Diversity of astrocyte functions and phenotypes in neural circuits

Morphological complexity

A rodent “typical” protoplasmic astrocyte consists of a soma, 4–10 major branches and thousands of branchlets and leaflets that lie close to synapses (Fig 1, Box 2). Astrocytes have highly complex spongiform shapes with diameters of ~40–60 µm and volumes of ~6.6 × 10⁴ µm³ (refs^{38, 39}). About 90–95% of an astrocyte’s area is formed by branches, branchlets and leaflets⁴⁰ (Fig 1), and only ~15% can be observed by GFAP staining^{38, 40}. Neighboring astrocytes display remarkably well delineated territories and interdigitate at the edges³⁸ by as little as ~5% (Fig 2). Comparisons between light and electron microscopy show that leaflets cannot be imaged with conventional light microscopy, and that they display dimensions on the tens-to-hundred nanometer scale at sites of interactions with synapses⁴¹. Hence, astrocytes are remarkably complex cells (Fig 1, Box 2). The functional implications of this complexity and their regional differences are beginning to be explored.

Astrocyte morphology, the number of synapses contained within individual astrocyte territories and the proximity of astrocyte leaflets to synapses varies between brain areas^{35, 36} and changes dramatically during disease and injury³⁷. For example, in hippocampal stratum radiatum individual astrocytes form territories that contain ~140,000 synapses^{38, 39}, and of these ~57% have juxtaposed astrocyte leaflets⁴². However, in neocortex this number may be as low as ~30%, whereas in cerebellum almost all synapses onto Purkinje neurons are contacted by leaflets of Bergmann glia³⁶, a specialized radial astrocyte. Astrocyte leaflets are generally irregular, thin and sheet-like in shape with a clear cytoplasm that contains glycogen granules and ribosomes³⁶. Astrocyte leaflets are devoid of GFAP, which is found only in major branches^{37, 40}, and leaflets lack organelles such as mitochondria, microtubules, and endoplasmic reticulum.

Real-time dynamics of astrocyte leaflets

Studies of hippocampal stratum radiatum show that most dendritic spines are contacted by astrocyte leaflets, which impinge on about one third of the available perimeter of individual synapses⁴². Synapses with astrocyte leaflets at their perimeters tend to be larger than those without⁴³, suggesting that interactions of astrocyte leaflets and synapses may be correlated with synaptic strength⁴⁴. Recent studies using time-lapse imaging methods in situ and in vivo suggest that astrocyte leaflet motility increases in response to synaptic activity and is associated with increased synapse coverage and spine stability⁴⁵. Hippocampal astrocyte leaflet motility was triggered by synaptic glutamate release and mediated by astrocyte metabotropic glutamate receptors and intracellular Ca²⁺ elevations⁴⁵. Moreover, stimuli that caused long-term synaptic potentiation (LTP) of synapses also caused astrocyte leaflets to increase their motility (over ~10 min), which then stabilized after ~30 min around enlarged synapses⁴⁵, recalling electron microscopy data⁴⁶. Key observations were extended to in vivo imaging of astrocytes in the barrel cortex during whisker stimulation, demonstrating that observations made in situ were relevant in the context of an intact brain⁴⁵. In accord, previous studies showed that whisker stimulation increased astrocyte coverage of synapses in the barrel cortex⁴⁷. Broadly similar conclusions were reached using different methods⁴⁸, however, these authors suggest that astrocyte leaflet structural changes may control astrocyte ability to regulate synapse function⁴⁸. Thus astrocyte morphology and leaflet proximity to synapses is dynamic and regulated by neuron-to-astrocyte signaling.

