Astrocyte regulation of neural circuit activity and network states

Abstract

Astrocytes are known to influence neuronal activity through different mechanisms, including the homeostatic control of extracellular levels of ions and neurotransmitters and the exchange of signaling molecules that regulate synaptic formation, structure, and function. While a great effort done in the past has defined many molecular mechanisms and cellular processes involved in astrocyte-neuron interactions at the cellular level, the consequences of these interactions at the network level in vivo have only relatively recently been identified. This review describes and discusses recent findings on the regulatory effects of astrocytes on the activity of neuronal networks in vivo. Accumulating but still limited, evidence indicates that astrocytes regulate neuronal network rhythmic activity and synchronization as well as brain states. These studies demonstrate a critical contribution of astrocytes to brain activity and are paving the way for a more thorough understanding of the cellular bases of brain function.

Graphical Abstract

Bidirectional astrocyte-neuron communication at tripartite synapses

Astrocytes were classically recognized as a major type of glial cells with important supporting roles for neuronal function, such as the control of extracellular levels of ions and neurotransmitters. Historically, they were not considered to actively participate in brain information processing, probably because, unlike neurons, they lack electrical excitability and they are not directly connected with peripheral sensory organs. However, in the 1990s, this view was challenged by the introduction of functional imaging fluorescence techniques in cell culture preparations that showed that astrocytes display a form of cellular excitability based on intracellular calcium variations, which can be manifested spontaneously or evoked by neurotransmitters. Further studies in more intact preparations, like brain slices, demonstrated that astrocytes sense synaptic activity through the expression of ion channels, transporters, and receptors (Araque et al., 2001, 2014; Volterra et al., 2014) and that the astrocyte calcium signal can be triggered by different synaptically released neurotransmitters, such as glutamate, GABA, acetylcholine, endocannabinoids, ATP, norepinephrine, and dopamine (Araque et al., 2014; Haydon & Carmignoto, 2006; Perea et al., 2009).

In turn, astrocytes can release neuroactive molecules, called gliotransmitters, such as glutamate, GABA, ATP, adenosine, or D-serine, that activate neuronal receptors (Araque et al., 2014; Halassa & Haydon, 2010; Volterra & Meldolesi, 2005). Several mechanisms, not mutually exclusive, have been proposed to mediate the release of gliotransmitters (Verkhratsky & Nedergaard, 2017), among which the calcium-dependent mechanism stands out as a major one as it is also linked with the astrocyte cellular excitability. Through the release of gliotransmitters, astrocytes have been found to modulate neuronal activity and synaptic transmission in several brain areas (Araque et al., 2014; Perea et al., 2009) and to impact animal behavior (Kofuji & Araque, 2021b; Oliveira et al., 2015).

The bidirectional communication between astrocytes and neurons is embodied in the concept of the tripartite synapse that includes astrocytes as integral elements of synaptic function (Araque et al., 1999; Perea et al., 2009). The functional implications of this concept on brain function are remarkable. A single astrocyte may contact more than 100000 synapses (Bushong et al., 2002) that can be independently regulated (Covelo & Araque, 2016), and an individual astrocyte may release different gliotransmitters with different regulatory effects at particular synapses (Covelo & Araque, 2018). Moreover, such complexity is further enhanced when considering that brain nuclei contain thousands of astrocytes that may regulate thousands of synapses in local circuits. Altogether, the regulatory role of astrocytes at individual synapses provides a huge number of degrees of freedom to the possible functional states of the synapses and hence the circuit, therefore increasing the computational power of the system enormously.

While the properties of tripartite synapses at the synaptic level and their consequences on animal behavior have been recently reviewed elsewhere (e.g., Araque et al., 2014; C. A. Durkee & Araque, 2019; Kofuji & Araque, 2021b; Oliveira et al., 2015), this review will focus on recent findings on the regulatory roles of astrocytes on neural networks. For that, we first discuss the latest studies reporting how astrocytes sense neuronal network activity by responding with Ca²⁺ activity to neural signals in vivo. Next, we discuss how astrocyte activation/manipulation results in the modulation of neural network activity.

Astrocytes display calcium-based cellular activity in vivo

The astrocyte Ca²⁺ signal manifested as intracellular Ca²⁺ variations is a crucial signaling event in the bidirectional communication between neurons and astrocytes. While the complexity of the astrocyte calcium signal and the availability and limitations of probing and analytical tools have been excellently discussed elsewhere (Rusakov, 2015; Semyanov, 2019; Semyanov et al., 2020; Shigetomi et al., 2016; Volterra et al., 2014; Yu et al., 2020), we here focus on the available evidence of the astrocyte activity in vivo.

