You are on page 1of 18

REVIEWS

Improved tools to study astrocytes


Xinzhu Yu1,3, Jun Nagai1,3 and Baljit S. Khakh   1,2 ✉
Abstract | Astrocytes are a type of glial cell that tile the CNS. They interact with multiple cell
types, including neurons, glial cells and blood vessels, and are involved or implicated in brain
disorders. Progress has been made in understanding astrocytes, but the field lacks detailed
information concerning how they perform their multifarious functions, and how and when
they influence the operations of the neural circuits with which they interact. One recognized
bottleneck to progress has been the paucity of reliable tools with which to explore astrocytes
within the adult vertebrate CNS in vivo. However, improved tools for molecular, genetic,
morphological and physiological assessments have been developed recently or have been
adapted from their original purposes to study neurons and are now being used to systematically
document and interrogate astrocyte biology in vivo. These tools, their uses and limitations, and
the insights that they afford are summarized in this Review.

Gliotransmission
Astrocytes tile the CNS and interact with multiple cell unexpected discoveries and testable hypotheses regard-
The release of transmitters types, including neurons, other glial cells and blood ing these fascinating cells, which, for far too long, have
from astrocytes owing to vessels1. Important strides have been made in under- been the neglected, less glamorous, siblings of neurons.
intracellular Ca2+ elevations, standing astrocytes, especially their crucial roles in syn- Here, we focus on recent studies with newer types of
through vesicular exocytosis or
apse formation and removal2,3. However, the field lacks tools that have been deployed in vivo. We explore four
by another regulated
mechanism. detailed understanding of how these versatile cells per- astrocyte-​relevant topics spanning the pertinent gamut
form their many functions, which vary depending on the of biology — ranging from molecules, cells and circuits
cells they contact, and how they influence the specialized to behaviour. We start with a brief introduction to the
operations of the neural circuits in which they exist4–7. key challenges in each topic and then describe recently
A major bottleneck has been the lack of reliable and developed tools used to address them, finishing each
selective tools with which to explore astrocyte functions section with some notes of caution and further con-
within the vertebrate adult CNS in vivo. sideration where necessary. One overriding emergent
Given their now-​recognized potential contributions theme is the need to choose the tool after carefully con-
to brain disorders8, interest has grown in studying fun- sidering the specifics of the question or questions being
damental astrocyte biology and evaluating the causa- explored; these types of decision will need to be made
tive and correlative roles of astrocytes in the operation on a case-​by-case basis.
of neural circuits, as well as astrocytic contributions to Although we do not discuss the evidence for or
behaviour and disease or disease-​related phenotypes. In against gliotransmission in depth (reviewed elsewhere15–17),
the astrocyte field, as in other areas of biology, global we do discuss some recent tools that have been used in
1
Department of Physiology, and conditional gene-​deletion mouse models have studies of the reported phenomenon. We do not consider
David Geffen School of provided powerful means to explore the functions of astrogliosis, because at this juncture it is hard to establish
Medicine, University of
these cells. Furthermore, tools to monitor and manipu- a clear definition for the process, and also because it has
California Los Angeles,
Los Angeles, CA, USA.
late cells precisely in vivo that are now broadly advanc- been excellently reviewed18. In the interests of space, we
2
Department of
ing neuroscience9–11 have also found applications in have concentrated mainly on work performed in the past
Neurobiology, David Geffen studying astrocytes. Major long-​standing questions 5 years in adult mice.
School of Medicine, University concerning whether, how and when astrocytes contri­
of California Los Angeles, bute to the functions of the nervous system12,13 are now Unmasking molecules and mechanisms
Los Angeles, CA, USA.
being addressed with methods that more directly permit The challenge. To perform hypothesis-​driven exper-
3
These authors contributed manipulation and interrogation of astrocytes. iments with astrocytes, it is valuable to have compre-
equally: Xinzhu Yu,
Jun Nagai.
A saying attributed to Sydney Brenner, one of the hensive profiles of their gene and/or protein expression.
✉e-​mail: bkhakh@ giants of biology, is that, “Progress in science depends on Such unbiased data are also useful to shed light on com-
mednet.ucla.edu new techniques, new discoveries and new ideas, prob- plex phenomena and to give initial direction to explora-
https://doi.org/10.1038/ ably in that order”14. This saying is apropos for astro- tory experimental work. Several methods have been used
s41583-020-0264-8 cytes: the use of new techniques and tools has resulted in to uncover astrocyte transcriptional and translational

Nature Reviews | Neuroscience


Reviews

Immunopanning
profiles, including earlier evaluations using cultured Not long thereafter, forebrain astrocytes were iso-
A method of cell purification astrocytes and more recent in vivo assessments using lated using FACS from mice expressing enhanced GFP
that uses cell-​type-specific designer tools and astrocyte-​selective gene targeting (EGFP) under the control of the human S100 calcium-​
antibodies immobilized in adult mice (Tables 1,2). binding protein-​β (S100β) gene promoter (S100B-​EGFP
to a solid surface (such as
a cell culture plate) to
reporter mice), from postnatal day 1 (P1) to P30 and
immunoprecipitate Molecular profiling of isolated astrocytes. The trans­ for more than 20,000 genes analysed with microarrays22.
specific cell populations. crip­tomes of purified astrocytes can be assessed From this study, Aldh1l1, which encodes aldehyde dehy-
using fluorescence-​a ctivated cell sorting (FACS), drogenase family 1 member L1 (ALDH1L1), emerged as
immunopanning19 and magnetic-​activated cell sorting in a pan astrocyte marker and was confirmed with Aldh1l1-
combination with reporter mouse lines and antibodies EGFP reporter mice that fortuitously already existed23.
against astrocyte-​specific molecules. Astrocytes from The RNA of cortical astrocytes from the Aldh1l1-EGFP
various brain regions, developmental stages and disease mice was sequenced, providing a platform for analys-
conditions have been studied in this way. ing and comparing transcriptomes in the brain24. Using
Soon after landmark explorations with astrocyte such methods, alterations in the gene expression pro-
cultures20, an antibody against the astrocytic glutamate files of purified astrocytes with ageing, ischaemic stroke
transporter GLT1 and transgenic mice overexpressing and neuro­inflammation have been documented25,26, and
green fluorescent protein (GFP) under the control of these data have been used to glean deeper insights into
the human glial fibrillary acidic protein (GFAP) pro- possible underlying mechanisms and implications27.
moter (GFAP-​GFP reporter mice) were used to docu- SOX9 is a more recently identified astrocyte-​specific
ment the transcriptional profiles of cortical astrocytes nuclear marker that identifies a population of astro-
from 10–12-week-​old mice21, representing the first cytes that overlaps highly with those identified by GLT1
analyses of adult astrocyte gene expression from a expression28. This merits specific mention, because
defined brain area. By comparing the transcriptional SOX9 could be useful to isolate nuclei from post-​
profiles of GLT1+;GFAP-​GFP+ astrocytes with those of mortem human brain tissue to assess astrocytes in
GLT1+;GFAP-​GFP– astrocytes, the authors found sub- pathophysiology. An innovative intersectional approach
tle differences in gene expression associated with GFAP in which astrocytes isolated from Aldh1l1-EGFP mice
expression, suggesting that GFAP may be of limited utility were grouped depending on their expression of cell-​
as a universal astrocyte marker. surface markers was used to document astrocyte

Table 1 | online database resources of astrocyte transcriptomes and proteomes


Name (website) Technique species (line) Ages Brain regions refs
Brain RNA-​seq RNA-​seq Mouse (Aldh1l1-EGFP) P7 Cortex 24

Human a
17–20 GW (fetal), Cortex 30

8–18 years (juvenile),


21–63 years (adult)
Adult Astrocyte RNA-​seq Mouse (RiboTag × Aldh1l1- P63–P82 Cortex, striatum and 39,40

RNA-Seq Explorer Cre/ERT2) hippocampus


LC–MS/MS Mouse (Aldh1l1-EGFP) P30 Striatum and 40

hippocampus
Astrocyte Aging RNA-​seq Mouse (RiboTag × Gfap-​Cre) 4 months, 2 years Visual cortex, motor 43

Transcription cortex, hypothalamus


and cerebellum
DropViz scRNA-​seq Mouse P60–P70 9 brain regions 53

Mouse Brain Atlas scRNA-​seq Mouse P12–P30, 6 weeks 19 regions of central, 52

and 8 weeks peripheral and enteric


nervous system
SCope scRNA-​seq Drosophila melanogaster 0, 1, 3, 6, 9, 15, 30 Brain 194

and 50 days
Mouse P21–P31 Somatosensory cortex 50

and hippocampus
P56–P80 Hypothalamus 195

snRNA-​seq Human 20–51 years Visual cortex, frontal 196

cortex and cerebellum


Allen Brain Atlas, scRNA-​seq Mouse P53–P59 Cortex and thalamus –
RNA-​seq data
snRNA-​seq Human 24–66 years Cortex and thalamus –
Macaque 2.3–17.9 years Thalamus –
Aldh1l1, encodes aldehyde dehydrogenase family 1 member L1; Cre, Cre recombinase; EGFP, enhanced green fluorescent protein;
ERT2, oestrogen receptor 2; GFAP, glial fibrillary acidic protein; GW, gestational weeks; LC–MS/MS, liquid chromatography-​tandem
mass spectrometry ; P, postnatal day; RiboTag, haemagglutinin-​tagged ribosomal protein; RNA-​seq, RNA sequencing; scRNA-​seq,
single-​cell RNA sequencing; snRNA-​seq, single-​nucleus RNA sequencing. aImmunopanning was used to isolate human astrocytes.

www.nature.com/nrn
Reviews

Table 2 | Transgenic mouse lines and viral vectors for targeting astrocytes
Line or virus Promoter or gene locus Proportion expression in refs
of astrocytes astrocytes relative
targeteda to other cellsb
Mouse lines
Gfap-​Cre Mouse Gfap High Low 39,197,198

GFAP-​Cre Human GFAP Very low Very low 60

GFAP-​Cre/ERT2 Human GFAP High Low 59,199

GFAP-​tTA Human GFAP Moderate Moderate 200,201

Aldh1l1-Cre/ERT2 Aldh1l1 Very high Very high 39,66

Slc1a3-Cre/ERT2 Slc1a3 (which encodes GL AST) Moderate Low 39,62,63

S100B-​Cre Mouse S100B Moderate Moderate 202

Slc6a11-Cre/ERT2 Slc6a11 (which encodes GAT3) Moderate Moderate 39

Gjb6-Cre/ERT2 Gjb6 (which encodes CX30) Very high Moderate 39,63

Fgfr3-Cre/ERT2 Fgfr3 High Low 203

Viral vectors
AAV5, AAV8 and AAVrh43 Gfa2 Very high High 204,205

AAV5, AAV9 and AAV-​PHP.eB GfaABC1D Very high Very high 41,45,46,
71–74,79,206

AAV5, AAV8 and AAV9 Rat Aldh1l1 Low Very low 78

AAV-​DJ Human ALDH1L1 Low Very low 77

Please note that this table is a summary: readers are strongly encouraged to consult the original articles cited herein and examine
the relevant data sets carefully before making a choice about which reagent to use. AAV, adeno-​associated virus; Aldh1l1, encodes
aldehyde dehydrogenase family 1 member L1; Cre, Cre recombinase; ERT2, oestrogen receptor 2; GFAP, glial fibrillary acidic
protein; S100β, S100 calcium-​binding protein-​β; tTA , tetracycline-​controlled transactivator protein. aUsed as a convenient
description to convey the proportion of targeted astrocytes in relation to the total number of astrocytes in the region of interest.
b
Used as a convenient description to convey the proportion of astrocytes relative to other cells such as neurons.

diversity between brain areas and to identify analogous be harvested using EGFP-​specific antibodies to enable
cell subtypes in gliomas29. sequencing of ribosome-​associated mRNAs. Multiple
Immunopanning19 was recently adapted to isolate bacterial artificial chromosome (BAC) transgenic mouse
astrocytes30. The method is useful when fluorescent lines exist to target TRAP to neurons, oligodendro-
reporter mouse lines are not available and/or in model cytes and astrocytes (bacTRAP lines). An Aldh1l1 bac-
species for which genetic modifications are not routine. TRAP line was used to identify brain-​region-dependent
Astrocytes from fetal, juvenile and adult human brain transcriptional differences in astrocytes across the
tissue were immunopanned with antibodies against mouse lifespan37 (see following section for discussion
hepatocyte cell adhesion molecule (HEPACAM), a cell-​ of mouse lines used).
surface protein expressed by human astrocytes. RNA RiboTag is a related method to isolate ribosome-​
sequencing (RNA-​seq) identified >2,000 genes differen- associated mRNAs from any cell type with an available
tially expressed in fetal versus mature human astro­cytes Cre driver38. A BAC transgenic mouse line was generated
and revealed functionally relevant molecular diff­erences to enable cell-​type-specific expression of haemaggluti-
between human and mouse astrocytes 30 that were nin (HA)-tagged RPL22 (a ribosomal protein) upon Cre
anticipated from careful neuroanatomical work31. recombinase-​mediated deletion of an upstream floxed
Bacterial artificial
chromosome (BAC) A similar approach called magnetic-​activated cell STOP cassette38. RPL22HA incorporates into polyribo-
transgenic sorting, which conjugates cell-​type-specific antibodies somes and can be immunoprecipitated with reliable
Describing a transgenic to magnetic beads, also enables isolation of different HA-​targeting antibodies to yield actively translating
organism generated by cell types from postnatal mice32,33, is relatively straight­ mRNAs from required cell populations. This RiboTag
random but stable integration
of large segments of DNA
forward to implement and has been used for RNA-​seq34. mouse line has been crossed with a tamoxifen-​inducible
(up to 300 kb) from BAC Magnetic-​activated cell sorting32–34 preserves astrocyte Aldh1l1-Cre/ERT2 mouse line for RNA-​seq of mature
vectors into the genome. morphology better than does FACS, does not involve cell astrocytes from the cortex, striatum and hippocam-
culture and overcomes some of the key limitations of pus39,40 (Fig. 1), and with Gfap-​Cre mice to enable analy-
Tamoxifen
other methods that use single cells. sis of molecular changes in astrocytes after spinal cord
A selective oestrogen receptor
modulator that specifically injury41 and in a mouse model of multiple sclerosis42.
activates tamoxifen-​inducible Molecular profiling of astrocytes in vivo. Newer strat- Adult and aged astrocyte transcriptomes from dif-
Cre recombinase (for example, egies have assessed translated mRNAs in specific cell ferent brain regions were also obtained using RiboTag
Cre/ERT2) and results populations of the brain without physical cell dissociation. mice crossed with Gfap-​Cre mice43. With ageing, genes
in the shuttling of the
Cre recombinase into
One approach, termed translating ribosome affinity related to synapse elimination and inflammatory path-
the nucleus, where the purification (TRAP)35,36, genetically targeted expression ways were upregulated, whereas genes related to choles-
recombination occurs. of an EGFP-​tagged ribosomal protein L10a, which can terol synthesis were downregulated. Recently, RiboTag

