Introduction

Every aspect of brain function relies on the delicately maintained water environment. It supports brain structural stability and molecular diffusion, laying the ground for information processing, metabolite shuttling and adaptation to living environments (Kimelberg, 2004). Water is the fundamental constituent of the cerebrospinal fluid infiltrating into the central nervous system and the interstitial fluid distributed in brain parenchyma (Agnati et al., 2017, Brinker et al., 2014). Brain fluid transport is suggested to support the diffusion of energetic fuels like glucose and lactate to warrant the metabolic circumstances in the parenchyma, and is also implicated in the clearance of waste molecules from the brain as described in the glymphatic system (Abbott et al., 2018, Louveau et al., 2017). In addition, brain water diffusion is the basis for diffusion-weighted magnetic resonance imaging (DW-MRI) for structural and functional neuroimaging (Le Bihan et al., 2006), and for clinical diagnostics widely applied for neurological diseases such as ischemia, brain tumor and edema (Gaddamanugu et al., 2022, Le Bihan & Iima, 2015).

Water equilibrium in the brain is maintained by dynamic transport between different cell entities. Aquaporin is a family of transmembrane channels facilitating bidirectional water flow, with aquaporin 4 (AQP4) being the main subtype expressed in the central nervous system. In the brain, AQP4 is predominantly expressed in astrocytes (Papadopoulos & Verkman, 2013, Xiao et al., 2023) that are a major type of glial cell in the parenchyma (Barres, 2008). This feature enables astrocytes to well balance osmotic oscillations imposed by the transmembrane transport of ions, metabolites, and signaling molecules, all mediated by water. Dysregulation in astrocyte aquaporin water transport is implicated in pathological scenarios such as cerebral edema and neuroinflammation (Verkman et al., 2011). Valuable insights have been obtained on the function of astrocyte aquaporin in specific physiopathological conditions including osmotically evoked volume responses, ischemia and neurodegenerative diseases (Verkman et al., 2017). Meanwhile, how astrocyte aquaporin contributes to brain water homeostasis in basal physiology has remained elusive. A better understanding will aid to discern astrocytic regulation of cerebral water states, thereby the potential maladaptations in neurological disorders.

Here, combining in vivo chemical targeting, optical and MRI imaging of water dynamics, we report that AQP4 sustains a tonic water efflux from brain astrocytes. This mechanism was found to be critical in maintaining astrocyte volume and signaling stability, as well as the extracellular space remodeling in mouse cortex during optogenetically induced cortical spreading depression. Using DW-MRI, we observed that acute inhibition of astrocyte water efflux in vivo heterogeneously alters water diffusion across brain regions. Our finding suggests that aquaporin acts as a water export route in astrocytes in physiological conditions, so as to counterbalance the constitutive intracellular water accumulation caused by constant transmitter and ion uptake, as well as the cytoplasmic metabolism processes. This mechanism hence plays a necessary role in maintaining water equilibrium in astrocytes, thereby brain water homeostasis.

Results

Astrocyte aquaporin mediates a tonic water efflux

To optically follow in situ astrocyte water transport, we performed fluorescence intensity-translated (FIT) imaging. To this end, mouse brain astrocytes were chemically labeled in vivo with a highly water-diffusible and astrocyte-specific red fluorescent dye, sulforhodamine B (SRB) (Fig. 1A) (Appaix et al., 2012, Pham et al., 2020). About one hour after its intraperitoneal injection, astrocytic SRB labeling was widely distributed in the mouse cortex, as seen in living acute brain slices (Fig. 1A) and validated by colocalization with EGFP-identified astrocytes in slices taken from GFAP-EGFP transgenic mice (Fig. 1A). Net water transport across the astrocyte membrane alters cytoplasmic SRB concentration and cellular volume, which can thus be followed by changes in fluorescence intensity when imaged in a fixed field of view. To validate the FIT imaging of astrocyte water transport, we employed wide-field fluorescence microscopy for single-plane time-lapse imaging in acute brain slices in the primary somatosensory (S1) cortex. This approach allowed us to collect fluorescence from both the focal plane and along the axial extensions, thereby imprinting volumetric fluorescence into the single image plane. Indeed, water influx induced by the applicaiton of hypoosmotic solution over different periods proportionally decreased astrocyte SRB fluorescence, which also reflected cell swelling (Fig. 1B); whereas water export and astrocyte shrinking upon hyperosmotic manipulation increased astrocyte fluorescence (Fig. 1B). Hence, FIT imaging enables real-time recording of astrocyte transmembrane water transport and volume dynamics.

Fig. 1.

