual processes in mouse hippocampus Ca 2+ signals encoding is known to be different in astrocytic cell body versus processes and involves different calcium sources such as internal stores release or external entry. Spatio-temporal Ca 2+ activity characteristics within astrocyte correlate to specific function associated with different territories. Astrocytes not only operate as individual cells but also take part in functional network through gap-junction coupling allowing remote communication in delimited functional brain area. We studied astrocytic Ca 2+ activity at once in the global astrocytic population and in single cell microdomains on mouse hippocampal slices (P17-P23) by using two complementary imaging techniques: i) Fluo-4 AM bulk loading to record calcium activity in the astrocytic population and ii) whole-cell patch-clamp technique to load individual astrocytes with Fluo-4 dye giving access to single cell processes territory (Additional file 2).
We analyzed signals by positioning regions of interest (ROIs) either in each cell body of Fluo-4 AM loaded astrocytes (20.0 ± 1.3 ROIs by frame; n = 43 slices from 14 animals in physiological condition; Fig. 1a, b) or by subdividing the entire patch-clamp loaded territory into subregions of similar area (1 μm 2 ; 111.7 ± 7.3 ROIs by astrocyte; n = 7 astrocytes from 6 animals in physiological condition; Fig. 1e, f ). These subregions corresponded to functional microdomains as previously defined by Di Castro et al. [8] since their size matched approximately the synaptic density in the neuropil region. We investigated some of the temporal properties of astrocytic Ca 2+ signals (such as the proportion of active cells, the proportion of active microdomains, the frequency of events) and some of the spatial properties of Ca 2+ signals within the astrocytic arbor (such as the events propagation).
In physiological condition, 49.1 ± 2.2% of bulk loaded astrocytes were spontaneously active during a 5 min recording period (n = 43 slices from 14 animals) with a mean frequency of 0.49 ± 0.01 event/min (n = 817 astrocytes; Fig. 1c). Frequency histograms revealed that the occurrence of Ca 2+ events in the astrocyte population was similar to a Poisson distribution with λ = 1.85 suggesting that spontaneous Ca 2+ event within stratum radiatum was a stochastic phenomenon (Fig. 1d). Based on an in vivo study performed in mouse cortical astrocytes [13], we classified astrocytes as inactive (0 event/ min), active (0.2-0.6 event/min) and hyperactive (> 0.6 event/min). We observed a minority of hyperactive astrocytes (11%) within mouse hippocampus stratum radiatum in physiological condition (Fig. 1d). This spontaneous activity is independent of neuronal activity since TTX (500 nM) had no effect on either the proportion of active astrocytes or the frequency of calcium activity or the proportion of active/hyperactive astrocyte ( Fig. 2a-c). This therefore confirmed that the global activity of the astrocytic population is totally autonomous within the mouse hippocampus stratum radiatum [18]. In single patch-clamp loaded astrocyte, we had access to Ca 2+ activity in a two-dimensional acquisition plane representing on average 1945.3 ± 140.1 μm 2 (n = 7 astrocytes; Fig. 1e). We analyzed spatio-temporal encoding of Ca 2+ signals in 111.7 ± 7.3 microdomains within the astrocytic arbor. In physiological condition, 57.9 ± 4.8% of these microdomains were active (i.e. at least 1 Ca 2+ event; Fig. 1f, g) during a 5 min recording period with a mean activity frequency of 0.59 ± 0.01 event/min (n = 782 ROIs). As previously described in mouse dentate gyrus astrocytes [8], we identified two types of Ca 2+ events: "focal events" that are confined to 1 to 4 microdomains and "expanded events" appearing in more than 4 contiguous subregions (Fig. 1h). An additional movie file shows these two types of events in more details (Additional file 3). Focal events were the large majority (67%) occurring randomly in all subregions. Expanded events arose less often (33%) and spread over 7.1 ± 0.4 μm 2 . Within stratum radiatum, these two types of Ca 2+ event were independent of neuronal activity since TTX application didn't affect either the proportion or the frequency of focal and expanded events (Fig. 2d, e).