Nature Reviews | Neuroscience


Reviews

a Aldh1l1-Cre/ERT2 × RiboTag mice b Top 500 astrocyte-enriched genes in each area

Cortex
128

40 49

283

78 69
mRNA from
all cells
Striatum 99 Hippocampus
Bulk RNA-seq

Anti-HA antibodies
Common genes
Magnetic beads Core functions:
• Potassium homeostasis
Magnetic IP • Neurotransmitter uptake
• Neurotransmitter degradation
• Gap junction signalling
mRNA from • Fatty acid oxidation
astrocytes • Serine biosynthesis
• Neuroinflammation signalling
Astrocyte RNA-seq

Fig. 1 | Transcriptomic profiling of astrocytes from cell-​specific rNA sequencing in adult mice. a | RiboTag mice
expressing the ribosomal subunit RPL22 with a haemagglutinin (HA) tag (RPL22HA) in a Cre-​dependent manner from the
Rosa26 locus were crossed with astrocyte-​specific Aldh1l1-Cre/ERT2 mice. The adult offspring that were positive for both
alleles were treated with tamoxifen to drive the expression of RPL22HA selectively in astrocytes. Subsequently , different
brain areas were dissected and the total mRNA from a ‘soup’ of all the included cells was sequenced (bulk RNA sequencing
(RNA-​seq)). Astrocyte-​specific RNA was also extracted using a well-​established immunoprecipitation (IP) technique, in
which magnetic beads were bound to HA epitopes of RPL22HA and its associated mRNA , which was then subjected to
sequencing (astrocyte RNA-​seq). b | In the example shown, the overlap between the top 500 genes showing astrocyte
enrichment in the cortex, hippocampus and striatum is shown, as are the numbers of genes specifically expressed in each
area. Functional analysis using Gene Ontology and Ingenuity Pathway Analysis of the 283 shared genes suggests the core
functions of astrocytes in these three brain areas. Original data are published in refs39,40 and are available at Adult
Astrocyte RNA-​Seq Explorer. Aldh1l1, encodes aldehyde dehydrogenase family 1 member L1; ERT2, oestrogen receptor 2.

was incorporated into viral vectors, including a floxed classified into three subpopulations with distinct spatial
RiboTag adeno-​a ssociated virus (AAV) vector and an distribution and signalling pathways: one subpopulation
astrocyte-​specific RiboTag AAV that can be deployed was enriched for genes involved in cell adhesion and
relatively easily in control and disease-​model mice to RNA processing, and another was enriched for genes
investigate astrocyte function44–46. involved in apoptosis regulation and cell proliferation51.
Two projects together used RNA-​seq of more than 1
Molecular profiles of astrocytes at a single-​cell level. million single cells to generate a detailed taxonomy of
Analyses of single-​cell transcriptional profiles have the cellular architecture of the entire mouse nervous sys-
unmasked cellular diversity within and between organs, tem52,53. Seven molecularly distinct astrocyte types were
species, developmental stages, and health and disease, identified, with discrete spatial distributions that corre-
with unparalleled clarity47–49. Briefly, in single-​cell RNA-​ lated with those of glutamate and glycine signalling52.
seq or single-​nucleus RNA-​seq, individual cells or cell Improved methods for single-​cell sequencing appear
nuclei, respectively, are isolated from a given tissue (such frequently and their use will probably accelerate, ena-
as a particular brain area) and sequenced. Subsequent bling the field to map, in detail, the molecular profiles of
analysis can yield information on complex and/or rare astrocytes within different brain areas, and to compare
cell populations, the developmental trajectories of cell them with astrocytes in other brain areas and with other
lineages and/or the functions of single cells in their cell types of the body54.
environment.
Single-​cell transcriptomic analyses have advanced Considerations. Studies with purified astrocytes have
our understanding of astrocyte diversity. For example, reported valuable information, but there are several
Adeno-associated virus mouse cortical astrocytes form two subclasses distin- aspects to consider.
(AAV) vector guished by differential expression of Gfap and Mfge8 First, a necessary step in assessing astrocytic profiles
A vector based on an AAV, with different laminar locations and morphologies50. is to purify these cells without introducing molecular
composed of exogenous DNA
(up to 5 kb) flanked by two
Similarly, using a form of single-​cell RNA-​seq that lim- alterations in the process. It seems implausible that iso-
145-nucleotide-​long inverted its transcriptional alterations typically caused by con- lated astrocytes are unaltered relative to their in vivo
terminal repeats. ventional cell preparations, amygdalar astrocytes were counterparts, given the stress of cell dissociation and

www.nature.com/nrn
Reviews

Tetracycline
sorting procedures. In addition, the complete transcrip- areas from which astrocytes are assessed are not known
An antibiotic used to either tional landscape of astrocytes in vivo may be impossible or accounted for experimentally or computationally.
repress (Tet-​Off) or activate to capture with purified astrocytes, because the latter do Nonetheless, early FACS and cell-​culture studies mark-
(Tet-​On) gene expression in not retain the archetypal bushy morphology of astro- edly advanced the field and promoted molecular and
a Tet system. Its derivative,
doxycycline, is also widely
cytes in vivo (Fig. 2). For the most highly expressed genes, mechanistic investigations of astrocyte biology, hand in
used. the effects of astrocyte dissociation and purification may hand with physiology.
be inconsequential; however, for functionally important Complementary tools will be needed to exploit the
genes with low expression, even subtle changes could be available single-​cell sequencing data to test specific
relevant for downstream analyses. hypotheses in silico, in vitro or in vivo. Furthermore,
Second, it is necessary to be confident that the pro­ more research is necessary to establish where and how
cedures are astrocyte-​s elective. Depending on the astrocyte populations are distributed in the brain and
procedures employed, certain subpopulations of astro- whether they have specific locations in relation to anato­
cytes may be preferentially selected, because many mically defined pathways, local microcircuits and areas.
reporter lines and antibodies are not completely specific As an example, chordin-​like protein 1-expressing astro-
and do not target all astrocytes ubiquitously (see the next cytes are enriched in layer II/III of the visual cortex,
section). where they regulate the plasticity of thalamocortical
Third, a further consideration when interpreting synapses58. Such avenues are probably best explored by
some past studies is that multiple brain areas seem combining gene expression analyses in anatomically
to have been pooled during isolation steps. As astro- intact preparations.
cyte transcriptomes vary depending on the brain Mass spectrometry-​b ased proteomics of FACS-​
region29,40,42,43,55–57, such pooling could potentially influ- isolated astrocytes from the striatum and the hippocam-
ence the interpretation of the resulting data, if the brain pus of Aldh1l1-EGFP mice has revealed similarities and
differences between brain areas that were consistent with
data gathered using RNA-​seq. However, mass spectrom-
Scale Methods etry detected fewer proteins than were predicted to be
Tissue • Multicolour intracellular expressed by RNA-​seq. To comprehensively address
~1 mm iontophoresis (dye-filling) these inconsistencies without bias, advances in pro-
• Transgenic mouse reporter lines
• Viral vector-mediated reporter teomics will be necessary to examine protein expression
expression in astrocytes that retain their complex morphology. As
• Plasmid vector-mediated an example of the mismatch between gene expression
reporter expression
and function, nearly all astrocytes show intracellular
increases in Ca2+ in response to extracellular ATP, but
RNA-​seq data suggest that expression of transcripts
Major encoding ATP P2Y G protein-​c oupled receptors
Cellular • Immunohistochemistry
~50 μm branch (GPCRs) is very low39,40,44–46. Detailed proteomics of the
• Intracellular iontophoresis
(dye-filling) various compartments that comprise complex astrocyte
Soma • Transgenic mouse reporter lines morphology (Fig. 2) in different brain regions will bring
• Viral vector-mediated reporter
expression us closer to uncovering physiological functions.
Territory
• Plasmid vector-mediated
reporter expression Genetic targeting
The challenge. Mouse lines for targeting and manip-
ulating astrocytes in vivo would be useful. However,
Subcellular Branchlets and leaflets • Intracellular iontophoresis
~0.1–1 μm (dye-filling) there is no ‘perfect’ mouse line for targeting astrocytes,
• Transgenic mouse reporter lines and every available line has its own limitations — some
• Viral vector-mediated reporter minor, others considerable. The properties of key astro-
expression
• Plasmid vector-mediated cyte mouse lines are summarized in Table 2 with a focus
reporter expression on those expressing Cre recombinase, because such lines
can be used in conjunction with other Cre-​dependent
Fine processes in territory mice or viruses to express designer reporters, sensors
and actuators.
Synaptic • Electron microscopy
~10–100 nm • Super-resolution microscopy
• NAPA Cre lines and drivers. When choosing a genetic strategy
to express or delete genes in astrocytes, two factors must
be considered: the proportion of astrocytes in the region
of interest that are targeted (which we term efficiency),
and the extent to which other cell types are excluded
from targeting (which we term specificity).
Much effort has been devoted to optimize both
Fig. 2 | Methods to study astrocyte morphology. The cartoon illustrates the scales at parameters with the use of GFAP, ALDH1L1, S100β,
which astrocyte morphology can be studied, with the methods currently in use listed to GLAST and GLT1 reporter and Cre lines. To date, sev-
the right. Please refer to the main text for a discussion of the strengths and weaknesses of eral Cre-​dependent or tetracycline-​controllable trans-
each individual approach. NAPA , neuron–astrocyte proximity assay. genic mouse lines have been applied to manipulate gene

Nature Reviews | Neuroscience


Reviews

Microglial reactivity
expression through promoters of astrocyte-​marker genes microinjections into the brain, but, of course, it is not
A microglial activation (Table 2). However, because the expression of astrocyte brain-​region-specific.
response to tissue damage or genes depends on the brain region, developmental stage
pathological insults, to mediate and physiological condition, and most astrocyte-​marker Considerations. It is worth noting that viral tropism may
inflammatory responses.
genes are often active in neural progenitor cells and vary depending on development, brain region, mouse
detected in non-​astrocytic cell types, one has to choose genetic background, cargo and administration routes.
and characterize each system based on experimental Until the precise mechanisms by which viruses target
need. For example, transgene expression driven by the specific cell types are better elucidated, careful character-
human GFAP promoter does not occur for all astrocytes ization and control experiments will be needed for each
and is often detected in large numbers of neurons59,60. study to empirically assess efficiency and specificity on
Mouse Gfap-​Cre lines are better than those that use a case-​by-case basis.
the human GFAP promoter, but neuronal expression is At this time, the most reliable mouse line seems to
expected and reported for both lines61. Slc1a3-Cre/ERT2 be Aldh1l1-Cre/ERT2 (refs39,66), and the most reliable
(also known as GLAST) lines label almost 100% of viral approach is the use of AAV2/5 with the GfaABC1D
Bergmann glia in the cerebellum, but only about promoter41,45,46,68,69,71–76, but better tools will probably be
20–30% of astrocytes in the cortex, hippocampus and developed before too long.
striatum39,62–65. As with all gene-​targeting strategies, one needs to
The best available transgenic model to target astro- assess astrocyte and microglial reactivity. This has not so
cytes across the entire brain is the tamoxifen-​inducible far proved a major problem, but parallel controls will
Aldh1l1-Cre/ERT2 mouse line, which permits temporal still be necessary — especially in brain areas not yet
control of gene expression and has demonstrated high studied with such approaches. Furthermore, depending
astrocyte specificity and efficiency when induced in on the question, even targeting strategies that are not
adulthood39,66. However, Aldh1l1, like the other marker astrocyte-​selective can be useful — for example, when
genes (including Gfap and Slc1a3), is expressed in mosaic expression of a reporter such as GFP is needed
postnatal neurogenic zones67 and thus early postnatal for morphological studies and neuron expression can
induction of this line will affect other cells. simply be ignored in downstream analyses, because neu-
rons and astrocytes can easily be discriminated based on
Viral methods. Although powerful, gene targeting with the expression of GFP.
transgenic approaches affects multiple brain regions
as well as some peripheral organs, making it difficult Exploring morphology
to investigate local circuit modulation and to interpret The challenge. A key task is to reliably measure, and
behavioural outcomes. One alternative approach to track changes in, astrocyte morphology. A stereotypical
overcome this limitation is to use viral vectors for gene mature astrocyte in the mammalian brain is remarkably
delivery to astrocytes in specific brain regions. complex, with an oval soma and several major branches
Lentiviruses and AAVs are the most commonly used that each split into secondary and tertiary branchlets,
viral vectors in astrocyte biology and have several attrac- which in turn ramify into fine processes called leaflets
tive properties compared with mouse lines, including (Fig. 2). The elaboration of this intricate morphology
their region specificity and flexibility of application in emerges postnatally and establishes boundaries between
different species and disease models. Lentiviruses can neighbouring mature astrocytes, a phenomenon called
carry larger cargoes (up to about 8 kb) than can AAVs ‘tiling’.
(up to about 5 kb), but AAVs do not integrate into the Astrocyte morphology has long been recognized to
genome and can be concentrated to higher titres for be non-​uniform. For example, protoplasmic astrocytes,
improved efficiency. Some viruses display natural selec- which reside in the grey matter, are highly morpholog-
tivity for astrocytes in vivo68,69 and astrocyte specificity ically complex, whereas fibrous astrocytes in the white
can be enhanced by combining with astrocyte-​specific matter are less complex. Numerous studies have con-
promoters (Table 2). firmed that astrocyte morphology is heterogeneous
The GfaABC1D promoter (681 bp) was derived from within and across brain regions, and dynamic in both
the human GFAP promoter gfa2 (2.2 kb). It is about twice physiological and pathological states. We review the
as active as gfa2, and its smaller size is better suited to approaches available to document the complexity and
viral vectors70. The expression of cargo genes under the track the plasticity of astrocyte morphology (Fig. 2).
GfaABC1D promoter in astrocytes is highly efficient We do not consider immunohistochemistry of
and specific in many studies41,45,46,71–76. However, leaky GFAP as this underestimates astrocyte complexity80,81
expression in other cell types has been reported with and is not detectable in many astrocytes. Furthermore,
Cre-​mediated recombination39,45. although several astrocyte markers, such as S100β,
Notably, the Aldh1l1 promoter preferentially drives glutamate transporters (GLAST and GLT1) and glu-
neuronal expression in many regions of the rodent tamine synthetase (GS), can reveal details of astro-
brain when used in viral vectors 77,78. The recently cytes, these markers essentially show the distribution
evolved AAV variant PHP.eB with the GfaABC 1D of the cognate protein and not the cellular architecture.
promoter enables widespread and efficient transduc- Furthermore, some marker proteins are not specific for
tion in astrocytes of the entire brain through intra- astrocytes in some brain regions; for example, S100β
venous injection 79, offering an alternative strategy also identifies oligodendrocytes and oligodendrocyte
to the use of transgenic mouse lines that avoids any progenitor cells, and GLT1 is detectable in some neurons.