Extensive studies performed in brain slices since the mid-1990s have demonstrated that astrocytes express a plethora of neurotransmitters receptors, many of which are G protein-coupled receptors (GPCRs) that, upon stimulation by neurotransmitters, elevate intracellular Ca²⁺ through its mobilization from internal stores (Kofuji & Araque, 2021a). The introduction in the 2000s of two-photon microscopy allowed monitoring astrocyte calcium activity in vivo. The pioneering study by Hirase et al. (2004) showed that cortical astrocytes display spontaneous calcium events similar to those observed in slices. Moreover, these events could be regulated by manipulating neural network activity, suggesting the existence of neuron-to-astrocyte communication in the intact brain. Indeed, subsequent studies demonstrated that astrocytes in the hippocampus, cerebellum, and cortex respond in vivo to sensory stimuli (Navarrete et al., 2012; Paukert et al., 2014; Reynolds et al., 2019; Takata et al., 2011; Wang et al., 2006; Zhao et al., 2012). Moreover, sensory-evoked cortical astrocyte network activity has been shown to depend on the frequency, duration, and intensity of peripheral stimuli (Lines et al., 2020). Astrocytes in other areas of the central nervous system have also reported to respond to sensory stimuli. In the olfactory bulb, astrocytes were found to respond with calcium elevations to odorants as physiological stimuli (Otsu et al., 2015). Astrocytes in the spinal cord show large-scale coordinated calcium responses to intense but not weak sensory inputs (Sekiguchi et al., 2016). It is worth mentioning that astrocyte activity is strongly depressed by anesthesia, which may be affecting neuronal and synaptic function or intrinsic astrocytic properties. Interestingly, the basal Ca²⁺ levels determine the scale of astrocytic Ca²⁺ upon activation, which is conserved across ex vivo and in vivo preparations (King et al., 2020).

Different neurotransmitters and neuromodulators have been shown to stimulate and regulate the astrocyte calcium signal in different brain areas, such as glutamate in the cortex and inferior colliculus (Kellner et al., 2021; Wang et al., 2006), acetylcholine in the cortex and hippocampus (Navarrete et al., 2012; Takata et al., 2011), norepinephrine in the cerebellum and visual cortex (Kjaerby et al., 2017; Nagai et al., 2021; Oe et al., 2020; Paukert et al., 2014; Ye et al., 2020), GABA in the prefrontal cortex (Mederos et al., 2021), ATP in the cortex and hippocampus (Delekate et al., 2014; Reichenbach et al., 2018), serotonin in the nucleus of the solitary tract (Mastitskaya et al., 2020) and dopamine in the nucleus accumbens (Corkrum et al., 2020).

While studies performed in brain slices have shown astrocyte calcium responsiveness to most neurotransmitters, evidence of their action in vivo for some of them, such as endocannabinoids or opioids that have been shown to induce astrocyte calcium elevations in brain slices (Corkrum et al., 2019; Navarrete & Araque, 2008), is still lacking.

Two-photon microscopy allows imaging astrocyte activity in vivo in anesthetized or head-restrained animals, yet it presents a major limitation to monitoring in vivo astrocyte calcium activity in deep brain structures. Alternative strategies have been successfully used to overcome this constraint, such as the use of GRIN lens for MINIscopes, which allowed monitoring astrocyte activity in the dorsal striatum (Yu et al., 2018), and a fiber-photometry system, which allowed recording astrocyte activity in the nucleus accumbens (Corkrum et al., 2020).

In summary, the astrocyte calcium signal evoked by neuronal and synaptic activity has been extensively documented in vivo in several brain areas, demonstrating the existence of neuron-to-astrocyte communication in vivo. Moreover, astrocytes are well suited to play an active role in the modulation of neuronal activity as they display several features to allow the local and long distance modulation: (1) astrocytes are ubiquitous in the central nervous system and establish their independent territories, forming a syncytium covering virtually the whole central nervous system (Bushong et al., 2002; Halassa et al., 2007); (2) single astrocytes establish contacts with hundreds of thousands of synapses in the rodent brain (Bushong et al., 2002), (3) they sense a large variety of signaling molecules as they express receptors that are activated by variety of neurotransmitters and neuromodulators (see above) known to regulate neuronal oscillations, synchrony and states; (4) they process and respond to incoming signals in a wide time window from tens of milliseconds to seconds, minutes, hours and days (Araque et al., 2014); and (5) the manipulation of astrocytic functions interferes with a wide array of behaviors (Kofuji & Araque, 2021b; Nagai et al., 2020; Oliveira et al., 2015).

The following sections will describe how such astrocyte activity impacts neural network function, discussing the available evidence that implicates the modulation of the power and synchronization of neuronal oscillations and brain state by astrocytes, potentially influencing behavior. While the modulation of neuronal activity by astrocytes at the single-synapse level has been discussed in several reviews (e.g., Araque et al., 2014; C. Durkee et al., 2021; Rusakov & Stewart, 2021; Semyanov & Verkhratsky, 2021), this review will focus on its extension to the network scale of local circuits (e.g., stratum radiatum of the hippocampus) or long-distance circuits (e.g., cortico-limbic networks). To address these complex levels of brain function, we will discuss studies that were mostly performed in vivo. We will emphasize novel approaches, and we will provide details on technical aspects that are important to data interpretation (e.g., studies in anesthetized VS. awake animals), as well as details on the brain circuits studied and implications to behavior.

