The diverse functions of glial cells prompt the question to which extent specific subtypes may be devoted to a specific function. We discuss this by reviewing one of the most recently discovered roles of glial cells, their function as neural stem cells (NSCs) and progenitor cells. First we give an overview of glial stem and progenitor cells during development; these are the radial glial cells that act as NSCs and other glial progenitors, highlighting the distinction between the lineage of cells in vivo and their potential when exposed to a different environment, e.g., in vitro. We then proceed to the adult stage and discuss the glial cells that continue to act as NSCs across vertebrates and others that are more lineage-restricted, such as the adult NG2-glia, the most frequent progenitor type in the adult mammalian brain, that remain within the oligodendrocyte lineage. Upon certain injury conditions, a distinct subset of quiescent astrocytes reactivates proliferation and a larger potential, clearly demonstrating the concept of heterogeneity with distinct subtypes of, e.g., astrocytes or NG2-glia performing rather different roles after brain injury. These new insights not only highlight the importance of glial cells for brain repair but also their great potential in various aspects of regeneration.
One of the most basic classifications of cells in the nervous system is the distinction between neurons and glia. Glia were originally defined as nonneuronal cells with support functions, as the word nervenkitt (the glue between the neurons/nerves) implies (128, 283). Accordingly, glial cells were implicated in supporting neurons as well as maintaining stability and the overall architecture of the nervous system. Such a role is still prevalent and is reflected in the diverse support functions of glial cells that have been allocated to the different glial subtypes that were subsequently discovered over the last centuries. These are the star-shaped astrocytes, largely implicated in ion and metabolic brain homeostasis, the highly branched surveilling microglia with key functions in phagocytosis and defense, the myelinating glial cells, oligodendrocytes in the central nervous system (CNS), and Schwann cells in the peripheral nervous system (PNS), forming the insulating myelin sheaths and the most recent member of the club, the NG2-glia discovered only recently by the antibody NG2 (152) recognizing the chondroitin sulfate proteoglycan 4 (CSPG4). NG2-glia are widely dispersed through the adult brain and CNS parenchyma, and their functions are not yet known.
Also in regard to neural information processing, neurons were considered as the main players because of their allegedly unique capacity to generate action potentials and communicate via chemical synapses. Conversely, glial cells were considered to clean up thereafter, by removing excess ions or transmitters after neuronal activity in the vicinity of synapses, a task primarily allocated to astrocytes or nonmyelinating Schwann cells enwrapping synapses in the CNS and PNS, respectively. This view has changed by now as it has become clear that glial cells actively contribute to synaptic communication as direct synaptic partners. The so-called tripartite synapse consists of either two neuronal elements (the pre- and postsynaptic endings) or a presynaptic neuronal element and cells of the target organ in the periphery, such as muscle cells, and a glial part (the astrocyte or terminal Schwann cell processes in direct contact with these elements that often ensheath the entire synapse). Moreover, also glial cells that form and receive synapses were discovered, such as the NG2-glia (29, 37, 243), the newly discovered glial cell type that is also involved in forming the perineuronal nets (PNNs), dense arrangements of extracellular matrix considered to firmly stabilize synapses (74, 286). In the meantime also other glial cells have been observed as synaptic partners with microglial processes found in close vicinity of synapses which have been suggested to play key roles in synapse remodeling (127, 244). Indeed, the multiple roles of several glial cell types in synaptic communication are well consistent with their particular increase in number in brains with higher complexity, eventually outnumbering neurons by severalfold (for recent review, see Refs. 98, 129). A close and life-sustaining cross-talk between glia and neurons also occurs at the neuronal axon, where myelinating glia exert trophic support to neurons (189). The axon is indeed a further site of neural computation, rather than a sheer propagating cable. In this context, it is of particular interest that specific glial cells, the NG2-glia, can also be a direct postsynaptic site for neuronal processes, e.g., axons, as is the case in various brain regions including the corpus callosum (243). Strikingly, however, the very same type of glial cells also acts as oligodendrocyte progenitor cells (OPCs) in the adult brain proliferating and generating oligodendrocytes particularly frequently in the corpus callosum (61, 282). This prompts a short comment on the confusing use of the term oligodendrocyte precursor/progenitor cells. The term progenitor refers to a proliferating cell, while precursor refers to an immature state, which is why we use the term oligodendrocyte progenitor cell as we are referring to proliferating cells. This is an important issue in regard to the key question of how such roles in synaptic communication on one side and as dividing progenitor cells on the other side can be combined by the same cells. Alternatively, there may be subtypes of NG2-glia, one acting as dividing progenitors and the other as synaptic partners.
Indeed, this is a central question which has contributed to ignore the role of glial cells as stem and progenitor cells. In part, the difficulty arose from the view that dividing cells retract all their processes during M-phase, in which case all the synaptic and/or other connections formed by glial processes with neurons would be retracted and would have to be newly established. One way out of this dilemma would be to postulate a division of labor with some subsets of glia performing stem and progenitor cell functions and the other specialising on other tasks; alternatively, however, all functions may be performed at the same time if proliferating cells maintain their processes including those involved in synaptic contacts (79, 143). A further intriguing possibility may be that glial cell division with process retraction could be a means to remodel the synaptic connections of glia and hence contribute to brain plasticity. Similar questions also arise for astrocytes that perform stem cell functions in specific brain regions and after injury (236), besides their classical roles executed by processes involved in synaptic communication and endfeet surrounding the blood vessels mediating the neurovascular coupling. How can such a cell divide or are astrocytes with stem cell functions a special kind? In the most extreme case, glial cells with stem or progenitor cell function may have just been mistaken for glial cells due to the expression of “a few markers” and there would be a deep divide between a mature glial cell and any type of stem or progenitor cell.
These will be the questions we shall discuss in this review. To set the stage for these central issues of glial functions and heterogeneity, we will proceed in a sequence following normal development and first review the progenitor and stem cell roles of glial cells in the developing brain, the stage at which the concept that glial cells also act as stem and progenitor cells was originally discovered (166, 185, 198). We shall review the evidence for radial glial cells, the ubiquitous glial cell type in the developing vertebrate CNS, acting as neural stem and progenitor cells in a rather wide-spread manner (87, 139) and then turn to the second type of glia to appear in the developing CNS, the OPCs. Thereafter, we shall describe the emergence of progenitors for the other macroglial cell types (for reviews on microglia, see Ref. 127). Afterwards we will turn to the adult stage and discuss glial cells that still act as stem and progenitor cells, which are largely radial glial cells or NG2-glia occurring at different frequency in different vertebrate groups. Finally, we shall turn to the injury condition and review glial cells acting as stem and progenitor cells in various injury paradigms. Notably, some glial populations that are postmitotic or quiescent in the healthy brain reactivate proliferation and acquire a broader potential after specific types of injury. This leads to the important discrimination of two terms that are often used in a confusing manner: lineage and potential. It is fundamental to discriminate between the progeny that these glial cells generate in vivo, their normal lineage, from their potential, i.e., the progeny the very same glial cells can generate when exposed to a different environment in vivo or in vitro.
The reactivation of stem and progenitor properties in some apparently differentiated glial cells will bring us back to the issue of heterogeneity, asking to which extent these glial cells with a broader potential may be a special subtype and which roles this tentative subtype and its proliferation may play in the context of brain injury. This will provide new insights into a profound and surprising heterogeneity within the same population of glial cells reacting to brain injury not only opening new approaches to improve the outcome of the glial reaction and minimize scar formation after injury, but also identifying novel avenues towards employing stem and progenitor cells right within the injury site for neuronal replacement. This will close the circle to the roles that glial stem and progenitor cells play during development, namely, generating neurons.
II. DEVELOPMENT: GLIAL STEM AND PROGENITOR CELLS
A. Radial Glial Cells During Development
Radial glial cells are the first glia to appear, in ontogeny and phylogeny. Radial glial cells have been described early in the deuterostome lineage, such as in echinodermata and the lancelet (for recent review, see chapter 5 in Ref. 129). In ontogeny, radial glial cells appear in all regions of the brain at the onset of neurogenesis when the first differentiated cell types are generated. In some regions, however, such as the rat spinal cord, radial glial cells appear only at later stages, at the onset of gliogenesis (175). In vertebrates, radial glial cells develop from the earlier neuroepithelial cells, the epithelial cells forming the neural plate and neural tube. Both these cell types are epithelial in nature as they possess a clear apico-basal polarity with an apical membrane domain that contains, e.g., the glycoprotein Prominin1 and is delineated by junctional complexes containing tight and adhesion junction proteins and a basolateral membrane domain with contact to the basement membrane (BM) at its most basal extension. For radial glial cells, this is the BM at the brain surface delineating the brain parenchyma from the meninges (Figure 1A). Given this structure, both neuroepithelial cells as well as the later differentiating radial glia provide a stable radial architecture throughout the developing brain. Radial glial cells differentiate with the onset of neurogenesis from neuroepithelial cells by acquiring glial hallmarks, such as expression of vimentin, the glial fibrillary acidic protein (GFAP), the two astrocyte-specific glutamate transporters (Slc1a3, also referred to as EAAT1 or GLAST, and Slc1a2 also referred to as EAAT2 or GLT-1), glutamine synthase (GS) as well as ultrastructural criteria of glycogen granules and filaments (see chapter 5 in Ref. 129). As the brain further grows in thickness during development, radial glial cells have to further extend their radial processes reaching out up to several millimeters in the human cerebral cortex. Accordingly, the first role of these cells to be discovered was their support function as stable radial structures also providing guidance for migrating neurons. Indeed, this role is important for various migration disorders and often mutations in genes resulting in neuronal dysplasia affect radial glia morphology and process stability (see, e.g., Ref. 39).
Given this functional importance of the long radial processes, it was difficult to perceive that these cells should divide at the same time despite their DNA synthesis as demonstrated by [3H]thymidine incorporation (184). Only when several labs provided evidence for radial glial cell division and contribution to neurogenesis did this concept catch on. This was demonstrated first by isolating radial glia by fluorescent-activated cell sorting (FACS) and following their progeny in vitro that included many neurons (164, 166), then by live imaging in slice cultures following radial glial cell divisions generating bipolar cells migrating towards the neuronal layers reminiscent of neurons (185, 198), and finally by in vivo tracing of radial glia progeny by genetic fate mapping (165). Genetic fate mapping allows tracing of the entire progeny of a given cell type by expressing a DNA recombinase (cre, flp, or others) in this cell type. This allows turning on a reporter gene encoding a marker protein, e.g., a fluorescent protein that is separated by a stop cassette flanked by sites recognized by the recombinase from a ubiquitous promoter. When expression of the recombinase is turned on, e.g., by the GFAP promoter in radial glial cells (165), it excises the stop cassette and thus the marker gene expression is turned on in this specific cell type. This is then inherited to all cell types that may ever be generated from this cell population as the excision of the stop cassette is irreversible and hence the complete progeny will also express this marker. Importantly, the recombinase can be further modified to increase temporal specificity which was essential for fate mapping the progeny of glial cells in the adult brain. Temporal specificity is achieved by fusing the recombinase to a ligand binding receptor such that it can only be active when the ligand is provided, allowing temporal control over the recombination (for review, see Refs. 113, 281). Regional and further cell type specificity can be achieved by combinatorial means, either combining different recombinases (104) or splitting the recombinase into two halves expressed under control of two different promoters (102, 103) such that recombination and subsequent expression of the reporter gene is only mediated in cells with both activated promoters. For example, this allowed the discrimination of adult neural stem cells (NSCs) expressing both GFAP and Prominin1 from surrounding astroglia (GFAP+, Prominin1-) and neighboring ependymal cells (GFAP−, Prominin1+; Ref. 16).
