Constitutive neurogenesis from glial cells in the vertebrate nervous system

In mouse and rats, the two prominent niches of constitutive neurogenesis are the subependymal zone of the lateral ventricle (SEZ) and the subgranular zone of the dentate gyrus of the hippocampus (SGZ) (Fig. 1). The latter niche is the most active in humans (Bergmann et al., 2015), although a recent study identified new neurons that are constitutively added to the human striatum from a source yet to be defined (Ernst et al., 2014). Both the SEZ and SGZ contain a large heterogeneous population of endogenous active progenitor cells, which is dominated by astroglial cells, a type of non-neuronal cell important for homeostasis and support within neural tissues (recently reviewed by Bond et al., 2015; Kempermann et al., 2015). Although the precise hierarchical relationship operating within rodent germinal niches is still being teased apart (Bonaguidi et al., 2012), genetic-tracing studies support the idea that the glial cells are NSCs, and generate neurons via a series of intermediate, amplifying and non-glial cell states (Fig. 2). Compared with mammals, teleost fish such as zebrafish and medaka show significantly greater rates of neurogenesis, as new neurons are generated in most brain regions throughout adult life (reviewed by Kizil et al., 2012b; Schmidt et al., 2013). The dorsal telencephalon, otherwise known as the pallium, is an area of the teleost fish brain that contains regions homologous to the mammalian SEZ and SGZ as well as a neocortex-like region, where in rodents neural progenitors are silent. In the teleost pallium, neurogenic radial glial cells act as self-renewing and multipotent progenitors at the single-cell level, behaving as bona fide NSCs (Rothenaigner et al., 2011; reviewed by Than-Trong and Bally-Cuif, 2015). But not all cells are equal: a recent lineage-tracing study (Box 1) revealed differences in NSC densities and activation frequencies across the anterior, medial and lateral pallium under normal physiological conditions. This regional diversity in NSC activity is important to consider, especially when analyzing regeneration (Dray et al., 2015). In addition to NSCs, non-glial cycling neuroblasts, postulated equivalents of mammalian transit-amplifying progenitors, are intermingled along the zebrafish ventricle (März et al., 2010) (Fig. 2). It has been shown that the neurons generated from the pallial neurogenic zone populate the olfactory bulb and the pallium proper (Kroehne et al., 2011; Dirian et al., 2014); however, a more detailed spatiotemporal characterization of the heterogeneity of this neurogenic niche in terms of cell subtypes, lineages, division modes and fate has not yet been reported. In the adult pallium of both the red spotted newt (Notophthalmus viridescens) and the axolotl, ventricular radial glial cells also exhibit neurogenic potential and the capacity to retain a bromodeoxyuridine (BrdU) label – indicative of slow cell cycling – under physiological conditions (Berg et al., 2010; Maden et al., 2013). Radial glial heterogeneity has also been described in the newt (Kirkham et al., 2014), but the exact lineage relationship between these cell types and their relation to zebrafish or mammalian adult progenitors remains to be defined.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Neuronal repair from niche progenitors. In the rodent subependymal zone (SEZ), glial cells give rise to neuroblasts that migrate along the rostral migratory stream into the olfactory bulb (OB) to generate unique types of interneurons (left panel, blue arrows). Stroke injury (red outline) results in localized cell death in the striatum and the proliferation of endogenous neural progenitor cells that migrate from the SEZ to the striatum to elicit regeneration (right panel, red arrows). This migration occurs at the expense of normal neuroblast migration from the SEZ to OB. In the rodent dentate gyrus (DG), radial glial cells produce transit-amplifying progenitors, called neuroblasts, which generate neurons (left panel). These newborn neurons migrate into the granule cell layer (blue arrows). Ischemia (center panel, red outline) induces the degeneration of pyramidal neurons. Following the ischemia the endogenous progenitors proliferate and subsequently migrate to regenerate new neurons (right panel, red arrows). The ventricular zone of the adult zebrafish pallium consists predominantly of radial glial cells, which act as self-renewing and multipotent progenitors (left panel). In addition, non-glial cycling neuroblasts are intermingled along the ventricle. Together, radial glia and neuroblasts generate pallial neurons (left panel, blue arrows). Reactive neurogenesis has been induced in the zebrafish adult pallium mostly by mechanical injury using stab lesion causing a circumscribed injury in the parenchyma of the telencephalon without injuring ventricular lining (center panel, red outline). In response, neuroblasts and radial glia increase their proliferation to produce neurons to compensate for the neuronal loss (right panel, red and green arrows, respectively). In the amphibian retina, the ciliary marginal zone (CMZ) continuously generates all neuronal subtypes (left panel, blue arrows). Upon extensive lesion in X. tropicalis (center panel, red outline), the CMZ is activated to elicit regeneration (right panel, red arrows). In the rodent and zebrafish schematics, only the left hemisphere is depicted.