Astrocytes modulate the activity of neuronal networks

Astrocytes have been reported to regulate neuronal oscillations in different brain areas and through different mechanisms (Figure 1). Extracellular K⁺ levels, which greatly influence neuronal firing, are homeostatically controlled by astrocytes that buffer the extracellular K⁺ concentration through the expression of K⁺ channels, like Kir4.1 (Kofuji & Newman, 2004; Olsen et al., 2015; Tong et al., 2014). In brain slices of the somatosensory cortex of mice, the blockade of K⁺ uptake has been shown to enhance neuronal oscillations in a wide range of frequencies. This increase in excitability was also mimicked by affecting the gap-junction connectivity of astrocytes through the blockade of connexin 43 (Bellot-Saez et al., 2018), suggesting an important role of the gap-junction-mediated astrocytic network. Moreover, the increase of neuronal oscillation power was concomitant with an increase in individual spike frequency that might contribute to the oscillation build-up (Bellot-Saez et al., 2018). In the lateral habenula (LHb), astroglial Kir4.1 channels have been shown to be upregulated in rat models of depression, which affects the extracellular K⁺ buffering, the membrane resting potential, and the amount of bursting activity of LHb neurons. Furthermore, astroglial-specific gain and loss of Kir4.1 in the LHb bidirectionally regulated neuronal bursting and depression-like symptoms, indicating that astrocytic homeostatic function regulates the neuronal firing in LHb, and that its dysfunction is involved in a psychiatric disease animal model (Cui et al., 2018).

Figure 1 -. Astrocytes modulate the activity of neuronal networks.

Example of a study showing that astrocyte activity modulates a form of neuronal network output. Left panel: above, schematic drawing of an experimental set to simultaneously record astrocyte Ca²⁺ signals in the primary somatosensory cortex (S1) and ECoG in vivo; below, schematic drawing depicting astrocyte ability to increase or decrease neuronal oscillation power. A, Representative pseudocolor Ca²⁺ images of basal astrocyte activity and during hind paw stimulation. Scale bar = 50 μm. B, ECoG recording (black top trace), spectrograms (middle), and astrocyte Ca²⁺ levels (green bottom traces). The delayed sensory-evoked cortical astrocyte calcium elevation, which occurs after the initial neuronal gamma activity surge, contributes to dampening the gamma activity from an initial peak to a steady-state. Modified from Lines et al., 2020 with permission.

Besides this homeostatic mechanism, direct stimulation of astrocytes has been suggested to tune neuronal electrical activity. A pioneering study using optogenetics to stimulate astrocytes in the respiratory chemoreceptor areas of the brainstem demonstrated that astrocyte activation stimulates neuronal firing rate and triggers respiratory responses in vivo (Gourine et al., 2010). In a series of subsequent elegant studies on astrocyte physiology in different brain stem nuclei, Gourine and Kasparov groups demonstrated that astrocytes are chemosensors of the microenvironment that respond with calcium elevations to changes in extracellular pH, CO_2, and O₂, and release the gliotransmitter ATP to control neuronal firing in those nuclei (for reviews see Gourine & Kasparov, 2011; Teschemacher et al., 2015). Recently, they showed that in the ventral respiratory group of the medulla, astrocyte signaling modulates the rhythm of pre-Bötzinger complex (preBötC) interneurons affecting the adaptive breathing of awake rats. The activation of G_q GPCR-dependent pathways in preBötC astrocytes facilitated the respiratory rhythm, while the blockade of vesicular release in preBötC astrocytes reduced the resting breathing rate and frequency of periodic sighs, decreased rhythm variability, impaired respiratory responses to hypoxia and hypercapnia, and dramatically reduced the exercise capacity (Sheikhbahaei et al., 2018).

In another brain stem nucleus, the trigeminal mesencephalic nucleus (MesV) involved in the masticatory rhythmic behavior, astrocytes were also shown to regulate the capacity to change neuronal firing patterns. This ability depends on the regulation of the extracellular Ca²⁺ concentration through astrocytic S100β (Morquette et al., 2015). Moreover, sensory inputs appear to regulate astrocytic coupling, which in turn is required for neuronal bursting in a region-specific manner (Condamine et al., 2018).

