Main

Sight is one of our most precious senses, and loss of sight exacts a large economic toll on both individuals and societies. Lost sight can result from traumatic injuries and diseases, such as glaucoma, diabetic retinopathy and macular degeneration. Several approaches for restoring sight to the blind are being pursued, including prosthetic devices, cell transplants and gene therapy1,2,3. Although each of these strategies has exhibited different degrees of success, they all rely on invasive surgeries and the introduction of foreign material into the eye. Ideally, one would like to develop a reparative strategy by which the retina could heal itself.

Although the idea of a self-healing retina may seem far-fetched, it is not unprecedented; teleost fish, such as zebrafish, have a remarkable capacity to regenerate their retina after damage and restore lost sight4,5,6. This regeneration relies on a single retinal cell type — the Müller glia — that is common to all vertebrate retinas. Müller glia are the major glial cell type in the retina and normally contribute to retinal structure and homeostasis7,8. However, after an injury to the retina, zebrafish Müller glia undergo a reprogramming event and acquire stem cell characteristics that enable them to generate progenitors for retinal repair9,10,11,12,13,14. Why zebrafish use this cell to regenerate a damaged retina and mammals do not remains unknown. It is possible that gaining a better understanding of retinal regeneration in teleost fish may hold the key for unlocking the regenerative potential of mammalian Müller glia.

In this Review, I summarize the responses of Müller glia to retinal injury in mammals, birds and fish. Because of the recent advances made in understanding how zebrafish Müller glia become reprogrammed for retinal repair, I have focused this Review on the signalling mechanisms underlying Müller glial cell reprogramming and the generation of Müller glial cell-derived progenitors in zebrafish. Finally, I describe future prospects for retina regeneration research in fish and mammals. The advances made in studying retina regeneration in fish are remarkable, and I suspect that these advances will inspire new strategies for stimulating retina regeneration in mammals. I hope that this Review helps to spur progress towards this goal.

Anatomy and function of Müller glia

The vertebrate retina is divided into three cellular layers: the outer nuclear layer (ONL), the inner nuclear layer (INL) and the ganglion cell layer (GCL) (Fig. 1). The ONL houses photoreceptors, which sense light and transduce this information to ganglion cells in the GCL through three types of interneurons (bipolar cells, amacrine cells and horizontal cells) that reside in the INL. Ganglion cells send their axons to the brain through the optic nerve and function to transfer visual information gathered in the eye to the brain.

Figure 1: Retinal anatomy.
figure 1

An illustration of the major retinal cell types and their organization in the retina. The retina is divided into three laminar layers: the outer nuclear layer (ONL), the inner nuclear layer (INL) and the ganglion cell layer (GCL). Six different neuronal cell types and one glial cell type are distributed among these layers: rod and cone photoreceptor cell bodies are located in the ONL; the cell bodies of bipolar, horizontal and amacrine interneurons, along with the cell bodies of Müller glia, are located in the INL; and the cell bodies of ganglion cells are located in the GCL. Ganglion cell axons run just beneath the GCL and comprise a nerve fibre layer (NFL). Synapses between photoreceptors and interneurons form in the outer plexiform layer (OPL), and synapses between interneurons and ganglion cells form in the inner plexiform layer (IPL). Müller glial processes span all retinal layers and contribute to the formation of the inner limiting membrane (ILM) and outer limiting membrane (OLM). The retinal pigment epithelium (RPE) consists of pigmented cells that absorb light and make contact with photoreceptors.

PowerPoint slide

Full size image

Müller glia arise from multipotent progenitors15 through a poorly understood process that includes Notch, retinal homeobox protein Rx (RAX) and Janus kinase (JAK) signalling pathways16,17,18,19,20. Although their cell bodies reside in the INL, Müller glia are the only cell type to span all retinal layers and have processes that contact neighbouring neurons and form part of the outer and inner limiting membranes21,22. Because of this, Müller glia are well positioned to monitor retinal homeostasis and contribute to retinal structure and function7,23. In doing so, they function as barriers and conduits for the transfer of a wide range of molecules between different retinal cells and compartments24,25,26. They also support neurons by releasing trophic factors, recycling neurotransmitters and controlling ionic balance in the extracellular space8,27,28,29. In addition, Müller glia phagocytose cone outer segments, contribute to outer segment assembly and participate in a cone-specific visual cycle that helps to recycle the retinal chromophore for photodetection30,31,32. Quite remarkably, it was recently found that, independent of their homeostatic function, Müller glia directly contribute to vision by acting as optical fibres to guide light to photoreceptors33. Although radial in structure, Müller glia differ from radial glia in the cortex in that they do not function as neural progenitors or serve as scaffolds for cell migration during retina development34. Nonetheless, progenitor characteristics have been noted in Müller glia, including the expression of progenitor-like genes, proliferative responses and the ability to generate neurons under special conditions35,36,37,38,39,40,41,42,43,44.

