Radial stem astrocytes (aka neural stem cells): Identity, development, physio‐pathology, and therapeutic potential

Adult neurogenesis is a striking example of neuroplasticity, which enables adaptive network remodelling in response to all forms of environmental stimulation in physiological and pathological contexts. Dysregulation or cessation of adult neurogenesis contributes to neuropathology negatively affecting brain functions and hampering regeneration of the nervous tissue while targeting adult neurogenesis may provide the basis for potential therapeutic interventions. Neural stem cells in the adult mammalian brain are at the core and the entry point of adult neurogenesis. By their origin and properties, these cells belong to astroglia, and are represented by stem radial astrocytes (RSA) which exhibit multipotent “stemness”. In the neurogenic niches, RSA interact with other cellular components, including protoplasmic astrocytes, which in turn regulate their neurogenic activity. In pathology, RSA become reactive, which affects their neurogenic capabilities, whereas reactive parenchymal astrocytes up‐regulate stem cell hallmarks and are able to generate progeny that remain within astrocyte lineage. What makes RSA special is their multipotency, represented by self‐renewing capacity capability to generate other cellular types as progeny. A broad understanding of the cellular features of RSA and parenchymal astrocytes provides an insight into the machinery that promotes/suppresses adult neurogenesis, clarifying principles of network remodelling. In this review, we discuss the cellular hallmarks, research tools, and models of RSA and astrocytes of the subventricular zone along the lateral ventricle and dentate gyrus of the hippocampus. We also discuss RSA in ageing, which has a great impact on the proliferative capacity of RSA, as well as the potential of RSA and astrocytes in therapeutic strategies aimed at cell replacement and regeneration.


| INTRODUCTION
More than a century ago Santiago Ramón y Cajal concluded that "Once the development was ended, the founts of growth and regeneration of the axons and dendrites dried up irrevocably. In adult centres the nerve paths are something fixed, ended, immutable… it is for the science of the future to change, if possible, this harsh decree"; since that moment everything in the brain was considered fixed and incapable of regeneration after the end of embryonic development. 1 Nevertheless, Cajal implicitly highlighted that the possibility of generating new cells in the adult brain remains an open question for future scientists to challenge his" harsh decree". In 1965, Joseph Altman and Gopal Das called Cajal's statement into a question by providing the first anatomical evidence and revealing adult neurogenesis in the adult rat hippocampus. 2 Since then a growing body of studies have uncovered adult neurogenesis in the majority of mammals, including human, and other animals such as birds, fishes, and reptiles. 3,4 Numerous research based on rodent models (especially mouse) documented and characterised adult neurogenesis in the ensuing decades, [5][6][7] however, low neurogenic rate, cross-species differences, and methodological limitations lead to caveats in detecting and studying neurogenesis in adult human brains and many unanswered questions remain. 3,5,8

| A NOTE OF THE NOMENCLATURE
The adult mammalian brain contains several subpopulations of cells that can both self-renew and produce progeny throughout life. These cells were initially defined as adult neural stem cells (aNSC). Lineage categories of these cells were proposed; for example, aNSC were defined as B/type I cells, intermediate progenitors as C/type II cells, and neuroblasts as A cells. A closer look at neural stem cells, however, revealed their astroglial nature and radial morphology, [9][10][11][12] therefore stem cells in the adult brain can be classified as radial stem astrocytes (RSA). It has to be noted that astroglia defines a highly heterogeneous class of neural cells, which includes parenchymal astrocytes (protoplasmic, fibrous, velate, marginal, etc.) radial astrocytes (Müller glia of the retina, cerebellar Bergmann glia, tanycytes, etc.) and cells lining the wall of ventricles (ependymocytes, choroids plexus cells). 13 Recently, astroglial nature of adult brain stem cells has been questioned, 14 without, in our view, strong rationale. Structurally, RSA present radial morphology, being thus distinct from protoplasmic astrocytes. Nonetheless, similar radial morphology is characteristic for many other astroglial cell types-tanycytes and Bergmann glia being the examples. RSA and other astroglial cells, including parenchymal astrocytes share numerous molecular and physiological features (discussed in detail in the following sections). Functionally, RSA perform all major functions of astroglia, supporting extracellular glutamate homeostasis, 15,16 covering synapses with terminal processes, [17][18][19] and establishing perivascular endfeet perimeter. 17,20 Radial stem astrocytes are concentrated in several brain regions known as "neurogenic niches", of which the subventricular zone (SVZ) along the lateral ventricle and the subgranular zone (SGZ) within the dentate gyrus (DG) in the hippocampus received major attention ( Figure 1A). A subset of tanycytes (another type of radial astroglia), lining the wall of the 3rd ventricle 21 also possess stem properties [22][23][24] ; the cellular nature of additional minor neurogenic niches, found in the hypothalamus, striatum, and prefrontal cortex, needs more detailed analysis. 25 In this review, we shall focus on SVZ and SGZ niches. In both neurogenic niches, RSA give rise to transit-amplifying progenitors that generate neuroblasts ( Figure 1B,C). In the SVZ, neuroblasts then migrate along the migratory olfactory stream to the olfactory bulbs where they differentiate into granule (inhibitory) neurones. In the SGZ, differentiated neuroblasts become immature neurones, which gradually mature to granule cell and migrate into granule cell layer ( Figure 1C).
In the ensuing chapters, we shall compare and summarise the similarities and differences between stem and protoplasmic astrocytes. We shall provide a brief introduction to research tools and models for studying these cells and discuss possible targeting of these cells in regenerative medicine.