NSCs in all species are relatively quiescent compared with most other dividing cells: a mechanism that helps to protect NSCs from exhaustion. However, recruiting endogenous NSCs for repair will, in part, necessitate an exit from quiescence, and thus the regulation of quiescence is a topic of great interest for neural repair. Molecular analyses in zebrafish have identified Notch3 signaling as a key pathway that maintains radial glial quiescence (Alunni et al., 2013). Notch signaling maintains NSC quiescence in constitutive niches (SEZ and SGZ) of the adult mouse as well (Imayoshi et al., 2010), although it is not yet clear which Notch receptor is involved. In the newt, blocking systemic Notch signaling has been shown to lead to an increased number of proliferating pallial radial glial cells (Kirkham et al., 2014). In both mouse and zebrafish, however, Notch1 signaling is necessary for the maintenance of activated (proliferating) NSCs, controlling either cell division or ‘stemness’ (Pierfelice et al., 2011; Alunni et al., 2013). Other molecular components of the NSC quiescence cascade identified in zebrafish include the transcription factors Id1 and Fezf2 (Berberoglu et al., 2014; Rodriguez Viales et al., 2015). Both factors are also expressed in adult mouse NSCs, and were specifically associated with increased quiescence (Nam and Benezra, 2009), although their functional role in quiescence control remains to be shown.

Recruiting niche glial progenitors for neuronal repair

The birth of new neurons via constitutive neurogenesis is not adequate to replenish the sudden loss of neurons that occurs following injury. Here, something greater is required: the recruitment of endogenous glial progenitors to undergo reactive neurogenesis. The zebrafish adult pallium can undergo reactive neurogenesis remarkably well, replacing lost neurons efficiently in all cases of mechanical injury using stab lesions (Ayari et al., 2010; Kroehne et al., 2011; März et al., 2011; Baumgart et al., 2012; Skaggs et al., 2014). The first response to this type of injury is immune cell activation: the number of microglia and leukocytes in the injured pallial hemisphere increases significantly for several days (Baumgart et al., 2012; Kyritsis et al., 2012). Next, ventricular cells are recruited to proliferate. Conditional Cre/lox lineage tracing in which radial glial cells and their progeny were permanently labeled demonstrated that radial glial cells give rise to neuroblasts that migrate to the site of injury, where they differentiate into long-lasting neurons (Kroehne et al., 2011). A recent lineage-tracing study (Box 1) showed how the division mode of NSCs partially switches upon mechanical lesion such that the proportion of symmetric neurogenic divisions, which is favorable to neuronal repair, increases. This is consistent with what is seen in the mouse SEZ (Ohab et al., 2006). These divisions consumed radial glial cells, generating either one cell maintaining gfap:gfp expression soon after division and one non-radial glial cell, or symmetric divisions in which two non-radial glial cells were produced (Barbosa et al., 2015). These studies show how radial glia from the endogenously active pallial NSC zone can be efficiently stimulated and re-routed towards brain repair (Fig. 2), a process that involves several distinct molecular pathways (Table 1).