An unexpected role for Cx30 in astrocyte morphology and leaflet-synapse proximity

Molecular mechanisms that determine astrocyte leaflet proximity to synapses are unclear, but may involve actin binding proteins profilin 1 and ezrin, which have been implicated in Ca²⁺ dependent astrocyte process outgrowth^{49, 50}. Moreover, glutamate causes astrocyte filopodial outgrowth⁵¹. Recent experiments reveal a new mechanism involving connexin 30 (Cx30), a gap junction protein found postnatally within astrocytes⁵². By using Cx30-deletion mice, Pannasch et al., found a Cx30 function independent of its ion channel pore: loss of Cx30 resulted in significantly greater astrocyte process elongation and ramification in the neuropil⁵². The altered astrocyte morphology explains decreased excitatory synaptic function in Cx30 mice because the closer proximity of astrocyte leaflets to synapses increased glutamate transporter numbers near synapses. The ensuing downstream effects include decreased AMPA receptor mediated synaptic transmission, reduced hippocampal LTP and reduced contextual fear memory⁵². Hence, differences in astrocyte leaflet proximity to synapses have strong effects on synapses, circuits and behavior⁵². Additionally, astrocyte leaflet motility may be needed for maturation of dendritic spines during development⁵³. Astrocyte morphology is controlled by FGF receptor signaling⁵⁴ in Drosophila, and by mGluR5-dependent glutamate signaling in the case of rodent cortical, but not hypothalamic, astrocytes during development⁵⁵.

Bergmann glia AMPA receptors and Purkinje cell synapses

Bergmann glia express GluA1 and GluA4 AMPA receptor subunits and functional AMPA receptors. Ca²⁺ permeable GluA1 containing AMPA receptors on Bergmann glia are needed for correct association of Bergmann glial processes with Purkinje cell synapses⁵⁶. Genetic disruption of AMPA receptor Ca²⁺ permeability in Bergmann glia resulted in prolonged kinetics of glutamate synaptic transmission and multiple innervations of Purkinje cells by climbing fibers⁵⁶. In adult mice, genetic deletion of Bergmann glia AMPA receptor subunits resulted in retraction of glial processes from the vicinity of synapses, altered synaptic transmission and changes in fine motor coordination of the mice⁵⁷. Together, these studies provide compelling evidence for the physiological relevance of a specific type of AMPA receptor-mediated Ca²⁺ signaling mechanism in vitro and in vivo, underscoring astrocyte morphological dynamics and the realization that Ca²⁺ signals are diverse and not all mediated by intracellular release⁵⁸. As far as we know, there is no evidence that cortical astrocytes express functional AMPA receptors.

Diverse functions of Ca²⁺ signals in astrocytes

What do diverse types of Ca²⁺ signals do within astrocytes? A complete answer to this question will require methods to mimic and block distinct types of Ca²⁺ signals precisely so that their consequences can be explored in specific circuits under specific settings⁶⁰. Additionally, a generic tool to block the vast majority of Ca²⁺ signals in astrocytes is needed. There is unlikely to be a single physiological role for astrocyte Ca²⁺ signals that will apply to the whole CNS. It seems likely that distinct Ca²⁺ signals will regulate the release of signaling molecules via a variety of possible pathways⁸³ (but see⁸⁴), control blood vessel diameter⁸⁵, control the release of synaptogenic/trophic factors^{1, 86–88}, regulate gene expression, modulate K⁺ uptake⁸⁹, regulate neurotransmitter uptake⁷¹, contribute to neuromodulation^{75, 80, 90}, control movement of astrocyte leaflets^{45, 48, 49}, contribute to axonal action potential broadening⁹¹, control neuronal synchronization⁹², contribute to UP states⁹³ and perhaps control the release of injury related molecules and inflammatory mediators⁹⁴. All of these phenomena are regulated by astrocyte Ca²⁺ signals and need to be explored with respect to the distinct types of recently discovered Ca²⁺ signals. It is implausible that all of these effects could be mediated by a single type of Ca²⁺ signal in a single type of astrocyte. From this perspective, the multitude of functional responses mediated by astrocytes provides the best indication yet for diversity within microcircuits. The key conceptual advance is that astrocytes display diverse Ca²⁺ signals in branches and branchlets that are spatially and temporally unrelated to those in the somata (Fig 2) and are distinct within circuits and microcircuits, emphasizing diversity of signalling within single cells and compartments. These insights⁷³ challenge the view that astrocytes display a single type of intracellular Ca²⁺ signal^{58, 95}.

In vivo and in situ tools to study Ca²⁺ signals in entire astrocytes

There has been tremendous progress in the generation and testing of strategies to express genetically-encoded Ca²⁺ indicators in astrocytes. These are now the tools of choice to study astrocyte Ca²⁺ signals (Box 3).