ORIGINS OF PARENCHYMAL AND RADIAL STEM ASTROCYTES
Astroglia, similarly to all other neural cells, originate from radial glia. [26][27][28][29] These cells with radial appearance were first visualised by Camillo Golgi, who defined them as epithelial cells, 30 and subsequently, they were named radial glia by Giuseppe Magini. 31 Ninety years later Pasco Rakic defined these cells as radial glial cells (RGC) and characterised in detail their role in the development of the nervous system 32 ; for the history of radial glia see also. 33 Radial glial cells are the pluripotent neural stem cells of the embryonic CNS. RGC produce various astroglial types either through asymmetric division, giving rise to astrocytic precursors, or through direct transformation ( Figure 2). The bulk of parenchymal (protoplasmic, fibrous, velate, marginal, etc.) astrocytes is, however, produced postnatally through symmetric division of relatively mature astrocytes. Direct transformation of RGC also produces the radial stem astrocytes. Similarly, RGC directly transform into ependymocytes and tanycytes, which both further propagate through symmetric division ( Figure 2).
Most of the time RSA in the adult brain are maintained in quiescence, featured by reversible cell cycle arrest and low metabolic activity. Exit from quiescence (known as "activation") of RSA is triggered by a variety of stimuli, and RSA proliferate by either symmetric or asymmetric division to renew RSA pool or generate differentiated progeny. In the SVZ, RSA arise from RGC in the developing forebrain. 34 Studies based on mouse models revealed tmajority of SVZ RSA emerge between E13.5 and E15.5 and remain quiescent until they are activated postnatally. 35,36 Instead of random selection, a subpopulation of embryonic RGC slowly dividing at E13.5-E15.5, is set aside as reserved pool of future RSA while other RGC divide rapidly to produce precursors for various neural cells that constitute the brain. 36 These slowly dividing RGC exhibit a high level of Notch signalling and constant Hey1 expression that help them sustain to postnatal period. 37 This likely reflects the fact that quiescence is essential for the life-long maintenance of RSA. Some RSA in the SVZ inherit the spatial code, that is, regional specification from their embryonic predecessors. For instance, progenitors producing pyramidal neurones in the developing cortex also generate RSA that give rise to superficial granule cells in the olfactory bulb postnatally. Unlike the RSA in the SVZ, the RSA in the SGZ seem to originate from highly proliferative RGC. Transient ablation of these dividing RGC not only depletes pool of stem cells but also compromises their neurogenic capacity. 38 In the mouse SGZ, RSA derive from a subset of sonic Hedgehog (Shh)-responsive RGC that appear in the ventral hippocampus at E17.5. 39,40 These Shh-responsive cells gradually migrate from ventral to dorsal DG along the longitudinal axis. 39 The first postnatal week is a critical time window for these cells to expand and constitute the RSA pool. Subsequently, RSA in the SGZ become quiescent and maintain this status until they are activated during the adulthood. 38,40 Parenchymal astrocytes similarly originate from RGC. 26,27 The first gliogenic wave takes place around E16-E18 immediately following the wave of cortical neuronogenesis. 26,27 The neuronogenic to gliogenic switch is controlled by transcription factors nuclear factor I-A (NFIA) and SOX9 (SRY-Box Transcription Factor 9). 41-44 Regulators F I G U R E 1 Adult neurogenic niche. Sagittal and coronal section view highlighting the main two neurogenic regions of adult rodents: SVZ along the lateral ventricle (LV) and DG of the hippocampus (A). Schematic diagram illustrating cellular components and lineage in the SVZ (B) and DG (C). See text for details. of gliogenic switch also include Notch and JAK/STAT and several auxiliary molecules associated with respective signalling cascades. 42,43,45,46 Activation of these pathways triggers methylation and expression of astrocyte-specific genes, such as GFAP (glial fibrillary acidic protein) and S100β. 42,47 Committed glial precursors originating from the asymmetric division of RGC give rise to immature astrocytes that migrate radially. 48,49 Embryonic astrogliogenesis contributes only a small portion of the entire astrocytic population in the CNS. The majority of astrocytes are generated through symmetric division of differentiated astrocytes during the first three postnatal weeks; while only some astrocytes derive from postnatal direct transformation of RGC. 28,29,50,51 Astrocytic heterogeneity is partially attributed to the brain regions/subareas where they are generated. Clonally related astrocytes disperse radially and remain within the functionally defined layers/columns. 52 Diverse subtypes of astrocytes could be directly inherited from the RGC or instructed by local neurochemical environment during the development. 53-56