www.nature.com/nrn
Reviews

Iontophoresis
Visualizing astrocyte morphology with intracellular observed to be dynamic and regulated by intracellular
An approach to deliver charged dye-​filling. Astrocytes in brain sections fixed with short Ca2+ and metabotropic glutamate receptors89.
molecules (such as fluorescent exposure to fixative have been visualized by delivering Postnatal electroporation of plasmids encoding
dye) into cells by applying fluorescent dyes into them through a sharp electrode, cytosolic and membrane-​tagged fluorescent proteins
electric current.
in an iontophoresis-​based protocol80; the dye diffuses was used to label astrocytes in mouse visual cortex,
Ezrin into the volume of the cell, revealing total astrocyte to measure and track the volume and complexity of
A protein that functions morphology. The method can be readily combined astrocytes and their processes. This approach revealed
as a linker between the with immunohistochemistry to confirm cell identity experience-​dependent increases in astrocytic complexity
plasma membrane and
and characterize protein expression in different cellular between P7 and P21, mediated by neuroligin–neurexin
the actin cytoskeleton.
compartments. The dye most frequently delivered in this interactions90.
Lck-GFP way is lucifer yellow, a bright fluorophore that is easily Lck-​GFP has been used to examine the effects of
A green fluorescent protein imaged with confocal microscopy. Using this method, brain-​derived neurotrophic factor (BDNF) on astro-
(GFP) tagged with a membrane-​ subtle differences were identified in the morphology of cyte morphology34. In another study, its expression was
targeting sequence of Lck (from
Src tyrosine kinase).
striatal and hippocampal astrocytes40. Remarkably, when induced through intracerebroventricular injections of
two adjacent hippocampal astrocytes were filled with virus at P0, to sparsely label astrocytes for morphological
Spaghetti monster differently coloured dyes, they formed discrete, non-​ and anatomical examination at different postnatal time
fluorescent proteins overlapping territories during postnatal development82 points91. The use of spaghetti monster fluorescent proteins
Modified fluorescent proteins
that remained intact following injury 83. could be powerful and improve signal-​to-noise ratios
with multiple epitope tags that
can be targeted using highly
Although intracellular dye-​filling is time consuming further for sparse labelling92, although they would
specific antibodies. and can sample only limited numbers of cells at a time, require antibody labelling and may not be appropriate
it nevertheless provides detailed, high signal-​to-noise for experiments such as those requiring live imaging.
images with, by definition, single-​cell resolution. It is
also relatively straightforward to implement84, is reliableVisualizing finer astrocyte morphology. Astrocytes
and, owing to the use of lightly fixed tissue (short expo- have extensive, ramified fine processes that measure
sure to fixative), preserves in vivo morphology. Lucifer on a scale of tens of nanometres in thickness; together,
yellow intracellular iontophoresis has elegantly shown these processes constitute up to 95% of total cell vol-
how astrocyte morphogenesis is compromised by neo- ume93. Therefore, to resolve the complete structure of
natal exposure to general anaesthesia via a mechanism astrocytes, evaluations with submicrometre resolution
involving ezrin and intracellular Ca2+ signalling85. are crucial.
Astrocytes can be identified by the presence of glyco­
Visualizing astrocyte morphology with genetic gen granules and intermediate filaments in the cyto-
approaches. Reporter expression — through viruses plasm using electron microscopy94,95. Three-​dimensional
or transgenic mouse lines — has been used for sparse reconstructions with serial-​section electron microscopy
labelling in vivo for single-​cell morphological stud- images have demonstrated that astrocyte processes con-
ies. Cytosolic fluorescent proteins (such as GFP and tact most, but not all, synapses in the rat hippocampal
tdTomato) and membrane-​tethered fluorescent pro- CA1 region: approximately 57% of the synapses were
teins (including Lck-​GFP) have been expressed using contacted by astrocyte processes and, in these cases,
astrocyte-​specific promoters. about 43% of the synaptic interface was surrounded
One study labelled astrocytes with EGFP through by astrocyte processes95,96. A similar phenomenon was
use of GFAP-​tTA mice, which express tetracycline-​ found in the cortex, where the extent of synapse cov-
controlled transactivator protein (tTA) under the control erage by astrocytes was greater in layers II/III than in
of the human GFAP promoter86. Single cortical astro- layer VI87.
cytes were seen to enwrap an average of four neuronal Super-​resolution microscopy refers to various appro­
somata and to contact 300–600 neuronal dendrites, aches that achieve spatial resolution below the diff­raction
forming synaptic ‘islands’ through which astrocytes limit of light microscopy. Combined with mathemat-
may synchronize and coordinate neuronal networks. ical reconstructions, super-​resolution microscopy has
Cortical astrocytes that were sparsely labelled using the enabled imaging of fine astrocytic processes and certain
Slc1a3-Cre/ERT2 mouse line were classified into four proteins within them, including those near synapses97.
distinct clusters on the basis of 24 morphological para­ Such approaches have great potential to visualize astro-
meters, including the territory volume, cell orientation cyte structures at a scale of tens of nanometres, includ-
and extent of ramification87. The distribution of these ing when combined with expansion microscopy 98,
subpopulations was also layer-​specific and depended on which physically expands fixed tissues. Stimulated
the laminar organization of cortical neurons87. emission depletion (STED) microscopy creates super-​
Another recently developed sparse labelling method, resolution images by quenching fluorophores in a
called mosaicism with repeat frameshift (MORF), takes doughnut-​shaped area around the excitation spot, to
advantage of frameshift mutations in a long mononucleo­ minimize the area of illumination at the focal point99.
tide repeat to stochastically label individual genetically Although super-​resolution methods have not yet been
defined neurons88, and could be used to label astrocytes. extensively used to study astrocyte fine structure, this
In a study in organotypic slices, membrane-​targeted is likely to change as the methods become increasingly
EGFP and mCherry were expressed by cell-​type-specific routine.
viruses in CA1 neurons and hippocampal astrocytes, Fluorescent proteins and a genetically targeted
respectively; putative astrocyte–spine contacts were neuron–astrocyte proximity assay (NAPA) have been

Nature Reviews | Neuroscience


Reviews

Förster resonance energy


used73,89 to measure the spatio-​temporal interactions microdomains of astrocyte processes93,109,110. These Ca2+
transfer between astrocyte processes and synapses. The NAPA signals are proposed to underlie important physiologi-
(FRET). Energy transfer depends on Förster resonance energy transfer (FRET) cal functions of astrocytes in multiple species, including
between two appropriate light-​ between GFP expressed in astrocytes and mCherry worms, flies, zebrafish, mice and, potentially, humans.
sensitive molecules, such as
expressed in presynaptic neurons. Thus, detecta- The cellular sources of astrocyte Ca2+ signals include
fluorophores, via dipole–dipole
coupling. The transfer typically ble FRET reflects interactions of astrocyte processes multiple pathways, including those involving various
reports distances on the tens and presynaptic compartments within approximately receptors, channels, exchangers and pumps on the plasma
of nanometres scale or lower 10 nm of each other. The NAPA could potentially be membrane, as well as within intracellular organelles such
and so can be used to assess
developed for imaging astrocyte–neuron interactions as mitochondria, endoplasmic reticulum, Golgi and acidic
the proximity between cognate
fluorophores that undergo
in awake behaving animals — for example, with two-​ organelles111. The relative contributions of these sources
Förster resonance energy photon fluorescence lifetime imaging microscopy100. In to overall astrocyte Ca2+ dynamics are not clear and may
transfer. addition, the NAPA could also be expanded to assess require a combination of experimental and modelling
spatial relationships between multiple cell types, such as approaches to delineate. Nevertheless, the available data
Microdomains
oligodendrocytes and microglia. clearly imply that Ca2+ signalling in astrocytes is as rich,
Highly localized, small (several
micrometres wide) territories of fascinating and complex as that in other cells of diverse
a cell. Considerations. Since the pioneering work that first doc- organisms111–114.
umented the shapes of astrocytes, advances in imaging The larger question of whether astrocyte Ca2+ signals
Photoisomerable and labelling techniques have expanded our knowledge cause changes in neural circuit function and behaviour
Able to undergo reversible
structural conversion between
of astrocyte biology. However, a single perfect solution has now been explored. These experiments have become
isomers in response to for imaging and tracking astrocyte morphology does not possible because of methodological advances (mostly
photoexcitation. exist, and many questions remain open. For instance, from the Genetically-​Encoded Neuronal Indicator and
do astrocytes with similar but separable morphologies Effector (GENIE) project at Janelia) in monitoring astro-
in a given brain area perform distinct functions? What cyte Ca2+ with genetically encoded calcium indicators
are the molecular mechanisms regulating astrocyte in vivo and in adult brain slices115–118 (Table 3). These
morphology within and between regions? Is structural advances have in turn necessitated the use of methods
rearrangement of astrocytes a cause or a consequence of to increase and decrease Ca2+ signals within astrocytes
neuronal plasticity and/or pathophysiology? in vivo (Fig. 3), and to evaluate the molecular and cellular
A major challenge is to develop methods that are consequences of such manipulations (Fig. 4). Here, we
innocuous and that can be used for live imaging and, consider tools for selectively stimulating and attenuating
subsequently, for three-​dimensional reconstructions astrocyte Ca2+ signalling in vivo and the insights they
by electron microscopy. Related to this point, the tissue have revealed. Studies using conventional pharmacolog-
fixation conditions used for electron microscopy affect ical approaches are not considered here but are reviewed
ultrastructural preservation of astrocytes and other cells101. elsewhere119,120.
Furthermore, advances in automated tracing of cellu-
lar elements are needed for fine-​scale reconstructions Increasing astrocyte Ca2+ with LiGluR. The light-​gated
of astrocytes, because existing methods that work for ionotropic glutamate receptor (LiGluR) is an ionotropic
neurons do not work reproducibly for astrocytes102. glutamate receptor type 6 isoform with a mutation
Machine-​learning approaches may solve this problem that enables the reversible, covalent attachment of a
before too long. photoisomerable molecule called maleimide–azoben-
Although the NAPA appears useful, a formal possi- zene–glutamate (MAG)121,122. Photostimulation of MAG-​
bility remains that, by detecting only the most proximal bound LiGluRs on astrocytes induced channel-​pore
astrocyte–neuron interactions, the method does not opening and increases in intracellular Ca2+. Interestingly,
permit assessment of the dynamics of astrocyte–neuron light-​stimulated astrocytes released glutamate through
spatial interactions at a scale of tens of nanometres73. anion channels in their membranes, resulting in prop-
Again, methods must be chosen after due considera- agation of Ca2+ signalling to neighbouring astrocytes
tion of the specifics of the question or questions being lacking the LiGluR123. LiGluRs have not yet been used
investigated. to stimulate astrocytes in vivo, perhaps owing to the
difficulty of innocuously delivering MAG to the brain.
Exploring calcium signalling
The challenge. Firing no action potentials and displaying Increasing astrocyte Ca2+ with DREADDs. Astrocyte
no known propagated electrical signals, astrocytes are GPCRs have been studied for more than two deca­
largely electrically silent in comparison with neurons13, des124. Pharmacological and gene expression studies
and were thus long viewed as lacking dynamic signals have suggested that astrocytes express a wide range of
for bidirectional communication with other astrocytes, GPCRs24,39,124, and much attention has focused on astro-
microglia and neurons. This view is beginning to change cyte GPCRs that promote inositol 1,4,5-trisphosphate
with the use of improved tools. About 30 years ago, intra- receptor (IP3R)-dependent Ca2+ release from intra­
cellular Ca2+ signalling was discovered in astrocyte cell cellular endoplasmic reticulum stores. However, as many
cultures103,104, and was subsequently reported in brain endogenously expressed GPCRs are also found on neu-
slices105, in anaesthetized106 and behaving animals107,108, rons and other cells, pharmacologically targeting such
and in human cells30,31. Since these landmark studies, mechanisms to explore causal roles within astrocytes,
dynamic and diverse astrocyte Ca2+ signals have been although valuable in cultures of purified astrocytes30,31,125,
widely reported, including those in highly localized has proved challenging in brain slices and in vivo120.

www.nature.com/nrn
Reviews

Table 3 | Genetically encoded single-​wavelength sensors useful for astrocyte work


Analyte sensor Kd Wavelengtha (nm) ref.
excitation emission
Intracellular analytes
Ca2+ GCaMP6f 375 nM 485 510 117

GCaMP6m 167 nM 485 510 117

GCaMP6s 144 nM 485 510 117

Lck-​GCaMP6f 375 nMb 488 510 39

jGCaMP7s 68 nM 485 510 116

jGCaMP7f 174 nM 485 510 116

jGCaMP7b 82 nM 485 510 116

jGCaMP7c 298 nM 485 510 116

jRCaMP1a 214 nM 575 595 207

jRCaMP1b 712 nM 570 595 207

jRGECO1a 148 nM 565 590 207

XCaMP-​B 71 nM 405 446 208

XCaMP-​G 200 nM 488 514c 208

XCaMP-​Gf 128 nM 488 514c 208

XCaMP-​Gf0 115 nM 488 514c 208

XCaMP-​Y 81 nM 488 527c 208

XCaMP-​R 97 nM 561 593 208

cAMP Flamindo2 2.1 µM 505 522 209

Pink Flamindo 7.2 µM 567 590 210

K+ GINKO1 0.42 mM 488 520 211

Extracellular analytes
Glutamate iGluSnFR 4–5 μM 470 550 212

SF-​iGluSnFR .A184S 0.6 μM 470 525 213

SF-​iGluSnFR .A184V 2.1 μM 470 525 213

SF-​iGluSnFR .S72A 34 μM 470 525 213

SF-​Azurite-iGluSnFR 9 μM 383 450 213

SF-​mTurquoise2-iGluSnFR ND 434 474 213

SF-​Venus-iGluSnFR 2 μM 515 530 213

ATP iATPSnFR1.0 350 μM 473 525 214

iATPSnFR1.1 138 μM 473 525 214

GABA iGABASnFR 30 μM 485 510 215

iGABASnFR .F102G 42 μM 485 510 215

iGABASnFR .F012Y.Y137L 106 μM 485 510 215

Dopamine dLight1.1 330 nM 490 516 216

dLight1.2 770 nM 490 516 216

dLight1.3a 2.3 μM 490 516 216

dLight1.3b 1.7 μM 490 516 216

dLight1.4 4.1 nM 490 516 216

dLight1.5 110 nM 490 516 216

GRABDA1m 130 nM 470 510 217

GRABDA1h 10 nM 470 510 217

Acetylcholine GACh 0.7–2 μM 480 520 218

Noradrenaline GRABNE1m 930 nM 488 525 219

GRABNE1h 83 nM 488 525 219

Not all these sensors have been used to explore astrocytes, but their availability means such experiments can now commence. Readers
are strongly encouraged to read the original articles cited here regarding the specific features, including the kinetic properties, of each
sensor. Kd, equilibrium dissociation constant; ND, not determined. aValues for peak excitation and emission wavelengths are approximate
and are from the cited papers based on the information provided (in some cases, the values have been estimated from the spectra in
the original papers — for example, for jRCaMP1a, jRCaMP1b, jRGECO1a, Flamindo2, GRABDA1m and GRABDA1h). All values are for single-​
photon imaging, because not all the sensors have been tested with two-​photon microscopy. Several sensors, such as Ca2+ and glutamate
sensors, have been iteratively improved over several years; given space constraints, reference is made to the most recent versions.
b
Kd of Lck-​GCaMP6f is assumed by that of GCaMP6f, based on comparisons between cytosolic-​targeted and membrane-​targeted
GCaMP2 (ref.110). cXCaMP-​G and XCaMP-​Y can be optically separated with linear unmixing, as suggested in the cited paper.