To date, recruiting endogenous progenitors is also the most successful strategy to restore neuronal function in rodents. Stroke injury in the rodent striatum leads to increased proliferation in the SEZ, an increased proportion of neurogenic divisions, the redirection of cell migration towards the striatum, and neuronal differentiation into striatal medium spiny neurons (Ohab et al., 2006). Light-induced apoptosis induction in corticospinal projection neurons also triggers the re-routing of SEZ neuroblasts towards the cortex, which is accompanied by some functional regeneration (Chen et al., 2004) (Fig. 2). Finally, ischemia-induced apoptosis of hippocampal CA1 pyramidal neurons stimulates the activation of endogenous SGZ neural progenitors in the dentate gyrus and their subsequent migration into the CA1 layer, leading to functional recovery, albeit incomplete (Nakatomi et al., 2002) (Fig. 2). Although it is clear that recruitment of endogenous neural progenitors can, in some cases, lead to functional recovery, the process remains inefficient, and there is still much to learn about the mechanisms that enhance the mobilization of these cells for repair. How is the migration of these progenitors channeled towards injury sites and how is their fate reoriented towards producing neuronal subtypes that match those of the missing neurons? How do these progenitors overcome anti-neurogenic influences when settling within non-neurogenic areas? The coordination of neurogenesis and angiogenesis appears to be crucial, and some of the factors controlling neuroblast recruitment and migration, such as brain-derived neurotrophic factor (BDNF), have been recently identified (Grade et al., 2013).

Shared and divergent processes in glial cell-mediated vertebrate neuronal repair

A novel regenerative perspective: a possible role for constitutive neuro-epithelial niches in neuronal repair

Non-glial cells with neuroepithelial characteristics can be found in some regions of the adult non-mammalian central nervous system and also serve as adult NSC-like progenitors under physiological conditions (for a review, see Than-Trong and Bally-Cuif, 2015). In Xenopus as well as adult zebrafish and medaka, these progenitors reside in the ciliary marginal zone (CMZ), as well as at the margin of the optic tectum of the adult zebrafish and medaka brain (Alunni et al., 2010; Ito et al., 2010) and the lateral edge of the zebrafish pallium (Dirian et al., 2014). Although the properties of these neuroepithelial cells have not been extensively tested, they do appear to rely on different pathways of maintenance compared with radial glial cells (Dirian et al., 2014), and they have been shown to be involved in regeneration. For example, the CMZ appears as the main source for retinal regeneration in Xenopus tropicalis (Miyake and Araki, 2014) (Fig. 2). At present, evidence to support the maintenance and functional relevance of such cells in the adult mammalian central nervous system is scattered, but suggestive: cells expressing progenitor markers can be induced in human retinal explants in a location homologous to the CMZ (Bhatia et al., 2009); a Notch1-independent neuroepithelial territory was observed at the posterior edge of the mouse optic tectum (Lutolf et al., 2002); and, finally, neuroepithelial progenitors able to generate a subset of adult SGZ NSCs at late embryonic stages were recently identified in the mouse hippocampus, in a location possibly homologous to the zebrafish pallial neuroepithelial pool (Li et al., 2013). Future studies should address and compare the highly important issue of regeneration from adult neuroepithelial pools in the brain and eye of both mammalian and non-mammalian species.