ANATOMICAL STRUCTURES OF PARENCHYMAL ASTROCYTES AND RSA
Astrocytes (the term introduced by Michael von Lenhossék 57 and much popularised by Ramón y Cajal 58 ), are represented by several categories, of which fibrous astrocytes of the white matter and protoplasmic astrocytes of the gray matter (named so by Rudolf Albert von Kolliker and Willem Lloyd Andriezen 59,60 ) are the most numerous. Fibrous astrocytes are organised along white matter tracts and their morphological appearance is manifested by straight and long processes with few branching points. 13 Processes of fibrous astrocytes contact vasculature with perivascular endfeet and axonal nodes of Ranvier with perinodal processes, thus contributing to the homeostatic maintenance of the local environment. 13 Protoplasmic astrocytes, which ubiquitously appear in the gray matter (including the hippocampus), exhibit a complex morphology with extremely ramified branches and terminal leaflets contacting numerous synapses within their domain. 61,62 In rodents, fine processes of hippocampal astrocytes fill the local environment in almost non-overlapping domain (less than 5% overlap), encompassing synapses and blood vessels. 61 Perisynaptic astrocytic leaflets form "synaptic cradle" around synaptic structures; astrocytes control synaptogenesis, synaptic maturation, homeostatic support critical for synaptic maintenance, and synaptic isolation. 62-64 Within its territorial domain a single astrocyte contacts 20 000-120 000 synapses in rodents and up to 2 000 000 synapses in the brain of humans. 13,65 This idiosyncratic cellular architecture enables astrocytes to contact and regulate local circuitry/blood vessel activities and coordinate synaptic activity, homeostatic tissue regulation, and blood flow. 66 F I G U R E 2 Developmental origins of astroglia. See text for further explanations.
Like their RGC ancestors, RSA retain a radial morphology characterised by prominent apical-basal polarity. Somata of RSA reside close to the ependymal wall in the SVZ; the apical processes extend toward ependymal cells and also contact ventricle lumen by a single (non-motile) process interdigitated between ependymocytes ( Figure 1B). 67 The long basal processes arising from the basal pole of the soma extend tangentially ( Figure 1C). 20 When the ventricle surface is viewed en face, the multiciliated/bilicliated ependymal cells are organised into a pinwheel formation, with the apical ending of RSA located in the center of this pinwheel. 20 In the adult hippocampus, cell bodies of RSA align the SGZ, with a long primary process extending through the granule cell layer and ending with elaborate branching in the inner molecular layer. The morphology of radial stem astrocytes is thus distinct from protoplasmic astrocytes also present in the neurogenic niche. Nevertheless, RSA share some common anatomic structures with mature astrocytes. Ultrastructural evidence revealed that RSA (similarly to protoplasmic astrocytes) form endfeet plastering the surface of neighboring blood vessels. 17,20 Moreover, blood vessel surface may serve as an adhesion point where RSA establish direct cell-to-cell contacts with niche protoplasmic astrocytes. 17 Again, and similarly to protoplasmic astrocytes, leaflets of radial stem astrocytes enwrap synapses thus forming synaptic cradle. 63,64 While in the DG leaflets of RSA ensheath asymmetric synapses in the inner molecular layer, in the SVZ RSA enwrap axonal serotonergic varicosities with apical microvilli. [17][18][19]

AND ASTROGLIA
There are various means to study cell-specific morphology including histological methods (such as immunohistochemistry or other staining techniques) and genetic strategies based around transgenic animals expressing makers under cell-specific promoters, inducible transgenic mouse lines (usually Cre-driver lines), and viral vector-based cell-specific transfection. In this section, we shall focus on the immunostaining of various types of astrocytes and on mouse models for specific targeting RSA and parenchymal astrocytes.

| Immunohistochemistry
Mature astrocytes and RSA share several common antigens for immunostaining, while there are also antibodies against molecules expressed specifically in only one of the two cell types ( Figure 3; Table 1). In general, GFAP and EAAT-1/2 (excitatory amino acid transporter) glutamate transporters (also known as GLAST and GLT-1, glutamateaspartate transporter, and glutamate transporter-1 in experiments in rodents) are usually used as markers for both RSA and parenchymal astrocytes. 10,11,68,69,70,71,72,73,74 Staining against EAAT-1/2 is punctate and confined to the plasmalemma, hence it does not reveal the full extent of the cell morphology. 69,72,73,74 Similarly, GFAP is far from being an ideal immunohistochemical marker. First, majority of parenchymal astrocytes in the healthy brain do not express GFAP at the level of antigen detection. 73,75,76 Second, GFAP immunoreactivity reveals only the primary branches of astrocytes while missing terminal leaflets or small endfeet altogether. 73,75,77,78,79,80,81 Vimentin is another member of F I G U R E 3 Labelling RSA and protoplasmic astrocytes by immunohistochemistry. Confocal micrograph showing staining with three antigens for RSA and/or astrocytes (A1): GFAP (green; A2), S100β (red; A3), and SOX2 (white; A4) in the adult DG (scale bars 20 μm). The schematic diagram shows immunostaining markers for astrocytes, RSA, and progenitors. Most of these markers do not specifically label astrocyte/RSA/progenitor (B). See text and Table 1 for a detailed explanation.
intermediate filament family expressed in astrocytes. Both nestin and vimentin are highly expressed in immature astrocytes and RSA; their expression in astrocytes is downregulated during maturation or differentiation. 82,83 Vimentin antibodies label a subset of astrocytes in some brain regions as well as proliferating RSA, which possibly choose gliogenic fate. 11,84,85 Immunostaining for nestin labels virtually all RSA 18,[86][87][88][89] ; in addition, it also stains pericytes and endothelial cells. 90,91 Transcription factor SOX9 is a nucleus marker expressed by both mature astrocytes and RSA in niche regions, while it is exclusively expressed by some astrocytes outside neurogenic niches. 92 Staining with brain lipid binding protein (BLBP) or SOX9 antibodies label the majority of RSA. [92][93][94][95] Immunohistochemical staining for calcium-binding protein S100β and glutamine synthetase (GS) labels most (if not all) protoplasmic astrocytes in the neurogenic niche; antibodies against GS also label a subpopulation of RSA. 76,[96][97][98] Glutamate transporter EAAT-2 is widely used as astrocytic marker, yet its immunoreactivity is also detected in the neurogenic region where RSA reside. [72][73][74] Although aquaporin 4 (AQP4, the water channel protein) is detected in some nestin-positive RSA in the SVZ, it is best known as a marker for mature astrocytes and highly expressed in astrocytic endfeet. [99][100][101] By and large, there is no specific marker to distinguish between protoplasmic astrocytes and RSA. It, therefore, requires at least two antigens or combination of immunostaining with transgenic mouse lines (or viral vectors) to label and segregate between protoplasmic astrocytes and RSA.