In hippocampal slices, the activation of channelrhodopsin (ChR2)-expressing astrocytes in the CA1 area specifically decreased the power of gamma oscillations mainly mediated by ATP-dependent inhibition of pyramidal neuronal activity (Tan et al., 2017). On the other hand, and independent of the gamma oscillation modulation, optogenetic activation of astrocytes increased the firing frequency of cholecystokinin (CCK)-positive, but not parvalbumin-positive, interneurons (Tan et al., 2017). The firing properties of another type of hippocampal neurons, neuropeptide Y (NPY) interneurons that display long-lasting trains of action potentials, have been reported to be influenced by astrocytes (Deemyad et al., 2018). This study showed that calcium signaling in astrocytes correlated with barrage firing of neighboring interneurons. Moreover, while barrage firing was enhanced by chemical or optogenetic activation of the astrocyte network, it was inhibited by chelating Ca²⁺ in the astrocytic network (Deemyad et al., 2018).

In the cerebral cortex, astrocytes were also described to modulate neuronal excitability, namely in the generation of population UP states (Poskanzer & Yuste, 2011). Specifically, in the mouse somatosensory area S1, the electrical activation of astrocytes while monitoring both electrophysiological and Ca²⁺ activities of the surrounding network allowed the authors to confirm that the stimulation of a single astrocyte activates other astrocytes in the local circuit and could trigger UP state synchronizations of neighboring neurons. Again, consistent with the prominent role of the astrocyte calcium signal, chelating Ca²⁺ in astrocytes inhibited spontaneous and stimulated UP states confirming a causal role in regulating the synchronized activation of cortical neuronal ensembles (Poskanzer & Yuste, 2011).

The astrocytic modulation of neuronal network activity observed in brain slices has also been confirmed in vivo. Earlier studies using anesthetized mice suggest that astrocytes can modulate basal neuronal activity, namely in slower frequencies. Using the dnSNARE model that displays an impairment of exocytosis specifically in astrocytes (Halassa et al., 2009; Papouin et al., 2017; Pascual et al., 2005; Sardinha et al., 2017; Sultan et al., 2015), Fellin and colleagues described a reduction of the power of neuronal oscillations slower than 1 Hz local field potentials recorded from the somatosensory cortex of urethane anesthetized mice (Fellin et al., 2009). These effects appear to be restricted to slower oscillations, as neuronal power is unchanged at faster frequencies in the same region or in the dorsal hippocampus (mice anesthetized with sevofluorane) (Sardinha et al., 2017). Interestingly, evoked astrocyte activity appears to influence neuronal power in slower frequencies, as the astrocyte-specific G_q or G_i/o GPCR activation is enough to raise delta power in the primary somatosensory cortex in urethane-anesthetized mice (C. A. Durkee et al., 2019). In line with these observations, the activation of astrocytic G_q GPCR in astrocytes of the centro medial nucleus of the amygdala was sufficient to decrease the neuronal firing rate also in this limbic region via selective potentiation and depression of inhibitory and excitatory inputs, respectively, to this nucleus (Martin-Fernandez et al., 2017). These results reveal the influential role of astrocytes on neuronal activity, which appears to be present across different brain regions.

In similar experimental settings, astrocytes were also reported to modulate the neuronal response to sensory stimulation. In the somatosensory cortex, the delayed sensory-evoked astrocyte calcium elevation, which occurs after the initial neuronal gamma activity surge, contributes to dampening the gamma activity from an initial peak to a steady-state (Lines et al., 2020). This phenomenon, which is mediated by astrocyte calcium signal G_q GPCR activation on astrocytes, has been proposed to contribute to cortical sensory adaptation (Lines et al., 2020). Finally, the specific deletion of astrocytic GABA_B receptors results in a decrease of theta and slow gamma power also after whisker stimulation, but not in resting conditions (Perea et al., 2016). Once again, this effect appears to be conserved across regions as the latter was observed in the dorsal hippocampus.

While slower neuronal oscillations might be prominent under anesthesia, recordings in awake mice further suggest the astrocyte modulation of neuronal power in faster frequencies. By recording the electrocorticogram of mice that express tetanus neurotoxin selectively in astrocytes, Lee and colleagues showed a marked decrease of gamma oscillations that was accompanied by an impaired novelty recognition performance (H. S. Lee et al., 2014). Consistent with this study, the reduction of astrocyte number in the medial prefrontal cortex of rats through the infusion of the gliotoxin L-AAA leads to a decrease of LFP delta, alpha, and gamma power, and impaired cognitive flexibility (Brockett et al., 2018). Finally, a recent elegant work using mice lacking GABA_B receptors, specifically in medial prefrontal cortex astrocytes, reported a reduction of slow gamma power in this region (Mederos et al., 2021). This electrophysiological deficit was paired with a worse working memory performance, a task that is highly dependent on this brain region. Both electrophysiological and working memory impairments were recovered by optogenetic activation of melanopsin in astrocytes (Mederos et al., 2021). While these works suggest a clear involvement of astrocytes in sustaining fast neuronal activity translated into oscillation power, astrocyte modulation may also lead to neuronal inhibition. On the one hand, the artificial reduction of astrocyte Ca²⁺-dependent signaling in the striatum led to a decrease of medium spiny neuron Ca²⁺ activity and excessive self-grooming in freely behaving mice with head-mounted MINIscopes (Yu et al., 2018). On the other hand, the simultaneous measurement of neuronal Ca²⁺ activity and recording of basal and sensory-evoked local field potentials upon optogenetic G_q GPCR activation of astrocytes in the somatosensory cortex revealed an adenosine mediated reduction of neuronal activity and depression of responses to whisker stimulation (Iwai et al., 2021). Further experiments are now required to assess the translation of neuronal Ca²⁺ activity in oscillation power and/or synchrony and to clarify the astrocytic role in the management of excitation/inhibition balance.