In basal conditions, flat fluorescence time course was recorded in brain slices suggesting equilibrated water transport across the astrocyte membrane (Fig. 1B) and volume homeostasis in the brain parenchyma. To inspect a tonic role of aquaporin in astrocyte water transport, we sought to acutely block astrocytic aquaporin channel, aquaporin 4 (AQP4). We used the synthetic compound 2-(nicotinamido)-1,3,4-thiadiazole (TGN-020) that is derived from the condensation of nicotinamide and thiadiazole derivatives (Burnett et al., 2015, Huber et al., 2009), whose specificity for AQP4 has been validated in vitro by ion channel heterologous expression system (Toft-Bertelsen et al., 2021) and in vivo using the AQP4 knockout (KO) mice (Harrison et al., 2020, Igarashi et al., 2013). This approach guaranteed the functional pinpointing of astrocyte aquaporin under physiological conditions, while avoiding the chronic compensations caused by genetic tools and mouse models that have been reported to alter brain water content, volume and extracellular architecture potentially confounding the functional readouts (Binder et al., 2004, Gomolka et al., 2023, Haj-Yasein et al., 2011, MacAulay, 2021, Yao et al., 2015).

AQP4 is a bidirectional channel facilitating passive water transport along the osmotic gradient (Papadopoulos & Verkman, 2013). As an osmotic equilibrium is maintained in the brain parenchyma in basal state, there might not be net water transport across AQP4. However, we observed in cortical slices that acute inhibition of astrocyte AQP4 by TGN-020 caused a gradual decrease in SRB fluorescence intensity, indicating an intracellular water accumulation (Fig. 1C). This observation suggests that astrocyte aquaporin mediates a tonic water efflux; its blocking causes intracellular water accumulation and swelling (Fig. 1C, right). To corroborate this observation in vivo, we performed fiber photometry recording in moving mice (Fig. 1D). An optical fiber was implanted into the mouse S1 cortex to follow fluorescence signals of local astrocyte population post the intraperitoneal injection of SRB (Fig. 1D, left). After about one hour, when astrocytes were well labeled, either saline or TGN-020 was injected intraperitoneally. While no significant effect was observed with the saline control, TGN-020 induced a decrease in astrocyte SRB fluorescence and a reduction in its oscillation range (Fig. 1D, right), mirroring an intracellular dilution of fluorescent molecules. This observation confirms that acute inhibition of astrocyte aquaporin leads to intracellular water accumulation thereby cell swelling.

Aquaporin inhibition perturbs astrocyte and neuron signaling

Astrocyte volume equilibrium not only determines brain architectural stability, but also associates with dynamic cellular signals. Astrocyte swelling has been shown to alter intracellular Ca²⁺ signaling (Benfenati et al., 2011, Eilert-Olsen et al., 2019). We performed Ca²⁺ imaging to confirm that acute aquaporin inhibition induces astrocyte swelling and then Ca²⁺ oscillation. The genetically encoded Ca²⁺ sensor GCaMP6f was selectively expressed in astrocytes by crossing the Glast-Cre^ERT2 and GCaMP6^floxP mouse lines (Herrera Moro Chao et al., 2022, Pham et al., 2020). Because astrocyte Ca²⁺ signals often occur in local domains, we adapted a light sheet microscope (Pham et al., 2020) for wide-field optical sectioning so as to image them with high signal-to-noise ratio (Fig. S1). We first confirmed that water accumulation, therefore astrocyte swelling, induced by hypotonic solution enhances the intracellular Ca²⁺ signal (Fig. 2A). Then we followed astrocyte Ca²⁺ in isotonic control condition while blocking astrocyte AQP4 with TGN-020, and observed that Ca²⁺ signaling was similarly augmented following the acute aquaporin inhibition (Fig. 2B). This result confirms that inhibiting astrocyte tonic water efflux leads to intracellular water accumulation and swelling, which causes alterations in astrocytic Ca²⁺ signaling. In addition, astrocyte swelling has been reported to induce the release of neuroactive molecules such as glutamate thereby influencing nearby neuron activity (Fiacco et al., 2007, Yang et al., 2019). Disrupting AQP4 tonic water outflow would not only cause astrocyte swelling but may also influence the activity of neighboring neurons. We then imaged Ca²⁺ signals as a surrogate for neuronal activity in cortical somatostatin (SST) interneurons which exhibit high excitability (Karagiannis et al., 2021) and express both ionotropic and metabotropic glutamate receptors (Cauli et al., 2000), rendering them ideal to sense astrocyte-released signaling molecules. We expressed GCaMP6f in SST interneurons by crossing homozygous SST-Cre (Taniguchi et al., 2011) and homozygous GCaMP6^floxP mice. As expected, TGN-020 inhibition of astrocyte AQP4 led to a global Ca²⁺ elevation in SST interneuron populations (Fig. 2C). These observations support that astrocyte aquaporin mediates a tonic water efflux, contributing to maintain both the volume and signaling homeostasis.

Fig. 2.