| Constitutive and inducible transgenic mouse line
A long and continually expanding list of transgenic mouse models enables selective targeting astrocytes or RSA for morphological analysis and visualisation of these cells in vivo or in acute brain slice for functional studies. We do not aim at covering every mouse line in existence; instead, we introduce several lines that are extensively used in studies of the adult neurogenesis (Table 2). Constitutive and inducible transgenic mouse lines allow the expression of a marker protein (usually a fluorescent protein such as green fluorescent protein, GFP, or tdTomato) 102 constantly or at a defined time point. The advantages of using such lines are (i) they exhibit global labelling of the targeting cells even in the brain regions where immunoreactivity of certain antigen, such as GFAP, is low or none; (ii) fluorescent protein is expressed over the entire cytoplasm, enabling visualisation and investigation of all subcellular compartments (for example astrocytic leaflets). We also T A B L E 1 Antigens for immunohistochemistry to visualise RSA and parenchymal astrocytes. evaluate and consider the efficiency (the population of targeting cells) and specificity (the proportion targeting cell to other cell types) when using these transgenic mouse models for studying astrocytes or RSA. Available mouse lines targeting RSA include, but are not limited to, GFAP-GFP, Glast-CreER/T2, Nestin-GFP, Nestin-CreER/T2, Sox1-GFP, Sox2-GFP, Sox2-CreER/T2; Gli1-CreER/T2, Acsl1-creER/T2, Hes5-CreER, and HopX-CreER (see Table 2 for full names and more strains; details were also reviewed in Refs. [104][105][106]). Of note, GFAP-GFP and Glast-CreER/T2 mice are frequently used in studies of other astroglial cells as well. 107,108 Additionally, tamoxifeninduced reporter protein expression in adult Gli1-CreER/ T2 mouse is also found in many non-proliferating astrocytes that are distributed broadly in the forebrain despite the scatter appearance in the hippocampus 109 ; this line is often used for studying cortical astrocytes. 110 Meanwhile, other Cre transgenic mouse lines, of which the recombination is induced in RSA, also label astrocytes to different extents. On the contrary, S100β-EGFP, Glt1-GFP, Aldh1l1-GFP, and Cx30-CreER/T2 are transgenic mouse lines for specific targeting of astrocytes. [111][112][113][114][115][116] Nevertheless, these lines either have astrocyte-specific recombination in limited brain regions or the recombination also labels other cell types such as oligodendrocytes, RSA, and neurones (reviewed in Refs. [117,118]). Recently, an inducible transgenic line, Aldh1l1-CreER/T2, was generated to efficiently and selectively target astrocytes in the CNS and several peripheral organs. 113,119 A recent study confirmed this mouse line can serve to specifically target protoplasmic astrocytes in the SGZ despite recombination was also observed in ependymal cells and RSA in the SVZ. 120

| Animal models and human neurogenesis
Mouse models dominate research into RSA, including their detection, cell identity, and switch from quiescence to proliferation. Rodent-centric views and associated bias are frequently questioned. 3 Adult neurogensis and RSA were characterised in fishes, reptiles, birds, and the majority of wild mammals. [120][121][122][123][124][125] Thus adult neurogenesis seems to be evolutionarily conserved, driven by the needs of environmental adaptations. The details of adult neurogenesis in humans (especially in the hippocampus), however, remain to be elucidated. Studies employing immunohistochemistry to identify neural progenitors or immature neurones are contradictory: some studies reported rare presence of neurogenesis after adolescence and the immature neurones are not newly generated whereas other reports indicated persistent neurogenesis in humans in adulthood with a decline in aging. [126][127][128][129][130][131] Recently, RNA-sequencing was adopted for investigating human adult neurogenesis in spite of the conflicting results. [132][133][134] The disagreement between those studies might, to some extent, might be due to sample stratification and methodology. 135 Nevertheless, the differences between humans and mice are substantial and should not be overlooked. The average life expectancy of a human is over 70 years, whereas laboratory mice live 26-30 months on average; it is, therefore, critical to align the specific timing of the neurogenesis trajectory across species. A shift in the timing of peak neurogenesis from mouse to human is well documented. 3,4,136 Further studies for interspecies comparison to identify equivalent developmental time points are essential to translate the rodent studies to humans. 3 Studies of adult RSA and neurogenesis in the hippocampus of mice were primarily focused on relatively young ages, whereas human research uses samples with a much wider age range; rodent and human studies may potentially focus on different stages over the lifespan, which may, at least partially, account for the conflicting reports describing adult hippocampal neurogenesis in humans. 3,136 Only 15.5% of transcriptomics of immature neurones in the adult human hippocampus overlapped with orthologous genes enriched in mouse hippocampal immature neurones. 134 Expression of doublecortin may drop to a very low level in adult human hippocampus so that it could be difficult to identify adult hippocampal neurogenesis. 4 It is very likely RSA in human brains also exhibit large differences in markers and epigenomics when compared with rodents. More advanced studies using transcriptomic and molecular analysis will advance our knowledge of the inter-species differences between human RSA and mouse RSA for further identification of the modulating mechanism of RSA activation/maintenance in humans for regenerative applications.