These techniques revealed that radial glial cells generate neurons and later glial cells (such as astrocytes, ependymal cells, and oligodendrocyte progenitors) in many brain regions and species (139, 218). In some regions, such as the spinal cord and possibly the retina, radial glia differentiate so late that they only contribute to gliogenesis (175). In regions where they contribute to neurogenesis, this regional specification is crucial to generate the neuronal diversity of the brain. In the ventral midbrain, for example, radial glial cells in the floor plate, the region defined by expression of the morphogen sonic hedgehog (SHH) and the transcription factor Foxa2, generate the ventral midbrain dopaminergic neurons (32, 130). Radial glial cells also differ profoundly in their regional specification and the neuronal subtypes they generate in the telencephalon. In the dorsal telencephalon, the anlage of the cerebral cortex, they generate largely glutamatergic pyramidal projection neurons governed, e.g., by the transcription factor Pax6 (for review, see Ref. 87). In the ventral telencephalon, radial glial cells differentiate gradually and have only acquired full radial glia identity by mid-neurogenesis (165, 215). They express the transcription factors Gsx2 and Ascl1 and contribute then to the generation of GABAergic projection neurons and interneurons (38, 70, 285). As the latter is a particularly large population of inhibitory neurons spreading throughout the telencephalon, radial glial cells generate this larger neuronal output by a series of intermediate progenitors that bears striking resemblance to amplifying lineages with a large output as observed in other species in development or the rodent olfactory bulb (OB) neurogenesis at postnatal and adult stages (215). Indeed, adult NSCs resemble radial glial cells not only in this region, the ventral murine telencephalon, but also in many other regions and species (see below).
In most regions of the mammalian brain, however, radial glial cells disappear by generating specialized glial cells that remain at the ventricle, the ependymal cells. While these remain anchored at the ventricle and maintain the junctional coupling and some degree of epithelial polarity, astrocytes and OPCs spread through the entire brain parenchyma where they homogeneously cover the neuropil with their fine processes. This is achieved by radial glial cells that do not directly generate postmitotic astrocytes or oligodendrocytes but rather proliferative intermediate progenitors. These progenitors then multiply and spread throughout the entire ventrodorsal extent (Figure 1B; Refs. 112, 125, 275, 289), while they sometimes respect rostro-caudal domains, such as the rhombomeres, segmental patterns in the developing hindbrain (180). Thus especially OPCs, which often originate in ventral regions of the brain and migrate, cover long-distances by both migration and continued proliferation, therefore populating large regions.
Surprisingly, this is rather different for astrocyte progenitors that also delaminate from the ventricle or the pial surface and continue to proliferate in the neonatal brain parenchyma, but still remain within regional constraints as they largely migrate radially possibly along the remaining radial glial cells (76, 77, 105, 280). These data therefore suggest that the spread of astrocytes at early postnatal stages results from local proliferation rather than long-distance migration. Importantly, this implies a regional heterogeneity of astrocytes that is of crucial functional relevance. For example, only astrocytes from the correct region are able to support dendrite development of regionally matched neurons in vitro (58, 239), and formation of specific synapses onto motoneurons was significantly decreased when astrocytes were reduced only by a third in this domain (280). So, neighboring astrocytes do not invade and compensate for the partial loss of astrocytes within a specific CNS region, highlighting the important concept of regional diversity of astrocytes, similar and/or in relation to the regional diversity of neuronal subtypes.
This is strikingly different for the other major set of glial progenitors, the OPCs, that readily compensate the loss of OPCs close by or in different domains as described below (124). Intriguingly, these differences between OPCs and astrocyte progenitors at early postnatal stages anticipate their respective behavior in the adult brain after injury, with astrocytes remaining very stationary with virtually no migration towards the injury site (8, 280), as opposed to glial progenitors of the OPC lineage that readily migrate and accumulate around the injury site (see below and Figure 4). Thus the behavior of these distinct glial progenitors differs profoundly already at early postnatal stages, with astrocytes showing a strongly regionalized diversity, while OPCs appear to be more promiscuous in this regard.
B. Oligodendrocyte Progenitor Cells During Development
OPCs are the progenitors of the myelinating cell type of the CNS, the oligodendrocytes, and appear in the developing brain almost as early as the radial glial cells discussed above. With only a few days delay after the onset of neurogenesis, OPCs were detected in ventral regions of the developing CNS from where they spread throughout all brain regions by migration and proliferation (Figure 1B). Their hallmarks are the expression of the platelet-derived growth factor receptor α (PDGFRα; Refs. 197, 220, 221), the proteoglycan NG2 (152, 153, 194, 195; for review, see Refs. 192, 193) as well as the key transcription factors Olig1/2 and Sox10 (177, 249, 272). OPCs were first discovered in dissociated cell cultures derived from optic nerve (224, 225), then from the spinal cord and the forebrain (223, 294). These bipolar cells were positive for several cell surface markers, such as the antibody A2B5, the proteoglycan NG2, or the receptor PDGFRα, and were shown to proliferate and then undergo a number of structural and biochemical changes as they mature (212) into myelin-forming cells enwrapping and myelinating axons in cocultures with neurons. During this differentiation process, cells lose the above-mentioned OPC specific proteins and begin to express surface antigens that bind the monoclonal antibody O4 (7, 260), galactocerebrosides (226), and early myelin proteins like the 2',3'-cyclic nucleotide phosphodiesterase (CNP; Ref. 212) and myelin basic protein (MBP; Ref. 267). This provided an extremely useful in vitro model to unravel factors that either inhibit or promote proliferation and differentiation of these cells, which could be verified in vivo as described below.
For some time, the spatiotemporal origin of these cells remained more controversial. The origin of oligodendrocytes was detected by some labs at multiple foci or throughout the embryonic ventricular zone (262, 265, 289), while others observed very restricted, specialized niches at the ventricular zone of the spinal cord and the forebrain (125, 203, 221, 264, 289; for review, see Ref. 240). Both hypotheses were supported by the expression of key factors and their receptors promoting OPC specification and proliferation such as SHH and PDGFRα that are initially localized ventrally and become more widespread at later stages (221, 263, 297) while the bone morphogenic proteins (BMPs) that inhibit OPC development are emerging from more dorsal sources (178). Finally, genetic fate mapping revealed that both views are correct, but temporally spaced. Indeed, at very early stages (E9/E10), plp/dm20-transcript positive OPCs emerge from hotspots in the ventral mouse forebrain (OPCs described in Ref. 262) (Figure 1B). Interestingly, these cells are not sensitive to PDGF and seem to mature later. Moreover, genetic fate mapping experiments using plp-Cre and plp-CreERT2 mice revealed that cells expressing plp-Cre at early stages contribute to neuronal and glial lineages, while cells expressing plp-Cre at later gestational stages (E13 and after) only generate oligodendrocytes (56).
In a second wave, cells responsive to PDGF and hence expressing PDGFRα emerge (around E14) from ventral regions in the spinal cord and the forebrain and their progeny comprises virtually exclusively more OPCs and mature oligodendrocytes (221, 289). Notably, the transcription factor Olig2 that is key for oligodendrocyte development (161, 162, 268) is also expressed at earlier stages by motoneuron progenitors (186, 305), but is at later stages rather specific for glial progenitors and yet later for OPCs. In the developing spinal cord, Olig2 is expressed in the ventral region, and inducible genetic fate mapping revealed that distinct sets of Olig2-expressing cells generate motoneurons at earlier stages, while a separate set of later appearing Olig2-expressing radial glial cells generates OPCs (188, 296). Notably, the same is the case for progenitors expressing the transcription factor Ascl1 (former named Mash1) with a distinct neuronal lineage first and a later glial lineage (13). Indeed, Ascl1 also labels OPCs originating from dorsal regions, and their dorsal origin has been further demonstrated by region-specific genetic fate mapping revealing cells spreading into ventral regions as, e.g., in the spinal cord (306). Such experiments using region-specific Cre lines even revealed a transient nature of the early ventrally derived OPCs at least in the telencephalon (Figure 1B). While these cells populate the entire telencephalon at neonatal stages, later OPCs arise at more dorsal positions (labeled by Emx1-Cre; Ref. 124) at postnatal stages. These cells gradually replace the earlier OPCs such that most mature oligodendrocytes in the dorsal telencephalon, the cerebral cortex, are derived from the dorsal telencephalic regions (Figure 1B). These data imply also some degree of regional diversity of mature oligodendrocytes as well as adult OPCs that continue to proliferate life-long at least in the rodent and human brains (80, 254). Indeed, some region-specific differences in the migratory and neurogenic potential have been observed (2), but other than this the functional relevance of this temporal replacement and/or regional origin and diversity has yet to be discovered, as no defects could be observed when one population replaces the other. This has been achieved by expressing diphtheria toxin in the dorsally derived OPCs to kill them selectively (124). Intriguingly, the remaining OPCs that were largely ventrally derived could readily replenish the depleted pool of dorsally derived OPCs and give rise to mature oligodendrocytes that could fully substitute the missing cell population (124). These data obtained in the developing brain already highlight the amazing compensatory and proliferative capacity that all of these OPC populations have and also maintain into adulthood. Even in the adult brain, OPCs can any time reactivate proliferation and replace lost NG2-glia or oligodendrocytes even in large numbers as discussed below.
In summary, one of the key differences between radial glial cells and OPCs is the persistence of the latter (at least the last wave of OPCs) throughout the mammalian brain while the former are largely transient and only restricted in few niches as adult neural stem cells where they continue to generate neurons (139). This raises not only the question to which extent very slow cycling, alias “quiescent” or “dormant” NSCs may still persist also in the adult mammalian CNS, possibly hidden by their altered morphology, such as the ependymal cells still lining the ventricle (10, 179) or the astrocytes within the grey or white matter parenchyma (35, 236, 255) but also whether the OPCs may have a broader potential and may serve as source of other cell types when needed.
C. Glial Progenitors With Multilineage Potential
The above concepts thus imply that only radial glial cells give rise to multiple cell types in vivo, while OPCs and astrocyte progenitors seemingly serve to multiply a single lineage, not withstanding further diversity among oligodendrocytes or astrocytes. Indeed, this is the case also within the neuronal lineage with radial glial cells sometimes generating further amplifying neuronal progenitors as mentioned above (87, 215). However, are radial glial cells really the only cell type with neural stem cell properties and multilineage potential, or can the OPCs, at least the early population, also give rise to more cell types than “just” oligodendrocytes?