Reactivation of latent progenitors for neuronal repair in the retina

The retina grows during the entire life of teleost fish by addition of new neurons, originating from retinal stem cells of the CMZ (Fig. 3). In addition, Müller glial cells, distributed all over the differentiated retina, divide very infrequently in the adult to produce new rod photoreceptors (Raymond et al., 2006) (Fig. 3). A key signaling pathway imposing Müller glia quiescence under physiological conditions is Notch (Conner et al., 2014). In lesional contexts, Müller glia are the source of retinal progenitor cells driving regeneration of retinal neurons (Bernardos and Raymond, 2006; Fausett and Goldman, 2006; Bernardos et al., 2007; Fimbel et al., 2007; Ramachandran et al., 2010b): they re-enter the cell cycle and divide asymmetrically to generate neurogenic clusters that then give rise to all the missing neurons (Nagashima et al., 2013). In this way, stimulated Müller glia cells display a much greater lineage repertoire than under homeostatic conditions, as they must produce an array of neuronal lineages that they do not usually make. Inhibition of the Müller glia cells during regeneration results in regenerative failure (Thummel et al., 2008) (Fig. 3).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Neuronal repair from latent progenitors. In the zebrafish retina, new retinal neurons are generated sequentially from retinal stem cells located in the ciliary marginal zone (CMZ). Under homeostatic conditions, the Müller glia cells (MGs) generate only rod precursors, which give rise to rod photoreceptors (left panel, green and blue arrows). Following lesion (center panel, red outline), MGs re-enter the cell cycle and divide once asymmetrically to generate neurogenic clusters that go on to produce all missing neurons (right panel, green arrows). In the rodent striatum, the astrocytes are not neurogenic (left panel). After a stroke (center panel, red outline), some striatal astrocytes generate neuroblasts that give rise to a limited number of new neurons (right panel, green arrows). In the newt midbrain under homeostatic conditions, ependymoglial cells are quiescent (left panel). A selective neurotoxin administered intraventricularly selectively eliminates midbrain dopaminergic neurons (center panel, red outline), inducing the proliferation of the ependymoglial cells, which generate new dopaminergic neurons (right panel, green arrows). Ependymoglial cells in zebrafish spinal cord are self-renewing and give rise to oligodendrocytes (left panel, red arrows). After a lesion (center panel, red outline), these cells divide, migrate and produce new motor neurons (right panel, green arrows).

The molecular cascades that drive Müller glia-mediated repair in zebrafish are certainly the most studied in the context of stimulating latent progenitors, and so deserve some detailed discussion in addition to recent reviews (Goldman, 2014; Lenkowski and Raymond, 2014). Several signaling pathways activated upon lesion stimulate Müller glia re-entry into the neural progenitor mode, including Tumor necrosis factor (Tnfα), which is produced by dying retinal neurons and further stimulates the production of other growth factors and cytokines at the injury site (Nelson et al., 2013). These factors converge onto the activation of the mitogen-activated protein kinase (MAPK) and phosphoinositide 3-kinase (PI3K) pathways, followed by activation of β-Catenin and Stat3. In parallel, two co-repressors, Tgif1 and Six3b, which are rapidly upregulated prior to the first Müller glia division, repress Tgfβ signaling in Müller glia through Smad2/3 to permit the proliferative response and limit gliosis, the non-specific pathological remodeling of glia cells in response to damage (Lenkowski et al., 2013; Lenkowski and Raymond, 2014). In addition, upregulation of another transcription factor, Ascl1, is a key event in Müller glia activation. One important downstream effector of Ascl1 is lin28, which contributes to the Müller glia response through a negative-feedback loop: it decreases let-7 miRNA levels, relieving repression of regeneration-associated mRNAs, including klf4, oct4 (pou5f3 – Zebrafish Information Network) and cmyc (myca – Zebrafish Information Network), which are components of well-known reprogramming cocktails (Ramachandran et al., 2010b). The Ascl1/lin28 pathway also results in induction of expression of Pax6, required upon repair for increased progenitor cell proliferation (Thummel et al., 2008; Rajaram et al., 2014). Finally, Ascl1 also induces expression of the transcription factor Insm1, which helps promote Müller glia re-entry into cycle (Ramachandran et al., 2011).