Nature Reviews | Neuroscience


Reviews

a Increasing astrocyte Ca2+ signalling via GPCR pathways


DREADDs

hM3Dq rM3Ds hM4Di Melanopsin Opto-XRs


Synthetic
b ligands Light
a Opsin

Intracellular
Major c Soma loop of GPCR
Gαq Gαs Gαi Gα Gα
branches

IP3
ER Ca2+

Territory IP3R
d comprising
branchlets Cytosol
and leaflets

b Increasing astrocyte Ca2+ signalling via channels c Decreasing astrocyte Ca2+ signalling d Decreasing astrocyte Ca2+
signalling via pumps
Light-gated Light-gated
channel channel
380 nm light Human Ca2+
K+ K+ Endogenous ligand PMCA2w/b
Tethered ligand (CalEx)

ChR2 LiGluR GPCR



Na+ Na+ PLC IP3 sponge
H+ Ca2+
(Ca2+) IP3 ADP ATP

IP3R2
deletion

Fig. 3 | cartoon of the tools used to manipulate astrocyte intracellular causes non-​selective cation flow through the plasma membrane, including
ca2+ signalling. Several genetically encoded tools and approaches exist to Na+, H+, K+ and Ca2+. However, the Ca2+ permeability of ChR2 is reported to be
selectively increase and decrease astrocyte intracellular Ca2+ levels. a | Five low220 and its mechanisms and physiological relevance are under debate
tools for stimulating astrocyte Ca2+ signalling through G protein-​coupled (also see Considerations in the section Exploring calcium signalling).
receptor (GPCR) pathways are shown. All downstream pathways of the tools Light-​gated ionotropic glutamate receptor (LiGluR) is a mutated ionotropic
involve inositol 1,4,5-trisphosphate receptor (IP3R)-mediated Ca2+ release glutamate receptor 6 that is Ca2+ permeable. Upon photostimulation, the
from the endoplasmic reticulum (ER). DREADDs (designer receptors photoisomerable tethered ligand maleimide–azobenzene–glutamate (MAG)
exclusively activated by designer drugs) are engineered GPCRs that can be opens the LiGluR channel pore to induce Ca2+ influx. Notably , both tools
activated by inert chemicals and not by endogenous ligands. Each DREADD induce K+ efflux. Increases in extracellular K+ resulting from ChR2 activation
interacts with distinct Gα proteins upon synthetic ligand binding: Gq-​coupled have been shown to be consequential for neuronal firing. c | Two strategies
hM3Dq, Gs-​coupled rM3Ds and Gi-​coupled hM4Di. Melanopsin, a type of for blocking GPCR–IP3R pathways to attenuate Ca2+ signalling are shown that
photopigment, can elicit intracellular signal cascades when stimulated with include genetic deletion of IP3R2, which is thought to be the most abundant
blue light (about 470–480 nm). Opto-​XRs are engineered chimaeras IP3R isoform in astrocytes, and a recombinant IP3 sponge that competes the
containing a photo-​activatable extracellular part from the pigment rhodopsin endogenous IP3Rs for IP3 binding. d | Human plasma membrane Ca2+ ATPase
(blue) and the intracellular part of endogenous GPCRs (red) that trigger splice variant w/b (PMCA2w/b) is a Ca2+ pump that has no endogenous
G protein-​mediated signalling. b | Two tools for increasing astrocyte Ca2+ levels expression in astrocytes and acts as a constitutive Ca2+ extruder (known as
through channels are shown. Photostimulation of channelrhodopsin 2 (ChR2) the CalEx approach).

In these regards, the development and use of DREADDs relatively non-​invasive stimulation of GPCR pathways in a
(designer receptors exclusively activated by designer genetically targeted manner in vivo.
drugs) has been extremely useful to explore astrocyte Several types of DREADD have been developed to tar-
biology in brain preparations, such as slices, and in vivo. get major Gα-​protein signalling pathways: Gq-​coupled
DREADDs are engineered GPCRs that have strongly hM3D126, Gi-​coupled hM4D126,128, Gs-​coupled rM3D129
attenuated responses to their endogenous ligands and and G12-coupled DREADD130. These engineered recep-
have been engineered to respond to synthetic, biolog- tors are chimaeras between a human or rat muscarinic
ically inert chemical actuators126,127. Because the syn- receptor and the turkey β1-adrenoceptor to confer speci­
thetic ligands are delivered in the animal’s water or food fic G protein pathway activation, and/or carry point
supply, or by systemic injection, DREADDs enable mutations to kill sensitivity to endogenous ligands.

www.nature.com/nrn
Reviews

That such receptors are so frequently used as biological CNO, evoked astrocyte Ca2+ elevations similar to the
tools is testimony to the deep value of the careful biochemi­ Ca2+ signals mediated by endogenous astrocyte GPCRs.
cal and biophysical studies of GPCRs that permitted Second, Gq, Gi and Gs DREADDs all evoked astrocyte
their design127,131–133. Ca2+ signalling40,134, despite the fact that they usually
AAVs for expressing DREADDs selectively in stri- have separable effects on neuronal membrane potential.
atal and hippocampal astrocytes were used to make However, astrocytes and neurons are distinct cells, and
three observations40. First, application of the ligand, so it is simplistic to think they would respond similarly

Astrocyte-specific
manipulation Effect on astrocytes Consequences Cellular mechanism

Circuits Behaviour
a
Gq-DREADD hM3Dq Increased Ca2+ signalling Increased FOS+ neurons Enhanced conditioned
recruited during fear fear memory acquisition
CNO conditioning

Enhanced LTP
Gαq via D-serine
FOS+

Ca2+ neuron
IP3R
Hippocampal CA1
ER

b
Gi-DREADD hM4Di Increased Ca2+ signalling Increased MSN firing Hyperactivity and
disrupted attention

Gαi
Synapse
formation
Ca2+ via TSP1
MSN

Dorsal striatum

c
Human PMCA2w/b (CalEx) Attenuated Ca2+ signalling Reduced MSN activity with Excessive self-grooming
increased correlation
Ca2+ Ca 2+

export

Regulation of
GABA-mediated
tonic inhibition
ADP ATP
Dorsal striatum

Fig. 4 | causative roles of astrocyte intracellular ca2+ signalling revealed release of the synaptogenic cue thrombospondin 1 (TSP1). Importantly ,
by new tools. a | Activation of the DREADD (designer receptor exclusively gabapentin, an anticonvulsant drug, antagonized the TSP1 synaptogenic
activated by designer drug) hM3Dq in astrocytes of the hippocampal CA1 effects and rescued the synaptic, circuit and behavioural changes
region causes increases in astrocyte Ca2+ signalling, as well as regulation of observed following astrocyte Gi pathway activation45. c | Strong astrocyte
synaptic transmission by d-​serine, and FOS expression in neurons recruited Ca2+ signalling attenuation by calcium extruder (CalEx; expression of a
during fear conditioning. Chemogenetic and optogenetic activation of the Ca2+-exporting pump) in the dorsal striatum caused excessive self-​grooming
astrocyte Gq pathway was used to delineate the time-​locked changes behaviour in mice reminiscent of behaviours in obsessive–compulsive
associated with the behavioural effect: enhanced memory acquisition71. disorder in humans. This was due to altered MSN activity in vivo and reduced
b | hM4Di activation in dorsal striatal astrocytes increased astrocyte Ca2+ tonic GABA inhibition resulting from increases in the functional expression
signalling, excitatory synaptic transmission and firing of medium spiny of astrocyte GABA transporter, GAT3. SNAP-5114, a pharmacological
neurons (MSNs) in vivo. The behavioural consequences were hyperactivity inhibitor of GAT3, rescued the excessive self-​grooming behaviour caused
with disrupted attention, mimicking certain symptoms of attention deficit by CalEx46. ER , endoplasmic reticulum; IP3R , inositol 1,4,5-trisphosphate
hyperactivity disorder in humans. The mechanisms involved astrocytic receptor ; PMCA2w/b, plasma membrane Ca2+ ATPase splice variant w/b.

Nature Reviews | Neuroscience


Reviews

Remote memory
to GPCRs or any other receptor. Indeed, multiple GPCR ChR2-mediated astrocyte excitation has been used to
The enduring memory of pathways converge on astrocyte Ca2+ signalling, includ- ‘excite’ astrocytes (but see Considerations) and explore
events that happened or were ing the Gi pathway124,133, and changes in Ca2+ signalling their roles in regulating circuits and behaviour, implicat-
learnt in the distant past. are not analogous to changes in membrane potential or ing them in breathing143, response selectivity in the visual
Retinal ganglion cells
excitability and rely on different molecular machinery. cortex144, regulation of blood flow145, sleep146,147 and cer-
Output neurons in the ganglion Third, although Gq-​DREADD-mediated Ca2+ signals ebellar motor function148 through various mechanisms.
cell layer of the retina, were similar for striatal and hippocampal astrocytes, A recent study demonstrated that ChR2 stimulation of
receiving visual input from Gi-​DREADD-mediated signals were substantially greater astrocytes caused transient increases in extracellular
photoreceptors and
in striatal astrocytes than in hippocampal astrocytes. K+ levels that affected neuronal firing in slices, in silico
interneurons.
Therefore, GPCR-​mediated astrocyte Ca2+ signalling and and in vivo149. In addition, changes in extracellular
its downstream effects may vary between brain areas. K+ levels affect not only neurons but also astrocytes,
Studies using DREADDs suggest that astrocyte Ca2+ microglia and blood vessels150. In another study, the
signalling is sufficient to modulate behaviour. Activation Arch opsin, which is inhibitory in neurons, was used
of Gq DREADDs in medial basal hypothalamic astro- to increase cortical astrocyte Ca2+ signalling76, again
cytes altered food intake by modulating the activity of highlighting the possibility that certain tools may have
neurons expressing agouti-​related protein (AGRP)135,136. different effects on neurons versus astrocytes.
Activating Gq DREADDs on astrocytes in the central
amygdala induced ATP or adenosine release, which Increasing astrocyte Ca2+ with melanopsin. Light-​
reduced neuronal firing, suppressing the gating of con- activated GPCRs have been used to manipulate astrocyte
ditioned fear expression137. Gq-​DREADD stimu­lation Ca2+ signalling. Melanopsin is a photoactivated GPCR
in hippocampal CA1 astrocytes induced astrocyte endogenously expressed in the plasma membrane of
d-​serine release and increased neuronal recruitment mammalian retinal ganglion cells but not in the brain.
during memory acquisition, enhancing recent memory71 Melanopsin can be activated by blue light (~470–480 nm
(Fig. 4a). In another study, activation of the Gq pathway wavelength) and deactivated by yellow light. Although it
in hippocampal astrocytes caused astrocytic release can couple to Gi/o, when examined in neurons melanopsin
of glutamate that directly activated neuronal NMDA almost exclusively induces Gq signalling151,152.
receptors134. Recently, AAV-​expressed melanopsin was deployed
A few studies have explored the consequences of acti- to selectively stimulate astrocyte Ca2+ signalling153.
vating the Gi pathway in astrocytes. In the dorsolateral Continuous photostimulation of melanopsin-​expressing
striatum of mice, activation of astrocyte Gi DREADDs hippocampal astrocytes for 1–20 s increased the fre-
induced the expression of thrombospondin 1 (TSP1), an quency of astrocyte Ca2+ signals in fine processes (micro-
astrocyte synaptogenic cue45. This, in turn, boosted fast domains), whereas stimulation for longer than 10 s was
excitatory synaptic transmission onto striatal medium required to evoke Ca2+ signals in astrocyte somata. These
spiny neurons, increased medium spiny neuron firing evoked Ca2+ signals were abolished in IP3R2-lacking
in vivo and resulted in behavioural hyperactivity with mice, confirming IP3-dependent signalling. When stim-
disrupted attention phenotypes45 (Fig. 4b). In the hippo­ ulated with a brief train of light pulses, stronger astro-
campus, activation of Gi DREADDs enhanced long-​ cyte Ca2+ signalling was associated with potentiation of
term potentiation of Schaffer-​collateral–CA1 pyramidal evoked excitatory postsynaptic currents in CA1 pyra­
neuron synapses and contextual memory acquisition138. midal neurons lasting about 10 minutes — that is,
Similar responses that are mediated by endogenous mimick­ing short-​term potentiation. By contrast, low-​
μ-​opioid receptors (MORs) on astrocytes may be rele- frequency optical stimulation of melanopsin-​expressing
vant to the acquisition of reward memories138. Activation astrocytes triggered long-​term potentiation and enhanced
of astrocytic MORs in the nucleus accumbens caused episodic memory, consistent with DREADD stimulation
the release of astrocyte glutamate that activated NMDA of the same population of astrocytes in vivo71. Melanopsin
receptors on neurons139, similar to that following acti- seems a very useful tool to optically stimulate astrocytes
vation of the Gi pathway in hippocampal astrocytes134. in a time-​controlled manner and to potentially mimic the
A preliminary report suggests that Gi-​DREADD acti- endogenous oscillatory patterns of GPCR pathways.
vation of hippocampal CA1 astrocytes does not affect
recent memory (unlike activation of the Gq pathway)71, Increasing astrocyte Ca2+ with opto-​XRs. Optical control
but impairs remote memory recall140, although the under- of GPCR pathways provides more rapid temporal con-
lying molecular and cellular mechanisms remain to be trol than do conventional pharmacological approaches,
explored. and is thus useful for assessing time-​locked changes in
syna­ptic function and behaviour71. In addition to naturally
Increasing astrocyte Ca with channelrhodopsin 2. occurring opsins, such as melanopsin (which is typically
2+