OPCs have indeed been described to generate also astrocytes first in vitro (18, 225) and then in vivo. Genetic fate mapping of NG2-expressing cells (using NG2-Cre transgenic mice) revealed that more than 40% of the protoplasmic astrocytes in the grey matter (GM) of the ventral forebrain were generated from NG2 expressing cells (307, 308), while this was not the case for astrocytes in the white matter (WM). Interestingly, a similar lineage was also observed in the spinal cord (309). Thus the lineage of NG2-expressing cells in the developing neural tube is not restricted to oligodendrocyte cells, and the term OPCs is not appropriate. In contrast, it is more appropriate to refer to these cells as NG2-glia, in light of their contribution also to other glial lineages and of their additional functions that will be described later. Importantly, other cell types in many organs also bear the proteoglycan NG2, but these are pericytes (204), cells residing in the perivascular space, which are generally not considered as glial cells as they are of mesodermal origin in most organs and regions of the neural tube. Here we use the term NG2-glia for CNS-derived glial cells expressing the proteoglycan NG2.
While the above genetic fate mapping experiments showed the contribution of NG2-glia to different glial cells, they could not resolve to which extent a single progenitor cell is bipotent generating both oligodendrocytes and astrocytes, or whether NG2+ glial progenitors comprise several pools of cells, some generating astrocytes and others generating oligodendrocytes. To separate such different sets of progenitor cells, various combinations of cell surface markers have been established, allowing to preferentially enrich, e.g., progenitor cells giving rise to both astrocytes and oligodendrocytes (90, 228). These cells have been called “glioblasts,” in analogy to the term neuroblasts, but are also often referred to as glia-restricted progenitors (GRPs) as their fate restriction to the glial lineage has been demonstrated in vitro as well as in vivo by transplantation experiments (99). As these cells share their bipolar morphology and many markers with NG2-glia, they may well represent the bipotent subset of these. Moreover, even tripotency, i.e., the potential to generate neurons, astrocytes, and oligodendrocytes, has been demonstrated under certain in vitro conditions when isolating cells enriching for NG2-glia from the developing early postnatal brain (2, 20, 137). Interestingly, these glial progenitors can be readily converted by application of BMP to an astrocyte-like cell in vitro which is then able to reveal the stem cell hallmarks of self-renewal and multipotency when grown in appropriate growth factor conditions (Figure 2; Ref. 137; see however Ref. 2 for the isolation of OPCs that naturally generate different lineages). Importantly, such multipotent cells can also be isolated from early postnatal rodent or even human brains by culturing cells first under conditions for astrocyte progenitors and then probing for stem cell hallmarks (149, 284). Taken together, at early postnatal stages, bi- and multi-potent progenitor cells as well as progenitors restricted to a single lineage exist in various regions of the rodent brain. Also note that this potency is often only revealed when cells are placed in vitro or in a different environment, while they only generate a single cell type in vivo, with the distinction between in vivo lineage and in vitro potential applying here. However, by 3 wk after birth in rats and mice, the plastic cell types with a larger potency appear to be much reduced or virtually lost (149), prompting the question as to whether remnants of multipotent cells may still exist in the fully adult mammalian brain and could possibly be activated by the appropriate stimuli.
III. ADULT BRAIN: GLIAL STEM AND PROGENITOR CELLS
A. Radial Glia in the Adult Brain: Adult Neural Stem Cells and Neurogenesis
1. Nonmammalian vertebrates
Generally, endogenous neurogenesis persists in brain regions, where radial glial cells persist. This is the case in the mammalian brain, where radial glial cells disappear in most regions, but persist in a few niches where also neurogenesis persists (Figure 3A). In most nonmammalian vertebrates, such as in amphibian, bony fish, rays and sharks, as well as reptiles, however, radial glial cells remain abundantly present in a widespread manner in the adult CNS (Figure 3B; Ref. 50). Most glial cells lining the ventricle possess long radial processes, have epithelial characteristics (92), and share many proteins with radial glial cells in the developing brain, such as Prominin1, brain lipid binding protein (BLBP), nestin, vimentin, GFAP, S100β, GS, and the transcription factors Hes1/5 and Sox2 (44, 75, 114, 140, 145, 170, 210). Notably, none of these proteins is specific for radial glia as Prominin1, vimentin, S100β, and Sox2 are also shared with ependymal cells and GFAP, S100β, GS, and Sox2 with parenchymal astrocytes in the murine brain (16, 236). The persistence of radial glial cells at the ventricular surface implies that no specialized cuboid ependymal cells are present (see also chapter 5 in Ref. 129). Likewise, also astrocytes are lacking in brain regions with persisting radial glia, such that radial glial cells encompass the function of both these mature glial cells types by lining the ventricle as ependymal cells do and enwrapping synapses and expressing K- and aquaporin channels for ion and water homeostasis as astrocytes do (92; see also chapter 5 in Ref. 129).
In addition, radial glial cells also continue to enact their function as stem and progenitor cells from development, namely, to divide and generate neurons (for review, see Refs. 42, 89). Indeed, neurogenesis continues in a widespread manner in the species of bony fish, reptiles, and amphibians which have been examined so far (89). For example, in the zebrafish telencephalon, a subset of radial glial cells (∼10%) divides typically in an asymmetric manner (238) giving rise to a radial glial cell and an intermediate progenitor. The intermediate progenitor loses its glial characteristics, but does not yet express neuronal traits. It continues to proliferate and then generates differentiating neurons (1, 43, 238). As the zebrafish and its brain continue to grow, it is of particular relevance that stem and progenitor cells persist and do not deplete. This is achieved by the asymmetric mode of division described above, and a reserve pool of radial glial cells with only some dividing while others remain quiescent. However, virtually all the quiescent radial glia can be activated to enter proliferation by inhibition of the Notch signaling pathway (43; see also Alunni et al. (4a), note added in proof) or after injury (see below), demonstrating that all of them have proliferative neural stem cell potential. This is also the case in regions with low or no proliferation and neurogenesis in adulthood, such as the zebrafish spinal cord (229, 231) or the salamander midbrain (26).
Moreover, some CNS regions, such as the dorsal midbrain and the retina, possess non-radial glia stem cells densely packed in a zone at the margin of the respective tissue, called the ciliary marginal zone in the retina. The stem cells in this region rather resemble the earlier neuroepithelial cells as they fail to express glial hallmarks (89). Summing up, many vertebrates are equipped with widespread adult NSCs that resemble the NSCs in the developing brain, either the neuroepithelial or the radial glial cells. This has important consequences for the reaction after brain injury, not only in regard to the repertoire of glial cells reacting to injury which is obviously different from mammalian brains (e.g., the zebrafish CNS largely lacks parenchymal astrocytes; see, e.g., Ref. 14) but also in regard to the capacity to replace degenerated neurons that is largely lacking in the mammalian CNS.
2. Mammalian (and avian) vertebrates
Interestingly, also in mammalian and avian brains radial glial cells persist into adulthood, but in a more spatially restricted manner mostly within the forebrain (for review, see Refs. 89, 190). For example, radial glial cells, rather than cuboidal multiciliated ependymal cells, line the ventricle in the hypothalamus (Figure 3A). These are often referred to as tanycytes, a name typically referring to radial glial cells lining the ventricle in adult brains. Tanycytes in the hypothalamus differ depending on their dorsoventral location with β-tanycytes located most ventrally, expressing FGF10 (fibroblast growth factor) and lacking GFAP and GLAST expression in mice, while α-tanycytes express GFAP and GLAST, but not FGF10 and are located more dorsally (93, 237). Most importantly, these different populations also exhibit functional differences recently revealed by genetic fate mapping showing that FGF10+ β-tanycytes generate neurons within the first two postnatal months and upon high-fat diet (93), while GLAST+ α-tanycytes generate few neurons and are rather gliogenic in normal adult mice (237). Most importantly, however, only the latter have the NSC hallmarks of long-term self-renewal and multipotency in vitro (237). Thus the hypothalamus certainly contains cells with NSC properties, while neurogenesis at resting state in the adult brain appears to be rather low (Figure 3A). Importantly, however, it can be activated by metabolic stimuli (93, 213), prompting the important question about the functional consequences of this induced adult neurogenesis.
In contrast to the hypothalamus, in the most active zone of adult neurogenesis in mice, at the ventricle, the progeny of the actively dividing radial glial cells are so many that they form a distinct layer, the subependymal zone (SEZ) located below the ependyma lining the ventricle. This progeny comprises transit-amplifying progenitors (TAPs) dividing rather fast and neuroblasts (NBs) that also continue for a while to proliferate and amplify while migrating through the rostral migratory stream (RMS) to their final destination, the olfactory bulb (Figure 3A). The radial glial cells act as slow dividing NSCs that are integrated with a process into the differentiated ependymal layer and their cell somata located below in the SEZ (183). This is why this zone is best referred to as SEZ to discriminate it from the subventricular zone (SVZ) in the developing brain when and where no ependymal cells are yet present.
Radial glial cells in this region have a shorter radial process compared with those in the developing brain reaching all the way to the BM underlying the meninges. The radial processes of adult NSCs in the SEZ typically contact the BM surrounding neighboring blood vessels and obtain key factors from this vascular niche, promoting their self-renewal (134, 135). As typical radial glia they also have epithelial hallmarks with junctional coupling to their neighboring radial glia or ependymal cells and a delineated apical membrane domain containing Prominin1 (16). They also express many of the radial glia proteins highlighted above (GFAP, GLAST, GLT-1, Sox2, Hes5, Prominin1, BLBP, etc.), and genetic fate mapping using the split-Cre technology clearly demonstrated that the cells coexpressing GFAP and Prominin1 act as NSCs self-renewing and contributing to neurogenesis over several months (16). Interestingly, there is some heterogeneity with long-term and short-term self-renewing NSCs that can be detected by genetic fate mapping targeting cells expressing some of the above radial glial proteins (83). Thus it appears that adult NSCs typically line the ventricle and possess or maintain neuroepithelial properties. Indeed, key signals derive from the cerebrospinal fluid (CSF) in the ventricle both during development (115, 116) and in adulthood (302), and the junctional complexes characteristic for epithelial cells act as crucial cellular signaling centers (115). However, also cells lacking direct ventricular contact and epithelial polarity, such as the radial astrocytes in the dentate gyrus (DG) act as adult NSCs and source of life-long neurogenesis in most mammalian species including humans (133, 261). These are astroglial-like cells with a radial morphology located in the subgranular zone (SGZ) of the DG (for review, see Refs. 139, 236), at some distance from the ventricle and no apicobasal polarity which is characteristic for epithelial cells. For example, Prominin1 that is sorted to the apical membrane domain in neuroepithelial and radial glial cells is spread along the entire membrane domain of DG NSCs including their long radial process (17). Interestingly, however, the combination of GFAP and Prominin1 expression targets specifically NSCs also in this region, highlighting still some molecular similarity to radial glial cells during development (217) and in the adult SEZ (16). Therefore, also glial cells resembling more parenchymal astrocytes with a radial morphology, the so-called radial astrocytes, can act as adult NSCs self-renewing for some time and giving rise to intermediate progenitors that then generate differentiating neuroblasts (181). Importantly, the DG NSCs reside in a vascular niche, the other frequent niche of NSCs (85). With regard to radial astrocytes, it is worth mentioning the Müller glia and the Bergmann glia, two prominent astroglia-like cell types with a characteristic radial morphology in the retina and the cerebellum, respectively. While they do not act as actively neurogenic NSCs in the mammalian species examined so far, especially the Müller glia is a very plastic cell type which reactivates a broader potential after injury (120). This prompts the question to which extent other radial or nonradial astrocytes may possess or acquire NSC hallmarks in the adult mammalian brain. In the normal healthy brain, parenchymal astrocytes do not proliferate and do not show any NSC potential upon dissociation and culture in the neural stem cell assay, the neurosphere conditions (Figure 2; see above and Ref. 255). Interestingly, however, there is a widespread proliferating glia population throughout most of the adult brain parenchyma in rodents and humans, the NG2-glia.