Apart from fish and urodeles, functional repair of damaged retinas has seldom been demonstrated in vertebrates. Rodent Müller glia proliferate and express genes associated with retinal stem cells in response to injury, but generally do not themselves function as retinal progenitors in vivo (Jadhav et al., 2009). Rather, they have been shown to undergo gliosis (Dyer and Cepko, 2000; Bringmann et al., 2009). Following retinal injury, mouse Müller glia do not express Ascl1; however, when forced to overexpress Ascl1 in culture they re-express progenitor genes, including Insm1, and re-enter the cell cycle to generate amplifying neuronal progenitors and neurons (Pollak et al., 2013). A recent report further demonstrates that forced expression of Ascl1 in Müller glia upon lesion in young mice is sufficient to drive them towards efficient neurogenesis (Ueki et al., 2015). Thus, a major difference in the regulation of Ascl1 expression in Müller glia upon lesion might underlie the difference in success of retinal regeneration between mouse and zebrafish.

Several important issues are raised by these data. First is the question of whether the lesion-induced molecular cascades in zebrafish Müller glia are truly lesion specific, or whether certain genes are expressed at low levels under basal conditions. Some Ascl1-induced pluripotency factors are endogenously expressed at low levels in zebrafish Müller glia (Ramachandran et al., 2012) and their promoters are hypomethylated even in the absence of lesion (Powell et al., 2013), whereas the promoters of progenitor genes are epigenetically silenced in mouse Müller glia (Pollak et al., 2013). A detailed comparison of the basal status of zebrafish versus mammalian Müller glia, or mammalian astrocytes, may bring to light some molecular targets for the recruitment of latent mammalian progenitors. Another issue is understanding how reactive zebrafish Müller glia are able to adjust their response to regenerate the appropriate number and subtypes of neurons. A possible hypothesis here is that, to permit proper neuronal patterning following reprogramming, Ascl1 induction must be transient, which may be permitted by negative-feedback loops involving Insm1 and Ascl1 (Ramachandran et al., 2012). If this is the case, then it might be possible to reconstruct such a molecular loop in mammalian Müller cells in order to promote retinal regeneration over gliosis. Another issue that must be addressed is the role of Tnfα in promoting progenitor survival (Conner et al., 2014), and the additional pathways and components that are involved, independent of Notch inhibition. Finally, an important aspect of the reparative reaction in the retina is also to limit it. This is crucial to prevent stem cell exhaustion, overgrowth and the development of tumors, and/or the establishment of a long-lasting gliosis reaction. In zebrafish, a second peak of Notch-dependent insm1 expression contributes to stopping the proliferative reaction by promoting differentiation (Ramachandran et al., 2012). In both reactive and homeostatic Müller glia, Notch itself also limits proliferation through the inhibition of Ascl1 and Stat3 expression (Conner et al., 2014). Whether and how this reaction is conserved across species remains unclear, as Notch has been shown to activate Müller glia in rodents and chicken (Hayes et al., 2007; Del Debbio et al., 2010).