Aquaporin water efflux regulates astrocyte volume response

We then examined the role of aquaporin water efflux in astrocyte volume response to osmotic changes. The AQP4 blocker TGN-020 or equimolar vehicle was applied throughout the recording, namely being continuously present before and after osmotic challenges. We first followed the evoked water efflux and shrinking induced by hypertonic solution while. As astrocyte AQP4 supports preferentially water efflux, its inhibition would attenuate hypertonicity-imposed water extrusion (Fig. 3A). Supporting this, application of TGN-020 slowed down the hypertonicity-induced water efflux and shrinking, reflected by the longer time to peak in the SRB fluorescence time course as compared to control (Fig. 3B), though the delay in the initial onset was not significantly prolonged. The maximal increase in astrocyte SRB fluorescence was lower in the presence of TGN-020, suggesting that AQP4 blocking reduced the overall amount of water efflux (Fig. 3B).

Fig. 3.

We next evoked water influx, thereby astrocyte swelling, by hypotonic solution in the presence or absence of TGN-020 (Fig. 3C). In contrast to the effects on hypertonicity-evoked water efflux, AQP4 acute inhibition was observed to accelerate both the initial water accumulation and the swelling rate in astrocytes, reflected by their earlier onset in SRB fluorescence decrease and faster reaching to the plateau value relative to the control condition (Fig. 3D). This observation cross-validates the role of astrocyte aquaporin in supporting water efflux, whereby its blockade facilitates the initial accumulation of the evoked water influx (Fig. 3C). In contrast, the late-stage maximum decrease in astrocyte SRB fluorescence was observed to be reduced in the presence of TGN-020 (Fig. 3D), reflecting a reduction in the total amount of water accumulation and swelling. As TGN-020 was present prior to hypotonic challenge, it would have slightly swelled astrocytes due to the blockade of tonic water efflux, thereby constraining the range of the further swelling induced the subsequent hypotonicity. Further, while astrocyte AQP4 sustains a water outflow in basal physiology, in conditions when transmembrane water gradient is altered as occurred in hypotonic solutions, the net water flow through AQP4 should be finally dictated by the osmotic gradient due to its bidirectional nature. Therefore, along the imposed hypotonicity, AQP4 started to instruct a water influx; its constant blocking by TGN-020 would have reduced the total amount of water influx and swelling.

Astrocytes are widely distributed throughout the brain parenchyma, orchestrating water transport and volume homeostasis (Ochoa-de la Paz & Gulias-Canizo, 2022). Water outflow via astrocyte AQP4 may play a role in the structural remodeling of parenchyma at the global level. We then induced general cell swelling by triggering cortical spreading depression (CSD) (Holthoff & Witte, 1996, Zhao et al., 2019). To be orthogonal to the pharmacological control of astrocyte aquaporin, the CSD depolarization wave was initiated by optogenetically stimulating ChR2-expressing glutamatergic pyramidal cells in the cortical slices of Emx1-Cre/Ai32ChR2 mice (Gorski et al., 2002, Madisen et al., 2012), termed Opto-CSD (Chung et al., 2019). As cell swelling increases the transmittance of infrared light, we imaged Opto-CSD by the intrinsic optical signal (IOS) derived from infrared illumination (Holthoff & Witte, 1996, MacVicar & Hochman, 1991), which was spectrally distinct from ChR2 photoactivation (detailed in Methods). The IOS displayed transient increase across cortical layers during the Opto-CSD (Fig. 4A-C), exhibiting propagating kinetics characteristic of spreading depression (Chung et al., 2019, Holthoff & Witte, 1996). The biphasic starting of IOS coincided with a sharp response in the extracellular potential indicating the CSD initiation (Fig. S2A). IOS signal contains a first peak reflecting the rapid CSD response (Fig. 4C, b) followed by a prolonged phase of general cellular swelling (Fig. 4C, c). By combining fluorescence imaging of SRB-labelled astrocytes, whose spectrum is separated from the IOS infrared signal, we observed that astrocytes swelling (i.e., decrease in SRB fluorescence) paralleled CSD swelling (Fig. S3). Consistent with our observation on astrocyte volume response (Fig. 3D), when pre-incubating slices with TGN-020 to block AQP4 tonic water outflow, the initiation of both the CSD and general swelling was accelerated while their maximum amplitude reduced (Fig. 4D-E; Fig. S2B). Blocking action potentials with tetrodotoxin (TTX) only reduced the amplitude of the initial CSD response while the effect on general swelling is insignificant (Fig. 4D-E). We further followed the swelling of individual SRB-labelled astrocytes during Opto-CSD. In consistence with the result obtained from hypotonic challenge (Fig. 3D), the presence of TGN-020 reduced the peak amplitude of astrocyte swelling (i.e., the maximal SRB fluorescence decrease; Fig. 4F). As a post hoc adaptation to the transiently induced Opto-CSD, astrocyte swelling is followed by a shrinking to recover cellular volume to the homeostatic level, which was recorded as a rebound in SRB fluorescence intensity. We noted that TGN-020 presence significantly restrained the recovery of SRB fluorescence back to the baseline (Fig. 4F), showing that AQP4 inhibition hinders astrocyte water efflux therefore the shrinking.

Fig. 4.

Collectively, our data show that astrocyte AQP4 sustains a tonic water outflow regulating the cellular volume response and the general cell swelling of the parenchyma.