B. NG2-Glia in the Adult Brain Parenchyma: OPCs or Stem Cells Outside the Neurogenic Niches?
Around 5–10% of the cells in the adult mouse brain parenchyma express NG2 (53, 158, 193, 219), PDGFRα (64, 234), and the transcription factors Olig2 (36, 61, 161) and Sox10 (249, 291), highly reminiscent of the profile of these glial progenitors during development. Indeed, these cells proliferate, as shown by the incorporation of bromodeoxyuridine (BrdU), but also by immunolabeling for Ki67 in actively proliferating cells (222, 254) and phosphorylated histone H3 in M-phase (Sirko, Dimou, and Götz, unpublished data). Interestingly, NG2-glia divide very slowly in the adult mouse forebrain with a cell cycle of several weeks (47, 222, 254), and previously proliferating cells reenter the cell cycle and reacquire Ki67 several weeks after previous BrdU incorporation (254). This is notably different from the fast division rate of these cells in the postnatal brain (see, e.g., Ref. 222). Upon injury however, NG2-glia in the adult brain reenter the cell cycle very fast (see below), suggesting that their cell cycle length and activation are subject to regulation by signaling. Importantly, virtually all proliferating cells outside the neurogenic niches in the adult mouse brain as described above are NG2/PDGFRα/Olig2+ (61, 79, 254), prompting the question to which extent their progeny in the adult brain may resemble the progeny in the developing brain.
Both BrdU-birthdating experiments (254) as well as inducible genetic fate mapping revealed that these proliferating cells expressing Olig2 (61), PDGFRα (119, 234), and NG2 (193) generate mature oligodendrocytes expressing CC1, GST-π, and myelin proteins in the adult brain. Using NG2-, PDGFRα-, and Olig2-CreER mouse lines allows inducing genetic recombination selectively in the adult brain by application of tamoxifen at this time and following the fate of the genetically labeled cells over weeks and months. Interestingly, there are profound region-specific differences with NG2-glia proliferating very little in some regions, such as the spinal cord (119), or generating more or less mature oligodendrocytes depending on their location. This difference is particularly pronounced between the cerebral cortex GM and WM with the majority of NG2-glia in the WM generating mature, myelinating oligodendrocytes, while NG2-glia in the GM generate significantly fewer mature oligodendrocytes (61, 118). Interestingly, recent homo- and heterotopic transplantation experiments revealed intrinsic differences between WM and GM NG2-glia in their propensity to generate mature oligodendrocytes (282). WM-derived cells generate mature oligodendrocytes very efficiently also in the less supportive GM environment, while GM-derived cells generate significantly fewer oligodendrocytes in the very same in vivo compartment. However, when GM-derived NG2-glial cells are exposed to a WM environment, they increase the generation of mature oligodendrocytes, even though many of these fail to acquire a full myelinating phenotype (282). Taken together, these experiments highlight for the first time major functional differences between NG2-glia from different, even though neighboring regions, with profound functional implications.
One important question emerging from these data is the role of NG2-glia that does not generate oligodendrocytes, but rather self-renews and generates more NG2-glia. This raises the hypothesis that NG2-glia in the adult brain has other functions beyond being progenitors for oligodendrocytes. Importantly, they differ from other glial cells by receiving synapses from neurons with clear ultrastructural specializations (29, 55, 243), a feature for which they are also named as “synantocytes.” These synapses can be GABAergic or glutamatergic, and NG2-glia possess the respective transmitter receptors, including GABAA or GluRB receptors (159, 174, 243, 279). Notably, the synaptic potentials of NG2-glia can be changed in a long-lasting manner with repetitive stimulation resulting in their long-term potentiation mediated by AMPA receptors on NG2-glia (78). Interestingly, although these synapses elicit clear postsynaptic potentials in the NG2-glia (140, 168, 169), they do not generate action potentials (according to all; Refs. 28, 29, 47, 78 except one study, Ref. 120, in the rat cerebellar WM). Therefore, NG2-glia appear to have some physiological hallmarks, previously thought to be unique for neurons in the CNS, such as direct synaptic innervation, but not all, such as action potential formation. As these properties, such as the direct synaptic input, are also not shared with other glial cells, it is appropriate to consider the NG2-glia as a distinct glial entity, the fifth glia population (besides astrocytes, oligodendrocytes, microglia, and ependymal cells) in the mammalian CNS (37). Whether NG2-glia are also present in other vertebrates, however, still remains to be elucidated (see below and Refs. 14, 171) like the question when this type of glia evolved during phylogeny which may also hold some clues in regard to their prime function.
While the function of this synaptic communication is not yet fully understood, an obvious hypothesis is that it may regulate proliferation or differentiation of these cells. Consistent with this hypothesis, sensory deprivation of the mouse barrel cortex during development leads to increased NG2-glia proliferation and results in their uniform distribution in the deprived barrels (60, 169), while physical activity of adult mice leads to an immediate cell cycle exit of NG2-glia and their differentiation into mature oligodendrocytes in the cerebral cortex (254). Based on the above described observations, these effects could indeed be mediated via the direct synaptic contacts that are formed between neurons and NG2-glia, even though this still remains to be proven (for review, see Ref. 243). Interestingly, NG2-glia lose their synaptic input when differentiating into oligodendrocytes (167), and myelination has long been described to be regulated in an activity-dependent manner (12, 57, 243, 300). Therefore, increased neuronal activity seems to lead to increased generation of myelinating oligodendrocytes via NG2-glia proliferation and differentiation even in the adult brain.
This has obvious functional consequences as increased myelination may improve the speed of information processing in the active brain regions (22, 67), and new generation of myelinating oligodendrocytes in the adult brain contributes to myelin remodeling (299). Also in nerves that are fully myelinated, such as the optic nerve, new myelinating oligodendrocytes are generated, suggesting some degree of myelin turnover (299). Intriguingly, this leads, at least in the optic nerve, to a change in internode length (299) with clear implications for conduction velocity (49, 60, 295). Thus remodeling of myelin and the ability to generate myelinating oligodendrocytes also in the adult mammalian brain provides new mechanisms for plasticity in the adult brain. Indeed, major plastic changes, such as the response of neurons in the visual cortex to largely one eye (ocular dominance, OD, plasticity), were previously thought to come to a permanent end after the critical period that was thought to terminate when myelination is completed (in mice and rats by about 2 months of age). However, recent studies demonstrated that OD plasticity can still be elicited, e.g., by silencing input from one eye in fully adult rodents at much later stages, albeit at a slower pace (106, 107, 216). The ongoing generation of new myelinating oligodendrocytes thus contributes to adult neuronal plasticity by myelin remodeling, possibly allowing sprouting and collateral formation during this process. Compared with the forebrain, other areas of the CNS, such as the spinal cord, show fewer proliferating NG2-glia and lower numbers of newly generated oligodendrocytes (109, 118), either due to fewer cells being able to perform these features or due to longer cell cycle length and/or differentiation duration. An interesting prediction would therefore be that, e.g., the spinal cord may exhibit less pronounced adult plasticity. A further fascinating concept emerging from the above considerations is that loss of myelinating oligodendrocytes, e.g., upon injury, may contribute to reopen a window of opportunity for rewiring, as observed for example after spinal cord injury where detour circuits allow some recovery of function (146). Most excitingly, the discovery of proliferating NG2-glia also in the adult human brain (80) suggests that these events also take place in the human brain, thereby allowing not only for neuronal plasticity but also revealing a source of actively dividing progenitor cells which may be useful for repair after injury. But do these cells have the capacity to generate more than oligodendrocytes?
C. From Lineage to Potential: Are There Neural Stem Cells Outside the Neurogenic Niches in the Adult Mammalian Brain?
1. In vivo
While all studies using birthdating and fate mapping of NG2-glia in the adult mouse brain experiments agree on them generating oligodendrocytes in several regions of the adult mouse brain, the evidence for contribution to other lineages has been more controversial. Rivers et al. (234) detected some labeled neurons appearing and increasing in number in the piriform cortex upon eliciting genetic recombination in PDGFRα-CreERT2 mice; however, this observation could not be reproduced by the same lab (47). Indeed, also other labs using the very same mouse line had not observed neurons emerging from PDGFRα+ cells (118), nor were they observed in the progeny of Olig2+ (61) or NG2+ [308; See also Huang et al. (109a), note added in proof] cells. Therefore, evidence from genetic fate mapping experiments converges on NG2-glia generating further NG2-glia or oligodendrocytes but no other cell types in the healthy adult brain and CNS.
Likewise, BrdU labeling experiments revealed only NG2-glia shortly after labeling, while longer survival times revealed the generation of mature oligodendrocytes (254). While neurons were sometimes found very close to BrdU-labeled NG2-glia in the neocortex (254), no double-labeled neurons were observed. While others reported the conversion of NG2-glia into GABAergic neurons (54, 269), it is important to bear in mind that BrdU incorporation also labels cells undergoing DNA repair (for the pitfalls of BrdU labeling, also see Ref. 63), and DNA repair processes have been implicated in normal regulation of transcription, i.e., not necessarily requiring cell damage. Likewise, genetic fate mapping has its own technical limitations such as side effects of tamoxifen, ectopic and low level of CreER expression in other cells (e.g., neurons in this case) becoming evident only after prolonged tamoxifen, and/or slow accumulation of the reporter gene activity. Therefore, it is best to obtain evidence by multiple approaches which is the case for NG2-glia generating oligodendrocytes or further NG2-glial cells which has been confirmed also by transplantation experiments (282). The lack of neurogenesis in most regions outside the dentate gyrus/hippocampus region has also been confirmed in the human brain by yet a different technique, namely, carbon isotope birthdating (30, 261; see also Ernst et al. (65a) note added in proof).
The above data then suggest that the glial progenitors outside the neurogenic niches either cannot generate other lineages (such as neurons) or that they do not do so in their local environment but could do so if they were not inhibited by the local signals. Thus the failure of these cells to generate neurons may be due to the potent antineurogenic environment to which these cells are exposed. Consistent with the latter scenario, transplantation experiments showed that neuroblasts or even neural stem cells, i.e., cells that are well capable to generate neurons normally, largely convert to gliogenesis when transplanted into the adult brain parenchyma (for recent review, see Ref. 190). This highlights the potent gliogenic environment NG2-glia are exposed to and prompts the question if they may also have a broader potential when exposed to more favorable conditions, e.g., the environment in the neurogenic niches.