Hippocampus

High resolution imaging of Ca²⁺ signals in astrocyte branches using 2-photon microscopy provides evidence for diversity of astrocyte functional responses within different hippocampal fields. Astrocytes in CA1 stratum radiatum respond with an increase in Ca²⁺ in major astrocyte branches during minimal stimulation of Schaffer collateral fibers⁷⁶. These responses are mediated by synaptic glutamate release acting on mGluR5 metabotropic glutamate receptors on astrocytes⁷⁶. The resultant increase in astrocyte Ca²⁺ is proposed to evoke the release of gliotransmitters (e.g. ATP), which most likely, via the production of adenosine, act on presynaptic A2A receptors to regulate neurotransmitter release probability⁷⁶. Astrocytes located in dentate gyrus (DG) display two types of Ca²⁺ signals that are termed focal and expanded; the latter are caused by action potential firing and the focal signals are mediated by spontaneous synaptic neurotransmitter release⁷⁷. Focal and expanded Ca²⁺ signals are mediated by synaptic ATP because both types of Ca²⁺ signal are reduced by a P2Y1 receptor antagonist and both are mediated by Ca²⁺ release from IP3R2-dependent intracellular Ca²⁺ stores⁷⁷. It is suggested that astrocyte Ca²⁺ signals in DG function to regulate the probability of excitatory neurotransmitter release onto granule cells perhaps via astrocytic release of a neuromodulator that acts on presynaptic nerve terminals⁷⁷. Hence, astrocytes are proposed to act as rapid local detectors of neuronal activity, and to respond to activity changes by locally regulating neurotransmitter release from the same synapses.

In contrast, astrocytes located in CA3 stratum lucidum region only respond during intense bursts of action potentials evoked within mossy fiber axons, and display Ca²⁺ signals that are mediated largely by mGluR2/3 receptors and encompass large parts of the astrocytes⁷². Hence, CA3 astrocytes do not respond to sparse neuronal input and seem incapable of regulating synapses locally because their Ca²⁺ signals cover entire cells. Instead, they may control neuronal network synchronization⁹². Serial block face electron microscopy provides one explanation for differences between CA1 and CA3 region astrocytes. The fine leaflets of astrocytes are much closer to excitatory synapses in the CA1 region than the leaflets of CA3 astrocytes are to mossy fiber synapses⁷², illustrating that astrocyte sub-cellular anatomy and function varies within hippocampal fields.

There is additional evidence that distinct types of Ca²⁺ signals may serve distinct functions. Long-term synaptic potentiation (LTP) in CA1 pyramidal neurons is largely reduced when astrocytes are dialysed with Ca²⁺ chelators⁹⁷, but is unaffected when intracellular store mediated astrocyte Ca²⁺ signals are reduced or increased⁹⁵. These findings may be explained by the variable dependence of LTP on the source of astrocyte Ca²⁺. Hence, basal, rather than elevated, astrocyte Ca²⁺ signals may be important for LTP by regulating the constitutive release of D-serine⁹⁸. The available data provide evidence that astrocyte functional roles are distinct within microcircuits that comprise the larger hippocampal circuit and are separable for distinct sources of Ca²⁺ within single astrocytes. These insights underscore the importance of studying specific and defined types of Ca²⁺ signal for specific neural responses under specified conditions.

Cerebellum

Bergmann glia display highly localised microdomain Ca²⁺ increases that can be evoked by electrical stimulation of afferent fibres⁹⁹. Studies using in vivo imaging in awake head fixed mice free to rest or run on a freely moving ball¹⁰⁰ revealed three types of Ca²⁺ signals within Bergmann glia of the cerebellum⁹⁰. Two types of ongoing and spontaneous Ca²⁺ signals within the cerebellum of mice at rest were termed “sparkles” and “bursts”, whereas a third form called “flares” occurred in hundreds of Bergmann glia in a concerted manner during locomotor activity. Sparkles occurred mainly in processes and were reduced by TTX and glutamate receptor antagonists⁹⁰. Bursts occurred spontaneously in mice at rest and appeared as radially expanding waves that encompassed ~55 Bergmann glia⁹⁰ and a volume of ~86,000 µm³. These signals were not sensitive to TTX, but were abolished by an ATP receptor antagonist, implying they were due to endogenous ATP⁹⁰ released from astrocytes themselves¹⁰¹. Flares occurred during voluntary running and encompassed hundreds of Bergmann glia and were abolished by TTX and a glutamate receptor antagonist⁹⁰. Thus, a single type of astrocyte can display a variety of responses and mechanisms in vivo depending on the behavior of the mouse.