Tonic astrocyte water transport underlies brain homeostasis

Water homeostasis sets the basis for molecular diffusion in the brain, which instructs neurotransmitter availability, ion recycling and metabolite trafficking. We then examined in vivo the role of astrocyte aquaporin outflow in brain water diffusion using DW-MRI (Le Bihan & Iima, 2015). It uses the diffusion of water molecules to generate contrast that can be quantified by the apparent diffusion coefficient (ADC) in the nerve tissue (Beaulieu, 2002, Le Bihan, 2014). We used a 7T MRI to image global brain water diffusion in mice lightly anesthetized with medetomidine that normalized the animals under the sedative state. Astrocyte AQP4 outflow was acutely perturbed by TGN-020 applied via intraperitoneal injection. Brain water diffusion was mapped every 5 min before and after TGN or saline (as control) administration (Fig. 5A). Each acquisition sweep was performed with three b values (0, 250, 1800) that accounted for ADC, so to fit the exponential curve (Fig. 5B). Compared to the control saline, TGN-020 increased basal water diffusion within multiple regions including the cortex, hippocampus and the striatum in a heterogeneous manner (Fig. 5C). The temporal features in water diffusion change appeared to be also different: an early elevation followed by a reversible tendency was observed for cortical areas, a rapid and long-lasting elevation for the striatum while a delayed and transient increase for the hippocampus (Fig. 5D). The in vivo neuroimaging results confirm that the tonic water efflux from astrocyte aquaporin contributes to maintain the homeostasis of brain water diffusion, and also suggest spatiotemporal heterogeneities in brain water handling.

Fig. 5.

Discussion

Water equilibrium underlies brain function, plasticity and dynamics. Astrocytes are the principal brain cell type expressing aquaporin water channels that are highly implicated in regulating local environments in the parenchyma (Xiao et al., 2023). We show that albeit being a bidirectional channel, astrocyte aquaporin sustains a tonic water outflow in basal states to ensure structural and functional stability, as well as the water homeostasis at the brain level.

The observation of a basal aquaporin water efflux implies there is a constitutive water accumulation in brain astrocytes. As a ubiquitous vehicle for transporting ions, transmitters and metabolites, water enters astrocytes via a wide range of ion channels, transporters and exchangers. For instance, they express Na⁺/K⁺ and Na⁺/HCO ^- cotransporters to dynamically regulate intra- and extracellular ion homeostasis (MacAulay, 2021). Standing as a major glial cell type controlling the neuropil environment, astrocytes take up synaptically released K⁺ through inwardly rectified K⁺ channels and neurotransmitters (e.g., glutamate, GABA) via high-affinity transporters to safeguard synaptic transmission (Dallerac et al., 2018). The transmembrane molecular exchange may lead to continuous water entry into astrocytes. Moreover, astrocytes juxtapose the cerebral vasculature, being a front relay station for brain metabolism. They express glucose transporters and lactate-permeable monocarboxylate transporters facilitating energy substrate uptake and transfer between blood vessels and neuron-glia networks (Cauli et al., 2023). As water is constantly produced during metabolic processes (Bonvento & Bolanos, 2021), astrocyte metabolism would also contribute to the constitutive water accumulation in the cytoplasm. An efficient efflux pathway appears necessary to counterbalance the constitutive water buildup, thereby maintaining astrocyte and brain homeostasis. Our data show that astrocyte AQP4 fulfills such a role by mediating tonic water outflow (Fig. S4), lending functional supports to the previous suggestion (Amiry-Moghaddam et al., 2003). It has been observed that knock out of AQP4 leads to water over accumulation in the brain, likely due to the reduction in water exit from the pial membrane (Haj-Yasein et al., 2011, Vindedal et al., 2016). Our result also echoes that in the principal cells of the kidney collecting duct, AQP4 is suggested to mainly mediate water exit to balance the AQP2-sustained water entry (Noda et al., 2010). Moreover, we observed that inhibiting astrocyte basal water efflux alters Ca²⁺ activities in both astrocytes and neurons, reminiscent of the early finding that inhibition of AQP4 by TGN-020 enhances cerebral blood flow in mouse cortex (Igarashi et al., 2013).