Indeed, when cells derived from regions like the spinal cord or the substantia nigra were transplanted into the region of their origin, they generated largely NG2-glia, but when transplanted into the neurogenic niche of the DG they seemingly gave rise to neurons (157, 250). These data suggest that multipotent adult progenitor cells exist in the adult CNS in a widespread manner and that these cells can generate different cell types depending on the local niche (neurogenic versus gliogenic) and the signals emerging from this environment. However, it is again important to bear in mind the technical limitations of these transplantation experiments. By now we know that transplanted cells can fuse with cells of their environment (see, e.g., Refs. 34, 51), but this has not been assessed in these previous transplantation experiments. Moreover, in these experiments, cells had been labeled by BrdU, such that their death may well have provided neighboring proliferating cells with BrdU to incorporate into their DNA when dividing. Indeed, more recent experiments revisiting whether glial cells (even cultured as neurospheres and exhibiting therefore some in vitro stem cell potential, such as self-renewal and multipotency; Ref. 252) can generate neurons when exposed to the neurogenic SEZ, failed to detect any neuronal progeny from these cells, suggesting that they may be unresponsive to the neurogenic cues, at least those provided by the SEZ environment. Thus the question of to what extent the widespread progenitors present in the adult brain parenchyma can contribute to other lineages in vivo still remains open.
2. In vitro
Conversely, in vitro evidence for some cells exhibiting the potential of long-term self-renewal and multipotency, i.e., the generation of neurons, astrocytes, and oligodendrocytes, has been achieved by several labs. As described briefly above, the neurosphere assay is a culture system well suited to probe for neural stem cell potential (Figure 2). Several decades ago Sam Weiss (232), Perry Bartlett (233), and colleagues placed dissociated cells from the adult rat brain at low density free-floating in medium with epidermal growth factor (EGF) and/or basic fibroblast growth factor (FGF2). Under these conditions, a few cells proliferated and formed small spheres, the “neurospheres,” especially when cells were isolated from the striatum. These cells can self-renew for a long time as they form new spheres when dissociated and plated in the same medium conditions for many passages (for recent review, see Ref. 208). To probe their differentiation potential, neurosphere cells can then be exposed to differentiation conditions by plating them on adherent substrates and withdrawing the mitogens. While most neurospheres can be obtained from the regions comprising adult NSCs, in particular the adult SEZ (see, e.g., Ref. 255 for direct comparison of SEZ and other regions), a few neurospheres also emerge when cells were isolated from different areas of the brain outside the neurogenic niches like the striatum (206), the cortex (205), optic nerve, corpus callosum, spinal cord, etc. (157, 251, 278, 292), with higher numbers reported for cells isolated from rat compared with mice (compare the above and Ref. 255). Interestingly, some of these neurosphere-forming cells have been suggested to express NG2 and Olig2, implying that some NG2-glia may yield a larger potential when exposed to these favorable conditions. However, so far no clear evidence for this was obtained, e.g., by genetic fate mapping.
Neurosphere forming multipotent stem or progenitor cells were also identified in the human and rodent WM (82, 199, 273). These cells mainly generate oligodendrocytes when cultivated in high density and in the presence of serum or are implanted into the adult brain in a demyelinating paradigm (293). In contrast, when these cells isolated from the subcortical WM are cultivated in a low density and depleted from serum, they are able to acquire multipotency and differentiate into neurons, astrocytes, and oligodendrocytes (199). Taken together, these data suggest that a pool of stem or progenitor cells with larger potential exists in the adult WM, but these cells are restricted to generate glia due to their local environment in vivo rather than their intrinsic fate restriction. The exciting question now is to which extent these cells may provide a useful source of cells for repair after brain injury when the environment changes dramatically.
IV. INJURY: GLIAL STEM AND PROGENITOR CELLS
A. Radial Glia in Vertebrate CNS Injury
Given the widespread neurogenesis persisting into adulthood in many vertebrates, these cells may provide the best local source for neuronal repair after brain injury. Indeed, the zebrafish as well as some other teleost and amphibian species are well-known models for regeneration of many organs (132), including scarless wound healing and neuronal repair after CNS injury (for review, see Refs. 15, 69, 131).
1. Regeneration of neurons
Upon traumatic injury inflicted into the forebrain or cerebellum by insertion of a small cannula through the skull or the nostrils, cells undergoing neurogenesis become activated and generate additional neurons that migrate to the site of injury (6, 14, 121, 140, 172). Genetic fate mapping of radial glial cells by CreERT2 expressed under the control of the regulatory elements of the Notch target gene Her4 suggests that the additional neurons derive from the radial glial cells or their immediate progeny still expressing sufficiently high levels of the Her4-driven Cre recombinase (140). Consistent with this, both radial glial cells and their progeny, the TAPs, transiently increase their proliferation (14, 172), a response mediated by inflammatory signals resulting in upregulation of the transcription factor Gata3 (131, 144). Thus cells lining the ventricle in the adult zebrafish brain and already undergoing neurogenesis become activated upon injury and increase proliferation to generate additional neurons migrating to the injury site.
Generally, this reaction appears comparable to the response to injury in the few sites in the mammalian brain where radial glial cells and adult neurogenesis persist into adulthood as described above. Also there, neurogenesis is increased in response to traumatic or ischemic injury close by, and various cell types are instructed to migrate from the neurogenic niche to the injury site (5, 23, 40, 156). However, there are major differences between this reaction in the mouse and the zebrafish forebrain. In the zebrafish model, many of the neurons generated in addition to the baseline neurogenesis survive long-term and the injury site disappears by scarless wound healing (6, 14, 140, 172). Conversely, few of the neuroblasts recruited to the injury site next to the SEZ differentiate into fully mature (DARPP32+) striatal projection neurons (5), and virtually none survives longer (274). Importantly, SEZ cells do not normally generate striatal projection neurons in these species but rather give rise to specific olfactory bulb interneurons in a highly region-specific manner (33, 160). It is thus conceivable that neuroblasts recruited into the striatum may not have the appropriate neuronal subtype specification and therefore eventually succumb to cell death. Moreover, ependymal cells that do not normally proliferate and generate neuroblasts become activated upon stroke (40) and may not complete the full neurogenic program, thereby generating immature or misspecified neurons during this injury-induced neurogenesis.
However, there is also evidence for plasticity of the type of neurons generated after injury (for review, see Ref. 236). In the few cases where new neurons with the adequate specification have been generated (or introduced by transplantation), their survival may be compromised by the environment faced in the site of injury, e.g., the scar. Indeed, direct evidence for this has been obtained by specific neuron ablation models in the mouse which elicit little scar formation and allow long-term survival of endogenous or transplanted new neurons (45, 163). These data further support a decisive role of the scarless wound healing observed in the zebrafish lesion models to allow long-term survival of the regenerated neurons.
Strikingly, the scar persisting for long time after mammalian brain injury continues signaling to recruit new neurons even months after the insult. The adjacent neurogenic SEZ region still supplies large numbers of neuroblasts migrating to the injury site (274), demonstrating persistent failure of these to survive and integrate even after the acute reaction has vanished. Thus one key difference between mouse and zebrafish brain injury is the scar formation which persists for months in the mammalian brain, while the injury site is no longer visible after days or weeks in the zebrafish brain depending on the size of the injury. An intriguing possibility for signaling the need for new neurons is the lack of functional neuron regeneration itself. Such a signaling feedback from intact neurons has been unraveled in the salamander midbrain where the lack of the neurotransmitter dopamine serves to activate the generation of new dopaminergic neurons (25, 26). Interestingly, longer range dopaminergic signals from the brain to the spinal cord are involved in regulating regeneration of spinal cord motoneurons (230), and other neurotransmitters also influence neurogenesis and gliogenesis (for recent review, see Ref. 25). Of note, in the regenerating spinal cord, the fibers releasing dopamine serve to further enhance the region-specific, developmental signaling pathway important for motoneuron generation and regeneration, the SHH signal. SHH is reactivated after spinal cord injury in the radial glial cells that then resume generation of motoneurons (229, 230).
Several important conclusions emerge from the above. First, signals from intact functional neurons provide a feedback mechanism in regard to the needs for replacement. Second, radial glia still lining the ventricle in nonmammalian vertebrates can reactivate adequate patterning cues resulting in neurogenesis of the appropriate neuronal subtypes (26, 229, 245) also in regions no longer undergoing neurogenesis, as is the case in the salamander midbrain and the zebrafish spinal cord or cerebellum. This is a profound difference to the mammalian brain, where outside the adult neurogenic niches no neurogenesis is initiated upon injury consistent with the absence of radial glial cells at these sites. However, these results observed in species with full regenerative potential prompt the search for a partial response in the mammalian species with activation of some developmental signals and partial reactivation of NSC properties (see below).
2. Scarless wound healing
Importantly, the persistence of radial glial cells has further consequences also for scar formation in nonmammalian vertebrates, as astrocytes are lacking in most of the regions exposed to injury. This is important as activated, reactive astrocytes surround the injury site in the mammalian CNS (see below), while such cells are absent upon injury in species with radial glial cells instead of astrocytes. Indeed, radial glia cells largely retain their cell somata at the ventricular surface also after injury, and little to no radial migration to the injury site is detected (14), except after very large injuries (140). Thus reactive astrogliosis as present in the mammalian brain fails to occur in these models, and this may contribute to create a favorable environment for neuronal replacement (see also Ref. 236 for further discussion).
However, a subset of reactive astrocytes in the mammalian spinal cord were recently observed to be derived from cells lining the ventricle, as occurs in zebrafish spinal cord, and these have beneficial effects on neuronal survival and wound contraction (242) reminiscent of the role of radial glia-derived emigrating cells in the zebrafish. Similarly, also NG2-glia appear to be largely absent or fail to be activated in these injury models. While cells expressing Olig2 are present in the adult zebrafish brain and spinal cord (14, 171, 207), few of these proliferate, and many seem to be rather differentiated oligodendrocytes. Even after injury few of these proliferate (14) with larger injuries eliciting an accumulation of Olig2+ cells around the injury site (172). It is noteworthy, however, that the CSPG4 recognized by the NG2 antibody and abundantly present on Olig2+ cells in mammalian brains, has not yet been detected on these cells in nonmammalian vertebrates which may have consequences for neurite regeneration in these species.
Conversely, microglial cells readily activate their proliferation and accumulate around the injury site also in the zebrafish model (14, 140). However, their type of reaction may well differ from those in the mammalian brain, where both beneficial and adverse types of microglia and macrophage reactions have been identified (127, 182, 248). In line with this, positive inflammatory signals after zebrafish brain injury promote the proliferative response of radial glial cells and seemingly promote neuronal repair in this model rather than impeding regeneration as in mammals (144). Thus the difference in the glial composition (no astrocytes, little activation of NG2-glia if at all) and the type of their reaction (microglia activation/macrophage roles/inflammatory response) in these models of scarless wound healing provides first insights into the possible culprits for scar formation in the mammalian CNS creating an adverse environment for neuronal survival and process regeneration after injury.