Bergmann glia and cortical astrocytes both respond during periods of enforced locomotion in ways that are indicative of arousal or heightened vigilance states, rather than locomotion per se^{75, 80}. Under these circumstances, the responses were coincident in these anatomically distinct brain areas, were global because all imaged cells displayed Ca²⁺ signals (Fig 4), and in both brain areas the responses were mediated by noradrenaline acting on a1 adrenoceptors located on astrocytes and Bergmann glia⁷⁵. Noradrenaline was likely released from varicose terminals of locus coeruleus neuron projections that extend throughout much of the brain, because cortical astrocytes displayed Ca²⁺ elevations when locus coeruleus neurons were electrically stimulated¹⁰². As such, astrocyte responses in cortex and cerebellum may be indicative of a global brain state associated with arousal or startle.

Thalamus

The thalamic reticular nucleus (nRT) is a source for endogenous benzodiazepine-mimicking peptides called endozepines¹⁰³. Recent work shows that endozepine peptides are found within astrocytes of nRT, and use of a gliotoxin revealed that astrocytes are needed for their positive allosteric effects on synaptic inhibition¹⁰⁴. The effect of gliotoxin was to reduce inhibition in nRT, because the toxin removed a source of endogenous endozepines within nRT¹⁰⁴. However, the effect of gliotoxin in the ventrobasal nucleus (VB) was to increase inhibition via a mechanism that involved GABA transporters. Hence, astrocytes located in adjacent nRT and VB microcircuits have distinct effects on inhibitory synapses mediated by endozepines and GABA transporters, respectively.

Medulla

The retrotrapezoid nucleus (RTN) of the medulla is an important site of chemoreception allowing hypercapnia (an increase in blood CO₂; pCO₂) to increase breathing. A subset of neurons in RTN are highly sensitive to CO₂/H⁺ and send excitatory glutamatergic projections to respiratory centers to regulate breathing. CO₂/H⁺-evoked ATP release from astrocytes in the vicinity of RTN influences respiratory drive¹⁰⁵ by regulating RTN neuron firing. Hence, astrocytes located near RTN are the cellular sources of ATP mediating the ATP component of the central chemosensory response to hypercapnia. Importantly, astrocytes within this area are particularly pH sensitive, responding with Ca²⁺ elevations and further release of ATP¹⁰⁵. This study suggests that a vital step in central chemoreception involves ATP release from astrocytes located on the ventral medulla surface, that this signal is propagated by astrocyte ATP release, ultimately arriving at RTN neurons to depolarize them and thus regulate breathing¹⁰⁵. As such, a potentially unique class of astrocytes within the respiratory chemoreceptor areas displays exquisite pH sensitivity and mediates direct effects on the ability to change breathing upon changes in pC0₂/H⁺.

Spinal cord

There is strong evidence for positionally distinct astrocyte subtypes in the spinal cord^{106, 107}. Evidence that astrocytes in distinct domains are functionally diverse has been provided by examining differential gene expression between dorsal and ventral spinal cord astrocytes³². Sema3a was the most highly expressed ventral gene and this pattern was confirmed by protein expression³². Loss of Sema3a in ventral astrocytes had consequences for a-motor neurons: defective axon initial segment orientation, abnormal synapse formation and function, and a-motor neuron loss³². Cell culture studies suggested that the positional identity of astrocytes is intrinsic, as the relevant properties were retained in co-cultures independent of local environment. These studies raise the possibility that distinct populations of astrocytes serve distinct functions within microcircuits, and provide approaches to tackle this issue also in other circuits. Irrespectively, studies of dorsal-ventral differences in spinal cord astrocytes provide evidence for regional neural circuit diversity of molecular machinery and function^{32, 106}.