Reactivation of latent progenitors for neuronal repair in the brain

The discovery that latent but reactivatable progenitor cells also exist within virtually all brain subdivisions in rodents further generated great hopes to mobilize these endogenous cellular sources for repair (reviewed by Robel et al., 2011), although it must be noted that regenerative neurogenesis in rodents remains far less efficient than in non-mammalian species. Among the major candidate cell types for repair are ependymal cells and astrocytes, both of which react to injury by re-expressing progenitor markers and by proliferating in vivo, and which exhibit multipotency at least in vitro (Buffo et al., 2008; Carlen et al., 2009; Barnabe-Heider et al., 2010; Sirko et al., 2013; Magnusson et al., 2014). Upon spinal incision in the adult mouse, ependymal cells located in the vicinity of the lesion site activate and re-orient their fate towards generating new ependymal cells and astrocytes, as well as a low number of oligodendrocytes (Barnabe-Heider et al., 2010). In the forebrain, ependymal cells lining the lateral ventricle also reactivate after stroke and re-orient their fate to generate astrocytes and neuroblasts in the lateral wall (Carlen et al., 2009). Both examples demonstrate the remarkable activation potential and fate plasticity of ependymal cells in vivo and identify them as interesting targets for manipulation of cell fate towards making neurons. In a similar way, astrocytes located in the cortical and striatal regions of the brain as well as in the spinal cord parenchyma also exhibit an impressive plasticity, as the fate of these cells can be re-oriented towards neurogenesis upon forced expression of some transcription factors, such as Sox2 (Niu et al., 2013), or a combination of Ascl1, Brn2 (Pou3f2 – Mouse Genome Informatics) and Myt1l (Torper et al., 2013). This plasticity can also be seen in some lesional contexts, in particular middle cerebral artery occlusion-induced stroke or mechanical lesions (Buffo et al., 2008; Sirko et al., 2013; Magnusson et al., 2014), although the local origin of these neurogenesis-competent astrocytes has recently been questioned (Faiz et al., 2015). Finally, the neurogenic potential of other macroglial cell types, such as a subset of parenchymal glial cells expressing the NG2 (chondroitin sulfate proteoglycan 4) antigen, known as NG2 glia, has recently been identified. NG2 glia are the main proliferating macroglial cell type in the intact adult brain, and are normally largely fated to the generation of NG2 glia and oligodendrocytes, as well as some astrocytes in restricted brain areas (Guo et al., 2014; reviewed by Dimou and Gotz, 2014). Importantly, the translational relevance of these mouse studies is reinforced by the fact that postmortem material from patients who suffered stroke, hemorrhage or some neurodegenerative diseases can display signs of reactive proliferation, re-expression of neural progenitor markers, and neurogenesis (reviewed by Robel et al., 2011), although the process is probably abortive (Huttner et al., 2014).

These studies demonstrate some heterogeneity within macroglial cells in response to a reactive context, and point to the different propensity for reactivation depending on the lesional context (Sirko et al., 2013; Guo et al., 2014). In spite of this, and as seen in the retina, a common denominator of successful neuroblast formation is Ascl1 (Magnusson et al., 2014). Furthermore, very recent data demonstrated that Ascl1 induces Insm1 expression during the forced conversion of adult mouse cortical astrocytes, human astrocytes or mouse fibroblasts into GABAergic neurons, and that Insm1 is a necessary component of this reprogramming process (Masserdotti et al., 2015). In fact, in combination with NeuroD4, Insm1 is also sufficient for reprogramming. These results are in line with the early role of Insm1 in reactive Müller glia in zebrafish, and further support the notion that activation of the nodal pair Ascl1/Insm1 is a fundamental reprogramming event upon lesion, conserved across species. Additional data from these studies highlight the importance of downregulating Notch1 signaling to permit reactivation and/or proliferation and/or acquisition of a neuroblast fate in both ependymal cells of the lateral wall and striatal astrocytes (Magnusson et al., 2014) (Fig. 3). Understanding the different possible functions of Notch in these contexts will be an important task for future studies.