Astrocytic Ca 2+ signaling involves both internal Ca 2+ stores and external Ca 2+ entry. Within the stratum radiatum astrocyte population, removal of external Ca 2+ from the bath (ACSF -0 Ca 2+ -EGTA 1 mM) had no effect on the proportion of spontaneously active astrocytes (48.2 ± 2.7% in Ca 2+ -free medium vs 49.1 ± 2.2% in ACSF; n = 9 slices from 5 animals; p = 0.97) but aborted approximately 35% of events within active cells Fig. 1 Detection of global and compartmentalized Ca 2+ events in astrocytes. a Fluo-4-loaded astrocytes in the stratum radiatum of a mouse coronal slice. Fluorescence variations were analyzed in astrocyte cell bodies (red square). b Example of typical fluorescence variations recorded in physiological condition. c Proportion of astrocytes displaying calcium activity during a 5-min recording. Each dot corresponds to individual value for each recorded slice (n = 43). d Frequency histogram showing the occurrence of Ca 2+ events in the astrocyte population and the related theoretical Poisson distribution (grey curve). Pie chart representing inactive (0 event per min), active (0.2-0.6 event per min) and hyperactive (> 0.6 event per min) astrocytes in physiological condition. e Single patch-clamp Fluo-4-loaded astrocyte in the stratum radiatum of a mouse coronal slice. Fluorescence variations within processes were analyzed in subregions of~1 μm 2 (red square) in a single z-plane. f Time-lapse Ca 2+ imaging in an astrocyte process showing an example of Ca 2+ event. Time between frames, 15 s. The black curve corresponded to the fluorescence signal recorded in a selected ROI (red square, first image). g Proportion of subregions displaying a calcium activity during a 5-min recording. Each dot corresponds to individual value for each recorded astrocyte (n = 7). h Artificial color superposition of focal Ca 2+ events (1-4 contiguous subregions, blue) and expanded Ca 2+ events (> 4 contiguous subregions, red) occurring during a 5-min recording in an astrocyte process, as defined by [8]. Traces showing occurrence of Ca 2+ peaks in contiguous subregions with blue and red highlights identifying focal (blue) and expanded (red) Ca 2+ events respectively. In physiological condition, focal events represent the major part of Ca 2+ events within the astrocytic processes (0.31 ± 0.02 event/min in Ca 2+ -free medium vs 0.48 ± 0.01 in physiological condition; n = 103 astrocytes; p < 0.0001) meaning that one third of somatic Ca 2+ events depended on a Ca 2+ entry (Fig. 2a-c). Particularly, the proportion of hyperactive/active astrocytes was strongly affected (3% of hyperactive astrocytes versus 11% in physiological condition; p < 0.001). Yet, location of internal stores is not uniformly distributed in the astrocytic territory since they are concentrated in cell body and thick processes, thin astrocytic processes being practically devoid of Ca 2+ stores [23]. Consistently, we observed that removing external Ca 2+ suppressed nearly all the compartmentalized Ca 2+ transients in the astrocytic arbor (5.1 ± 1.3% active territories vs 57.8 ± 4.8% in physiological condition; n = 7 astrocytes from 6 animals; p = 0.0006) and reduced the frequency of the residual signals (0.41 ± 0.01 event/min vs 0.59 ± 0.01 in physiological condition; n = 687 ROIs; p = 0.04) (Fig. 2d, e).