Two broadly defined types of Ca²⁺ signal and responses in circuits – local and global

Recent in vivo studies represent a milestone because they provide a basis to explore and understand the physiological settings under which astrocytes respond to behaviors that are engaged within the intact brain. Although this is a nascent field, several general points are worth noting. First, during locomotion the responses of astrocytes in the cortex and Bergmann glia were correlated, implying that the underlying causes may be widespread⁷⁵. Second, when astrocytes responded during locomotion, their responses were largely global and not localized to single cells or areas of single cells^{73, 75, 102}. Third, astrocyte responses were mediated by a slow neuromodulator and not by a fast neurotransmitter^{75, 80}. Together, these data suggest that astrocytes respond to slow neuromodulators that are released as volumetric transmitters⁸⁰, and when astrocytes respond they do so over large parts of their territories. Moreover, in vivo responses and those in the CA3 region indicate that astrocytes may only modulate neuronal function on slow time scales and over large distances^{72, 73, 75, 80, 90, 102}. These slow and global responses are distinct from local responses shown for CA1 and DG astrocytes^{76, 77}. Notably, responses of layer 2/3 visual cortex astrocytes to light were enhanced markedly by enforced locomotion⁷⁵, which implies that local and global signals act synergistically. Synergism for astrocyte Ca²⁺ signals evoked by noradrenaline and VIP receptors has been demonstrated in vitro¹⁰⁸, and these experiments are relevant in the context of in vivo work where both noradrenaline and VIP are likely to be released during startle. There is also evidence that astrocytes display Ca²⁺ elevations during endogenous release of acetylcholine, which acts on astrocyte muscarinic receptors^{109, 110}. It seems likely that astrocytes may respond to local signals and to volumetric signals depending on the particular conditions and brain region.

Astrocyte effects in Huntington’s disease (HD) models

HD is caused by an expanded chain of polyglutamines in the huntingtin protein (HTT) leading to intracellular accumulation and aggregation of mutant huntingtin (mHTT). HD patients and mouse models of HD exhibit nuclear inclusions of mHTT in striatal astrocytes¹¹³. Recent findings show that the onset and progression of neurological disease in HD mouse models is associated with specific astrocyte dysfunction in striatum, but not hippocampus, suggesting a neural circuit-specific nature of astrocyte roles in HD. Tong et al¹¹⁴ found, using two HD mouse models, that in comparison with pre-symptomatic stages, the onset of neurological symptoms was associated with a significant increase in astrocytes with mHTT nuclear inclusions and significant reductions in astrocyte functional proteins including GLT-1 (EAAT2) and Kir4.1. At symptom onset, and in spite of clear evidence of astrocyte dysfunction, there were no major phenotypic changes associated with astrocyte reactivity such as proliferation, hypertrophy and increased GFAP expression. These findings indicate that mHTT is associated with disruption of the expression of important astrocyte functional proteins that initially alter astrocyte function without noticeable astrogliosis. Loss of the glutamate transporter, GLT-1 (EAAT2), has been extensively studied in the context of HD¹¹⁵. Kir4.1 potassium channels regulate extracellular K⁺ levels and are expressed in the CNS primarily by astrocytes¹¹⁶. Loss of Kir4.1 currents in striatal astrocytes in HD mouse models¹¹⁴ led to higher ambient K⁺, which increased medium spiny neuron (MSN) excitability, a hallmark of HD mouse models¹¹⁷, and these deficits could be rescued by virus mediated delivery of Kir4.1-GFP channels selectively to astrocytes, thereby providing evidence that subtle increases in extracellular K⁺ due to astrocyte Kir4.1 loss can phenocopy MSN changes in HD mouse models. Together, these findings support the hypothesis that key aspects of altered MSN excitability in HD are caused by the specific primary dysfunction of a key astrocyte homeostatic function, maintenance of extracellular K⁺, and are not merely a generic or non-specific secondary consequence of astrocyte reactivity. This notion is relevant in the context of human HD, which displays a progressive increase in astrocyte reactivity over time. Thus, astrocyte contributions to underlying HD pathological mechanisms are diverse and change with the progression of disease, such that early stages of the disease are associated with functional deficits in Kir4.1 and GLT-1 (EAAT2), and later stages are also associated with increasing astrocyte reactivity. Therapeutic strategies targeting astrocytes need to embrace the concept that astrocytes display a range of specific dysfunctions as well as astrogliosis.