Channelrhodopsin 2 (ChR2) is a light-​gated, non-​ coupled to Gq), conopsin (Gi/o) and the non-​mammalian
selective cation channel originally found in algae; when rhodopsin–guanylyl cyclase (Gs)154, several engineered
stimulated with blue light (~470 nm), ChR2 mediates light-​sensitive GPCRs, called opto-​XRs, have been used.
the flow of cations such as protons and sodium into the In these opto-​XRs, the intracellular loops of vertebrate
cell. As ChR2 can cause membrane potential depolar- rhodopsins are replaced by those of α-​adrenoceptors and
ization and action potentials when expressed in neu- β-​adrenoceptors to create light-​sensitive mutant proteins
rons, the channel and its derivatives have immeasurably that recruit Gq-​mediated and Gs-​mediated pathways,
advanced systems neuroscience141,142. respectively155. Several opto-​XRs, including opto-​α1-AR

www.nature.com/nrn
Reviews

(Gq), opto-​β2-AR (Gs), opto-A2A-​R (Gs) and opto-​MOR Gq-​coupled hM3D DREADD in the same cells136. How
(Gi/o), have been validated in neurons154. p130PH affects basal and spontaneous Ca2+ signals in
Stimulation of opto-​α1-ARs in astrocytes recapitu- astrocytes is not clear, because their mechanisms may be
lated the enhanced fear memory effect that was triggered independent of IP3 signals. Notably, p130PH also binds
by Gq-​DREADD stimulation of hippocampal astro- other phospholipids, including PtdIns(4,5)P2, possibly
cytes71 — presumably through the same mechanism. As complicating the interpretation of such studies165.
far as we know, other opto-​XRs have not yet been tested
in astrocytes. Decreasing astrocyte Ca2+ with calcium extruder. Given
Although opto-​XRs allow temporally precise opti- the diversity of astrocytic Ca2+ signalling, a genetic
cal stimulation, light from the tip of an optical fibre is approach was developed to attenuate such signalling
expected to spread into a restricted volume of brain regardless of the Ca2+ source46. An efficient Ca2+ pump,
tissue and to scatter back, thus complicating anatom- called plasma membrane Ca2+ ATPase isoform 2 splice
ically precise stimulation based on fibre position156. variant w/b (PMCA2w/b), constitutively extrudes Ca2+
Light-​dependent tissue heating can also cause neuronal out of cells and is not endogenously expressed in astro-
firing156,157 and, in most cases, fewer astrocytes are stim- cytes. AAVs overexpressing human PMCA2w/b under
ulated with light than using chemogenetics where chem- the control of the GfaABC1D promoter were used to
ical actuators spread in the tissue, possibly complicating induce expression of the pump on the surface of stri-
direct comparisons between these methods. From these atal astrocytes. This calcium extruder (CalEx) approach
perspectives, it is interesting to consider that chemo­ reduced the amplitude and duration of spontaneous
genetic and opto-​α1-AR Gq pathway activation in hippo­ and GPCR-​mediated Ca2+ signals by approximately 80%
campal astrocytes produced almost identical effects on without causing overt changes in their frequency.
fear memory71. Attenuating astrocyte Ca2+ signalling in the dorsolateral
striatum of adult mice resulted in repetitive behaviour,
Decreasing astrocyte Ca2+ by targeting IP3. Astrocyte reminiscent of certain symptoms of obsessive–compulsive
Ca2+ signalling is complex and mediated by multiple disorder in humans (Fig. 4c). Available data suggest that
mechanisms111,113. In one major pathway leading to ele- PMCA2w/b does not detectably alter extracellular Ca2+
vated intracellular Ca2+ in astrocytes, IP3 generated by concentrations sufficiently to alter neurotransmitter
activated cell-​surface GPCRs stimulates endoplasmic release probability46, which has a fourth power depend-
reticulum IP3Rs, resulting in Ca2+ release. IP3 recep- ency on extracellular Ca2+. The pronounced behavioural
tor type 2 (IP3R2) is the predominant IP3R isoform in phenotype was mediated by reductions in tonic GABA
astrocytes in the rodent brain158–160. Mice genetically inhibition of striatal neurons through upregulation of
lacking IP3R2 display strongly attenuated spontaneous astrocyte GABA transporter GAT3. Further examina-
and Gq-​GPCR-mediated Ca2+ signals in the somata of tion using astrocyte-​specific RNA-​seq revealed that
astrocytes from multiple brain regions. These mice and reducing astrocyte Ca2+ levels leads to a decrease in
the interpretations of these data have been extensively Rab11a expression, which in turn results in increased
reviewed16,17. GAT3 function166.
Another approach to disrupt IP3-dependent Ca2+
signall­ing is to use molecules that buffer cytosolic IP3, to Instrumentation and software for assessing astrocyte
prevent IP3-mediated Ca2+ release from the endoplasmic Ca2+. Several software packages have become available
reticulum. An ‘IP3 sponge’ based on the IP3 ligand-​ to analyse astrocyte Ca2+ signals167,168. Of these, Astrocyte
binding site of IP3R1 was constructed to compete with Quantitative Analysis (AQuA) software seems poten-
endogenous IP3Rs for IP3 binding161. In a transgenic tially most useful168 as it allows quantification of the
mouse line using the Tet system with a GLT1 promoter, spatio-​temporal patterns of Ca2+ signalling.
IP3 sponges were expressed in astrocytes in most brain Furthermore, there have been advances with instru-
regions, except the cerebellum, resulting in an approxi- mentation to complement single-​photon and multi-
mately 30% reduction in the integrated area of metabo- photon microscopy. For example, fluorescence lifetime
tropic glutamate receptor (mGluR)-mediated Ca 2+ imaging can measure the absolute concentration of
responses in hippocampal astrocytes and reduced astro- Ca2+ in different astrocyte compartments169, and three-​
cytic coverage of synapses. These changes were associated dimensional imaging can monitor Ca2+ signals in parts
with deficits in hippocampus-​dependent behaviours162. of astrocytes that are larger than those that can be
However, leaky expression of the GLT1-promoter-​driven imaged in a single plane170. Wearable miniature micro-
transgene was also detected in hippocampal neurons162,163, scopes have been used to image astrocyte Ca2+ in freely
possibly complicating the interpretation of these data. behaving mice171, and an optical fibre imaging system
Tet system The pleckstrin homology domain of PLC-​like pro- was developed to monitor astrocyte Ca2+ signals from
An inducible gene expression
tein p130 (p130PH) acts as a mobile cytosolic IP3 buffer two different brain areas simultaneously172.
system that reversibly activates
(Tet-​On) or represses (Tet-​Off) and, when expressed by a recombinant AAV with the Various computational methods and models for
transcription in the presence of astrocyte-​specific promoter GfaABC1D, suppresses analy­sing data and making predictions for new exper-
a tetracycline transactivator ATP-​induced Ca2+ elevation in cortical astrocytes by iments are also emerging173–178, although these will need
(tTA) protein, a DNA sequence about 70% in vitro and in vivo164. Selective expression modifying for astrocytes in specific brain areas. Overall,
called the tetracycline
response element (TRE) and
of p130PH in hypothalamic astrocytes considerably developments in software, modelling and instrumenta-
treatment with tetracycline or reduced the food intake of mice. By contrast, feeding tion are allowing finer-​scale investigation of astrocytes at
its derivatives. behaviour was increased by selectively activating the multiple spatial and temporal scales in vivo and in silico.

Nature Reviews | Neuroscience


Reviews

Acidification
Considerations. The studies above demonstrate that longer. It may be that ChR2 stimulation of astrocytes
The act of something such activation of astrocytic GPCR pathways results in gene may not mimic physiological astrocytic processes in
as a cellular compartment expression changes that affect synaptic function, as well astrocytes — at least, not those that have so far been
becoming more acidic as more rapid changes from the release of glutamate, readily identifiable— but might reflect pathophysi-
(that is, becoming protonated).
ATP or adenosine, leading to slow and fast modulation ology when  acidification occurs182. Further points to
of neural circuits with discernible behavioural conse- consider when using optogenetics for astrocyte stud-
quences. Differences between the effects of Gq and Gi ies include the non-​specific effects of light157,183, the
modulation suggest that these pathways may serve sep- effects of tissue warming157 and, in the case of silencing
arable functions. However, several considerations are approaches, off-​target circuit-​related effects184,185. Some
worth noting. of these concerns will undoubtedly be addressed with
Whether glutamate gliotransmission exists under future variants of ChR2 and its derivatives.
physiological conditions is debated owing to contradic- The findings with IP3 sponges and with CalEx are
tory findings16,17, and further demonstration of this pheno­ interesting, as they begin to tackle questions related to
menon with DREADDs (see earlier sections), although causation. However, until we recognize the molecular
supportive, does not address this point. That is, DREADD identity of astrocyte Ca2+ sources, our understanding
activation allows one to trigger a specific pathway in about the nature of the attenuated signals and their
astrocytes but, without additional evidence, this does endogenous engagement will be limited. As more gene
not necessarily mean that those pathways are employed. and protein expression data become available, the field
Furthermore, although activation of GPCRs in astro- could move to more specific approaches to delete or
cytes increases intracellular Ca2+ levels, caution is needed modify certain molecules. Reversible optical activa-
before concluding that Ca2+ is the effector responsible for tion and silencing of endogenous GPCRs186 seems to
the observed synaptic, circuit and behavioural effects. be a promising and emerging approach for future use;
Activation of the Gi pathway is expected to cause there are some obvious candidates to explore, such as
multiple intracellular signalling events, including α-​adrenoceptors167,172,187.
changes in cyclic AMP levels and gene expression. Some Perhaps the best way forward is to capitalize on
or all of these changes may contribute to the reported insights about astrocyte Ca2+ signalling in flies and
neural-​circuit and behavioural changes, in a brain-​ zebra­f ish166,188–191 that were revealed by performing
area-dependent, context-​dependent and experimental-​ genetic screens to identify molecular pathways agnos-
approach-dependent manner. Furthermore, to interpret tically. As it is not restricted to a specific Ca2+ pathway,
the results usefully, several parallel controls for off-​target CalEx enables circuit-​specific suppression of astrocyte
effects of CNO and its breakdown products must be Ca2+ signalling in the mature brain. However, CalEx
performed when using DREADDs179. Studies that do could be further developed to become more useful — for
not report ‘CNO-​only’ controls should be interpreted example, by making it ligand-​activated through designer
cautiously179. engineering, or by modifying it to enable fine temporal
Future studies may benefit from the DREADD control that could enable it to reveal rapid Ca2+-triggered
agonist compound 21 (ref.180), which is not degraded. physiological changes. Of course, intracellular dialysis
Another inhibitory κ-​opioid receptor-​based Gi DREADD with the Ca2+ buffer BAPTA can attenuate astrocyte
(KORD) is activated by the inert ligand salvinorin B128. Ca 2+ signalling. However, this approach is hard to
To our knowledge, KORD has not been tested in astro- deploy in vivo, and recent work suggests that BAPTA
cytes, but could be useful in combination with other has off-​target effects, including effects on ion homeo-
DREADDs in a multiplexed approach, to enable Gi and stasis regulated by Na,K-​ATPase192, that may complicate
Gq GPCR pathways to be stimulated concomitantly in interpretation of synaptic and circuit responses.
the same animal. Many genetically encoded sensors for Ca2+, neuro-
A final concern with the use of DREADDs, especially transmitters and neuromodulators have been developed
in vivo, is that the effects require at least tens of min- in the past few years. Genetically encoded single-​
utes to occur and can last hours — largely depending wavelength sensors are summarized in Table 3, with
on the pharmacodynamics of the DREADD ligands127,181. appropriate references to their development and/or
Thus, DREADDs are not appropriate for every astro- use in astrocytes. Additional sensors will probably be
cyte experiment, and due consideration is needed about developed in the next few years for most neurotrans-
the specific question being addressed and the kinetics mitters and neuromodulators, and will allow research-
of DREADD-​mediated responses in vivo relative to the ers to directly monitor the downstream consequences of
metrics used to evaluate them in brain slices. astrocyte Ca2+ manipulations.
In regard to ChR2, how ion flux and depolarization
of photostimulated ChR2-expressing astrocytes affect Finishing remarks
their physiology is not fully understood, because astro- The past few years have witnessed inspiring advances in
cytes express few voltage-​gated ion channels and do several complementary approaches and tools to explore
not display propagated voltage signals. Furthermore, astrocyte molecular mechanisms, signalling and func-
the effects of ion fluxes on the multiple ion-​coupled tions in adult mice. Although important progress has
transporters that astrocytes express have not been occurred, further technical advances are necessary.
addressed experimentally and will not be straightfor- With currently available tools, it is possible to design
ward to deduce empirically. These concerns are great- experiments to explore astrocyte biology from mole-
est for long-​duration optical stimuli lasting seconds or cules to systems. Additional tools are needed to explore

www.nature.com/nrn
Reviews

how various cell types of the nervous system, including tools is starting to provide evidence that astrocytes
astrocytes, neurons and microglia, interact structurally contribute to complex behavioural and cognitive pro-
and functionally in health and in neuroimmune and cesses45,46,71,72,191,193. Building on fruitful partnerships
brain–periphery interactions. Such tools may crack open between tool generators, modellers and experimental-
completely new areas of biology related to how neurons ists will be crucial for the growth of glial biology and
and glia work synergistically in the brain and how the could catalyse major breakthroughs in cell biology
brain responds to sensory input and controls the body. and medicine.
From this perspective, emerging work with zebrafish
and mice may be portentous. There, the use of improved Published online xx xx xxxx