In summary, accumulating evidence supports a clear involvement of astrocytes in the modulation of neuronal network activity. Moreover, although astrocyte-induced reduction of excitability has been observed, most of the evidence described above suggests that astrocyte stimulation results in an enhancement of neuronal excitability. The enhancement of the neuronal excitability may result in the enhancement or decrease of the network activity, depending on the neuronal type, i.e., excitatory or inhibitory, affected (see below). Further studies including different astrocyte populations, other cell types, and brain areas are required to establish whether this is a general phenomenon.

Astrocytes modulate neuronal synchronization

Astrocyte manipulation was also shown to interfere with neuronal synchronization (Figure 2), a feature that is independent of neuronal power. Astrocytic GABA_B receptors were shown to be critical for the theta phase-gamma amplitude coupling within the dorsal hippocampus taking place after whisker stimulation (Perea et al., 2016). More recently, we observed a basal desynchronization of theta oscillation between the dorsal hippocampus and the medial prefrontal cortex in dnSNARE mice, in which astrocytic exocytosis is impaired. This desynchronization was accompanied by a poor performance in cognitive tasks computed by these brain regions, namely in the memory consolidation phase (Sardinha et al., 2017).

Figure 2 -. Astrocytes modulate neuronal synchronization.

Example of a study showing that astrocyte activity modulates a form of neuronal synchronization. Left panel: above, schematic drawing of an experimental set to simultaneously record local field potentials (LFPs) in the dorsal hippocampus (dHIP, blue) and the medial prefrontal cortex (mPFC, orange); below, schematic drawing depicting astrocyte ability to increase neuronal synchronization. A, Overlap of representative theta filtered LFP traces of mPFC and dHIP, recorded from control (WT), mice with astrocytic signaling impairment by blockade exocytosis (dnSNARE) showing a decrease of theta synchronization, and dnSNARE mice after administrating D-serine intraperitoneally, which restored dHIP-mPFC theta synchronization (scale bars: 50 μV, 500 ms). B, Representative heatmaps of dHIP–mPFC spectral coherence for control (WT, top) and mice with astrocytic signaling impairment (dnSNARE, bottom) over time; dashed line indicates the moment of D-serine administration; each spectrogram represents the dHIP–mPFC coherence calculated in intervals of 5 min for 4–40 Hz (theta, 4–12 Hz; beta, 12–20 Hz; low gamma, 20–40 Hz); color range: 0, dark blue; 1, red; D-serine administration rescues the astrocyte-related synchrony deficit in dnSNARE mice. Modified from Sardinha et al., 2017 with permission.

While these studies were conducted in anesthetized animals, a few studies have also provided further evidence of astrocytic involvement in neuronal synchronization in awake animals. The pharmacological treatment with the gliotoxin L-AAA caused a reduction of neuronal gamma power in the medial prefrontal cortex of rats without affecting the phase-amplitude coupling between theta and gamma or alpha and gamma oscillations in this region (Brockett et al., 2018). Notably, the dialysis of the astrocyte-specific Ca²⁺ binding protein S100β increases the phase-amplitude coupling between theta and gamma oscillations and improves cognitive flexibility (Brockett et al., 2018), suggesting that different astrocytic mechanisms might control different features of the neuronal oscillations. Recently, Pereás group showed that along with the reduction of slow gamma power, the deletion of astrocytic GABA_B receptors affects theta phase-gamma amplitude coupling in this region. This reduction of theta-gamma coupling was also paired with a poorer working memory performance. Following the recovery of neuronal slow gamma power described above, both phase-amplitude coupling and working memory impairments were recovered by optogenetic activation of melanopsin in astrocytes (Mederos et al., 2021). In summary, recent studies showed astrocytes are required for the sustainability of neuronal synchronization, at least in the hippocampus. Further studies are required to assess their involvement in neuronal synchronization in different brain areas and behavioral contexts.