Astrocyte changes in Alzheimer’s disease models

A mouse model of familial Alzheimer’s disease displayed elevated basal cortical astrocyte Ca²⁺ levels and increased Ca²⁺ transients termed hyperactive Ca²⁺ signaling¹¹⁸. Neither normal nor hyperactive Ca²⁺ signaling was the result of action potential firing in neurons¹¹⁹. Hyperactivity arose because reactive astrocytes near b-amyloid plaques displayed enhanced P2Y1-receptor mediated Ca²⁺ signaling¹¹⁹. The hyperactive phenotype could be rescued by blocking ATP release from connexin channels and by blocking P2Y1 receptors¹¹⁹. Together, these data indicate that astrocyte hyperactivity in Alzheimer’s disease may be caused by a combination of ATP release and/or increased P2Y1 receptor expression by a population of astrocytes near plaques, implying that blocking P2Y1 receptors represents a novel therapeutic strategy for Alzheimer’s disease. These observations also point towards diversity of astrocyte responses to different CNS insults. P2Y1 receptor mRNA expression and Ca²⁺ signaling was markedly reduced in in vitro models of astrocyte reactivity triggered by inflammatory mediators²⁷, suggesting that the increase observed around β-amyloid plaques may represent a novel cellular phenotype specific to Alzheimer’s disease.

Diversity of astrocyte secondary responses to injury and disease

Considerable information has accrued regarding molecular and structural aspects of reactive astrogliosis^{17, 94, 111, 120, 121}. Recent progress demonstrates that in contrast to long held beliefs, reactive astrogliosis is not a uniform set of stereotypic astrocyte changes regulated by a simple single genetic on-off switch. Instead, molecular dissection in vivo has revealed that diverse signaling events tailor astrocyte responses to specific CNS insults^{94, 111}. For example, transcriptional changes induced by either ischemia or inflammation differ substantially^{27, 28}, and trauma induced changes exhibit clear gradients that vary with distance from lesions and with injury intensity^{17, 122}. Thus, reactive astrogliosis is both complex and heterogeneous and can be defined as a finely graded continuum of multiple potential changes that range from reversible alterations in gene expression and cell hypertrophy within preserved individual astrocyte domains, to scar formation that involves substantial cell proliferation and permanent rearrangement of tissue structure ^{17, 94, 111}. Thus, reactive astrocytes exhibit substantial diversity at multiple levels, including with respect to gene expression, cell structure, cell signaling and cell function. This diversity exists not only across different CNS regions such as cerebral cortex, hippocampus and spinal cord, but also topographically with respect to distance from lesions and locally among neighboring cells ^{27, 122, 123}.

Two functionally distinct categories of reactive astrocytes

Much of the evidence that delineates heterogeneity among reactive astrocytes pertains to differences in molecular expression and cell structure. Although less is known about functional diversity, current information points towards a working model that recognizes at least two fundamentally distinct broad functional categories of reactive astrocytes, which we refer to as scar-forming reactive astrocytes and hypertrophic reactive astrocytes. (Fig. 5, Table 1).

Distinguishing reactive astrogliosis from astrocyte dysfunction

Transgenic loss-of-function studies applied to in vivo models of CNS insults show that under normal circumstances, astrogliosis and scar formation serve essential functions that limit tissue damage, restrict inflammation and preserve function after CNS traumatic injury, stroke, infection, autoimmune attack and degenerative disease^{17, 94, 111, 112, 120, 121, 130}. Emerging data also indicate that primary astrocyte dysfunctions can contribute to neuronal dysfunction and neurological symptoms prior to the onset of detectable astrogliosis¹¹⁴. Thus, astrocyte dysfunction and astrogliosis are not synonymous and can occur independently of each other. Together with the emergence of a concept of ‘normal astrogliosis’ critical for preservation of tissue and function after CNS insults, is the recognition that astrogliosis can also be prone to dysfunction as a result of genetic defects, genetic polymorphisms, autoimmune attack, or chronic exposure inflammatory stimuli caused by secondary infections⁹⁴. Recognizing and understanding the differences between beneficial and dysfunctioning astrogliosis is likely to be fundamental to dissecting the cellular mechanisms underlying many CNS disorders.