Astrocytes become hyperactive in presence of amyloid-β oligomers
To investigate the effect of soluble Aβ 1-42 oligomers (Aβo) on astrocytic calcium signaling, we perfused hippocampal slices with 100 nM of oligomeric forms of the peptide during 5 min before recording. In the global astrocytic population, application of Aβo resulted in a Fig. 2 Spontaneous astrocytic Ca 2+ events are fully autonomous and partly depend on external Ca 2+ entry. a, b Within the astrocytic population, proportion of astrocytes displaying calcium activity and frequency of astrocyte calcium activity in physiological condition (grey; n = 43), under 500 nM TTX application (dark blue; n = 6) or in Ca 2+ -free medium (0 Ca; light blue; n = 9). Each dot corresponds to individual value for each recorded slice (proportion) or each recorded ROI (frequency). c Frequency histogram revealing that 500 nM TTX had no effect on the frequency distribution whereas significantly less astrocytes were hyperactive in Ca 2+ -free medium (0 Ca 2+ ; light blue). d, e Within single astrocyte arbor, proportion of subregions displaying calcium activity and frequency of astrocyte calcium activity in physiological condition (grey; n = 7), under 500 nM TTX application (dark blue; n = 7) or in Ca 2+ -free medium (0 Ca 2+ ; light blue; n = 7). Each dot corresponds to individual value for each recorded astrocyte (proportion) or each recorded ROI (frequency). Results are compared with the physiological condition with *, p < 0.05; **, p < 0.01 and ***, p < 0.001 significant increase in the proportion of active astrocytes (69.7 ± 2.1% vs 49.1 ± 2.2% in physiological condition; n = 12 slices; p < 0.0001; Fig. 3a). Within active cells, the frequency of Ca 2+ events also significantly rose (0.83 ± 0.04 event/min vs 0.48 ± 0.01 event/min in physiological condition; n = 237 astrocytes; p < 0.0001; Fig. 3b). Frequency histogram was switched to high frequency and developed differently from the basal Poisson distribution (λ = 4.39; Fig. 3c). This suggested that Ca 2+ event was not anymore a stochastic phenomenon but was managed by Aβo. The fraction of hyperactive astrocyte (i.e. > 0.6 event/ min) was significantly larger (41% in Aβo condition vs 11% in physiological condition; p = 0.012; Fig. 3c). Thus, under application of 100 nM Aβo, there is an increase in the population of hyperactive astrocytes at the expense of basal-active and -inactive astrocytes.
Thus, 100 nM Aβo triggered astrocytic calcium hyperactivity within the stratum radiatum as soon as 5 min in the global astrocytic population, manifested as an increase of the proportion of active cells along with an increase of the frequency of individual Ca 2+ events. Interestingly, this global hyperactivity went along with a sustained increase of Ca 2+ activity within astrocytic processes including an increase of the proportion of active microdomains together with an increase of the expanded events size in the astrocytic processes.
Aβo-induced astrocytic Ca 2+ hyperactivity is independent of neuronal activity or microglia activation and involves external Ca 2+ entry Aβo could potentially affect directly and/or indirectly Ca 2+ signaling within astrocytes. In order to ascertain the neuronal part in the astrocyte hyperactivity, we coapplied 500 nM TTX with 100 nM Aβo. In bulk loading condition, TTX had no effect on the Aβ-induced astrocytic Ca 2+ hyperactivity either in the proportion of active astrocytes (61.9 ± 2.2%; n = 8 slices from 6 animals; Fig. 3a) or in the frequency of Ca 2+ events (0.89 ± 0.04 event/min; n = 145 astrocytes; Fig. 3b). The proportion of hyperactive astrocytes was similar (p = 0.16) and the frequency distribution was not impacted by TTX application (Fig. 3d). In the same way, in individual patch-clamped astrocyte, the blockade of action potentials with TTX did not affect the Aβo effect on the proportion of active territories (n = 5 astrocytes from 4 animals; p = 1.0; Fig. 3f) and on the frequency of Ca 2+ events (n = 870 ROIs; p = 0.9; Fig. 3g). The enlarged size of expanded events was not reduced with TTX (9.0 ± 0.5 μm 2 ; p = 0.68; Additional file 4c). These data suggested that, within 5 min, Aβo acts directly on astrocyte signaling independently of any neuronal activity.
Activation of microglia can be a relatively early event in the Aβ mediated pathological process and can in turn activate astroglia cells [24,25]. In order to assess the role of microglia in the Aβo-induced astrocytic Ca 2+ hyperactivity, we pretreated slices with minocycline, an inhibitor of microglia activation [24][25][26]. Minocycline 50 nM pre-incubated during 15 min did not prevent the ability of Aβo to trigger astrocytic Ca 2+ hyperactivity increasing to the same extent the proportion of active astrocytes (61.0 ± 3.3%; n = 6 slices from 4 animals; p = 0.0043) and the frequency of Ca 2+ events (0.79 ± 0.08 event/min; n = 69 astrocytes; p = 0.02; Additional file 5a). Thus, within 5 min, Aβo acts directly on astrocyte signaling independently of any microglia activation.