This study also provides mechanistic hints to understand AQP4-relevant pathologies in specific contexts. For instance, astrocyte water accumulation and swelling are implicated in the development of brain edema. Different types of edema have been delineated depending on the pathological contexts, including such as cytotoxic, vasogenic and hydrocephalic subtypes. The presence of AQP4 has been found to ameliorate vasogenic (vascular leak) and hydrocephalic edema (Jeon et al., 2021, Tait et al., 2010), where the excessive water has been suggested to exit the brain via AQP4-dependent route. This is consistent with our observation that astrocyte AQP4 contributes to basal water extrusion. Hence aquaporin facilitators would help to control vasogenic and hydrocephalic edema. Differently, inhibiting AQP4 has been observed to ameliorate cytotoxic edema encountered during cerebral ischemia (Verkman et al., 2011, Tait et al., 2010). Beneficial effects of TGN-020 on edema have been reported in spinal cord injuries (Li et al., 2019), in ischemia (Sun et al., 2022) and in retinal edema in diabetic animal models (Oosuka et al., 2020). These observations suggest that the role of AQP4 in pathological conditions is context-dependent. Indeed, while our results suggest astrocyte AQP4 mediates a water outflow in basal states, the direction of water transport in specific contexts will be determined by the actual transmembrane water gradient. As the origin of edema is variable and water is among the many other ions involved in edema formation (Stokum et al., 2016), astrocyte aquaporin may either be the primary cause for cytosolic water accumulation or secondarily sustain protective water extrusion. In neuromyelitis optica spectrum disorders (NMOSD), an unusual nervous autoimmune disorder often featured with circulating IgG autoantibody against AQP4, the impairment in astrocyte water homeostasis has been observed (Hinson et al., 2012, Lucchinetti et al., 2014). The inhibition of astrocyte AQP4 by autoantibody may initially affect the basal water efflux. Yet chronic pathological evolution could convert astrocytes into reactive states, with either or both the aquaporin expression, intra- and extracellular environments being altered, the role of AQP4 in water transport might be accordingly shifted (Mireles-Ramirez et al., 2022). Additionally, pro-inflammatory cytokines and chemokines during the immunological maladaptation could also contribute to shape astrocyte functions including aquaporin-mediated water transport (Lucchinetti et al., 2014, Mireles-Ramirez et al., 2022). Recent data show that TGN-020 targeting of AQP4 mitigates the inflammation during ischemia-reperfusion injury, by enhancing glymphatic amyloid clearance and inhibiting the ERK 1/2 pathway (Li et al., 2023), suggesting that in pathological conditions, multiple factors could be recruited by astrocyte aquaporin targeting.

We observed that acutely blocking aquaporin caused basal water accumulation leading to swelling in astrocytes, and facilitated the initial occurrence of the evoked swelling by hypoosmoticity and during cortical spreading depression. This recalls early observation by electron microscopy that knocking out the anchoring protein alpha-syntrophin to disperse AQP4 clustering induces an enlargement of astrocyte endfeet (Amiry-Moghaddam et al., 2003). It likely reflects astrocyte local swelling caused by the disruption of AQP4-mediated water efflux. Our observation also parallels the report that AQP4 knock out facilitates astrocytes swelling induced by hypoosmotic solution (Murphy et al., 2017). Nevertheless, genetic inactivation of AQP4 also reveals inhibitory effects on astrocyte swelling (Benfenati et al., 2011, Mola et al., 2016, Woo et al., 2018). Such discrepancy might be due to the variable compensations in brain water content and structure integrity during chronic genetic manipulations (Binder et al., 2004, Gomolka et al., 2023, Haj-Yasein et al., 2011, MacAulay, 2021, Yao et al., 2015). In addition, though our data suggest that under basal condition, astrocyte AQP4 sustains a tonic water outflow, in conditions when massive water influx persists as exemplified by the application of hypoosmotic solutions, the direction of water transport of AQP4 would follow the imposed water gradient.

Astrocyte aquaporin modulates brain water transport (MacAulay, 2021), though its role in the glymphatic system is under deliberation (Mestre et al., 2018, Rasmussen et al., 2022, Smith & Verkman, 2019, Smith et al., 2017). Our data suggest that astrocyte aquaporin-mediated outflow helps to maintain local water environment in the brain parenchyma. Moreover, besides affecting water transport, targeting AQP4 also impacts astrocyte volume therefore the size of extracellular space. These two factors would need to be considered when evaluating AQP4 involvement in brain fluid transport.

We show by DW-MRI that water transport by astrocyte aquaporin is critical for brain water homeostasis. When blocking AQP4 with TGN-020, therefore its basal water outflow, we observed spatially heterogeneous elevations of diverse temporal kinetics in brain water diffusion rate. Our current data are derived from 5 min apart acquisitions, providing information over the early phase of AQP4 inhibition therefore extending our early report (Debacker et al., 2020). The regional heterogeneity likely reflects the various levels of AQP4 expression; its relative enrichment in the cerebral cortex (Clarke et al., 2018) corresponds to the pronounced effect observed here by DW-MRI. An overall increase in brain water diffusion rate was observed when blocking AQP4 water efflux. This treatment causes water accumulation inside astrocytes and their swelling, which reduces the extracellular space but increases the intracellular space. Water diffusion would be enhanced inside astrocytes and decreased in extracellular space, respectively. DW-MRI maps global water diffusion of both intra- and extracellular space. Likely, a net increase in brain water diffusion may reflect its intracellular increase exceeding the extracellular decrease. In addition, convective brain fluid flow has been suggested to be present in the perivascular space (Smith & Verkman, 2019), a scenario extended by the glymphatic system to the extracellular space of the parenchyma (Mestre et al., 2018). In this sense, transiently squeezing extracellular space might increase the rate of the water flow. The current MRI data yet lack sufficient spatial resolution to delineate the spatial compartmentalization of brain fluid flow.