OF PROTOPLASMIC AND RADIAL STEM ASTROCYTES
Ionic composition of the cytoplasm and associated ionic gradients, as well as selective membrane permeability dominated by an expression of multiple K + channels, defines the membrane potential of RSA and niche protoplasmic astrocytes. The RSA in the forebrain neurogenic niches of mouse are characterised by hyperpolarised resting membrane potential (V m , −75 to −85 mV) and low input resistance (R in, ~30 MΩ in SVZ; ~70 MΩ in SGZ), which is indistinguishable from electrophysiological properties of mature protoplasmic astrocytes. 88,137,138 Highly negative V m of RSA reflects large resting K + conductance. Immunostaining and electrophysiological recordings demonstrated the expression of functional K + channels on RSA, including inwardly rectifying K + (K ir ) channels and voltage-gated K + channels. 138,139 High resting K + conductance of RSA and protoplasmic astrocytes is mainly mediated by Ba 2+ -sensitive inward rectifying K ir 4.1 channels. [139][140][141] Decreased K ir channel activity is associated with cell cycle progression (from G1/S checkpoint to S phase), whereas inhibition of K ir channel by Ba 2+ induced membrane and subsequent cell proliferation of both astrocytes and RSA. 139,142 Depolarisation of astrocytes and RSA induced by other stimuli such as neurotransmitters or mechanical displacements also associates with cell proliferation. 18,143,144 In addition to various K + channels that sustain passive membrane properties of astrocytes, connexion-based gap junctional coupling provides for syncytial isopotentiality 146 and second messenger signalling due to connexin permeability to various small molecules with molecular weight ≤1 KDa. Similarly to astrocytes, RSA are coupled through gap junctions. 147 This kind of connectivity allows intercellular propagation of Ca 2+ signals, which occurs within astrocytic or RSA networks either spontaneously or in response to various stimuli. 89,148,149,150 Since astroglia are non-excitable cells, Ca 2+ transients, and oscillations are used as an indicator of astrocytic and RSA physiological activity. 18,89,151 Pharmacological inhibition of gap junction suppresses intercellular Ca 2+ waves in astrocytic and RSA syncytia. [152][153][154] Astroglial Ca 2+ signalling is associated with numerous cellular processes and gene activation cascades; in particular in RSA intracellular Ca 2+ signals regulate proliferation and differentiation. 149,155,156 Whether RSA possess Na + signalling, which plays a key role in coordinating homeostatic responses of protoplasmic astrocytes with neuronal activity 157,158 remains to be studied. Connexins, gap junction-forming proteins, are classified based on their molecular weight (e.g., Cx30 and Cx43) are widely expressed in neuroglia. 139,159,160,161 Both astrocytes and RSA predominantly express Cx43 with lesser expression of Cx30. 147,162,163 Deletion of Cx43 and/or Cx30 results in decreased RSA proliferation and compromises physiological functions of astrocytes, in particular affecting isopotentiality of the syncytium. 147,164,165,166 In contrast, pharmacological inhibition of gap junctions causes conversion of passive to rectifying current profile and increased input resistance.

| Astroglial reactivity
Fundamentally, astrocytes in pathological contexts are protective; astrocytic pathological metamorphoses aim at an increased maintenance and survival of the nervous tissue. [166][167][168] Failure in astrocytic defence exacerbates damage and facilitates propagation of neuronal death leading to neurological and cognitive deficits. Astroglial contribution to neuropathology is multifaceted and complex; different pathological phenotypes may coexist or develop in sequence depending on the disease severity or stage. Glial cells in pathological contexts acquire many heterogeneous phenotypes, which may increase or limit neuroprotection, or result in the emergence of aberrant forms of glial cells that may drive the disease progression. 169,170 Astroglial pathology can be broadly classified into: (i) astrocytic reactivity, represented by anisomorphic, proliferative irreversible reactive astrogliosis and isomorphic, allostatic reversible astrogliosis; (ii) astrocytopathies, characterised by the emergence of aberrant astrocytes driving neuropathology; (iii) astrocytic atrophy with loss of function; (iv) astrodegeneration and death. 13,169,170,171 Astrocytic reactivity ranges from stereotypic anisomoprhic reactive astrogliosis in response to traumatic brain injury (neurotrauma, neuroinfection, stroke, or autoimmune attack) to a highly heterogeneous isomorphic astrocytic reactivity mainly deduced from upregulation of GFAP expression, observed in various neuropathologies not associated with the breach of blood-brain barrier (BBB). 166,172,173,174,175 Astrocytic reactivity is an evolutionary conserved response to various types of lesions to the nervous system when astrocytes engage in changes in transcriptional regulation as well as biochemical, morphological, and physiological remodelling associated with immediate reaction to the lesion and functional adaptation to the post-injury environment. 174,175 It is almost universally accepted that cellular hallmark of reactive astrogliosis is an overexpression of GFAP and morphological hypertrophy. 173 Increase in GFAP expression, however, is not an ideal marker for reactive astrocytes since (i) GFAP expression in astrocytes, as mentioned previously, differs between brain regions; (ii) GFAP expression is modulated by various extrinsic and intrinsic factors and often indicates adaptive plasticity. 73,75,78,176,177 Furthermore, an increase in GFAP does not report full cellular morphology, and often increase in GFAPpositive profiles does not coincide with actual cellular hypertrophy. 178 In the brain trauma, stroke, immune, or infectious attacks reactive astrocytes become proliferative. 179,180 These proliferative astrocytes possess various genetic and molecular hallmarks of RSA, indicating the potential for turning astrocytes into stem cells/progenitors and even further into different types of neurones. 181 Indeed, forced expression of SOX2 induced conversion of astrocytes into neural progenitors. [182][183][184] Isomorphic reactivity of astrocytes is highly heterogeneous and context-specific with a great diversity in morphological and molecular states in different CNS regions and various diseases. [185][186][187][188] Astrocytic reactivity is represented by a broad spectrum of plastic and context-dependent changes, and further assessment of multiple parameters (epigenetics, proteomics, morphology, interactions with other cells) is required to determine the subtype and states of reactive astrocytes. 174,175 Astrocytic atrophy with loss of function results in indirect neuronal damage due to the reduced homeostatic support and neuroprotection; it may also lead to a malfunction of neuronal networks. 189 Such astroglial asthenia is particularly prominent in neuropsychiatric diseases. 190 Astrocytopathies refer to an emergence of aberrant forms of astrocytes, which contribute or lead to pathophysiology of various neurological diseases, though either direct effects, or reduced homeostatic support, such as, for example, observed in Alexander disease. 191 Astrodegeneration is the most extreme form of astrogliopathology; in conditions of severe insults, astrocytes exhibit degeneration and death featured by fragmentation and loss of processes with swelling and vacuolisation of the soma, known as clasmatodendrosis. Clasmatodendrosis was described by Alzheimer and Cajal and subsequent studies have reported the occurrence in ageing and various neurological disease. [192][193][194] A subpopulation of RSA also displays reactive remodelling under certain circumstances. 196,197 The reactive RSA are hypertrophic with thicker processes and increased morphological complexity. 196,197 In young animals, reactive hippocampal RSA either directly transform into reactive astrocytes or give rise to reactive astrocytes as daughter cells in response to strong neuronal hyperactivity (e.g., high-dose kainic acid-induced epileptic form of neuronal hyperactivity accompanied with seizures) at the expense of neurogenesis. 197 It is intriguing whether reactive astrocytes derived from reactive RSA undergo similar epigenetic modifications and perform similar functions as reactive parenchymal astrocytes. 197 In aged hippocampus, however, reactive RSA remain mostly quiescent even when faced with neuronal hyperactivity following injection of kainic acid. 196 Reactivity of RSA lead to their depletion and impaired neurogenesis, as reactive RSA lose their neurogenic ability. 196,197