Astrocytes modulate brain states

The existence of major changes in brain activity underlies the switch between behavior states (Figure 3). The most striking changes occur between sleep and wakefulness (Gervasoni et al., 2004), but multiple states may also be identified within these phases. States characterized by arousal and attention appear to enhance sensory processing, while sleep periods and quiet wakefulness are relevant for learning and memory (S.-H. Lee & Dan, 2012). Combining electrophysiological techniques to record cortical electroencephalogram (EEG), electromyogram (EMG), or deep local field potentials (LFPs) (Kajikawa & Schroeder, 2011; Steriade et al., 1993) with brain imaging and behavior stratification allowed the identification of different phases during sleep and wakefulness. During sleep, EEG and/or EMG patterns allow the identification of states such as rapid eye movement (REM) and non-REM sleep (Hobson, 2005). Moreover, the brain activity during wakefulness may also wander between inattentive and vigilant states, influencing not only the way sensory inputs are processed but also how information is flowing, how decisions are made, and how actions are taken (McCormick et al., 2020). As discussed above, astrocytic features place them in a pivotal position to integrate neuromodulatory systems underlying brain states (Zagha & McCormick, 2014). The involvement of astrocytes in brain state modulation has been recently discussed in astrocyte-focused reviews (Kastanenka et al., 2020; Santello et al., 2019). The available evidence linking astrocytes to this higher level of neuronal integration is consistent but rather scarce. Consequently, the current understanding of the mechanisms underlying brain states usually disregards the participation of astrocytes (e.g., S.-H. Lee & Dan, 2012; McCormick et al., 2020; Zagha & McCormick, 2014), which, however, should be more thoughtfully considered.

Figure 3 -. Astrocytes modulate brain states.

Example of a study showing that astrocyte activity modulates brain state switch. Left panel: above, schematic drawing of an experimental set to simultaneously record astrocyte Ca²⁺ signals in the primary visual cortex (V1) and local field potentials in vivo; mice expressing astrocytic GCaMP6f were head-fixed on a horizontal treadmill to record astrocyte Ca²⁺, LFP, EMG, and locomotion; below, schematic drawing depicting how brain activity may wander between states. Despite their likely multidimensionality, they are depicted as bidimensional for the sake of simplicity. A, Example of neuronal slow-wave activity (SWA) obtained by filtering LFP (0.5–4 Hz) recordings (top), with corresponding astrocyte Ca²⁺ events (bottom). Behavioral state denoted by color: sleep (red) and wake (locomotory in blue, stationary in cyan). B, Left, the proportion of time mice spend sleeping after CNO administration (during Ca²⁺ suppression period) is increased relative to saline controls, and time in the wake is decreased, suggesting Ca²⁺ suppression is sufficient to increase sleep (paired t-test, n = 11 mice; analyses are performed in the 1 hr, 52-min period of Ca²⁺ suppression). Right, sleep-to-wake transitions and wake-to-sleep transitions are decreased with CNO activation of G_q-DREADD relative to saline. This shows that astrocyte Ca²⁺ events characterize transitions from low to high SWA and that the G_q-DREADD activation of astrocytes is necessary for sleep-wake transitions. Modified from Vaidyanathan et al., 2021 with permission.

In the cortex, two states are generally defined as the more synchronized, slow-oscillation–dominated state and the fairly desynchronized, attentive state (S.-H. Lee & Dan, 2012). The slow-oscillation–dominated state is a neocortical rhythm characterized by synchronized neuronal firing, which is linked with sleep and mnesic processes. Back in 2016, Poskanzer and Yuste described a link between astrocyte activity and state transition by employing a combination of two-photon imaging of calcium dynamics in populations of astrocytes and recording of neuronal electrophysiological and calcium activities under anesthesia. They found that astrocytic calcium activity precedes spontaneous circuit shifts to the slow-oscillation–dominated state and that this state could be induced in local neuronal networks by optogenetic activation of astrocytes. Moreover, they found that astrocytic transients in glutamate co-occurred with shifts to the synchronized state and that optogenetically activated astrocytes could trigger these glutamate transients. This set of experiments provided a causal link between astrocyte activity, glutamate release, and cortical state shift (Poskanzer & Yuste, 2016). In a different study, the simultaneous imaging of astrocytic Ca²⁺ activity and recording of local field potentials in the V1 region of anesthetized mice revealed that astrocyte networks display synchronized recurrent activity coupled to neuronal UP states. Moreover, the extensive synchronization of the astrocytic network preceded the spatial build-up of neuronal synchronization. The potential link of causality was suggested by the fact that neurons surrounded by active astrocytes were more likely to join slow-wave activity (SWA) and that the blockade of astrocytic gap junctional communication or inhibition of astrocytic Ca²⁺ transients reduced the ratio of both astrocytes and neurons involved in SWA (Szabó et al., 2017). Although the use of anesthesia in these studies allowed stable experimental conditions and resembled features of natural sleep, further studies using naturally sleeping animals were required to confirm this link.