Our study sheds light on the mechanisms by which astrocyte aquaporin contributes to the water environment in brain parenchyma, and will help to understand the processes underlying brain homeostasis and adaptation to life conditions.

Materials and methods

Animals and acute brain slice preparation

Experiments were undertaken in accordance with European Community guiding principles on the care and use of animals (86/609/CEE), and the rules of the host institute. For water transport and volume imaging, coronal slices comprising somatosensory cortex were acutely prepared from C57BL/6 mice of both sexes at ages of 4-6 weeks, unless otherwise indicated. Mice were deeply anesthetized by isoflurane (Axcience) evaporation in a closed plexiglass box. The brain was taken out and placed in a modified artificial cerebrospinal fluid (aCSF) for slicing (in mM: 30 NaCl, 4.5 KCl, 1.2 NaH2PO4, 1 MgCl2, 26 NaHCO3, 10 D-Glucose, and 194 sucrose) maintained at 4 °C during sectioning. The brain was cut into 300-μm thick slices with a vibratome (Leica VT1200S). Brain slices were recovered in standard aCSF (mM: 124 NaCl, 4.5 KCl, 1.2 NaH2PO4, 1 MgCl2, 2 CaCl2, 26 NaHCO3, and 10 D-Glucose) at 37 °C for about 1 h, and the same aCSF was used for brain slice imaging at room temperature. For optogenetic cortical spreading depression (opto-CSD) experiments, 18 to 21 day-old mice were deeply anesthetized with 150 μL of isoflurane (ISOVET, isoflurane, Piramal/healthcare). After the decapitation, the brain was quickly removed and placed into ice-colded oxygenated aCSF containing the following (in mM): NaCl 125; NaHCO3 26; Sucrose 15; Glucose 10; KCl 2,5; CaCl2 2; NaH2PO4 1,25; MgCl2 1; kynurenic acid 1 (nonspecific glutamate receptor antagonist, Sigma-Aldrich). Coronal slices from mouse somatosensory cortex (300μm thick) were cut with a vibratome (VT1000, Leica) and allowed to recover at room temperature for at least 45 min in aCSF saturated with O₂/CO₂ (95%/5%).

In vivo chemical and genetic targeting of astrocytes

Mice of similar age were used also for in vivo labeling of astrocytes, where the astrocyte-specific red fluorescent dye sulforhodamine B (10 mg/ml, Sigma) was intraperitoneally injected into awake mice at a dose of 10 μL/g. The genetically encoded Ca²⁺ indicator GCaMP6 was expressed in astrocytes in vivo by crossing a Cre-dependent GCaMP6f mouse line (Ai95, The Jackson Laboratory) with an astrocyte-specific Glast-Cre^ERT2 mouse line (Slezak et al., 2007). Tamoxifen (T5648, Sigma) was dissolved in ethanol 100 % (at 1 mg/250µL) and vortexed until tamoxifen is fully dissolved. Corn oil (sigma) were then added to have a final solution at 10% ethanol and 90% oil. Solution is then heated at 37°c for 15 min and vortexed, and put in sonicator for 15 min to reach transparency, and then aliquoted. All procedures were made under light-proof protocols. Aliquots were stored at -20 °C until use. GCaMP6f/Glast-Cre^ERT2 mice were injected with tamoxifen at ∼3–4 weeks of age, once a day (1 mg) for two consecutive days. Genotyping of the Cre-dependent GCaMP6f mouse line (Ai95, The Jackson Lab) used the standard primers and polymerase chain reaction protocols provided by the supplier. Genotyping of GLAST-Cre^ERT2 mouse line was performed using the primers for Cre recombinase (TK139/141) as reported (Slezak et al., 2007). For Ca²⁺ imaging in somatostatin interneurons, homozygous mice B6J.Cg-Sst^{tm2.1(cre)Zjh}/MwarJ (SST-Cre, Jackson Laboratory) (Taniguchi et al., 2011) have been crossed with homozygous Ai95^{GCaMP6f/GCaMP6f} mice to obtain heterozygous SST-cre^cre/WT::Ai95^GCaMP6f//WT mice. For opto-CSD, homozygous mice B6;129PEmx1tm1lr/J ((Gorski et al., 2002), Emx1-Cre^cre/cre, Jackson Laboratory) have been crossed with homozygous mice B6;129SGt(ROSA)26Sortm32.1(CAG-COP4*H134R/EYFP)Hze/J ((Madisen et al., 2012) Ai32^ChR2/ChR2, Jackson Laboratory) to obtain heterozygous Emx1-cre^cre/WT::Ai32^ChR2/WT mice. All procedures using animals were carried out in strict accordance with French regulations (Rural Code R214/87 to R214/130) and conformed to the ethical recommendations of the European Economic Community (Directive 86/609/EEC) and the National Charter French on ethics in animal experimentation. All protocols were approved by the Charles Darwin ethics committee and submitted to the French Ministry of Education and Research (Approval 2015 061011367540 APAFIS#573-2015061011367547 v1).