TRPA1 channels underlie Ca 2+ hyperactivity induced by Aβo
Within astrocyte, Ca 2+ entry can occur through astrocytic ligand-gated Ca 2+ channels (ionotropic receptors), transient receptor potential (TRP) receptors and reverse operation of Na + /Ca 2+ exchanger (NCX). Astrocytic cationic ionotropic receptors (such as AMPA, NMDA or P2X receptors) have relatively low single channel conductance (~1 to 3 pS) and accounted for~4% of fractional Ca 2+ current [6]. Among TRP receptors, TRPA1 channels have been recently involved in regulating astrocyte basal Ca 2+ levels in the hippocampus stratum radiatum [10,11]. Interestingly, TRPA1 channels have a relatively high single channel conductance (~110 pS) accounting for~17% of fractional Ca 2+ current for the constitutively open channel [27] referring them as important actors involved in astrocytic Ca 2+ entry. Within the CA1 region of mouse hippocampus, functional TRPA1 were only detected in astrocytes [10]. Based on immunohistochemistry of fixed brain sections, we found that TRPA1 channels staining appeared as discrete punctates predominantly co-localized with astrocytes (GFAP-positive cells; Additional file 6). Higher-resolution images showed that TRPA1 channels were expressed in both cell body and thick processes of stratum radiatum astrocytes objectivized by the GFAP co-staining. Some staining also appeared in adjacent territories (Fig. 4a) continuous with GFAP-positive stem processes. An additional 3D-reconstruct movie file shows this in more details (Additional file 7) and suggested that TRPA1 channels were expressed in astrocyte thin processes. These observations agreed with immunogold electron microscopy studies performed in rat trigeminal caudal nucleus showing that TRPA1 channels are localized in astrocyte peripheral processes [28].
Thus, while having a negligible impact on physiological spontaneous Ca 2+ activity, blocking TRPA1 channels cancelled astrocyte Ca 2+ hyperactivity induced by Aβo leading most of the spatiotemporal Ca 2+ signal properties back to physiological state in either the astrocytic arbor microdomains or the global astrocytic population.
APP/PS1-21 mice display early astrocytic hyperactivity that involves TRPA1 channels APP/PS1-21 mice co-express the human KM/67/671NL mutation in the amyloid precursor protein (APPswe) and the human L166P-mutated presenilin 1 (PS1) under the control of Thy1 promotor. These transgenic mice display high transgene expression with Aβ 1-42 levels being significant at 1 month and plaques appearing at 6 weeks of age in the neocortex and at 4-5 months in the CA1 hippocampus [15]. As external application of Aβo induced an immediate astrocytic hyperactivity in hippocampal healthy brain slices, we wonder about the effect of Aβ in APP/PS1-21 mice at the beginning of its secretion within the brain. Thy1 is known to be expressed from~P14 in mouse CA1 hippocampus [31] thus Aβ 1-42 might be produced from this stage. However, while being present at this early stage [15], Aβ were not yet aggregated in plaques in APP/PS1-21 mice at 1-month-old (Fig. 5a) while Aβ plaques were present in the hippocampus at 6-month-old as testified with Thioflavin S labeling [21]. Likewise, astroglial marker such as GFAP showed a similar profile in WT versus APP/PS1-21 mice at 1-monthold confirming the lack of reactive astrogliosis in these early stages (Fig. 5b, d). Thus, studying astrocyte activity at these early stages allowed us to measure the impact of not aggregated Aβ on healthy astrocytes in an AD transgenic mice model.
We first investigated the expression of TRPA1 channels in the hippocampus of young APP/PS1-21 mice and their littermates (WT). We found that the protein level of TRPA1 channels was similar in P19 mice (n = 6 hippocampus from 3 animals for each; p = 0.9; Fig. 5c) and twice higher in P30 APP/PS1-21 than WT mice (n = 8 hippocampus from 4 animals for each; p = 0.02; Fig. 5c) suggesting an early upregulation of this channel from the onset of Aβ secretion. At the same time, GFAP protein level was stable (Fig. 5d) testifying the lack of astrogliosis in these early AD stages.