| Heterogeneity of protoplasmic and radial stem astrocytes
Both protoplasmic astrocytes and RSA are heterogeneous comprising many distinct subpopulations. 13,198,199 Identification of cellular heterogeneity and deciphering the functional/behavioral differences of protoplasmic astrocytes and RSA are critical for understanding brain plasticity in the healthy and diseased brain. A great diversity of signals and factors influences the cellular responses of RSA and their progeny through specific signalling pathways or transcriptional modifications. Lineage-tracing and fate-mapping revealed a mosaic of different stem cells occupying distinct domains, and correlating with the regional expression of specific transcription factors in different areas along the dorso-ventral or rostro-caudal axis of the SVZ. These diverse cells give rise to distinct subtypes of interneurones in the olfactory bulbs during adulthood. [200][201][202][203][204] Similarly, studies focusing on the SGZ indicated that different RSA subsets exhibit distinct morphology and behaviour, thus being associated with regional functionality. 205,206 The RSA in the SGZ are categorised into two subsets-type α and type β cellsdistinguished by morphology and proliferative behaviour. 205 Anatomically the hippocampus, including all its subareas, is divided into different regions along the septotemporal axis, with a gradient distribution of neuronal types and afferent terminals. 206 So does the expression of molecules that are associated with the regulation of RSA quiescence and activation: there is a graded expression of secreted frizzled-related protein 3, which is secreted by mature neurones along the septotemporal axis of the DG. 207 Elimination of this molecular gradient results in increased RSA proliferation preferentially in the temporal division of the DG. 207 Ablation of RSA subsets by X-ray irradiation in different areas along the septotemporal axis demonstrated that neurogenesis in the septal region plays a critical role in contextual discrimination task while neurogenesis in the temporal DG is associated with anxiolytic/ antidepressant effects of fluoxetine. 208,209 Additional studies are required to identify specific markers for RSA subsets with distinct proliferative properties and responses to niche signalling. Nevertheless, transcriptomic studies have unveiled molecular signatures of quiescent and activated RSA as well as the molecular cascade underlying transitions between the two states. 210,211 Further linking transcriptional modifications and functional alterations may advance our knowledge of adult neurogenesis, and provide the basis for the potential application of RSA in regeneration medicine.
Parenchymal astroglia in mammals are similarly heterogeneous in their morphology, physiology, functions, epigenetic signatures, and metabolism amongst regions. 198,211 Golgi and Ramón y Cajal provided the first evidence of astrocytic heterogeneity by showing the different morphological features of various types of parenchymal astrocytes. 30,58,212 Protoplasmic astrocytes in particular, exhibit diverse molecular, morphological, and functional features in a region-specific manner. [213][214][215][216] Transcriptomic profiling of astrocytes showed various gene expression and selective modulation of synaptogenesis in cortical and subcortical regions along the dorsoventral axis. 216 Astrocytes display intra-regional differences in their morphological, molecular, and epigenetic features. 52,53,56,213 Combined morphological analysis, physiological assessment, transcriptomic profiling, and proteomic data, revealed significant differences in electrophysiological properties, Ca 2+ signalling, and proximity to synapse between striatal and hippocampal astrocytes, suggesting specialisation of astrocytes within different neural circuits. 214 Moreover, a recent study suggested that astrocytes within the DG exhibit layer-specific features in molecular expression, morphology, and coupling scale, implicating diverse functions of astrocytes in different microenvironment. 217 The interregional heterogeneity of astrocytes in different areas is, at least in part, attributed to the microenvironment where they were born; inherited from their radial glial ancestors or instructed by the local neurones. 54,218 Specialisation of astrocytes based on their position occurs together with spatial patterning of neurones, suggesting that maturation processes and distribution of neurones and astrocytes may be under the control of similar epigenetic mechanisms. 211 Astrocyte diversity not only enables different connection and support to surrounding neurones, but also potentially contributes to different astrocytic responses and reactivity in distinct pathological conditions. Hence, there is a significant therapeutic potential of targeting/manipulating astrocyte subsets in neuropathology.