B. NG2-Glia Reaction to Injury in the Mammalian CNS
Given the above considerations, it is crucial to understand the glial cell reaction after brain injury in the mammalian brain to delineate beneficial from adverse roles. Indeed, the major proliferative glial cell type in the mammalian brain, the NG2-glia, has only recently been investigated in the context of traumatic injury, as they were so far largely studied in the context of demyelinating conditions given their predominant view as OPCs (41, 59, 123). However, NG2-glia also show a striking reaction after traumatic injury in the CNS with a very fast and huge increase in proliferation within 1–3 days up to almost 100-fold their baseline proliferation rate (254). Such a traumatic CNS lesion is the stab wound injury model that results in highly reproducible injury of a well-defined size. The size of the injury can be chosen by the dimension of the cannula or small blade that is penetrated into the respective CNS region. For example, a punctate wound of 200 μm length or a stab wound of 1 mm length can be generated by inserting different-sized blades into the cerebral cortex (8). The ensuing cellular response is limited to the injury tract and the immediate surrounding tissue. Already within hours after injury proliferation of microglia at the lesion edges and infiltration of the wound site by activated macrophages occurs (84), followed by the activation of NG2-glia undergoing hypertrophy, cell division, and migration within 24–48 h (254; von Streitberg and Dimou, unpublished data; Figure 4). At later stages, astrocytes react by hypertrophy and delayed proliferation (8, 254 and below). While some of these proliferating NG2-glia differentiate into mature oligodendrocytes, the majority of them remains as NG2-glia (61, 136), indicating some degree of heterogeneity in their reaction to an injury insult (see below and Figure 4). Interestingly, NG2-glia increase in number and density at the site of injury (254), which is particularly intriguing as their number and density is tightly controlled by cell death and repulsive behavior in the healthy brain, thereby maintaining a constant network of NG2-glia under normal conditions (110).
The original density is largely restored several weeks after injury by a progressive decrease of the NG2-glia with increasing postlesion times (151, 254). Notably, not all injury conditions are sufficient to overcome this strict homeostasis (19, 110), prompting the search for the identity of the signals involved in NG2-glia homeostasis. Moreover, under some lesion paradigms, NG2-glia also contribute to generate astrocytes (94, 272), as also observed when reactive NG2-glia are cultured in vitro (unpublished observations). This means that under certain conditions NG2-glia generate GFAP+ astrocytes in ectopic positions. It will be interesting to determine when and how, e.g., which subtype of NG2-glia, generates this specific type of reactive astrocytes as well as if and how these astrocytes derived from NG2-glia differ from other astrocytes especially in regard to their contribution to scar formation. Indeed, reactive astrocytes derived from ependymal cell lining the spinal cord also perform specific and largely beneficial functions (242). In addition, the role of NG2-glia in neurite outgrowth and synapse formation has yet to be studied after injury, as these cells are covered by the CSPG4 that has been shown to block neurite outgrowth (for review, see Ref. 271). However, this protein also interacts with scaffolding proteins anchoring transmitter receptors, such as GluR2/3 subunits in the postsynaptic membrane (266), suggesting that it may also act as an important factor for synapse formation or function. Despite all the data published so far, much remains to be understood about the contribution of this novel class of abundant glial cells to the traumatic injury reaction.
C. Astrocyte Reaction to Injury in the Mammalian CNS
In contrast to the NG2-glia described above, astrocytes are well studied after brain injury and their functional contribution to the injury reaction has been probed by various cell ablation and gene deletion experiments (236, 259). Astrocytes serve as biosensors for almost any type of brain damage in neuropathology, based on their upregulation of intermediate filaments, such as GFAP, vimentin, nestin, and synemin; proteins that are normally expressed at low levels or entirely absent in adult astrocytes (for review see Refs. 209, 236, 259). Indeed, this is the most sensitive reaction of astrocytes that occurs already in virtually all astrocytes upon very small traumatic brain injury (8) and under all injury conditions examined so far (255). The profound upregulation of GFAP also led to the concept that astrocytes were the scar-forming cells as they increase in the vicinity of the injury site, given that only few GFAP+ cells are present prior to injury and many more thereafter. However, this increase may rather reflect upregulation of GFAP than an actual increase in astrocyte numbers as many, if not most astrocytes do not express GFAP in many regions of the healthy brain. Indeed, when monitored by immunohistochemistry for proteins also present in astrocytes prior to injury (254), such as GS, S100β, or Aldh1l1, by genetic fate mapping (35) or live imaging (8), the total increase in astrocyte number is only 10–20% in the adult mouse cerebral cortex even after large stab wound injury. Importantly, chronic live imaging of astrocytes after stab wound injury revealed that astrocyte numbers increase solely by proliferation as reactive astrocytes did not migrate to the injury site in vivo (Figure 4; Refs. 8, 187). Importantly, also the proliferative reaction of astrocytes is rather limited with only few astrocytes (10–20%) dividing a single time (8, 187). This occurs in a rather delayed manner with proliferation starting only at 3 days post injury (dpi) and peaking between 5 and 7 dpi (see also Ref. 254). Moreover, the few dividing astrocytes reside in a special niche, as their soma is in direct contact with the BM surrounding blood vessels, suggestive of highly specific functions at this interface (8). Thus a special subset of juxtavascular astrocytes divides and increases astrocyte numbers largely in direct apposition to the vasculature, while no astrocytes migrate towards the site of GM injury in the cerebral cortex (for spinal cord however, see Ref. 242). In contrast to the proliferating astrocytes showing a juxtavascular localization, proliferating NG2-glia do not show any preference in their localization (Dimou, unpublished observations). This is of interest as after injury vascular endothelial growth factors (VEGF) released by endothelial cells have been linked with the proliferation and migration of NG2-glia. Indeed, it has been shown in vitro that the behavior of postnatal OPCs can be influenced by VEGFs, with VEGF-A to promote migration and VEGF-C to enhance proliferation of OPCs (96, 150).
These data challenge the concept that astrocytes would considerably increase in number around the site of injury and form the scar, even though they may of course contribute by indirect means, e.g., via signaling mechanisms attracting other cells. In support of this, a small reduction in astrocyte numbers at the injury site, due to a decrease in their proliferation (8), is already sufficient to increase the number of microglial cells in the injury site (235). Indeed, higher numbers of microglia and macrophages were also observed upon genetic ablation of proliferating astrocytes in various injury models using mice expressing thymidine kinase under the GFAP promoter (for review, see Ref. 259). In these mice, proliferating astrocytes will be selectively killed when ganciclovir is provided and is converted to an agent toxic for proliferating cells by the enzyme selectively present in reactive astrocytes. Intriguingly, this led to a significant increase in infiltrating macrophages and inflammatory processes, as well as a larger injury size and later blood-brain barrier (BBB) closure (259). All of these effects are reminiscent of the location of the proliferating astrocytes at juxtavascular positions that are well positioned to regulate macrophage invasion and inflammation as well as BBB closure. Interestingly, recent work highlighted different waves and sources of invading macrophages. An earlier wave of the classically activated macrophages (M1) performs more adverse roles, while a later wave of invading M2 macrophages with acquired deactivation or alternative activation exert more beneficial effects (182, 248). While the later wave derives largely from the choroid plexus, the earlier wave is recruited from the circulation (248), i.e., at the interphase where the proliferating astrocytes are located (8). Notably, other means to reduce astrocyte proliferation after injury, such as by deleting STAT3 (100, 202, 288), cdc42 (235), or RAS (242), also result in increased macrophage and/or microglia numbers, scar formation, and lesion size. Interestingly, in the spinal cord, a specific subpopulation of reactive astrocytes has been observed to originate from the cells lining the ventricle which has potent beneficial effects on injury size and compaction (242). Thus specific subsets of astrocytes perform apparently distinct functions after brain and spinal cord injury.
These considerations highlight two important new concepts. One is to which extent the subsets of astrocytes observed to exert distinct behaviors after cerebral cortex injury may differ intrinsically due to a distinct origin. Such as the tanycyte/ependymal-derived reactive astrocytes in the spinal cord differ from other parenchymal astrocytes (242), also the juxtavascular astrocytes that constitute only a third of all astrocytes but are the most proliferative subset after stab wound injury in the cerebral cortex (8) may have a distinct developmental origin as juxtavascular astrocytes have been observed in distinct clones during early postnatal development (76). This prompts the concept that, e.g., also the further heterogeneity in the astrocyte reaction as observed by live imaging with some subsets not reacting at all, even though located right next to the injury site (8) may have a distinct developmental origin differing from, e.g., the subset of astrocytes forming long polarized processes towards the injury site, the so-called palisading astrocytes (200, 288). Alternatively or in addition, the local niche to which these subtypes are exposed, like the close vicinity of juxtavascular astrocytes to blood vessels, may expose them to specific signals influencing their subtype-specific behavior [See also Martín-López et al. (169a) and note added in proof]. In this regard it is interesting that both astrocyte subtypes with NSC in vitro potential identified in the cerebral cortex (255) and the spinal cord are (242) exposed to specific niches reminiscent of the environment of NSCs in the neurogenic zones, such as the vascular niche (255) and the CSF (242).
The second and most important new concept is the functional relevance of astrocyte's heterogeneity in reaction to injury. If each of these astrocyte subsets serves a specific functional role, identifying the signals or cell types regulating this subset may allow selectively influencing specific types of reactive astrocyte functions and not others. Given that reactive astrocytes with NSC properties seemingly exert beneficial functions as discussed above (235, 242, 255), the CSF composition and vascular niche signals may provide prime candidates to foster these beneficial populations selectively. Interestingly, specific sets of pericytes have been recently implicated in formation of the fibrotic scar (86), prompting ideas about reciprocal signaling between pericytes and juxtavascular astrocytes such that the latter could interfere with pericyte infiltration and fibrotic scar formation in the brain and conversely pericyte subtypes may inhibit juxtavascular astrocyte proliferation. Thus the vascular niche as an important regulator of adult NSC proliferation in the SEZ and DG (for recent review, see Ref. 85) also holds promises for mediating key properties for subsets of reactive astrocytes including activation of their NSC-like properties.
Indeed, reactive astrocytes reactivate several hallmarks of radial glial cells that have been downregulated during their differentiation, such as the intermediate filament nestin, the extracellular matrix component tenascin-C, or the DSD1-proteoglycan, all of which are normally absent in mature astrocytes, but present in the radial glial cells of the developing and adult brain (209, 236, 259). Moreover, some astrocytes proceed from a quiescent or postmitotic state to a proliferative state as described above. However, despite these changes, most genetically traced astrocytes remain in their lineage in vivo (10, 35). However, when the potential of these cells is probed by the neurosphere stem cell assay mentioned above, there is a clear and significant increase in in vitro multipotent and self-renewing NSCs derived from genetically fate mapped cerebral cortex GM astrocytes (35, 255, 256) or tanycytes/ependymal cells lining the spinal cord ventricle (10, 68). While this is only a very small fraction of reactive astrocytes in the cerebral cortex and ventricular cells in the spinal cord (68), neurosphere formation from the stab wound injured or ischemic cerebral cortex reaches about 1/5 to 1/10 of the NSC frequency in the adult SEZ where endogenous actively proliferating stem cells reside (255). Moreover, the appearance of these reactive astrocytes with NSC potential correlates with the degree of proliferation in vivo, with a peak at 5 dpi in stab wound injury and a significant decline between 3 and 6 mo of age (255). Indeed, SHH entering the brain from plasma and CSF after traumatic injury has recently been identified as a key factor regulating both astrocyte proliferation and NSC potential (255). SHH acts directly on astrocytes, as conditional deletion of the receptor smoothened only in astrocytes potently reduced both their proliferation as well as their neurosphere formation (255). In summary, these data highlight different sets of glial cells, juxtavascular astrocytes in the cerebral cortex (255) and ventricular tanycytes in the spinal cord (242), with the capacity to reactivate NSC potential after invasive injury. This is important as these cells with NSC hallmarks are local to the injury site and hence equipped with the adequate region-specific identity, thereby providing exciting sources for improving the injury outcome in various aspects.