In this AD mouse model, no signs of neuroinflammation were apparent in 1-month-old mice and microgliosis has been described to be concomitant with plaques appearance [15]. Indeed, at these early stages, microglia phenotype was not different in APP/PS1-21 mice from WT littermate (Iba-1 immunostaining; Additional file 5b). However, to confirm that microglia was not involved in these early astrocytic events, we pretreated slices with minocycline in order to inhibit potential microglia activation [24][25][26]. Minocycline 50 nM, pre-incubated during 15 min, had no effect on either the proportion of active cells or the frequency of astrocytic Ca 2+ events within APP/PS1-21 mice (n = 6 slices from 4 animals; p = 0.88 and 0.82 respectively; Additional file 5c) testifying that microglia activation was not involved in the astrocytic hyperactivity setting up at the beginning of Aβ secretion in this AD mouse model. In patch-clamp loaded astrocyte, the proportion of active microdomains increased in APP/ PS1-21 mice (65.7 ± 2.0% vs 55.0 ± 4.8% in WT; n = 4 astrocytes from 4 animals for each; p = 0.052; Fig. 6c) together with the frequency of events within these microdomains (0.45 ± 0.02 event/min vs 0.40 ± 0.01 event/ min in WT; n = 255 and 426 ROIs respectively; p = 0.04; Fig. 6c). The proportion of expanded/focal events was slightly affected in APP/PS1-21 mice in favor of expanded event and the size of expanded events was similar (6.8 ± 0.3 μm 2 in WT vs 6.7 ± 0.3 μm 2 in APP/PS1-21; p = 0.42; Additional Histogram showing quantification of GFAP expression normalized to protein loading levels (n = 7 hippocampus in each group at P19 and P30). Results are compared with the WT mice with *, p < 0.05; **, p < 0.01 and ***, p < 0.001 file 4b, d). Thus, in young APP/PS1-21 mice, Ca 2+ activity in the global astrocytic population and the compartmentalized activity within the astrocytic arbor started to be affected from the beginning of Aβ overproduction.
In these conditions, blocking TRPA1 channels with HC 030031 (40 μM) in APP/PS1-21 mice strongly decreased the frequency of Ca 2+ events (0.37 ± 0.03 event/ min with HC 030031 vs 0.64 ± 0.07; n = 113 astrocytes from 7 slices; p = 0.013; Fig. 6a) resulting in a clear redistribution of Ca 2+ events frequency (λ = 1.35) and a decrease of the proportion of hyperactive cells that look alike the WT situation (4% in APP/PS1-21 + HC 030031 vs 21% in APP/PS1-21; p = 0.0025; Fig. 6b). In patch-clamp loaded astrocyte, HC 030031 induced a significant decrease of the proportion of active microdomains in APP/PS1-21 mice (29.3 ± 6.2%; n = 7 from 4 animals; p = 0.01; Fig. 6c) together with a restrained effect on the frequency (0.43 ± 0.06 event/min; n = 1213 and 1577 ROIs respectively; p = 0.058). The proportion of expanded/focal events was also reduced with HC 030031 treatment (Additional file 4b). Overall, these observations in young APP/PS1-21 mice matched the above related effect of Aβo application in healthy slices consistently with a TRPA1 channel involvement. Merely, they were more subtle and initially restrained to the astrocytic arbor.