Ex vivo fluorescence imaging

Light sheet and epifluorescence imaging were both performed using a wide-field upright microscope (Zeiss Axioskop 50, Germany) equipped with water-immersion objectives. Epifluorescence illumination was provided by a monochromator light source (Polychrome II, TILL Photonics, Germany) directly coupled to the imaging objective via an optical fiber. We used water immersion imaging objectives (ZEISS): ×10 NA0.3 used to have a global view, ×20 NA0.5 used to perform light sheet imaging, and ×63 NA1.0 to image individual astrocytes under epifluorescence configuration so as to collect their whole-cell SRB fluorescence to follow astrocyte water transport and volume dynamics. The green fluorescence of GCaMP6 was separated from the red fluorescence of SRB, by spectrally exclusive double-band filters (Di03-R488/561-t3 and FF01-523/610, Semrock). SRB-labeled astrocytes were excited at 550 nm. Fluorescence signal was collected using a digital electron-multiplying charge-coupled device (EMCCD Cascade 512B, Photometrics).

A light sheet imaging system was adapted from our previous publication (Pham et al., 2020), where a chamber with a slightly tilted surface was used so as to spare the agarose embedding and allow the light sheet to access to any parts of the brain slice (Fig. S1). Briefly, the in-focus light sheet was introduced laterally from an independent optical module (Alpha3 light sheet add-on, Phaseview, France) equipped with an air objective (Zeiss EC EPIPlan ×10, 0.25NA). This optical module was orthogonal to the upright imaging pathway and coupled via a wavelength combiner (Thorlabs) to two continuous wave lasers 473-nm and 561-nm (CNI, China), for the excitation of GCaMP6 and SRB, respectively. The imaging chamber (20 × 30 × 46 mm width, height and length) was custom-designed using 3D modeling software (Trimble SketchUP), produced by a Mojo 3D printer (Stratasys) where the horizontal slice-supporting surface in the chamber was tilted by 8° along the laser entry direction. In this way, the brain slice could be laid on the slightly tilted surface while being stabilized with a conventional Harp slice anchor grid (Warner Instruments), whereby the laterally generated laser light sheet could reach local regions throughout the brain slices. The chamber was mounted on a motorized PI stage (Physik Instrumente GmbH, Germany) for axial micro-manipulation. Laser excitation and image acquisition were controlled by MetaMorph (Molecular Devices). Fluorescence dynamics of individual astrocytes was analyzed with ImageJ software (NIH). Background averaged from non-fluorescent subregion was subtracted from time-lapse images. Because single astrocytes typically occupy an area of ∼50-100 µm in radius (Appaix et al., 2012), regions of interest of 50-µm diameter were delineated surrounding identified astrocyte footprint, to comprise the extended area of individual astrocytes while minimizing interference with neighbors. Time courses extracted from single regions of interest were normalized to the baseline (F0) expressed as dF/F₀.

Dynamic imaging of the extracellular space and astrocytes volume changes

Individual slices were transferred to a submerged recording chamber and perfused with oxygenated aCSF (∼2 ml/min) lacking kynurenic acid. Slices were observed using a double port upright microscope (BX51WI, Olympus, Japan) equipped with Dodt gradient contrast optics (DGC; Luigs and Neumann), collimated light emitting device (LED; 780 nm, ThorLabs) as transmitted light sources, ×4 (PlanN, 0.10NA, Olympus), ×20 (UMPlanFL N, 0.50 W, Olympus) or ×40 (LUMPlanF/IR, 0.80 w, Olympus) objectives, and a digital camera (OrcaFlash 4.0, Hamamatsu) attached on the front port of the microscope. The observation with ×4 allows to delimit the cortical layers and to confirm the presence of “barrels” in layer IV. The ×20 objective was used to observe infrared intrinsic optical signals (IOS) (Holthoff & Witte, 1996, MacVicar & Hochman, 1991) of the cortical layers I to IV. The ×40 objective was used to image SRB labelled astrocytes combined with IOS in layers I and II/III.

After one minute of baseline period, a 10-s continuous photostimulation at 470 nm of the ChR2_H134R (30% of the maximal power of the LED) was carried out via the epifluorescence port of the microscope using a collimated LED system (CoolLED, PreciseExcite) and a set of multiband filters consisting of an excitation filter (HC 392/474/554/635, Semrock), a dichroic mirror (BS 409/493/573/652, Semrock) and an emission filter (HC 432/515/595/730, Semrock) to observe the induced changes of light transmittance. The infrared and epifluorescence images were acquired each at 1 Hz (exposure 100 ms, binning 2×2) using the Imaging Workbench 6.1 software (Indec Byosystems) for a total duration of 15 min.