D. Different Injuries Lead to Different Glial Reactions in the Tissue
While we focused above on traumatic injury paradigms, it is important to consider the profound difference of glial reaction in distinct injuries, especially as there is a most profound effect on the proliferative response of reactive glia. As discussed above, glial cells begin to proliferate in vivo in response to various insults such as ischemia, trauma, and models of Alzheimer's or Parkinson's disease (19, 35, 66, 254, 255, 298), but there are profound differences amongst the different types of glial cells that proliferate. While microglial cells readily proliferate in virtually all injury conditions so far examined, the proliferation of NG2-glia and astrocytes appears to be triggered largely by lesion conditions disrupting the tissue and BBB. Most strikingly, ablation of virtually half of all neurons in the adult mouse cerebral cortex by inducible expression of p25 in neurons fails to elicit a proliferative response from either astrocytes or NG2-glia (255). This happens despite an otherwise profound reactive gliosis with microglia activation and proliferation, astroglia upregulating GFAP, vimentin, nestin, and many other hallmarks of reactive astrocytes as well as becoming hypertrophic together with NG2-glia (255). However, literally no reactive astrocyte proliferates and the NG2-glia only continue to divide at their slow baseline rate despite the death of so many neurons and the clearance of their debris by highly proliferative activated microglia.
Indeed, also other conditions eliciting the typical hallmarks of astrogliosis (hypertrophy, GFAP upregulation, etc.) but failing to activate their proliferation have been described, such as amyloid plaque deposition in mouse models of Alzheimer's disease (19, 118, 255). A similar reaction is observed in inflammatory conditions induced, e.g., by lipopolysacharide injection mimicking bacterial infection with astrocytes failing to activate proliferation (257). Interestingly, in both these conditions, NG2-glia react with increased proliferation (19, 257), demonstrating that the signals regulating the proliferative response of reactive astrocytes and NG2-glia are at least in part distinct. Conversely, chronic inflammatory conditions, such as experimental autoimmune encephalomyelitis (EAE), activate both astrocyte and NG2-glia proliferation. Interestingly, in EAE conditions, the BBB also becomes permeable for cells of the innate and adaptive immune system (27, 258). This raises the important point of cellular invasion into the brain beyond diffusible signals such as SHH that can enter in case of BBB disruption and is sufficient to elicit pronounced reactive astrogliosis without any signs of neurodegeneration or inflammation (246). Cells invading either via the CSF, the meninges, or the blood vessels act as the source for distinct chemokine signals that activate specific parenchymal glial cells and exert distinct effects on neuronal survival (see below). Taken together, the type of injury, the signals, and the cells recruited into the brain are crucial for eliciting and regulating the proliferative response of astrocytes and NG2-glia. Their proliferation is distinctly regulated in specific contexts only, while proliferation of microglia is activated in a broader spectrum of injury paradigms. These selective signals regulating, e.g., NG2-glia but not astrocyte proliferation in some disease settings allow a novel entry point for highly innovative therapeutic approaches. For example, if NG2-glial cell accumulation around the injury site contributes to scar formation, it may be beneficial to selectively inhibit their proliferation. On the other hand, blockage of NG2-glia proliferation could at the same time be detrimental as these cells generate new oligodendrocytes, that are responsible for the remyelinating capacity of the tissue. Taking into consideration these both faces of the medal, the best approach would be to selectively inhibit proliferation and promote differentiation of NG2-glia. Importantly, selective influences on distinct glial subsets no longer apply only to the main glial subtypes, but also to their respective heterogeneous subpopulations that may further help to delineate beneficial from adverse effects.
V. REPAIR AND REGENERATION: DISTINCT CONTRIBUTIONS OF REACTIVE GLIAL SUBSETS
A. Wound Closure and Extracellular Homeostasis
The amazing specificity and heterogeneity in the glial cell behavior after brain injury described above raises the importance of distinct functional contributions made by these specific glial subtypes with the final aim of promoting specific beneficial subpopulations and reactions, e.g., those promoting wound healing, while inhibiting the adverse roles, such as scar formation.
Indeed, proof of concept for distinct functional contributions of cellular subsets has been validated for the microglia and macrophage reactions with beneficial and adverse signaling pathways and subtypes being more and more clearly discernible. For example, microglia activation can either favor the release of neuroprotective factors or activate phagocytotic and cell death-promoting subtypes and signaling pathways (95, 127, 248). Likewise, two major types of macrophages invading the CNS after injury have been identified as discussed above (248) which influence not only neuronal survival but also remyelination and differentiation of NG2-glia into mature oligodendrocytes in demyelinating conditions (182 and below). These are important new insights to open novel avenues of limiting the responses promoting adverse inflammatory signals and cell death while promoting anti-inflammatory and prosurvival signals after injury. Likewise, distinct functions seem to emerge for the proliferative NG2-glia and astrocytes.
1. Fast NG2-glia reaction and function
NG2-glia ensheath a wound in the mouse cerebral cortex very fast with their processes and NG2-glia numbers increase relatively fast within 2–4 days by fast reentry into the cell cycle (110, 254). This leads to a considerable increase in their number surrounding the injury site in the cerebral cortex GM. Similarly, an accumulation of NG2-glia surrounding the fibrotic scar has been observed in the injured spinal cord, implicating these glial cells into functions previously allocated to reactive astrocytes: wound closure, inhibition of neurite outgrowth by inhibitory chondroitin proteoglycan presentation, and scar formation by creating a cell dense region enriched in glial cells.
2. Astrocyte reaction & function at early stages after injury
Also astrocytes show a fast reaction but neither by directing processes towards the injury site nor by proliferation, which are both reactions occurring at later stages. Rather, the very first reactions of astrocytes to injury consist of cellular hypertrophy and gene expression changes (301). These imply upregulation of the phagocytic protein machinery implying astrocytes in helping to engulf damaged cell processes or released proteins as well as helping to reinstall the extracellular milieu, e.g., by upregulation of glutamate transporters. Importantly, the most obvious and well-examined reaction of reactive astrocytes, the intermediate filament upregulation, plays a role in this context promoting delivery of proteins to the cell surface (154, 236). This has been demonstrated by lower glutamate transport levels in mice lacking the intermediate filament proteins GFAP and vimentin (154). Thus one may speculate that the early reaction of astrocytes common to all injury conditions (hypertrophy and IF upregulation) largely encompasses functions involved in removing aberrant cell debris, ions, amino acids, etc., and reinstalling ion and protein homeostasis, possibly an ancient function of glial cells (see chapt. 5 in Ref. 129). Indeed, this is also a role exerted by astrocytes in noninvasive injury conditions, such as amyloid plaque deposition where they contribute to reduce the plaque load (214). Interestingly, this function is also severely impaired when GFAP and vimentin are genetically deleted and cannot be upregulated by reactive astrocytes (138). Several other fast contributions of astrocyte activation have been suggested after inflammatory and ischemic stimuli (301), including inhibition of bacterial growth, peptidase inhibitors etc.
3. Reactive astrocyte functions at later stages
In regard to the focus of this review on glial progenitors, it is important to note that genome-wide expression analysis of reactive astrocytes found genes promoting proliferation activated with some delay in reactive astrocytes with a peak between 3–5 days (301). This is in full agreement with the observations made by BrdU or Ki67 immunostaining and live imaging that show a peak of astrocyte proliferation in this time window (8, 35, 254, 255, 288). As mentioned above, this is also the time window when most neurosphere-forming astrocytes are present after stab wound injury. Given the predominant location of proliferating astrocytes at juxtavascular positions in the vascular niche, this is suggestive of a role at this interphase, either regulating vascular remodeling, reestablishment of certain BBB properties or invasion of blood-born cells, such as lymphocytes or macrophages as discussed above. Most importantly, however, the discovery of subsets of astrocytes in unique locations undergoing specific changes (e.g., cell proliferation) not only allows to better nail down their exact function but also to delineate separate subsets of reactive astrocytes performing distinct functions (242). Thus the approach to modulate the outcome of brain injury by improving wound healing, reinstalling the extracellular milieu, and promoting neuronal cell survival and process outgrowth has become more realistic by the delineation of distinct glial subsets proliferating at different times and supposedly devoted to very different roles. These new insights from basic research open new therapeutic avenues by modulating specific glial subtypes and functions.
B. Contribution of Reactive Glial Progenitors to Neurite/Axonal Regeneration
In addition to helping neurons to survive an insult, also their processes need to survive and reconnect beyond or around the site of damage. During development, newly generated neurons are able to grow an axon while mature neurons after injury in the mammalian CNS are incapable of self-repair and regeneration. When an axon is damaged, the distal segment undergoes Wallerian degeneration (48, 123) and demyelination, while the proximal segment dies by apoptosis or undergoes chromatolysis as an attempt for repair, often by forming retraction bulbs, plastic structures that never extend beyond the injury site. Unfortunately, and in contrast to PNS injury, injury in the CNS is not followed by extensive regeneration, in particular when exposed to a fibrotic scar and the extracellular matrix factors that altogether create an unfavorable environment for axonal regeneration, counteracting the repair of axons and myelin. To exclude a limited intrinsic ability of mature neurons to grow an axon after injury, permissive PNS bridges (e.g., sciatic nerve grafts or Schwann cells; Ref. 21) have been implanted in an injured spinal cord showing the regrowth of axons and pointing to a regeneration inhibition solely by the environment. This inhibitory environment is comprised of the glial scar and many extracellular matrix inhibitory molecules like chondroitin sulfate proteoglycans (for review, see Refs. 4, 247) as well as inhibitory molecules expressed by myelin such as Nogo-A, MAG, OMgp, ephrins, etc. (24, 46, 62, 287) or NG2-glia such as CSPG4 (270, 271) or the astrocytic and/or fibrotic scar (86, 97). Genetic deletion or blocking of several inhibitory molecules showed an improvement in axonal regeneration, but this was never sufficient to allow the majority of injured axons to regenerate, pointing to a rather complex inhibitory machinery elicited by the glial cells of the CNS. Interestingly, most of these inhibitory molecules are localized in mature, differentiated glial cells demonstrating that one aspect of maturation of glial cells from a progenitor state is also to downregulate neurite outgrowth-promoting aspects. This idea could be endorsed after transplantation of glial progenitor cells, the GRPs, which exerted a beneficial effect on axonal regeneration after spinal cord injury (3, 101), suggestive of additional roles of progenitor cells in instructing a more plastic environment (as discussed above).