Discussion
In this study, we investigated the contribution of astrocytes in early Aβo toxicity by studying calcium signaling in different parts of the whole astrocyte territory. We found that astroglia is a frontline target of Aβo exhibiting a global and local Ca 2+ hyperactivity that involves TRPA1 channels. This TRPA1 channel-dependent astrocytic Ca 2+ hyperactivity exerts regulatory influences on synaptic function and is linked to the glutamatergic synapse hyperactivity recorded in CA1 neurons. Concurrently to these acute Aβo-induced effects, astrocytes in young APP/PS1-21 mice hippocampus elicit a similar pattern of calcium hyperactivity in close relationship with the setting up of a precocious neuronal hyperactivity that are both reversed when TRPA1 channel is blocked. Moreover, the TRPA1 channel is gradually overexpressed at the onset of Aβ production in this AD mouse model. Intracellular Ca 2+ transients are considered as the primary signal by which astrocytes interact with each other and with neighboring neurons. Ca 2+ has been extensively studied within the astrocytic cell body and thick branches. More recently, local Ca 2+ dynamics in distal fine processes has been investigated emphasizing a highly compartmentalized signaling, interconnected with physiological transmission at neighboring synapses [8,9]. Compartmentalization of astrocytic Ca 2+ dynamics needs to be attentively considered in order to understand how astrocytes may contribute to brain information processing [8]. We thus chose to study both levels of information (i.e. global population signaling and local microdomain signaling) combining bulk loading and single cell astrocyte loading. Genetically encoded Ca 2+ indicators (GECIs) have been recently used to study Ca 2+ signals in distal thin processes [9]. Alternatively, patch pipette loading give access to the whole territory of a single astrocyte and, currently, Fluo-4 is far more sensitive than GECIs therefore enabling to track smaller signals [8]. Accordingly, we observed a similar and even better diffusion of Fluo-4 in a single astrocyte compared to SR101. We first characterized the physiological calcium activity of mouse CA1 stratum radiatum astrocytes and showed that this activity is fully autonomous, i.e. independent of neuronal activity, both at the astrocytic population level and at the microdomain level. This is in agreement with data obtained in mouse CA1 hippocampus [30,33] but not with astrocyte behavior in the dentate gyrus where expanded Ca 2+ events were partly dependent on neuronal activity [8]. Interestingly, external Ca 2+ entry is the main source of Ca 2+ within thin processes whereas it only partly contributes to somatic signaling. This discrepancy between Ca 2+ sources in astrocyte soma and distal processes has already been described in brain slices [34]. This might be supported by the subcellular location of calcium stores that are concentrated in the cell body and thick processes but are almost absent from thin processes [23].
A central element of the pathogenesis of AD is the progressive accumulation of Aβo species, ultimately resulting in the formation of plaques. Yet, small soluble Aβ oligomers are sufficient to induce several features of the AD phenotype [1]. The paths by which Aβo leads to neurodegeneration are probably multifactorial but all converge all towards synaptic dysfunction. The major challenge of AD research is to understand the complex cellular reaction underlying the long prodromal phase of AD [14]. Astrocytes are an integral part of synaptic transmission and are therefore critical for the establishment and maintenance of neuronal health [5]. They contribute to neuronal dysfunction by being proinflammatory [35] but also play a protective role, e.g. through the release of gliotransmitters [36] and Aβ clearance [37,38]. It is therefore of major importance to distinguish the beneficial from the deleterious impact of Aβo on astrocyte function. Aβo has already been involved as a direct effector on astrocytes in primary cultures [39], in hippocampal slices [40] and in vivo [12] but here we showed a peculiar rapid action on compartmentalized calcium activity, activating a membrane Ca 2+ permeable channel. Indeed, Aβo application triggers global Ca 2+ hyperactivity in CA1 hippocampus astroglial population together with an intensification of the compartmentalized Ca 2+ activity in the astrocytic processes and a spatial extension of the size of the expanded Ca 2+ events within microdomains. These effects are specific to oligomeric forms of Aβ since application of the monomeric form in the same conditions had no impact on astrocytic calcium activity. We reported that the effect of Aβo on astrocyte excitability is fully independent of neuronal activity since TTX application does not prevent the Aβo effect neither on the global hyperactivity nor on the compartmentalized hyperactivity in processes. Concurrently, microglia activation does not participate in this astrocytic hyperactivity, at least in the time scale studied, whereas longer applications of Aβ activate microglia together with astroglia [24]. Thus, astrocytes seem to express a distinctive precocious detector involved at the onset of Aβo appearance. Removal of external Ca 2+ largely inhibits Aβo-induced astrocyte hyperactivity at the population level while it has no effect in physiological conditions. Removal of external Ca 2+ also inhibits the majority of the compartmentalized Aβo-induced hyperactivity. Thus, transmembrane Ca 2+ entry carries most of the Aβoinduced hyperactivity. Remarkably, we showed that both global and compartmentalized hyperactivities are driven by TRPA1-dependent Ca 2+ entry since HC 030031, a specific TRPA1 channel inhibitor [29], strongly abolishes Aβo-induced astrocyte hyperexcitability and totally restores the spatiotemporal properties of Ca 2+ events back to a physiological level.