| RSA in ageing
Ageing is a complex process, in which multiple factors and various cellular elements are involved. Ageing is accompanied with gradual decline adaptive capabilities of the organism accompanied with reduced physiological functions and cognitive capacity. Research based on rodent models indicated the ability of RSA to proliferate and generate new neurones is significantly diminished in the SVZ and SGZ regions during ageing. [219][220][221][222][223] This reduction might be attributed to decreased RSA pool, decreased neuronal differentiation, and increased RSA dormancy. [224][225][226] Both murine and human studies reported a decrease in RSA pool in neurogenic niches during ageing, 224,227,228,229 which arguably translates into the decline in neurogenesis. Moreover, mouse studies indicated that some subpopulations of RSA deplete faster and are more vulnerable to ageing whereas other subsets of RSA sustain in aged brains, although they become more reluctant to be activated with lengthening quiescence 195,230,231,232 ; RSA in aged rodent brain undergo several intrinsic changes such as alterations in epigenetic, proteomic, and metabolomic states alongside adaptions to extrinsic alterations such as local inflammation and signals released by other cells in the niche region. 233,234 These factors and processes may account for the compromised proliferative/self-renewal ability of RSA (reviewed in Ref. [233]). It is critical to investigate whether RSA in aged human brain exhibit heterogeneous populations showing different susceptibility to ageing characterised by specific alterations in genetic, molecular, and metabolic states as well as in cell cycle. Studies probing interventions that are beneficial for sustainable RSA pool maintenance and activation will advance the development of RSA-based therapeutic strategies for neurodegeneration.

| Astrocytes and RSA as targets for endogenous cell replacement
Several CNS pathologies that are associated with synaptic elimination and neuronal death and degeneration may trigger structural plasticity aimed at network reconstruction and compensation for malfunctional circuitry. The most prominent feature of such adaptive plasticity is the substitution of damaged neurones with new cells. Strategies for neuronal replacement were pursued through recruitment of endogenous adult neurogenesis, reprogramming of local glial cells (especially astrocytes) to differentiate into neurones, and transplantation of neurones derived from exogenous sources (e.g., induced pluripotent stem cells or fetal stem cell). [236][237][238][239][240][241][242][243][244] Numerous reports explored and discussed potential applications as well as the benefits and disadvantages of transplantation of exogenous engineered cells for therapeutic interventions. 238,245 Here we focus on the endogenous approaches, that is, recruitment of postnatal neurogenesis and reprogramming of glial cells, which take advantage of the proliferative properties of RSA and astrocytes. These two strategies evade the major difficulties associated with transplantation, such as appropriate timing for transplantation, immune responses, ethnical and legislative issues of the cell source, technical problems of clinical grade cell handling, complexities and limitations for non-focal neurological conditions. 238,246 The multipotent properties of adult RSA highlight the recruitment of endogenous RSA to repair the brain damage after acute insults or in chronic pathology as a target for the development of therapeutic applications. Accumulating evidence suggests the role for aberrant RSA activity and impaired neurogenesis in many brain diseases, [247][248][249] thus elevating RSA to a foremost position for cell replacement strategy. Many factors enhancing endogenous neurogenesis in the adult brain were identified, including manipulations with RSA proliferation, differentiation, survival of newly generated neurones, and functional network integration. It is nonetheless difficult to target RSA specifically and efficiently despite of the fact that several viral vectors for this purpose were developed. 105,250,251 Recently, the highly selective tropism of a recombinant adenoassociated virus serotype 4 (rAAV4)-based vector targeting RSA was discovered, making it a supreme tool for specific activation (or silencing) of RSA proliferation and manipulating RSA gene expression in the hippocampus. 252 Combining this technique with genomic and single-cell sequencing approaches targeting distinct subsets of RSA with heterogeneous gene expression in response to brain pathology permits to control the expression of candidate genes in RSA to restore and/or improve neurogenesis. 253 For example, single-cell RNA sequencing of murine RSA suggest groups of genes related to glycolytic metabolism and Notch signalling are critical for RSA quiescent state and pool maintenance, 210,211 and targeting these genes in RSA may serve as a potential strategy to fine-tune the balance between RSA activation and pool maintenance, thus mitigate the reduction of RSA in ageing and neurological diseases. The rAAV4 can be used in different contexts to facilitate RSA proliferation and maintenance in cell replacement therapies for brain diseases. Further studies focusing on the neurogenic steps after activation of RSA, including fate decision, survival, functional integration, and functional integration are required to test the sufficiency and efficiency of recruitment of RSA for cell replacement.
The balance between RSA quiescence and activation is critical not only for neurogenesis but also for the maintenance of the RSA pool that supports long-term regenerative capacity of the neurogenic niche. Hence, it is crucial to explore the effects of regulatory factors of RSA activity on neurogenesis in addition to direct RSA targeting. Quiescent RSA are highly sensitive to signalling from various cellular components in the niche region. RSA quiescence and activation states are regulated by local and systemic factors. 248,254,255,256 Emerging evidence indicated that RSA listen to the neuronal network very much like parenchymal astrocytes, and respond with changes in their quiescent/activated states. 18,90,257,258 A prominent instance is mossy cell (MC) regulation on RSA behavior. MC represent one of the major excitatory neuronal populations in the DG and are critical for the overall excitability of the DG. Commissural projections of MC structurally and functionally interact with RSA and dynamically control the balance of quiescence and activation of RSA in the adult DG. Depending on the activity state, MC regulate RSA behavior through the balance of glutamatergic and GABAergic signalling onto RSA: moderate activation of MC promotes RSA quiescence through indirect GABA input, whereas strong activation of MC increases RSA proliferation through direct glutamatergic pathway. Inhibition of MC and long-term disruption of MC circuitry leads to transient increases in RSA proliferation (Figure 4). 18 Other neurotransmitters also affect RSA quiescence, activation, and proliferation in the DG and/or SVZ (reviewed by Ref. [259]). For instance, serotonin induces RSA proliferation in both SVZ and SGZ of rodents 259,260 ; cholinergic input from medial septum promotes RSA proliferation in the DG of adult rats, whereas GABAergic inputs from the same region facilitate RSA quiescence. 90,261 Noradrenergic signalling increases RSA proliferation in the DG but not in the SVZ, [262][263][264] whereas effects of dopamine on RSA activation remain controversial due to its interactions with local circuitry and different subtypes of receptors expressed by the niche cells. 259,265 Meanwhile, RSA are also sensitive to environmental and behavioral stimuli, such as environmental enrichment, physical exercise, stress, and diet. 254,256,266,267 Of note, modulation of intrinsic and extrinsic signals on RSA activity can be affected by gender, age, and circadian rhythms. 256,267,268 Circadian clock in particular plays a key role in regulating RSA quiescence and pool maintenance by controlling the progression of cell cycle. 268,269 An in vivo Ca 2+ imaging characterised Ca 2+ signatures of RSA quiescence and activation, and demonstrated that circadian-related hormone melatonin promotes RSA proliferation through Ca 2+ signaling. 270,271 Details about the mechanism underlying regulation of RSA states by various systemic signals have been exquisitely discussed in other review articles. 248,254,255,256 Reprogramming parenchymal astrocytes into neurones is another prospective strategy for network repair and functional recovery in neuropathology. As mentioned above, cortical astrocytes display remarkable diversity, which is arguably inherited from their radial glia ancestors and/ or instructed by the local neurones, based on their laminar positions. 53-56 This makes cortical astrocytes a strong candidate for reprogramming into different subtypes of neurones essential for the reconstruction of the damaged circuitry. 272 Various vectors carrying transcription factors (such as NeuroD1, Ascl1, neurogenin 2, etc.) were used to successfully convert astrocytes into neurones. 240,273,274 Combining neurogenin 2 and nuclear receptor-related 1, increases the reprogramming efficiency, and cortical astrocytes are reprogrammed into pyramidal neurones with different morphological and molecular identities according to the lamination, and hence forming functional synaptic connections. 274 Using AAV vector to express NeuroD1 in reactive astrocytes in a stab-injury mouse model effectively converts reactive astrocytes to functional neurones and repopulate astrocytic pool by the proliferation of remaining local astrocytes. 275 This neuronal regeneration is accompanied with several beneficial effects, including rebalancing-neurone-astrocyte ratio, significant reduction in neuroinflammation, as well as restoration of BBB, suggesting that in vivo reprogramming of astrocytes might be the core of future therapeutic strategy for neuronal regeneration and neural tissue repair in stab-lesion or ischemic injury. 275 Many studies focused on promoting the proliferative activity of endogenous RSA and reprogramming of local glial cells to achieve high efficiency and specificity. Nevertheless, it should be kept in mind that the ultimate goal is to repair the network connections and function. It is, therefore. vital to pin down the long-term survival and functional integration in acute or chronic brain insults with more evidence. 274,276 Further studies are required for the assessment of behaviour relevance in various injury or disease models. Moreover, the development of less invasive and more diffusive materials, nanoscale particles, for instance, may be beneficial to cell replacement therapy and calls for more attention and investigations. 277