Müller glial cell response to injury

Müller glia are remarkably resilient to damage, a property that might be attributed to their unique physiology8,22,23,45,46. Müller glia respond to retinal injury and disease by changing their morphology, biochemistry and physiology23. This injury response is often referred to as reactive gliosis. Depending on the severity of damage, this response may include proliferation of Müller glia. However, the triggers for proliferative gliosis are not well understood. Both proliferative and non-proliferative responses to injury are accompanied by changes in gene and protein expression and are often associated with Müller glial cell hypertrophy. This reactive gliosis can be beneficial to neurons, as it prevents glutamate neurotoxicity and triggers the release of a range of factors that protect neurons from cell death23. However, prolonged gliosis is detrimental because it interferes with retinal homeostasis and the ability of Müller glia to support retinal neurons and therefore often leads to neurodegeneration. Furthermore, the deposition of cell masses as a consequence of proliferative gliosis impedes normal retina function.

As noted above, Müller glia share some characteristics with retinal stem cells, and in some species Müller glia can regenerate neurons. Thus, if one could tip the balance from a gliotic response to one that is reparative, it might be possible to use Müller glia for endogenous repair. In order for Müller glia to participate in retinal repair, one can envisage three important steps that must occur (Fig. 2): reprogramming of Müller glia to adopt stem cell characteristics, generation of a proliferating population of multipotent Müller glial cell-derived progenitors and progenitor cell cycle exit and neuronal differentiation.

Figure 2: Generation of multipotent Müller glial cell-derived progenitors for retinal repair.
figure 2

Adult Müller glia in zebrafish respond to retinal injury by reprogramming their genome (illustrated by a change in the colour of the cell) so that they can acquire stem cell properties9,10. This reprogramming results in interkinetic nuclear migration to the outer nuclear layer and asymmetrical cell division near the outer limiting membrane13. This asymmetrical cell division generates multipotent progenitors that transiently proliferate and restore the original Müller glia. Multipotent progenitors migrate to all cell layers, exit the cell cycle and regenerate all major retinal cell types11,80.

PowerPoint slide

Full size image

Models for studying retina regeneration

Several model systems have provided important insights into the process of retina regeneration and the role of Müller glia in this process (Table 1). Three animals have dominated the field: teleost fish, which naturally regenerate a damaged retina; postnatal chicks, which exhibit a limited regenerative capacity; and mice, which normally do not regenerate but are an important model for devising and testing strategies for mammalian retinal repair.

Table 1 Factors affecting Müller glial cell reprogramming and proliferation
Full size table

Birds. Adult birds do not regenerate a damaged retina. However, postnatal chicks respond to retinal injury with proliferation of Müller glia and a small amount of neural regeneration39. This proliferation is stimulated by Notch signalling47,48, and proliferating cells express progenitor markers, such as paired box 6 (PAX6), achaete-scute homologue 1 (ASCL1) and Ceh-10 homeodomain-containing homologue (CHX10; also known as VSX2)39. Although the endogenous factors that mediate the proliferation of Müller glia remain unknown, candidates include growth factors, such as fibroblast growth factor 2 (FGF2), insulin and insulin-like growth factor 1 (IGF1)49,50,51. Remarkably, these factors can also stimulate Müller glial cell proliferation in the uninjured chick retina49,50 and seem to act through Notch as well as FGF receptor, mitogen-activated protein kinase (MAPK) and extracellular signal-regulated kinase (ERK) signalling pathways47,52.

Mammals. Although mammalian Müller glia can respond to injury, proliferate and express genes that are associated with retinal stem cells16,37, they do not function as retinal progenitors in vivo. Nonetheless, these characteristics suggest that, under the right circumstances, Müller glia might be persuaded to adopt characteristics of a retinal progenitor that can be used for repair. Indeed, in rodent and human cell cultures, Müller glia have been observed to generate both neurons and glia53,54. Importantly, primary human Müller glial cell cultures can generate photoreceptors and retinal ganglion cells that have some reparative potential when transplanted into a damaged rodent retina55,56,57. These studies suggest that human Müller glia are capable of generating neurons under appropriate conditions and that they may be able to participate in repair.

Several studies have attempted to enable mammalian Müller glia to mount a regenerative response in vivo with limited success. Pharmacological damage to ganglion and bipolar cells or photoreceptors can induce a small amount of Müller glial cell proliferation and neuronal regeneration in rodents, but these regenerative events are very rare41,44. Other pharmacological and genetic engineering approaches indicate that WNT–β-catenin-, sonic hedgehog-, epidermal growth factor (EGF)–EGF receptor (EGFR)-, glutamate- and ASCL1-dependent signalling events can stimulate some proliferation of Müller glia and neural regeneration in the injured mammalian retina40,41,42,43,58,59.

One of the most potent methods for stimulating the proliferation of Müller glia in the mouse retina is a combination of NMDA-induced retinal damage and EGF treatment40. EGFR expression in Müller glia is suppressed during postnatal development and this, along with increased transforming growth factor-β (TGFβ) signalling, correlates with the reduced proliferative capacity of Müller glia60,61. Although it is not clear whether TGFβ signalling is suppressed in the NMDA-damaged mouse retina, EGFR expression is observed after damage has occurred60. EGF and NMDA treatment seems to stimulate the proliferation of Müller glia by activating MAPK, phosphoinositide 3-kinase (PI3K) and bone morphogenetic protein (BMP) signalling pathways62. Interestingly, subretinal delivery of low non-toxic doses of glutamate also stimulates Müller glial cell proliferation and a small amount of neural regeneration43. This raises the intriguing possibility that glutamate itself may be a secreted factor that stimulates the proliferation of Müller glia in the injured retina.