Regeneration from latent progenitors in the adult brain in non-mammals has not been extensively studied, except for the loss and subsequent regeneration of dopaminergic and cholinergic neurons in the newt midbrain. Stereotaxic injection of the neurotoxin 6-hydroxydopamine (6-OHDA) into the brain ventricle, which eliminates midbrain dopaminergic neurons, was followed by complete regeneration in adult newts (Fig. 3) (Berg et al., 2010). The source of the new neurons was shown to be ependymoglial cells, which are present along the midbrain ventricle, and have radial glia morphology and express GFAP. Constitutive neurogenic niches in the adult newt are restricted to the forebrain, and thus ependymoglial cells are normally quiescent. However, both sham and 6-OHDA injections triggered the re-entry of midbrain ependymoglial into the cell cycle, and in the latter case, proliferation was sustained and led to neurogenesis and functional repair (Parish et al., 2007). In a separate study, ablation of cholinergic neurons upon injection of ethylcholine aziridinium resulted in regeneration of the neurons after 7 weeks, again following a phase of induced ependymoglial proliferation in the midbrain ventricles (Berg et al., 2011). In this latter study, the authors demonstrated the importance of dopamine in the regenerative response. Following the destruction of dopaminergic neurons in the newt midbrain, injection of the dopamine mimic L-dopa blocked the proliferative response of ependymoglial cells (Berg et al., 2011). Interestingly, this did not affect the ependymoglial cell response to the destruction of cholinergic neurons, and appears to have been exerted directly on midbrain ependymoglial cells, probably via their expression of the dopamine receptor D2. The authors proposed that the role of dopamine was to act within a negative-feedback mechanism to match the production of dopaminergic neurons with the size of the existing midbrain dopaminergic pool (Berg et al., 2011, 2013). Importantly, the effect of dopamine is not limited to the regenerative context, as under physiological conditions dopamine acts to promote ependymal cell quiescence (Berg et al., 2011), and in mammals, dopamine promotes neurogenesis in active neurogenic zones (Kippin et al., 2005; O'Keeffe et al., 2009). As in the newt, the destruction of midbrain dopaminergic neurons by 6-OHDA in rats also leads to the activation of dormant midbrain progenitors, but these cells fail to regenerate dopaminergic neurons in situ (Lie et al., 2002).

Another important player in this context is Sonic Hedgehog (Shh). Shh expression was also induced in ependymoglial cells at the midbrain ventricle upon ablation of dopaminergic neurons by 6-OHDA in the newt. Furthermore, cyclopamine inhibition of Shh signaling reduced dopaminergic neuron regeneration, possibly through targeting progenitor specification or differentiation (Berg et al., 2010). Shh signaling plays multiple context-dependent roles during neurogenesis in vertebrate embryos and adults, impacting patterning, proliferation, cell division mode, cell cycle exit and progenitor migration (Ferent and Traiffort, 2015). In the context of rodent brain regeneration, it has so far only been implicated as a proliferation-inducing factor in mouse cortical astrocytes (Sirko et al., 2013).

Recruiting latent progenitors for repair in the spinal cord

Under normal conditions, the adult zebrafish spinal cord shows very little, if any, cell proliferation and neurogenesis, which is consistent with what is observed in the mammalian spinal cord (Yamamoto et al., 2001b). In contrast to mammals, however, neurons are generated in high numbers in the zebrafish spinal cord after a lesion (Fig. 3). A possible source of these cells is the ependymoglial cell population. These cells line the central canal in the adult zebrafish spinal cord and have a radial glial morphology, expressing Blbp (Fabp7a – Zebrafish Information Network) (dorsally) and/or the transcription factor Olig2 (in a more ventral domain) under normal physiological conditions. These ependymoglial cells normally divide slowly and asymmetrically to self-renew and produce oligodendrocytes (Park et al., 2007), but following a lesion, Olig2-positive ependymoglial cells re-enter the cell cycle, migrate, and differentiate into mature motor neurons (Reimer et al., 2008). Lineage-tracing studies performed during spinal cord regeneration in axolotl led to similar conclusions (McHedlishvili et al., 2007).