Exploiting astrocyte diversity

The realization that astrocytes are diverse and mediate separable responses in the context of disease raises the possibility that this diversity could be exploited to derive therapeutic effects. One immediate way forward is to understand the full complement of proteins that astrocytes express; a good starting point is G-protein coupled receptors. Receptors with astrocyte specific expression could be exploited to produce neuromodulatory effects within circuits with potentially minimal effects on neurons. For example, a receptor pathway that leads to elevated Kir.4.1 expression may be useful for HD therapeutics and a P2Y1 antagonist may be of benefit in Alzheimer’s disease. Perhaps the release of endozepines from astrocytes in nRT can be exploited in the context of epilepsy. At this juncture what is needed is a deeper understanding of fundamental astrocyte biology and diversity.

Emerging questions

A first class of question concerns molecular mechanisms. What is the molecular makeup and signaling machinery of astrocytes in distinct circuits? By systematically comparing between different brain circuits it should be possible to assemble a set of molecular metrics with which to classify astrocytes as distinct or the same in different regions. Do astrocytes utilize signals other than Ca²⁺ for cellular communication? New families of highly specific genetically encoded indicators for neurotransmitters and neuromodulators are needed. We need also to measure the nature and amounts of substances released from astrocytes in specific brain areas and then explore the functions of key signaling species on a case-by-case basis in the context of specific circuits and behaviors.

A second class of question concerns how astrocytes contribute to the formation and function of synapses and circuits. Do individual astrocytes signal locally to single or small sets of synapses to allow formation or removal, or do single astrocytes signal globally to entire dendritic arbors within their territory or perhaps even to entire populations of neurons that reside within extended astrocyte volumes connected by gap junctions? Is synapse formation and removal triggered by activity-dependent and bidirectional signaling between astrocytes and neurons? If so, what chemical signals are involved, are they unique and can they be exploited pharmacologically? What do the various types of astrocyte Ca²⁺ signals encode within circuits? Is the main function of astrocytes to listen locally to synapses and then to regulate them locally, or is it to listen to widespread volumetric neuromodulator release and respond globally over large volumes of brain? Studies demonstrating motility of astrocyte leaflets are important because altering astrocyte proximity to synapses has the potential to alter synaptic and circuit function by altering ion and neurotransmitter homeostasis.

A third class of question concerns astrocyte functional diversity in CNS disorders. Astrocytes display a plethora of changes in CNS disorders, but which of these are correlative, beneficial or harmful will need to be determined on a case-by-case basis. Studies will need to explore astrocyte contributions to specific conditions in specific brain areas and at specific stages of pathology and avoid the temptation to generalize across the spectrum of disorders ranging from traumatic injury to neurodegenerative diseases. In this regard it is noteworthy that astrocytes perform specialized functions in higher primates¹³⁵ and it is possible that astrocyte diversity is greater in humans than in rodents. It is likely that in vitro studies of astrocytes derived from patient induced pluripotent stem cells will be fruitful. Much could be learned about the regional and functional diversity of astrocytes and their contributions to pathology and neural circuit mechanisms within organoid systems as models for human disorders¹³⁶.

Finally, how should one classify diverse astrocytes? We suggest that the answer will emerge from a broad understanding of functional diversity, rather than relying solely on single criteria such as morphology, connectivity, marker expression or gene profiles. Our emphasis is to define astrocyte diversity based on direct measurements of function, dynamics and molecular mechanisms in specific circuits. This captures recent progress with the classification of interneurons¹³⁷.

Summary

We conclude by advancing the concept that astrocytes display a range of functional attributes that are astrocyte compartment, circuit, microcircuit and disease specific. As such, previous notions that astrocytes represent a homogenous population of cells throughout the CNS need to be revised. A deeper understanding of astrocyte diversity and its implications is necessary in order to fully understand neural circuits³ in health and disease, and to provide new opportunities to treat CNS disorders.