The TRPA1 channel is a Ca 2+ permeable non-selective cation channel initially known to be expressed in primary afferent nociceptive neurons [29]. In mouse CA1 hippocampus, TRPA1 channels are found to be preferentially expressed in astrocytes [10]. However, their involvement in physiological astrocytic Ca 2+ signaling is highly debated. It has been shown that TRPA1 channels contribute to maintain basal Ca 2+ levels and regulatẽ 20% of spontaneous Ca 2+ signals within astrocyte branches [11] but in the end, poorly take part in basal astrocytic Ca 2+ signaling [30,33]. Our data assume that TRPA1 channels are only slightly involved in the astrocytic Ca 2+ signaling in physiological conditions. However, we highlighted they are quickly and largely involved in case of Aβo presence. The absence of an obvious involvement of these channels in astrocyte physiological Ca 2+ signaling is startling since we evidenced a TRPA1 channel expression in thick and in adjacent thin astrocytic processes. Thus, TRPA1 channels might only behave as an "aggression sensor". Indeed, TRPA1channel gating is particularly regulated by numerous electrophilic activators -such as reactive oxygen species, reactive nitrogen species or oxidized lipids -and also functions as a mechanosensor [27]. Hence, TRPA1 channels might be directly targeted by Aβo or might be secondarily activated through an Aβo-induced oxidative stress and/or through its mechanosensor properties if Aβo binds to the astrocytic cholesterol-rich plasma membrane [14].
Strikingly, in young APP/PS1-21 mice (~3-4 weeks), we observed a similar pattern of astrocytic hyperactivity starting at the beginning of Aβ overproduction in the hippocampus, long before its aggregation into plaques. These early repercussions in young APP/PS1-21 mice were restricted to the frequency of astrocytic Ca 2+ events in either the astrocytic population and microdomains of astrocytic processes with fewer impacts on the proportion of active cells or microdomains. This suggests a gradual impact of surrounding Aβ on astrocyte signaling, increasing the frequency of compartmentalized Ca 2+ events and, to a lesser extent, the proportion of active territories within the astrocytic processes. These impacts on astrocytic processes go along with a noteworthy redistribution of the frequency of Ca 2+ events within the astrocytic population. Interestingly, blockade of TRPA1 channels with HC 030031 abolished the astrocyte Ca 2+ hyperactivity. Overall, TRPA1 channel signaling seems to be at the frontline in mediating these Aβo progressive effects in early stages of AD. Data obtained in an advanced AD transgenic model showed an astrocyte network hyperactivity in cortical areas close to Aβ plaques and an involvement of metabotropic purinergic signaling in this astrocyte hyperactivity [13]. This suggests a differential evolution of astrocyte engagement in AD pathogenesis depending on the stage, the structure and the physiopathological state of the astrocyte. Likewise, it has been reported that the TRPA1 channels' protein level was increased in hippocampal astrocytes of 8 month-old APP/PS1 mice at where it mediated inflammation through astrocyte activation [41]. Here, we showed that TRPA1 channel expression in hippocampus is increased much earlier, as soon as 1 month of age, in a more aggressive AD mouse model (APP/PS1-21). These data point towards a TRPA1 channel contribution in early stages of pathophysiology, that is as soon as the Aβo level increases and long before the setting up of astrogliosis or inflammatory mechanisms.