1. Barres, B. A. The mystery and magic of glia: 26. Orre, M. et al. Acute isolation and transcriptome 43. Boisvert, M. M., Erikson, G. A., Shokhirev, M. N. &
a perspective on their roles in health and disease. characterization of cortical astrocytes and microglia Allen, N. J. The aging astrocyte transcriptome from
Neuron 60, 430–440 (2008). from young and aged mice. Neurobiol. Aging 35, multiple regions of the mouse brain. Cell Rep. 22,
2. Allen, N. J. & Eroglu, C. Cell biology of astrocyte– 1–14 (2014). 269–285 (2018).
synapse interactions. Neuron 96, 697–708 (2017). 27. Liddelow, S. A. et al. Neurotoxic reactive astrocytes 44. Diaz-​Castro, B., Gangwani, M. R., Yu, X., Coppola, G.
3. Eroglu, C. & Barres, B. A. Regulation of synaptic are induced by activated microglia. Nature 541, & Khakh, B. S. Astrocyte molecular signatures in
connectivity by glia. Nature 468, 223–231 (2010). 481–487 (2017). Huntington’s disease. Sci. Transl Med. 11, eaaw8546
4. Khakh, B. S. & Sofroniew, M. V. Diversity of astrocyte 28. Sun, W. et al. SOX9 is an astrocyte-​specific nuclear (2019).
functions and phenotypes in neural circuits. marker in the adult brain outside the neurogenic 45. Nagai, J. et al. Hyperactivity with disrupted attention
Nat. Neurosci. 18, 942–952 (2015). regions. J. Neurosci. 37, 4493–4507 (2017). by activation of an astrocyte synaptogenic cue. Cell
5. Khakh, B. S. Astrocyte–neuron interactions in the 29. John Lin, C. C. et al. Identification of diverse 177, 1280–1292.e20 (2019).
striatum: insights on identity, form, and function. astrocyte populations and their malignant analogs. This study finds that chemogenetic activation of
Trends Neurosci. 42, 617–630 (2019). Nat. Neurosci. 20, 396–405 (2017). the astrocyte Gi-​GPCR pathway in the dorsal
6. Allen, N. J. & Barres, B. A. Neuroscience: glia — more This study identifies five astrocyte subpopulations striatum causes attention deficit hyperactivity
than just brain glue. Nature 457, 675–677 (2009). that are each present across five brain regions disorder-​like behaviours in mice through
7. Allen, N. J. & Lyons, D. A. Glia as architects of central using cell sorting of ALDH1L1-GFP+ astrocytes. reactivation of astrocyte-​induced synaptogenesis.
nervous system formation and function. Sci. 362, The subpopulations are molecularly and 46. Yu, X. et al. Reducing astrocyte calcium signaling
181–185 (2018). functionally diverse, and highlight a synaptogenesis- in vivo alters striatal microcircuits and causes
8. Chung, W.-S., Welsh, C. A., Barres, B. A. & Stevens, B. promoting subpopulation in the context of glioma. repetitive behavior. Neuron 99, 1170–1187 (2018).
Do glia drive synaptic and cognitive impairment in 30. Zhang, Y. et al. Purification and characterization of This study develops the novel genetic approach of
disease? Nat. Neurosci. 18, 1539–1545 (2015). progenitor and mature human astrocytes reveals CalEx to attenuate astrocyte Ca2+-dependent
9. Jorgenson, L. A. et al. The BRAIN Initiative: transcriptional and functional differences with mouse. signalling, and describes its application in the
developing technology to catalyse neuroscience Neuron 89, 37–53 (2016). dorsal striatum, which resulted in obsessive–
discovery. Philos. Trans. R. Soc. Lond. B Biol. Sci. 31. Oberheim, N. A. et al. Uniquely hominid features of compulsive disorder-​like behaviour in mice
370, 20140164 (2015). adult human astrocytes. J. Neurosci. 29, 3276–3287 through GABA-​mediated neuromodulation.
10. Deisseroth, K. & Schnitzer, M. J. Engineering (2009). 47. Svensson, V., Vento-​Tormo, R. & Teichmann, S. A.
approaches to illuminating brain structure and 32. Holt, L. M. & Olsen, M. L. Novel applications of Exponential scaling of single-​cell RNA-​seq in the
dynamics. Neuron 80, 568–577 (2013). magnetic cell sorting to analyze cell-​type specific gene past decade. Nat. Protoc. 13, 599–604 (2018).
11. Rajasethupathy, P., Ferenczi, E. & Deisseroth, K. and protein expression in the central nervous system. 48. Poulin, J. F., Tasic, B., Hjerling-​Leffler, J., Trimarchi, J. M.
Targeting neural circuits. Cell 165, 524–534 (2016). PLoS One 11, e0150290 (2016). & Awatramani, R. Disentangling neural cell diversity
12. Kettenmann, H. & Verkhratsky, A. Neuroglia: the 33. Holt, L. M., Stoyanof, S. T. & Olsen, M. L. Magnetic using single-​cell transcriptomics. Nat. Neurosci. 19,
150 years after. Trends Neurosci. 31, 653–659 (2008). cell sorting for in vivo and in vitro astrocyte, neuron, 1131–1141 (2016).
13. Kuffler, S. W. Neuroglial cells: physiological properties and microglia analysis. Curr. Protoc. Neurosci. 88, e71 49. Chen, X., Teichmann, S. A. & Meyer, K. B. From tissues
and a potassium mediated effect of neuronal activity (2019). to cell types and back: single-​cell gene expression
on the glial membrane potential. Proc. R. Soc. Lond. 34. Holt, L. M. et al. Astrocyte morphogenesis is analysis of tissue architecture. Annu. Rev. Biomed.
B Biol. Sci. 168, 1–21 (1967). dependent on BDNF signaling via astrocytic TrkB.T1. Data Sci 1, 29–51 (2018).
14. Brenner, S. Life sentences: Detective Rummage eLife 8, e44667 (2019). 50. Zeisel, A. et al. Brain structure. Cell types in the
investigates. Genome Biol. 3, 1013.1–1013.2 35. Heiman, M. et al. A translational profiling approach mouse cortex and hippocampus revealed by single-​cell
(2002). for the molecular characterization of CNS cell types. RNA-​seq. Science 347, 1138–1142 (2015).
15. Haydon, P. G. & Nedergaard, M. How do astrocytes Cell 135, 738–748 (2008). 51. Wu, Y. E., Pan, L., Zuo, Y. N., Li, X. M. & Hong, W. Z.
participate in neural plasticity? Cold Spring Harb. 36. Doyle, J. P. et al. Application of a translational profiling Detecting activated cell populations using single-​cell
Perspect. Biol. 11, a020438 (2014). approach for the comparative analysis of CNS cell RNA-​seq. Neuron 96, 313–329 (2017).
16. Fiacco, T. A. & McCarthy, K. D. Multiple lines of types. Cell 135, 749–762 (2008). 52. Zeisel, A. et al. Molecular architecture of the mouse
evidence indicate that gliotransmission does not occur 37. Clarke, L. E. et al. Normal aging induces A1-like nervous system. Cell 174, 999–1014.e22 (2018).
under physiological conditions. J. Neurosci. 38, 3–13 astrocyte reactivity. Proc. Natl Acad. Sci. USA 115, 53. Saunders, A. et al. Molecular diversity and
(2018). E1896–E1905 (2018). specializations among the cells of the adult mouse
17. Savtchouk, I. & Volterra, A. Gliotransmission: beyond 38. Sanz, E. et al. Cell-​type-specific isolation of brain. Cell 174, 1015–1030.e16 (2018).
black-​and-white. J. Neurosci. 38, 14–25 (2018). ribosome-​associated mRNA from complex tissues. 54. Schaum, N. et al. Single-​cell transcriptomics of 20
18. Escartin, C., Guillemaud, O. & Carrillo-​de Sauvage, M. A. Proc. Natl Acad. Sci. USA 106, 13939–13944 mouse organs creates a Tabula Muris. Nature 562,
Questions and (some) answers on reactive astrocytes. (2009). 367–372 (2018).
Glia 67, 2221–2247 (2019). 39. Srinivasan, R. et al. New transgenic mouse lines for 55. Khakh, B. S. & Deneen, B. The emerging nature of
19. Barres, B. A., Silverstein, B. E., Corey, D. P. selectively targeting astrocytes and studying calcium astrocyte diversity. Annu. Rev. Neurosci. 42, 187–207
& Chun, L. L. Immunological, morphological, and signals in astrocyte processes in situ and in vivo. (2019).
electrophysiological variation among retinal ganglion Neuron 92, 1181–1195 (2016). 56. Kelley, K. W., Nakao-​Inoue, H., Molofsky, A. V. &
cells purified by panning. Neuron 1, 791–803 (1988). This study systematically compares several Cre-​ Oldham, M. C. Variation among intact tissue
20. Bachoo, R. M. et al. Molecular diversity of astrocytes dependent transgenic mouse lines for genetically samples reveals the core transcriptional features
with implications for neurological disorders. targeting astrocytes and characterized the Aldh1l1- of human CNS cell classes. Nat. Neurosci. 21,
Proc. Natl Acad. Sci. USA 101, 8384–8389 (2004). Cre/ERT2 mouse line as an astrocyte-​selective, 1171–1184 (2018).
21. Lovatt, D. et al. The transcriptome and metabolic gene pan-​astrocytic and inducible Cre line. 57. Cuevas-​Diaz Duran, R., C. Y., W., Zheng, H., Deneen, B.
signature of protoplasmic astrocytes in the adult 40. Chai, H. et al. Neural circuit-​specialized astrocytes: & Wu, J. Q. Brain region-​specific gene signatures
murine cortex. J. Neurosci. 27, 12255–12266 (2007). transcriptomic, proteomic, morphological and revealed by distinct astrocyte subpopulations unveil
22. Cahoy, J. D. et al. A transcriptome database for functional evidence. Neuron 95, 531–549 links to glioma and neurodegenerative diseases.
astrocytes, neurons, and oligodendrocytes: a new (2017). eNeuro https://doi.org/10.1523/ENEURO.0288-
resource for understanding brain development and This study compares astrocytes in the adult 18.2019 (2019).
function. J. Neurosci. 28, 264–278 (2008). hippocampus and striatum using RiboTag 58. Blanco-​Suarez, E., Liu, T. F., Kopelevich, A. & Allen, N. J.
23. Heintz, N. Gene expression nervous system atlas RNA-​seq, FACS-​based proteomics, serial Astrocyte-​secreted chordin-​like 1 drives synapse
(GENSAT). Nat. Neurosci. 7, 483 (2004). section electron microscopy electrophysiology maturation and limits plasticity by increasing synaptic
24. Zhang, Y. et al. An RNA-​sequencing transcriptome and and calcium imaging to reveal circuit-​specific GluA2 AMPA receptors. Neuron 100, 1116–1132
splicing database of glia, neurons, and vascular cells of astrocyte properties. (2018).
the cerebral cortex. J. Neurosci. 34, 11929–11947 41. Anderson, M. A. et al. Astrocyte scar formation aids 59. Park, Y. M., Chun, H., Shin, J. I. & Lee, C. J. Astrocyte
(2014). central nervous system axon regeneration. Nature specificity and coverage of hGFAP-​CreERT2 [Tg(GFAP-​
This study is the first transcriptomic analysis of 532, 195–200 (2016). Cre/ERT2)13Kdmc] mouse line in various brain
purified astrocytes and other cell types in the 42. Itoh, N. et al. Cell-​specific and region-​specific regions. Exp. Neurobiol. 27, 508–525 (2018).
mouse cortex by immunopanning and RNA-​seq. transcriptomics in the multiple sclerosis model: 60. Zhuo, L. et al. hGFAP-​cre transgenic mice for
25. Zamanian, J. L. et al. Genomic analysis of reactive focus on astrocytes. Proc. Natl Acad. Sci. USA 115, manipulation of glial and neuronal function in vivo.
astrogliosis. J. Neurosci. 32, 6391–6410 (2012). E302–E309 (2018). Genes 31, 85–94 (2001).