Notably, the relatively low extent of regeneration cannot explain the level of recovery in function as observed in rodents or primates after spinal cord injury. This could be the result of newly generated multisynaptic connections between cortical and spinal motor neurons that could restore function. Interestingly, following spinal cord injury, new axonal branches (sprouts) extend from corticospinal axons above the lesion site or from the intact contralateral side that establish contacts with interneurons forming a relay circuit restoring the input to segments below the injury site (9, 273). These results suggest that rearrangements in the spared circuitry provide an indirect mechanism to achieve endogenous repair to some extent. Importantly, glial progenitors, in particular NG2-glia, may not only help to remyelinate damaged axons, but also allow sprouting and plastic changes by remodelling myelin as discussed above.
C. Remyelination and Axonal Survival
NG2-glia play a prime role in conditions resulting in the destruction of myelin (demyelination) where they mediate successful remyelination providing a good example of regeneration also in the adult mammalian CNS. The most common demyelinating disease in human is multiple sclerosis (MS), which is believed to occur due to a chronic inflammatory response that causes focal plaques in the brain and spinal cord (148) as the result of an autoimmune response towards myelin proteins (108) leading to the death of oligodendrocytes. Also after spinal cord injury, one of the pathological outcomes is the demyelination of the axons at the lesion site due to oligodendrocyte death through necrosis or apoptosis (65, 277). Although these two demyelination paradigms are very different, the reaction of NG2-glia being recruited to the site of injury to remyelinate denuded axons is common (81, 122, 176). Watanabe et al. (290) could show that endogenous NG2-glia have the ability to proliferate and differentiate in and around a chemically induced demyelination lesion in the spinal cord, and similar results could be obtained after transplantation of the very same cells into a demyelination lesion or into myelin-deficient mice (91, 304). Interestingly, also after spinal cord injury, NG2-glia strongly proliferate and produce a high number of new oligodendrocytes (111, 176). In general, one can conclude that endogenous NG2-glia proliferate after demyelination or injury to enhance remyelination. However, this new myelin formation following demyelination is not normal, with thinner and not so compacted myelin sheaths, abnormal g-ratio (ratio of axon circumference to myelin circumference), and shorter internode length (211; for review, see Ref. 73). Although remyelination is not optimal, it is still beneficial by helping to protect the axons from further damage or even degeneration and leads to an increased conductance velocity even if it never reaches physiological levels (147). Thus, even in this example of successful regeneration, there remains ample room for improvement.
A better understanding of the signals regulating NG2-glia proliferation, migration, and differentiation provides an obvious entry point to improve full myelination. In this regard, it is important to remember the profound regional heterogeneity in myelination capacity between WM and GM derived NG2-glia (282). For example, if many NG2-glia surrounding the demyelination foci migrated from the GM, the cells that are not intrinsically very potent to myelinate, then myelination cannot proceed to completion and the NG2-glia may well fail to repair the damaged myelin sheaths in a satisfactory manner, prompting the question which could be the factors overcoming this limitation and/or recruit NG2-glia from the WM. Possible additional explanations may be an inhibitory environment at the site of demyelination that does not promote differentiation or the lack of growth factors and signals that would lead to the complete remyelination. An additional pitfall is that although remyelination in acute demyelinating lesions is pretty effective, this is not the case in chronically demyelinated lesions. Again, different hypotheses have been suggested for the limited remyelination after chronic demyelination: 1) depletion of progenitors by time (173), 2) axon damage due to the chronic demyelination (31, 227), and/or 3) age-dependent block of NG2-glia proliferation and differentiation (253). The latter idea of the age-associated decrease in remyelination efficiency is associated with an impairment of NG2-glia recruitment as well as their limited differentiation into remyelinating oligodendrocytes (253). These age-dependent changes could be due to the different expression of growth factors influencing the behavior of NG2-glia and to the changes in the inflammatory processes associated with demyelination (72). Further studies are needed to resolve the inability of NG2-glia to efficiently repair myelin after demyelination that could then provide possible therapeutic approaches.
Notably, in addition to the NG2-glia, also Schwann cells, the myelin-forming cells of the PNS, have been suggested as a source of remyelinating cells (11, 71, 155). A recent study could show that after a chemically induced demyelination injury Schwann cells do not migrate from the PNS, as speculated before, but are directly generated by PDGFRα+ NG2-glia resident around the injury site (303), further supporting the important role of NG2-glia for the remyelinating properties of the CNS.
D. Neuronal Replacement From Glial Stem and Progenitor Cells
Glial cells and their progenitors play multiple roles helping to alleviate neurological defects after brain injury, not the least by promoting axonal and neuronal function and thereby fostering neuronal survival. However, once neurons have degenerated and been lost, could any of the above glial progenitors actually be used to replace the lost neurons? And if so, can glial progenitors help to allow the regenerated neurons connecting to their appropriate target sites?
As discussed above, reactive glial cells by and large do not generate new neurons endogenously in vivo in the injured brain. This is the case for both NG2-glia as well as the proliferating reactive astrocytes that all largely remain within their respective lineage in vivo (35, 61, 118, 236). However, they provide a pool of stem and progenitor cells that are local to the injury site and could hence be instructed towards neurogenesis. Indeed, genome-wide expression analysis of adult NSCs and astrocytes isolated from various forebrain regions showed that adult NSCs are more similar to mature astrocytes and ependymal cells from neighboring regions than to neurogenic radial glial cells from the developing brain (16). Thus cells with a predominant astroglial gene expression profile can act as neurogenic NSCs. The crucial differences are that the NSCs within the neurogenic niches are primed towards neurogenesis by expressing already, albeit at low levels, key neurogenic fate determinants, such as Pax6, Dlx2, Sox11, Sox4, and others (16, 191). Accordingly, upregulation of such factors in reactive astrocytes that resumed proliferation already should be sufficient to instruct them towards neurogenesis. This was first shown by forced expression of the transcription factor Pax6, which surprisingly succeeded to instruct some proliferating glial cells towards neurogenesis within the stab wound injured cerebral cortex (36) or the ischemic striatum (141). Moreover, factors efficient in converting even fibroblasts into neurons also proved successful in the striatum in vivo when targeted to astrocytes in the healthy brain (196, 276). Likewise, transcription factors of the proneural basic helix-loop-helix family Neurog2 or Ascl1 are sufficient to instruct local glial progenitor cells (most likely NG2-glia as they are the only population proliferating in the uninjured CNS) towards neurogenesis in the spinal cord and cerebral cortex (88, 201). Combination with factors promoting proliferation and dedifferentiation, such as EGF and FGF2, further potentiates this response both in the spinal cord and the ischemic or stab wound injured forebrain (88, 201), supporting the concept that proliferation and neurogenic fate determinants are key in instructing neurogenesis from local glial stem and progenitor cells. Thus proliferating glial cells within the injury site provide a local source of cells for repair including the replacement of neurons.
This is particularly exciting given the regionalization of glial cells, in particular the astrocytes as described above. Proliferating reactive astrocytes may hence be easier to instruct towards the neuronal subtype appropriate for the respective brain region compared with other NSCs derived from the SEZ or the SGZ that are specialized to generate different types of neurons, namely, olfactory bulb interneurons or dentate gyrus granule cells. Moreover, glial cells may hold the key to overcome also the other major problem still faced in such attempts to replace degenerated neurons, namely, the long-term survival of these newly regenerated neurons. Especially after brain injury, the number of the originally elicited neurons profoundly decreases over weeks with few of them surviving (36, 88). But is longer survival and successful integration of new neurons into the adult brain parenchyma at all possible? The answer is yes, as shown when an environment beneficial for survival and integration of new neurons is provided. This is the case when neurons die in an apoptotic manner and little reactive gliosis is elicited (45, 163). In such conditions, both endogenously recruited cells or neuroblasts isolated from the same region (the developing cerebral cortex) and transplanted after ablation of, e.g., corticospinal neurons, readily differentiate into mature neurons that send their axons for long distances throughout the entire brain even to the spinal cord (45). Indeed, young neurons can extend axonal processes also in the adult brain and white matter as they are not repelled, e.g., by the growth inhibitory components of myelin (52). These neurons survive for many months, clearly demonstrating that young neurons can reconnect even over long distances and integrate and survive in an adult brain, if the environment is sufficiently supportive. However, in the context of severe brain injury with reactive gliosis, this has not yet been achieved. Thus a key task for the future is to understand which glial subtypes create an adverse environment for survival of new neurons and best utilize these to turn them into neurons by neurogenic fate determinants. Taken together, glial stem and progenitor cells at the injury site not only provide a local progenitor source amenable to fate conversion towards neurogenesis allowing replacement of lost neurons, but also hold the key to provide a favorable environment for neuronal survival (see also Ref. 242) and successful integration. And this key lies in understanding and utilizing the diversity of glial subtypes.
VI. OUTLOOK: DIVERSITY OF GLIAL CELLS
Given the above exciting new avenues for repair, it is now important to unravel which subsets of reactive glia are closest to neurogenesis and can be best turned into the appropriate neuronal subtype. Considering the similarity between adult NSCs and astrocytes or radial glia and tanycytes, these may be the proliferating subset of reactive astrocytes or cells lining the ventricle. Moreover, these cells retain their regional specification. However, utilizing these cells for neuronal replacement further supports the need to understand their endogenous role after brain injury, e.g., in wound or BBB repair or helping neuronal survival. Given that they indeed exert a beneficial role, one strategy could be to amplify these proliferative reactive glia with NSC potential first, such that sufficient cells are still available even after turning some into new neurons.
An alternative strategy is to turn reactive glial subtypes with adverse and scar forming functions into new neurons, thereby improving the environment for neuronal survival and integration and eliciting neurogenesis locally. To which extent NG2-glia may be this cell type remains to be experimentally tested. If the early reacting NG2-glia perform important functions in wound closure, a strategy could be to target them with neurogenic factors with some delay first allowing them to mediate wound closure and then turning the additional NG2-glia that are anyhow subject to later reduction into neurons. Importantly, the enormous capacity of NG2-glia for homeostasis will most likely lead to a repopulation of the brain even after turning a considerable fraction of these cells into a different cell type, in this case neurons. Thus a sufficient number of NG2-glia will be again available to fulfil their normal functions in the healthy brain, such as activity-dependent communication and providing newly generated oligodendrocytes to myelinate the axons of regenerated long-distance projection neurons.
Taken together, the new insights into glial cell diversity now open new avenues towards best utilizing the regional, temporal, and subtype diversity of glial cells for beneficial influences and repair purposes, but much still remains to be understood about this diversity to put it to its best use.
No conflicts of interest, financial or otherwise, are declared by the authors.
NOTE ADDED IN PROOF
Address for reprint requests and other correspondence: M. Götz, Physiological Genomics, LMU, Pettenkoferstr. 12, 80336 Munich, Germany (e-mail:).
- Copyright © 2014 the American Physiological Society