Numerous laboratory studies in the past decade have shown that Aβo impairs synaptic function and synaptic structure [42]. However, how soluble Aβo initiates these effects remains to be determined. Each astrocyte deploys many fine processes to contact up to 140,000 synapses in the CA1 region [43]. As we highlighted an intense and early effect of Aβo on astrocyte Ca 2+ activity within processes, we assessed the link with spontaneous neuronal activity. Indeed Aβo is also known to enhance spontaneous neuronal excitability in CA1 [32,44,45]. Consistently, we showed here that Aβo induces a rapid and strong increase of spontaneous EPSCs frequency in CA1 neurons. Strikingly, blocking TRPA1 channels totally prevents this Aβo neuronal impact. Yet, when we blocked neuronal activity with TTX, we did not affect the Aβo-induced astrocyte hyperactivity which would be partly the case if a neuronal TRPA1 was involved. This precocious Aβo impact thus seems to trigger a one-way communication from astrocyte to neuron related to TRPA1 activation. This TRPA1-dependent neuronal hyperactivity was similarly observed in APP/PS1-21 mice at the onset of Aβ overproduction testifying its physiopathological relevance in the AD initiation process. It has been shown that Aβ can increase astrocytic release of glutamate to the extrasynaptic space resulting in the activation of extrasynaptic NMDARs and the disruption of neuronal signaling [40,46,47]. Besides, astrocytes can regulate synaptic and extrasynaptic neurotransmitter concentrations, such as glutamate, in a Ca 2+ -dependent manner e.g. via vesicular release, bidirectional transport or hemichannel opening [48]. We will further decipher pathways implemented by the Aβinduced TRPA1-mediated Ca 2+ entry that consequently affect neuronal transmission.
It has been demonstrated that soluble Aβo can affect astrocyte signaling properties in various ways in mouse hippocampal CA1 astrocytes [40,49,50]. To some extent, the involvement of TRPA1 channels superimposed to these effects, directly affecting local synaptic function in a distinctive precocious manner. This actor might thus contribute to the complex cellular phase of AD, upstream of symptomatic neurodegeneration [14].

Conclusions
In this work, we have shown that intricate global and compartmentalized astroglial Ca 2+ signaling disturbances induced by Aβo are mediated by a TRPA1 channeldependent Ca 2+ signaling. This new highlighted mechanism is at work with both external supply of Aβo and at the onset of Aβ production in an AD transgenic mouse model. This astrocytic pathway is promptly implemented and involved in the well-known and characteristic Aβoinduced synaptic dysfunction. Blockade of TRPA1 channel, that appears to be preferentially expressed in astrocytes within the hippocampus [11], is sufficient to counteract the impact of Aβo on spontaneous neuronal activity. This suggests that astrocytes can be considered as a particularly precocious target in Aβo toxicity consequently affecting nearby synapses. Up until now, altered astrocyte activation was usually associated with late AD, i.e. when amyloid plaques are already developed, and the role of astrocytes in the initial toxic effect of Aβo was not yet evidenced. One of the major findings of our study is to suggest that astrocytes are implicated far before astrogliosis and inflammatory processes. Thus, focusing on this astrocyte involvement, for instance by modulating TRPA1 channel activity, may represent a novel target to hamper early dysfunction in AD. condition (grey; n = 43), under 50 nM minocycline (dark grey; n = 5); 100 nM Aβo application (orange; n = 12) and under 50 nM minocycline +100 nM Aβo application (dark orange; n = 6). (b) Immunohistochemistry of mouse stratum radiatum microglia showing that Iba1-positive cells were not hypertrophic in one-month-old APP/PS1-21 mice when compared to WT littermates. Higher magnification of representative microglia is shown in the lower panels. (c) Proportion of astrocytes displaying calcium activity and frequency of astrocyte calcium activity in APP/PS1-21 mice (orange; n = 8), under 50 nM minocycline application (dark orange; n = 6) or in WT littermates (grey; n = 8). Minocycline is pre-incubated 15 min before recording. Results are compared with the physiological condition, with or without minocycline, or with the WT littermates with *, p < 0.05; **, p < 0.01 and ***, p < 0.001. (TIFF 1680 kb) Additional file 6: TRPA1 is expressed in stratum radiatum astrocyte cell body and processes. (a) Immunohistochemistry of mouse stratum radiatum astrocytes showing that TRPA1 channels (green) are expressed within astrocytic domains certified by GFAP staining (magenta).