Cortical states are tightly related to sleep-wake transitions, as well as with different phases of sleep. For instance, non-REM sleep is characterized by SWA, while during REM sleep or wakefulness, the neuronal activity appears to be fairly desynchronized. Astrocytic signaling was previously shown to reduce sleep pressure and SWA (Halassa et al., 2009), but the details of the astrocytic involvement in the transition between sleep and wake states would require the paired analysis/manipulation of astrocyte Ca²⁺ and neuronal activities during these different phases. A handful of very recent and elegant works using that combinatorial approach suggest the involvement of astrocytes in the modulation of neuronal activity during sleep states. First, Bojarskaité and colleagues reported that astrocytic Ca²⁺ signals in the barrel cortex are reduced during the sleep phase compared to wakefulness. Moreover, they showed that an increase in astrocytic Ca²⁺ signaling precedes transitions from slow-wave sleep to wakefulness. This Ca²⁺ signaling appears to be required to control this transition as the genetic ablation of IP₃R2 (that mediates global Ca²⁺ release) impairs slow-wave sleep and results in an increased number of microarousals, abnormal brain rhythms, and SWA periods and sleep spindles (Bojarskaite et al., 2020). These results are in line with a study conducted in the frontal cortex showing that astrocytic Ca²⁺ changes dynamically with sleep, wake, and sleep loss, and it encodes changes observed in sleep need. Moreover, the reduction of astrocyte Ca²⁺ decreased the sleep drive after sleep loss (Ingiosi et al., 2020). Cortical astrocytes appear to also play a pivotal role in regulating different features of NREM sleep via GPCR signaling. By combining the imaging cortical astrocytic populations, electrophysiological, and chemogenetic manipulations, Vaidyanathan and colleagues found that endogenous Ca²⁺ activity is inversely correlated with SWA and exhibits bidirectional changes before sleep-wake transitions. Furthermore, they showed that astrocytes actively regulate both NREM features via different pathways, confirming a specialization of neurotransmitter-elicited activation of astrocytes. Astrocytic G_i-induced Ca²⁺ leads to an increase in SWA, related to sleep depth, while sleep-wake transitions, related to sleep duration, are dependent on G_q GPCR-induced astrocytic Ca²⁺. Interestingly, in this study, the authors also report that manipulating astrocyte Ca²⁺ in the visual cortex has an impact not only in SWA not only locally but also in the contralateral frontal cortex, which suggests a cortex-wide regulation of neural oscillations (Vaidyanathan et al., 2021). Finally, the recording of astrocytic Ca²⁺ levels in additional brain regions confirmed that Ca²⁺ levels in astrocytes substantially decrease during rapid eye movement (REM) sleep and increase after the onset of wakefulness. These experiments also suggested that the astrocytic involvement in the features of NREM and REM sleep and the transition between sleep/wake states is region-dependent with a more prominent role for cortical and hippocampal astrocytes (Tsunematsu et al., 2021).

The ability of astrocytes to sense and influence neuronal network activity with a relevant behavioral outcome has been elegantly obtained in zebrafish (Mu et al., 2019). Brain stem astrocytes of zebrafish show noradrenergic-evoked cumulative calcium responses during swimming episodes associated with ineffective movement, and, after reaching a threshold, they activate the GABAergic neuronal network to suppress swimming (Mu et al. 2019). Hence, astrocytes integrate and regulate neural network signaling to produce behavioral state changes.

It is noteworthy that the astrocytic modulation of brain states is not exclusively translated into electrophysiological signatures. For instance, astrocytic clocks were shown to drive circadian molecular oscillations in the suprachiasmatic nucleus by reinstating clock gene expression. This way, astrocytes modulate multiple networks, ruling circadian rhythms and behaviors in mice (Brancaccio et al., 2017, 2019).

In summary, this recent body of evidence described the astrocytic modulation of sleep states and sleep/wake transitions. This influence of brain states is extremely relevant for a better understanding of the whole-brain function, and these seminal works paved the way for further studies.

Concluding remarks and outstanding questions

Accumulating evidence clearly supports a direct regulatory effect of astrocytes on the activity of neuronal networks not only in slice preparations but in vivo as well. Most of the studies have been performed in certain paradigmatic brain areas, mainly the cortex, hippocampus, and brain stem nuclei. While it is expected that these regulatory processes are a general feature of the astrocytic function, further studies considering different cell types and other brain areas are needed to confirm if this is the case. The available evidence of astrocyte contribution to neuronal and network activity in vivo is still relatively scarce, yet these studies describe surprising roles for astrocytes as active integrators of neuronal networks. Particularly striking is the fact that astrocyte activation seems to be usually associated with the enhancement of network activity. Moreover, astrocytic signaling appears to be required to sustain intra- and inter-regional neuronal synchronization. Finally, solid evidence demonstrated that astrocytes modulate sleep states and sleep/wake transitions. The last years have witnessed an increasing number of studies combining simultaneous assessment of multiple network activity and behavior upon astrocytic manipulation in vivo. These studies are pioneering a new era in the neuroscience field, in which astrocytes play prominent physiological roles in the neuronal network activity, brain function, and behavior. Yet, as expected in an emerging field, several outstanding questions remain. For example:

1. Is the modulation of neuronal oscillation power and/or synchrony a general capacity off all astrocytes, being dependent on the circuit it is inserted in? For instance, a similar astrocytic mechanism would control the excitatory/inhibitory equilibrium differently depending on the neuronal type modulated (excitatory or inhibitory). Similarly, do astrocytes play similar roles (e.g., positive feedback, threshold setters) in several brain regions, resulting in behavioral consequences (e.g., sleep, memory, feeding, circadian control) that do not reflect functional specialization of astrocytes but of the circuits they regulate?