In vivo fiber photometry

The optical fiber (200 µm core, NA = 0.37, Neurophotometrics) coupled to a stainless-steel ferrule (1.25 mm) was stereotaxically implanted in mouse S1 cortex (rostro-caudal 0 mm, lateral +/-2.5 mm, vertical -1.2mm), and fixed to the skull with dental cement (SuperBond, Sun medical). Mice were habituated to the cage and fiber photometry setup for 30 minutes. Then, fluorescence was recorded in mobile mice. 5-min baseline was recorded before SRB injection, then SRB was applied to label brain astrocytes via intraperitoneal injection and the fluorescence time course was continuously recorded. To examine in vivo the effect on astrocyte water transport of acute AQP4 blocking, either TGN-020 (200 mg/kg; Tocris or MedChemExpress) or saline control was intraperitoneally injected 1 h after SRB injection, a time point when astrocytes were well labeled as confirmed by brain slice imaging.

Fluorescence signals were recorded using a Neurophotometrics fiber photometry system (FP3002). A branched fiber optic patch cord (BFP_200/230/900-0.37_FC-MF1.25, Doric Lenses) connected to the fiber photometry apparatus was attached to the implanted fiber optic cannula using a cubic zirconia sleeve. To record fluorescence signals from SRB, excitation light from 560-nm LED was band-pass filtered, collimated, reflected by a dichroic mirror and focused by a ×20 objective. The excitation light power at the tip of the patch cord was 50 µW. Emitted fluorescence was band-pass filtered and focused on digital camera. Signals were collected at a rate of 1 Hz for SRB and visualized using the open-source software Bonsai 2.4 (http://bonsai-rx.org) and analyzed in Igor software (Wavemetrics).

In vivo diffusion-weighted MRI

To test the effect of astrocyte AQP4 acute inhibition on brain water homeostasis, we used diffusion-weighted magnetic resonance imaging (DW-MRI) to calculate the apparent diffusion coefficient (ADC) of tissue water molecules in the brain. MRI experiment was performed with medetomidine anesthesia (0.6 mg/kg/h, i.v.). The mice were injected with either the water-soluble TGN-020 sodium salt (Key Organics, LR-0041) or saline. For the crossover test, the order of injection of TGN-020 or saline was randomly determined. The respiration was monitored, and the body temperature was maintained at 36 °C throughout the measurement.

The MRI experiments were conducted on a Bruker 7T scanner with a cryogenic radio frequency coil. After 10 min from the start of the scanning, either TGN-020 (200 mg/kg body weight) or Saline was injected. Then, scanning was continued for 60 min. The respiration rate was monitored and was confirmed within the range of 80-104 /min throughout the experiment. Mice were injected with 200 mg/kg of TGN-020. We used a standard diffusion-weighted spin echo EPI sequence, with the following parameters: spatial resolution = 175 × 175 × 500 μm³/voxel, 3 b-values (b = 0, 250 and 1800 s/mm²), 6 directions, 1 segment, echo time = 37.1 ms / repetition time = 5769 ms, bandwidth = 300 kHz, δ = 3 ms, Δ = 24 ms, scan time = 5 min. The DWI scan was continued for 70 min (total 14 scans).

The ADC at each time point was calculated from DWI data using DSIstudio (https://dsi-studio.labsolver.org/). The percentage change in ADC was then calculated according to the equation: ΔADC_i = (ADC_i / ADC₂ - 1) × 100 (%). The ADC_i is the ADC at the i-th timepoint. We define the ADC at the 2^nd time-point as the baseline because saline injection was performed after the 2^nd DWI scan. The averaged ADC changes within the regions of interests (ROIs) were calculated using a homemade program. The ROIs of the cortex, the hippocampus, and the striatum were created with reference to Allen mouse brain atlas.

Statistics

Experimental data are expressed as mean ± standard error unless otherwise mentioned. The t-test was performed for two-group comparisons and significant difference was determined by p values less than 0.05. Mann-Whitney U test was also used to ascertain the significance level, in considering the deviation of the experimental data from the theoretical symmetrical distribution. Bonferroni-Holm correction was used for multiple comparisons. Statistical tests were carried out with Matlab (The MathWorks) and Statistica 6 (Statsoft). The significance levels are shown in figures by *p < 0.05, **p < 0.01, ***p < 0.001.

Acknowledgements

We thank the animal and imaging facilities of the IBPS (Sorbonne Université, Paris, France). We thank Nathalie Rouach for the GFAP-EGFP mice, Hervé Le Corronc and Thomas Panier for the discussion and the 3D printing of the light sheet chamber, respectively. This work was supported by the Agence Nationale de la Recherche (ANR-17-CE37-0010-03; ANR-20-CE14-0025-02), France Alzheimer individual fellowship (2021-2022), Japanese society for promotion of science (JSPS) 2022 summer grant, Grant-in-Aid for Challenging Research (Exploratory) in Japan (grant number 21K19464), the i-Bio initiative grant of Sorbonne Université. BLG was supported by a fellowship from the Fondation pour la Recherche sur Alzheimer.

Competing interest

The authors declare no competing financial interests.

Data sharing plans

The authors commit to share data, documentation, and code used in analysis.