Nature Reviews | Neuroscience


Reviews

61. Sofroniew, M. V. Transgenic techniques for cell ablation unique domains upon reaction to injury. Proc. Natl 105. Duffy, S. & MacVicar, B. A. Adrenergic calcium signaling
or molecular deletion to investigate functions of Acad. Sci. USA 103, 17513–17518 (2006). in astrocyte networks within the hippocampal slice.
astrocytes and other GFAP-​expressing cell types. 84. Moye, S. L., Diaz-​Castro, B., Gangwani, M. R. & J. Neurosci. 15, 5535–5550 (1995).
Methods Mol. Biol. 814, 531–544 (2012). Khakh, B. S. A protocol for visualizing astrocyte 106. Hirase, H., Qian, L., Bartho, P. & Buzsaki, G. Calcium
62. Mori, T. et al. Inducible gene deletion in astroglia and morphology using Lucifer yellow iontophoresis. JoVE dynamics of cortical astrocytic networks in vivo.
radial glia — a valuable tool for functional and lineage 151, e60225 (2019). PLoS Biol. 2, E96 (2004).
analysis. Glia 54, 21–34 (2006). 85. Zhou, B. et al. Astroglial dysfunctions drive aberrant 107. Nimmerjahn, A. & Bergles, D. E. Large-​scale recording
63. Slezak, M. et al. Transgenic mice for conditional gene synaptogenesis and social behavioral deficits in mice of astrocyte activity. Curr. Opin. Neurobiol. 32,
manipulation in astroglial cells. Glia 55, 1565–1576 with neonatal exposure to lengthy general anesthesia. 95–106 (2015).
(2007). PLoS Biol. 17, e3000086 (2019). 108. Wang, X. et al. Astrocytic Ca2+ signaling evoked by
64. Jahn, H. M., Scheller, A. & Kirchhoff, F. Genetic This study uses iontophoresis of fluorescent dye to sensory stimulation in vivo. Nat. Neurosci. 9,
control of astrocyte function in neural circuits. show that exposure to general anaesthetics during 816–823 (2006).
Front. Cell Neurosci. 9, 310 (2015). brain development reduces astrocyte Ca2+ signalling 109. Grosche, J. et al. Microdomains for neuron–glia
65. Santello, M., Toni, N. & Volterra, A. Astrocyte function and morphogenesis, resulting in neural circuit interaction: parallel fiber signaling to Bergmann glial
from information processing to cognition and cognitive malformation. cells. Nat. Neurosci. 2, 139–143 (1999).
impairment. Nat. Neurosci. 22, 154–166 (2019). 86. Halassa, M. M., Fellin, T., Takano, H., Dong, J. H. & 110. Shigetomi, E., Kracun, S., Sofroniew, M. V. & Khakh, B. S.
66. Winchenbach, J. et al. Inducible targeting of CNS Haydon, P. G. Synaptic islands defined by the territory A genetically targeted optical sensor to monitor
astrocytes in Aldh1l1-CreERT2 BAC transgenic mice. of a single astrocyte. J. Neurosci. 27, 6473–6477 calcium signals in astrocyte processes. Nat. Neurosci.
F1000Res 5, 2934 (2016). (2007). 13, 759–766 (2010).
67. Foo, L. C. & Dougherty, J. D. Aldh1L1 is expressed by 87. Lanjakornsiripan, D. et al. Layer-​specific morphological 111. Shigetomi, E., Patel, S. & Khakh, B. S. Probing
postnatal neural stem cells in vivo. Glia 61, 1533–1541 and molecular differences in neocortical astrocytes and the complexities of astrocyte calcium signaling.
(2013). their dependence on neuronal layers. Nat. Commun. 9, Trends Cell Biol. 26, 300–312 (2016).
68. Ortinski, P. I. et al. Selective induction of astrocytic 1623 (2018). 112. Clapham, D. E. Calcium signaling. Cell 131, 1047–1058
gliosis generates deficits in neuronal inhibition. In this paper, the authors sparsely label astrocytes (2007).
Nat. Neurosci. 13, 584–591 (2010). in the somatosensory cortex using the Slc1a3-Cre/ 113. Bazargani, N. & Attwell, D. Astrocyte calcium
69. Merienne, N., Le Douce, J., Faivre, E., Deglon, N. & ERT2 mouse line, and detail the cortical layer-​ signalling: the third wave. Nat. Neurosci. 19,
Bonvento, G. Efficient gene delivery and selective specific morphology and the molecular machinery 182–189 (2016).
transduction of astrocytes in the mammalian brain involved therein. 114. Volterra, A., Liaudet, N. & Savtchouk, I. Astrocyte
using viral vectors. Front. Cell Neurosci. 7, 106 88. Lu, X. H. & Yang, X. W. Genetically-​directed sparse Ca2+ signalling: an unexpected complexity.
(2013). neuronal labeling in BAC transgenic mice through Nat. Rev. Neurosci. 15, 327–335 (2014).
70. Lee, Y., Messing, A., Su, M. & Brenner, M. GFAP mononucleotide repeat frameshift. Sci. Rep. 7, 43915 115. Hires, S. A., Tian, L. & Looger, L. L. Reporting neural
promoter elements required for region-​specific and (2017). activity with genetically encoded calcium indicators.
astrocyte-​specific expression. Glia 56, 481–493 89. Bernardinelli, Y. et al. Activity-​dependent structural Brain Cell Biol. 36, 69–86 (2008).
(2008). plasticity of perisynaptic astrocytic domains promotes 116. Dana, H. et al. High-​performance calcium sensors
71. Adamsky, A. et al. Astrocytic activation generates excitatory synapse stability. Curr. Biol. 24, 1679–1688 for imaging activity in neuronal populations and
de novo neuronal potentiation and memory (2014). microcompartments. Nat. Methods 16, 649–657
enhancement. Cell 174, 59–71 (2018). 90. Stogsdill, J. A. et al. Astrocytic neuroligins control (2019).
This study shows that activation of astrocyte astrocyte morphogenesis and synaptogenesis. Nature 117. Chen, T. W. et al. Ultrasensitive fluorescent proteins
Gq-GPCR signalling in the hippocampal CA1 with 551, 192–197 (2017). for imaging neuronal activity. Nature 499, 295–300
Gq-DREADDs and opto-​α1-AR approaches induces 91. Testen, A. et al. Region-​specific differences in (2013).
synaptic long-​term potentiation and enhances morphometric features and synaptic colocalization of 118. Tian, L. et al. Imaging neural activity in worms, flies
memory acquisition in mice. astrocytes during development. Neuroscience 400, and mice with improved GCaMP calcium indicators.
72. Cui, Y. et al. Astroglial Kir4.1 in the lateral habenula 98–109 (2019). Nat. Methods. 6, 875–881 (2009).
drives neuronal bursts in depression. Nature 554, 92. Viswanathan, S. et al. High-​performance probes for 119. Agulhon, C. et al. What is the role of astrocyte calcium
323–327 (2018). light and electron microscopy. Nat. Methods 12, in neurophysiology? Neuron 59, 932–946 (2008).
73. Octeau, J. C. et al. An optical neuron–astrocyte 568–576 (2015). 120. Fiacco, T. A., Agulhon, C. & McCarthy, K. D.
proximity assay at synaptic distance scales. Neuron 93. Shigetomi, E. et al. Imaging calcium microdomains Sorting out astrocyte physiology from pharmacology.
98, 49–66 (2018). within entire astrocyte territories and endfeet with Annu. Rev. Pharmacol. Toxicol. 49, 151–174 (2009).
74. Tong, X. et al. Astrocyte Kir4.1 ion channel deficits GCaMPs expressed using adeno-​associated viruses. 121. Volgraf, M. et al. Allosteric control of an ionotropic
contribute to neuronal dysfunction in Huntington’s J. Gen. Physiol. 141, 633–647 (2013). glutamate receptor with an optical switch.
disease model mice. Nat. Neurosci. 17, 694–703 94. Maxwell, D. S. & Kruger, L. The fine structure of Nat. Chem. Biol. 2, 47–52 (2006).
(2014). astrocytes in the cerebral cortex and their response 122. Szobota, S. et al. Remote control of neuronal activity
75. Bonder, D. E. & McCarthy, K. D. Astrocytic Gq-GPCR- to focal injury produced by heavy ionizing particles. with a light-​gated glutamate receptor. Neuron 54,
linked IP3R-dependent Ca2+ signaling does not mediate J. Cell Biol. 25, 141–157 (1965). 535–545 (2007).
neurovascular coupling in mouse visual cortex in vivo. 95. Ventura, R. & Harris, K. M. Three-​dimensional 123. Li, D., Herault, K., Isacoff, E. Y., Oheim, M. &
J. Neurosci. 34, 13139–13150 (2014). relationships between hippocampal synapses and Ropert, N. Optogenetic activation of LiGluR-​
76. Poskanzer, K. E. & Yuste, R. Astrocytes regulate cortical astrocytes. J. Neurosci. 19, 6897–6906 (1999). expressing astrocytes evokes anion channel-​mediated
state switching in vivo. Proc. Natl Acad. Sci. USA 113, This pioneering study investigates the structural glutamate release. J. Physiol. 590, 855–873 (2012).
E2675–E2684 (2016). relationship between astrocyte processes and 124. Porter, J. T. & McCarthy, K. D. Astrocytic
77. Koh, W., Park, Y. M., Lee, S. E. & Lee, C. J. synapses in the rat hippocampal CA1 region with neurotransmitter receptors in situ and in vivo.
AAV-mediated astrocyte-​specific gene expression serial section electron microscopy. Prog. Neurobiol. 51, 439–455 (1997).
under human ALDH1L1 promoter in mouse thalamus. 96. Gavrilov, N. et al. Astrocytic coverage of dendritic 125. Hamby, M. E. et al. Inflammatory mediators alter the
Exp. Neurobiol. 26, 350–361 (2017). spines, dendritic shafts, and axonal boutons in astrocyte transcriptome and calcium signaling elicited
78. Mudannayake, J. M., Mouravlev, A., Fong, D. M. & hippocampal neuropil. Front. Cell Neurosci. 12, 248 by multiple G-​protein-coupled receptors. J. Neurosci.
Young, D. Transcriptional activity of novel ALDH1L1 (2018). 32, 14489–14510 (2012).
promoters in the rat brain following AAV vector-​ 97. Heller, J. P., Odii, T., Zheng, K. & Rusakov, D. A. 126. Armbruster, B. N., Li, X., Pausch, M. H., Herlitze, S.
mediated gene transfer. Mol. Ther. Methods Clin. Dev. Imaging tripartite synapses using super-​resolution & Roth, B. L. Evolving the lock to fit the key to create a
3, 16075 (2016). microscopy. Methods Mol. Biol. https://doi.org/ family of G protein-​coupled receptors potently activated
79. Chan, K. Y. et al. Engineered AAVs for efficient 10.1016/j.ymeth.2019.05.024 (2019). by an inert ligand. Proc. Natl Acad. Sci. USA 104,
noninvasive gene delivery to the central and 98. Chen, F., Tillberg, P. W. & Boyden, E. S. Optical 5163–5168 (2007).
peripheral nervous systems. Nat. Neurosci. 20, imaging. Expansion microscopy. Science 347, 127. Roth, B. L. DREADDs for neuroscientists. Neuron 89,
1172–1179 (2017). 543–548 (2015). 683–694 (2016).
80. Bushong, E. A., Martone, M. E., Jones, Y. Z. & 99. Willig, K. I. et al. Nanoscale resolution in GFP-​based 128. Vardy, E. et al. A new DREADD facilitates the
Ellisman, M. H. Protoplasmic astrocytes in CA1 microscopy. Nat. Methods 3, 721–723 (2006). multiplexed chemogenetic interrogation of behavior.
stratum radiatum occupy separate anatomical 100. Tang, S. & Yasuda, R. Imaging ERK and PKA activation Neuron 86, 936–946 (2015).
domains. J. Neurosci. 22, 183–192 (2002). in single dendritic spines during structural plasticity. 129. Guettier, J. M. et al. A chemical-​genetic approach to
This landmark study reveals tiling properties of Neuron 93, 1315–1324.e3 (2017). study G protein regulation of β cell function in vivo.
adjacent astrocytes in the rat hippocampal CA1 101. Korogod, N., Petersen, C. C. & Knott, G. W. Proc. Natl Acad. Sci. USA 106, 19197–19202 (2009).
region using iontophoresis of two different Ultrastructural analysis of adult mouse neocortex 130. Inoue, A. et al. Illuminating G-​protein-coupling selectivity
fluorescent dyes. comparing aldehyde perfusion with cryo fixation. of GPCRs. Cell 177, 1933–1947.e25 (2019).
81. Reeves, A. M., Shigetomi, E. & Khakh, B. S. Bulk eLife 4, e05793 (2015). 131. Roth, B. L. How structure informs and transforms
loading of calcium indicator dyes to study astrocyte 102. Kasthuri, N. et al. Saturated reconstruction of a chemogenetics. Curr. Opin. Struct. Biol. 57, 9–16
physiology: key limitations and improvements using volume of neocortex. Cell 162, 648–661 (2015). (2019).
morphological maps. J. Neurosci. 31, 9353–9358 103. Cornell-​Bell, A. H., Finkbeiner, S. M., Cooper, M. S. 132. Rosenbaum, D. M., Rasmussen, S. G. & Kobilka, B. K.
(2011). & Smith, S. J. Glutamate induces calcium waves in The structure and function of G-​protein-coupled
82. Bushong, E. A., Martone, M. E. & Ellisman, M. H. cultured astrocytes: long-​range glial signaling. Science receptors. Nature 459, 356–363 (2009).
Maturation of astrocyte morphology and the 247, 470–473 (1990). 133. Pierce, K. L., Premont, R. T. & Lefkowitz, R. J. Seven-​
establishment of astrocyte domains during postnatal 104. Charles, A. C., Merrill, J. E., Dirksen, E. R. & transmembrane receptors. Nat. Rev. Mol. Cell Biol. 3,
hippocampal development. Int. J. Dev. Neurosci. 22, Sanderson, M. J. Intercellular signaling in glial cells: 639–650 (2002).
73–86 (2004). calcium waves and oscillations in response to 134. Durkee, C. A. et al. Gi/o protein-​coupled receptors
83. Wilhelmsson, U. et al. Redefining the concept of mechanical stimulation and glutamate. Neuron 6, inhibit neurons but activate astrocytes and stimulate
reactive astrocytes as cells that remain within their 983–992 (1991). gliotransmission. Glia 67, 1076–1093 (2019).