| CONCLUSION
Multipotent radial stem astrocytes (which belong to a wider class of astroglia 10,13,278 ) capable of generating neural precursors are concentrated in discrete regions in adult mammalian CNS and contribute to life-long brain plasticity. Protoplasmic astrocytes and RSA share many common properties in molecular, physiological, and functional aspects. 182 RSA contribute to both homeostatic neuroplasticity (through astrocyte-specific homeostatic pathways) and network plasticity (through generating new neural cells). Some reactive astrocytes are also capable of symmetric proliferation to generate more astrocytes. 180,181 Furthermore, in response to certain stimulation astrocytes can be converted into neurones for potential cell replacement in CNS diseases. Many studies focusing on cellular properties of RSA and astrocytes and the regulatory factors provide the basis for a better understanding of postnatal neurogenesis itself and how we can take advantage of it F I G U R E 4 MC modulation on RSA quiescence and activation. Schematic illustration depicting regulation of RSA by MC activities through a dynamic balance of direct and indirect pathways. MC commissural projections provide both direct glutamatergic and indirect GABAergic inputs onto RSA. When MC are activated at a moderate level, the indirect GABA pathway is dominant and promotes RSA quiescence. When MC undergo high activation, the direct glutamatergic pathway becomes dominant and facilitates RSA activation. Inhibition of MCs results in a profound reduction in GABAergic tone, thus causing a transient RSA activation. Similarly, loss of MC leads to a profound decrease in the indirect GABAergic pathway, followed by pool depletion. The figure is modified from Ref. [18].
for the brain repair. However, artificial boost of neurogenesis should be treated with caution. Over-activation may deplete RSA pool, and thus it is crucial to delicately tune the balance between neurogenesis and self-renewal. 197,279 Another risk is the emergence of malignant glioblastoma, a type of brain cancer. Recent analysis indicated that glioblastoma stem cells may be derived from RSA in the SVZ. 280,281 Further studies are required to uncover the mechanism of triggering mutagenesis of RSA, which may avoid the formation of glioblastoma and ensure the safety of RSA-targeting applications in therapeutic interventions. Both astrocytes and RSA exhibit phenotypical and functional heterogeneity. Further studies combining multiple parameters, such as gene profiles and morphometric analysis, are required to characterise distinct subgroups of astrocytes/RSA that show heterogeneous responses to different brain pathology. Selective targeting of a specific subpopulation of astrocytes/RSA might be a potentially efficient strategy for repairing brain plasticity in pathological conditions. It is fundamental to develop new tools for selective activation/manipulation of RSA/astrocytes or reprogramming of astrocytes, which will aid the advance in regenerative medicine.

CONFLICT OF INTEREST STATEMENT
Authors declare no conflict of interest.