Recent studies have uncovered some of the key molecular players involved in activation of ependymoglial cells and subsequent spinal cord regeneration. One of the earliest factors expressed in activated zebrafish ependymoglial cells following injury is Sox2, which is required for re-initiation of proliferation (Ogai et al., 2014). A similar function for Sox2 was also demonstrated in the axolotl (Fei et al., 2014). Sox11b is also upregulated in zebrafish ependymoglial cells lining the central canal, and is necessary for their reactive proliferation and for locomotor recovery after spinal cord injury (Guo et al., 2011). Notch pathway genes are also upregulated after zebrafish spinal lesion in ependymoglial cells, but in this case they function to reduce progenitor cell proliferation and motor neuron generation. This is reminiscent of its reaction-antagonizing function following injury in the pallium and retina (Conner et al., 2014; Rodriguez Viales et al., 2015). Importantly, however, in the spinal cord, Notch manipulations have no effect on ependymoglial cells under normal physiological conditions (Dias et al., 2012), suggesting that quiescence is maintained in these cells by a mechanism that differs from other central nervous system areas and/or cell types. As previously discussed, spinal cord lesions in the adult rodent are followed by the reactivation of ependymoglial cells lining the central canal, which go on to generate astrocytes and oligodendrocytes but no neurons (Barnabe-Heider et al., 2010). Previous studies demonstrated an upregulation of Notch signaling in these ependymoglial cells post-lesion, and a role for Notch in inhibiting the acquisition of a neuronal fate (Yamamoto et al., 2001a). Whether Notch also plays a role in limiting ependymoglial cell activation in this system remains to be tested.

As previously demonstrated in the brain, another important player of spinal cord regeneration is dopamine. During regeneration in zebrafish, dopaminergic axons that descend from the brain were shown to undergo sprouting rostral to the lesion site (Reimer et al., 2013). In line with this, in situ hybridization indicated lesion-induced upregulation of the dopamine D4a receptor gene in the ependymal progenitor zone of the spinal cord segment located rostral to the lesion. Injection of the dopamine agonist NPA to mimic dopaminergic innervation posterior to the lesion increased motor neuron regeneration. The results from this study highlight a proliferation-promoting and/or neurogenic role for dopamine on quiescent ependymoglial cells during the spinal regeneration process. This action is clearly distinct from the effect of dopamine in the regenerating newt midbrain, and it will be very important to understand how dopamine can exert such disparate effects on similar cell types in the spinal cord and midbrain, and how specificity is achieved to control neuronal subtype generation. In the spinal cord, dopamine acts at least in part through activating the Hedgehog pathway (Reimer et al., 2013). The endogenous, albeit weak, expression of Shh and dorsoventral patterning markers in ependymoglial cells of the zebrafish adult spinal cord suggests that a latent embryonic positional information program persists in these cells. A similar conclusion was obtained in the axolotl and the newt (Schnapp et al., 2005). In addition, Shh expression is strongly increased in the ventral ependymoglial cells from which motor neurons regenerate, located in the vicinity of a spinal cord lesion in the adult zebrafish and adjacent to a more dorsally induced domain of strong Pax6 expression. Unlike in the midbrain, however, where blocking Shh signaling in vivo did not affect the lesion-induced proliferation of ependymoglial progenitors, it does so in the spinal cord, and further impairs motor neuron regeneration. As such, Shh is clearly an important factor for promoting the activity of ependymoglial cells as motor neuron progenitors (Reimer et al., 2009). Whether the primary function of Shh in this context is repatterning, induction of proliferation and/or reprogramming, remains to be assessed.

Many other cellular events contribute to regeneration after spinal cord injury including inflammation, cell death, proliferative response, neurogenesis and axonal regrowth (Hui et al., 2014). Given the capacity for rodent spinal ependymoglial cells to generate neurons in vitro but not in vivo (Barnabe-Heider et al., 2010), it is interesting to consider not only cell-intrinsic factors but also environmental cues that may bias progenitor fate in zebrafish. Several studies showed that during the spinal regeneration process, inhibitory components, such as chondroitin sulfate proteoglycans (Fawcett et al., 2012) and myelin-associated inhibitory molecules normally present in mammals, are absent in zebrafish (Becker and Becker, 2014). This may contribute to generating an environment that is permissive for neurogenesis and/or neuronal survival in zebrafish. In addition, pericytes are prominently involved in scar formation in the injured rodent spinal cord, a process that impairs regeneration (reviewed by Sabelstrom et al., 2013). The nature and fate of pericytes in the zebrafish spinal cord upon lesion remains to be studied.