www.nature.com/nrn
Reviews

135. Yang, L., Qi, Y. & Yang, Y. Astrocytes control food intake 161. Uchiyama, T., Yoshikawa, F., Hishida, A., Furuichi, T. & 186. Pittolo, S. et al. Reversible silencing of endogenous
by inhibiting AGRP neuron activity via adenosine A1 Mikoshiba, K. A novel recombinant hyperaffinity receptors in intact brain tissue using 2-photon
receptors. Cell Rep. 11, 798–807 (2015). inositol 1,4,5-trisphosphate (IP3) absorbent traps IP3, pharmacology. Proc. Natl Acad. Sci. USA 116,
136. Chen, N. et al. Direct modulation of GFAP-​expressing resulting in specific inhibition of IP3-mediated calcium 13680–13689 (2019).
glia in the arcuate nucleus bi-​directionally regulates signaling. J. Biol. Chem. 277, 8106–8113 (2002). 187. Ding, F. et al. α1-Adrenergic receptors mediate
feeding. eLife 5, e18716 (2016). 162. Tanaka, M. et al. Astrocytic Ca2+ signals are required coordinated Ca2+ signaling of cortical astrocytes in
137. Martin-​Fernandez, M. et al. Synapse-​specific astrocyte for the functional integrity of tripartite synapses. awake, behaving mice. Cell Calcium 54, 387–394
gating of amygdala-​related behavior. Nat. Neurosci. Mol. Brain 6, 6 (2013). (2013).
20, 1540–1548 (2017). 163. de Vivo, L., Melone, M., Rothstein, J. D. & Conti, F. 188. Ma, Z., Stork, T., Bergles, D. E. & Freeman, M. R.
138. Nam, M. H. et al. Activation of astrocytic μ-​opioid GLT-1 promoter activity in astrocytes and neurons Neuromodulators signal through astrocytes to alter
receptor causes conditioned place preference. of mouse hippocampus and somatic sensory cortex. neural circuit activity and behaviour. Nature 539,
Cell Rep. 28, 1154–1166 (2019). Front. Neuroanat. 3, 31 (2010). 428–432 (2016).
139. Corkrum, M., Rothwell, P. E., Thomas, M. J., Kofuji, P. 164. Xie, Y., Wang, T., Sun, G. Y. & Ding, S. Specific 189. Melom, J. E. & Littleton, J. T. Mutation of a NCKX
& Araque, A. Opioid-​mediated astrocyte–neuron disruption of astrocytic Ca2+ signaling pathway in vivo eliminates glial microdomain calcium oscillations
signaling in the nucleus accumbens. Cells 8, E586 by adeno-​associated viral transduction. Neuroscience and enhances seizure susceptibility. J. Neurosci. 33,
(2019). 170, 992–1003 (2010). 1169–1178 (2013).
140. Kol, A. et al. Astrocytes contribute to remote memory 165. Várnai, P. et al. Inositol lipid binding and membrane 190. Weiss, S., Melom, J. E., Ormerod, K. G., Zhang, Y. V. &
formation by modulating hippocampal–cortical localization of isolated pleckstrin homology (PH) Littleton, J. T. Glial Ca2+ signaling links endocytosis to
communication during learning. Preprint at bioRxiv domains. Studies on the PH domains of phospholipase K+ buffering around neuronal somas to regulate
https://doi.org/10.1101/682344 (2019). Cδ1 and p130. J. Biol. Chem. 277, 27412–27422 excitability. eLife 8, e44186 (2019).
141. Bernstein, J. G. & Boyden, E. S. Optogenetic tools (2002). 191. Mu, Y. et al. Glia accumulate evidence that actions are
for analyzing the neural circuits of behavior. 166. Zhang, Y. V., Ormerod, K. G. & Littleton, J. T. Astrocyte futile and suppress unsuccessful behavior. Cell 178,
Trends Cogn. Sci. 15, 592–600 (2011). Ca2+ influx negatively regulates neuronal activity. 27–43 (2019).
142. Deisseroth, K. Optogenetics: 10 years of microbial eNeuro https://doi.org/10.1523/ENEURO.0340- 192. Smith, N. A. et al. Fluorescent Ca2+ indicators directly
opsins in neuroscience. Nat. Neurosci. 18, 1213–1225 16.2017 (2017). inhibit the Na,K-​ATPase and disrupt cellular functions.
(2015). 167. Srinivasan, R. et al. Ca2+ signaling in astrocytes Sci. Signal. 11, eaal2039 (2018).
143. Gourine, A. V. et al. Astrocytes control breathing from Ip3r2–/– mice in brain slices and during startle 193. Diaz Verdugo, C. et al. Glia–neuron interactions
through pH-​dependent release of ATP. Science 329, responses in vivo. Nat. Neurosci. 18, 708–717 (2015). underlie state transitions to generalized seizures.
571–575 (2010). 168. Wang, Y. et al. Accurate quantification of astrocyte and Nat. Commun. 10, 3830 (2019).
144. Perea, G., Yang, A., Boyden, E. S. & Sur, M. neurotransmitter fluorescence dynamics for single-​cell 194. Davie, K. et al. A single-​cell transcriptome atlas
Optogenetic astrocyte activation modulates and population-​level physiology. Nat. Neurosci. 11, of the aging drosophila brain. Cell 174, 982–998
response selectivity of visual cortex neurons in vivo. 1936–1944 (2018). (2018).
Nat. Commun. 5, 3262 (2014). This study provides Astrocyte Quantitative Analysis 195. Chen, R., Wu, X., Jiang, L. & Zhang, Y. Single-​cell
145. Masamoto, K. et al. Unveiling astrocytic control of (AQuA) software enabling event-​based detection RNA-seq reveals hypothalamic cell diversity. Cell Rep.
cerebral blood flow with optogenetics. Sci. Rep. 5, of astrocyte Ca2+ or optical signals and the 18, 3227–3241 (2017).
11455 (2015). quantification of their spatio-​temporal dynamics. 196. Lake, B. B. et al. Integrative single-​cell analysis of
146. Pelluru, D., Konadhode, R. R., Bhat, N. R. & 169. Zheng, K. et al. Time-​resolved imaging reveals transcriptional and epigenetic states in the human
Shiromani, P. J. Optogenetic stimulation of astrocytes in heterogeneous landscapes of nanomolar Ca2+ in adult brain. Nat. Biotechnol. 36, 70–80 (2018).
the posterior hypothalamus increases sleep at night neurons and astroglia. Neuron 88, 277–288 (2015). 197. Garcia, A. D., Doan, N. B., Imura, T., Bush, T. G. &
in C57BL/6J mice. Eur. J. Neurosci. 43, 1298–1306 170. Bindocci, E. et al. Three-​dimensional Ca2+ imaging Sofroniew, M. V. GFAP-​expressing progenitors are
(2016). advances understanding of astrocyte biology. Science the principal source of constitutive neurogenesis in
147. Yamashita, A. et al. Astrocytic activation in the anterior 356, eaai8185 (2017). adult mouse forebrain. Nat. Neurosci. 7, 1233–1241
cingulate cortex is critical for sleep disorder under 171. Sekiguchi, K. J. et al. Imaging large-​scale cellular (2004).
neuropathic pain. Synapse 68, 235–247 (2014). activity in spinal cord of freely behaving mice. 198. Gregorian, C. et al. Pten deletion in adult neural stem/
148. Sasaki, T. et al. Application of an optogenetic Nat. Commun. 7, 11450 (2016). progenitor cells enhances constitutive neurogenesis.
byway for perturbing neuronal activity via glial 172. Paukert, M. et al. Norepinephrine controls astroglial J. Neurosci. 29, 1874–1886 (2009).
photostimulation. Proc. Natl Acad. Sci. USA 109, responsiveness to local circuit activity. Neuron 82, 199. Hirrlinger, P. G., Scheller, A., Braun, C., Hirrlinger, J. &
20720–20725 (2012). 1263–1270 (2014). Kirchhoff, F. Temporal control of gene recombination in
149. Octeau, J. C. et al. Transient, consequential 173. Handy, G., Taheri, M., White, J. A. & Borisyuk, A. astrocytes by transgenic expression of the tamoxifen-​
increases in extracellular potassium ions accompany Mathematical investigation of IP3-dependent calcium inducible DNA recombinase variant CreERT2. Glia 54,
channelrhodopsin2 excitation. Cell Rep. 27, dynamics in astrocytes. J. Comput. Neurosci. 42, 11–20 (2006).
2249–2261.e7 (2019). 257–273 (2017). 200. Pascual, O. et al. Astrocytic purinergic signaling
150. McDonough, A. A. & Youn, J. H. Potassium 174. Taheri, M., Handy, G., Borisyuk, A. & White, J. A. coordinates synaptic networks. Science 310, 113–116
homeostasis: the knowns, the unknowns, and the Diversity of evoked astrocyte Ca2+ dynamics quantified (2005).
health benefits. Physiology 32, 100–111 (2017). through experimental measurements and mathematical 201. Lin, W. et al. Interferon-​γ induced medulloblastoma
151. Koizumi, A., Tanaka, K. F. & Yamanaka, A. The modeling. Front. Syst. Neurosci. 11, 79 (2017). in the developing cerebellum. J. Neurosci. 24,
manipulation of neural and cellular activities by 175. Cresswell-​Clay, E., Crock, N., Tabak, J. & Erlebacher, G. 10074–10083 (2004).
ectopic expression of melanopsin. Neurosci. Res. 75, A compartmental model to investigate local and global 202. Tanaka, M. et al. Lack of Connexin43-mediated
3–5 (2013). Ca2+ dynamics in astrocytes. Front. Comput. Neurosci. Bergmann glial gap junctional coupling does not affect
152. Spoida, K. et al. Melanopsin variants as intrinsic 12, 94 (2018). cerebellar long-​term depression, motor coordination,
optogenetic on and off switches for transient versus 176. Savtchenko, L. P. et al. Disentangling astroglial or eyeblink conditioning. Front. Behav. Neurosci.
sustained activation of G protein pathways. Curr. Biol. physiology with a realistic cell model in silico. 2, 1 (2008).
26, 1206–1212 (2016). Nat. Commun. 9, 3554 (2018). 203. Young, K. M. et al. An Fgfr3-iCreERT2 transgenic
153. Mederos, S. et al. Melanopsin for precise optogenetic 177. De Pittà, M., Volman, V., Levine, H. & Ben-​Jacob, E. mouse line for studies of neural stem cells and
activation of astrocyte–neuron networks. Glia 67, Multimodal encoding in a simplified model of astrocytes. Glia 58, 943–953 (2010).
915–934 (2019). intracellular calcium signaling. Cogn. Process. 10, 204. Drinkut, A., Tereshchenko, Y., Schulz, J. B., Bahr, M.
154. Spangler, S. M. & Bruchas, M. R. Optogenetic S55–S70 (2009). & Kugler, S. Efficient gene therapy for Parkinson’s
approaches for dissecting neuromodulation and GPCR 178. Denizot, A., Arizono, M., Nägerl, U. V., Soula, H. & disease using astrocytes as hosts for localized
signaling in neural circuits. Curr. Opin. Pharmacol. 32, Berry, H. Simulation of calcium signaling in fine neurotrophic factor delivery. Mol. Ther. 20, 534–543
56–70 (2017). astrocytic processes: effect of spatial properties on (2012).
155. Airan, R. D., Thompson, K. R., Fenno, L. E., Bernstein, H. spontaneous activity. PLoS Comput. Biol. 15, 205. Lawlor, P. A., Bland, R. J., Mouravlev, A., Young, D.
& Deisseroth, K. Temporally precise in vivo control of e1006795 (2019). & During, M. J. Efficient gene delivery and selective
intracellular signalling. Nature 458, 1025 (2009). 179. Gomez, J. L. et al. Chemogenetics revealed: DREADD transduction of glial cells in the mammalian brain by
156. Stujenske, J. M., Spellman, T. & Gordon, J. A. occupancy and activation via converted clozapine. AAV serotypes isolated from nonhuman primates.
Modeling the spatiotemporal dynamics of light and Science 357, 503–507 (2017). Mol. Ther. 17, 1692–1702 (2009).
heat propagation for in vivo optogenetics. Cell Rep. 180. Chen, X. et al. The first structure–activity relationship 206. Vagner, T., Dvorzhak, A., Wojtowicz, A. M., Harms, C.
12, 525–534 (2015). studies for designer receptors exclusively activated by & Grantyn, R. Systemic application of AAV vectors
157. Owen, S. F., Liu, M. H. & Kreitzer, A. C. Thermal designer drugs. ACS Chem. Neurosci. 6, 476–484 targeting GFAP-​expressing astrocytes in Z-​Q175-KI
constraints on in vivo optogenetic manipulations. (2015). Huntington’s disease mice. Mol. Cell Neurosci. 77,
Nat. Neurosci. 22, 1061–1065 (2019). 181. Smith, M. A., Zhang, H. & Robinson, A. M. The effects of 76–86 (2016).
158. Sharp, A. H. et al. Differential cellular expression of excitatory and inhibitory social cues on cocaine-seeking 207. Dana, H. et al. Sensitive red protein calcium indicators
isoforms of inositol 1,4,5-triphosphate receptors in behavior. Front. Behav. Neurosci. 10, 217 (2016). for imaging neural activity. eLife 5, e12727 (2016).
neurons and glia in brain. J. Comp. Neurol. 406, 182. Beppu, K. et al. Optogenetic countering of glial 208. Inoue, M. et al. Rational engineering of XCaMPs, a
207–220 (1999). acidosis suppresses glial glutamate release and multicolor GECI suite for in vivo imaging of complex
159. Holtzclaw, L. A., Pandhit, S., Bare, D. J., Mignery, G. A. ischemic brain damage. Neuron 81, 314–320 (2014). brain circuit dynamics. Cell 177, 1346–1360 (2019).
& Russell, J. T. Astrocytes in adult rat brain express 183. Rungta, R. L., Osmanski, B. F., Boido, D., Tanter, M. & 209. Odaka, H., Arai, S., Inoue, T. & Kitaguchi, T.
type 2 inositol 1,4,5-trisphosphate receptors. Glia 39, Charpak, S. Light controls cerebral blood flow in naive Genetically-​encoded yellow fluorescent cAMP indicator
69–84 (2002). animals. Nat. Commun. 8, 14191 (2017). with an expanded dynamic range for dual-​color
160. Hertle, D. N. & Yeckel, M. F. Distribution of inositol- 184. Otchy, T. M. et al. Acute off-​target effects of neural imaging. PLoS One 9, e100252 (2014).
1,4,5-trisphosphate receptor isotypes and ryanodine circuit manipulations. Nature 528, 358–363 (2015). 210. Harada, K. et al. Red fluorescent protein-​based cAMP
receptor isotypes during maturation of the rat 185. Sudhof, T. C. Reproducibility: experimental mismatch indicator applicable to optogenetics and in vivo
hippocampus. Neuroscience 150, 625–638 (2007). in neural circuits. Nature 528, 338–339 (2015). imaging. Sci. Rep. 7, 7351 (2017).

Nature Reviews | Neuroscience


Reviews

211. Shen, Y. et al. Genetically encoded fluorescent 219. Feng, J. et al. A genetically encoded fluorescent Author contributions
indicators for imaging intracellular potassium ion sensor for rapid and specific in vivo detection of All authors contributed equally to all aspects of the
concentration. Commun. Biol. 2, 18 (2019). norepinephrine. Neuron 102, 745–761 (2019). manuscript.
212. Marvin, J. S. et al. An optimized fluorescent probe for 220. Lin, J. Y., Lin, M. Z., Steinbach, P. & Tsien, R. Y.
visualizing glutamate neurotransmission. Nat. Methods Characterization of engineered channelrhodopsin Competing interests
10, 162–170 (2013). variants with improved properties and kinetics. The authors declare no competing interests.
213. Marvin, J. S. et al. Stability, affinity, and chromatic Biophys. J. 96, 1803–1814 (2009).
variants of the glutamate sensor iGluSnFR. Nat. Publisher’s note
Methods 15, 936–939 (2018). Acknowledgements Springer Nature remains neutral with regard to jurisdictional
214. Lobas, M. A. et al. A genetically encoded single-​ The authors are supported by the US National Institutes of claims in published maps and institutional affiliations.
wavelength sensor for imaging cytosolic and cell Health (NS111583, DA047444, NS060677 and MH104069),
surface ATP. Nat. Commun. 10, 711 (2019). a Paul G. Allen Distinguished Investigator Award and the CHDI
215. Marvin, J. S. et al. A genetically encoded fluorescent Foundation (to B.S.K.). B.S.K. was also partly supported by the Related links
sensor for in vivo imaging of GABA. Nat. Methods 16, Ressler Family Foundation. X.Y. was supported partly by an Adult Astrocyte RNA-​seq explorer: http://astrocyternaseq.org/
763–770 (2019). American Heart Association Postdoctoral Fellowship Allen Brain Atlas, RNA-​seq data: https://celltypes.brain-​map.
216. Patriarchi, T. et al. Ultrafast neuronal imaging of (16POST27260256). J.N. was partly supported by a Japan org/rnaseq
dopamine dynamics with designed genetically Society for the Promotion of Science (JSPS) Overseas Research Astrocyte Ageing Transcription: http://igc1.salk.edu:3838/
encoded sensors. Science 360, eaat4422 (2018). Fellowship (H28-729) and the Uehara Memorial Foundation astrocyte_aging_transcriptome/
217. Sun, F. et al. A genetically encoded fluorescent sensor Overseas Postdoctoral Research Fellowship (201730082). The Brain RNA-​seq: http://www.brainrnaseq.org/
enables rapid and specific detection of dopamine in authors regret that many papers could not be cited (especially Dropviz: http://dropviz.org/
flies, fish, and mice. Cell 174, 481–496 (2018). early studies), because of space limits and the requirement to Mouse Brain Atlas: http://mousebrain.org
218. Jing, M. et al. A genetically encoded fluorescent focus primarily on the past 5 years. The authors thank mem- sCope: http://scope.aertslab.org
acetylcholine indicator for in vitro and in vivo studies. bers of the Khakh laboratory for useful discussions, and the
Nat. Biotechnol. 36, 726–737 (2018). anonymous reviewers for their comments. © Springer Nature Limited 2020

www.nature.com/nrn

You might also like