2. What is the role of astrocytes in circuits that regulate brain states, namely those lying in sub-cortical areas, such as basal forebrain/preoptic area, thalamus, hypothalamus, or brainstem nuclei (S.-H. Lee & Dan, 2012; McCormick et al., 2020)?

3. Does each brain region display a palette of populations of astrocytes, as suggested by recent single-cell transcriptomic studies indicative of cell heterogeneity (Batiuk et al., 2020; Bayraktar et al., 2020), which display specific machinery to control particular circuits?

4. Can astrocytes control the electrical activity of the circuit, despite being devoid of voltage-gated ion channels? Can they work as activity oscillators? Are they regulators of set points/activity threshold/activity accumulators? Do they act as relay nodes in the circuit for positive/negative feedback? Are there other possible activity profiles?

5. What are the additional and emergent properties provided by a non-neuronal cell type - astrocytes - to contribute to brain oscillations, states, and behavior?

We are experiencing the exciting advent of multiscale monitoring of neuronal and astrocytic electrical and Ca²⁺ activities, combined with cell manipulation via opto- and pharmaco-genetics in freely behaving animals. This reveals the unexpected roles of astrocytes in complex features of brain activity that will be revolutionary in our understanding of the cellular basis of brain function and animal behavior. We have described here the early findings on this exciting topic, and we are thrilled and expectant with what the future holds for this emerging field.

Glossary

Neural oscillation

Rhythmic or repetitive continuous oscillatory signals generated by neurons in the central nervous system. Oscillations result from the sum of synchronized neuronal firing and feedback signaling produced by excitatory and inhibitory cells of a given brain circuit. The features primarily characterized in neuronal oscillations are their frequency, amplitude, and phase

Oscillation frequency

The number of oscillations per period

Oscillation amplitude

The size of the signal change in one oscillation

Oscillation phase

The measure of the timing of a given oscillation. It can be measured in angular units, as a full circle represents the entire period of the oscillation, and the angle corresponds to the instant deflection of an oscillation

Oscillation power

A measure of the intensity of neuronal activity in a given frequency band, given by the square of the signal amplitude in that band

Frequency bands

Frequency ranges used to distinguish the rhythms of neural oscillations, generally distributed as Delta (<4 Hz), Theta (4–12 Hz), Alpha (8–12 Hz), Beta (12–20 Hz), Gamma (20–100 Hz)

Slow-Wave Activity (SWA)

A classical EEG signature of slow (0.5 to 4 Hz), synchronized, oscillatory neocortical activity that is typically displayed during non-rapid-eye-movement (non-REM) sleep

Single-neuron activity:

Action potential

Fluctuation of the membrane potential of one single neuron, characterized by a fast depolarization leading to overshoot, followed by rapid repolarization, generally followed by a hyperpolarizing refractory period

Neuronal spike

Extracellular recording that reflects voltage changes generated by a single action potential

Firing rate

Frequency of observation of neuronal action potentials or spikes

Barrage firing

High-frequency trains of action potentials that can last for tens of seconds

Ensemble, local or regional activity:

Local Field Potential (LFP)

The voltage fluctuation recorded in the extracellular space, reflecting the sum of neuronal oscillations in all frequencies

Electroencephalogram (EEG)

Recording of the electrical activity of the brain using metal electrodes placed at the scalp

Cortical electroencephalogram (ECoG)

EEG recorded intracranially, placing the electrodes in contact with the surface of the brain

Synchronization:

Phase synchronization

Measure of synchrony between the phase component of neural oscillations in different bands or brain regions

Spectral coherence

A classical measure of synchronization between two brain regions that considers the magnitude of neural oscillations, quantifying both the phase synchronization and the co-variation of power

Spike–field coherence

The preferential firing of action potentials during a specific phase range of the local field potential oscillations recorded in the same or a different brain region

Phase-amplitude coupling

The alignment/synchronization of phase and amplitude of oscillations recorded in different regions

States:

Brain states

Classically considered as major changes in brain activity that underlie the switch between behavior states such as between sleep and wakefulness, attentive and inattentive states, and so forth

UP/DOWN states

the synchronized activation/depolarization (UP) or inhibition (DOWN) of cortical neuronal ensembles, typically recorded in brain slices, that could last hundreds of milliseconds