Although Müller glia proliferated and activated the expression of progenitor genes in EGF- or NMDA-treated mouse retinas40, there was a notable lack of expression of Ascl1, a gene previously shown to be induced in proliferating Müller glia in the injured chick retina39 and which is essential for the reprogramming and proliferation of Müller glia in zebrafish10,63,64. Remarkably, forced overexpression of Ascl1 in combination with EGF treatment stimulated the reprogramming and proliferation of Müller glia and the generation of bipolar neurons in postnatal mouse retinal explants59.

Fish. Unlike birds and mammals, teleost fish such as zebrafish can regenerate a damaged retina that restores visually mediated behaviours4,5,6. This regenerative response, along with its amenability to genetic manipulation, has made zebrafish a favoured model for studying retina regeneration. Although the teleost retina shares structure and function with the mammalian retina, distinguishing features include: a ciliary marginal zone in which retinal progenitors reside and add new neurons and glia as the retina expands throughout the animal's life65; rod precursors in the ONL that selectively generate rods as the retina grows66; and Müller glia that generate rod progenitors and can be stimulated to generate multipotent progenitors for retinal repair11,67,68,69. Together, these features suggest that teleost fish possess a unique retinal environment that supports progenitor cell formation and maintenance. Next, I describe the mechanistic underpinnings of retina regeneration in fish.

Mechanisms of regeneration in fish

Early studies, which predominantly used goldfish as a model system, firmly established that retina regeneration stemmed from the actions of injury-responsive cells that are intrinsic to the central retina70,71,72,73,74,75. Although the nature and origin of these progenitors were unknown, rod precursors, Müller glia and neuroepithelial cells were considered as candidates for this unknown cell type72,73,76,77,78. Müller glia were finally identified as the source of these progenitors using transgenic zebrafish in which Müller glia or Müller glial cell-derived progenitors were specifically labelled with green fluorescent protein11,14,67,69,79. By taking advantage of bromodeoxyuridine (BrdU) lineage tracing strategies and Cre–loxP technology that enabled researchers to generate transgenic fish in which Müller glial cell-derived progenitors were permanently labelled, it was shown that these progenitors produced all major retinal cell types and remained stably integrated into the retinal architecture11,80. When thinking about how Müller glia respond to retinal injury, we need to consider not only how injury signals are sensed and transmitted to the genome to reprogramme Müller glia but also how Müller glial cell-derived progenitors exit the cell cycle and differentiate.

How do Müller glia sense injury? Fish Müller glia undergo reprogramming that enables regeneration in response to various retinal injuries, including those caused by intense light67,81, chemicals69, mechanical damage11,79 and cell type-specific expression of toxic genes82 (Fig. 3). It is likely that these different onslaughts converge on similar signalling pathways to stimulate reprogramming of Müller glia. Although photoreceptor damage was previously thought to be necessary to stimulate a regenerative response72, more recent studies suggest that this is not the case39,69.

Figure 3: Injury paradigms and the communication of injury to Müller glia.
figure 3

Various injury paradigms have been used to induce retinal damage and stimulate regeneration in zebrafish. These include: prolonged exposure to intense bright light and short exposure to ultraviolet (UV) light; intravitreal injection of toxins (such as ouabain and NMDA); expression of a toxic gene (such as bacterial nitroreductase, which, in combination with a pro-drug, generates a cytotoxic product); and mechanical injury (such as that resulting from a needle poke)11,12,69,81,82. Light-based damage paradigms generally destroy a population of photoreceptors, whereas toxins can cause widespread damage. Cytotoxic gene products can be directed to specific retinal cell types by using appropriate promoters to drive their expression. Mechanical injury generally destroys all retinal cell types in a circumscribed region of the retina. The figure illustrates the ways in which injured cells might communicate with Müller glia to stimulate their reprogramming. These include secretion of signalling molecules (arrows) from damaged cells, Müller glia or infiltrating microglia; altered contact between damaged cells and Müller glia; and phagocytosis of injured cells by Müller glia. Recent studies have suggested that growth factors, such as heparin-binding epidermal growth factor (EGF)-like growth factor a (Hbegfa), and cytokines, such as tumour necrosis factor-α (Tnfα), are necessary for Müller glial cell reprogramming and progenitor formation in the injured retina83,84. These factors are produced in Müller glia at the injury site and therefore may act in an autocrine and paracrine manner. Tnfα and ADP are also released from injured retinal neurons84,87.

PowerPoint slide

Full size image

Secreted factors, such as heparin-binding EGF-like growth factor (Hbegf), tumour necrosis factor-α (Tnfα), Wnts, ADP and ciliary neurotrophic factor (CNTF) have been reported to promote injury-induced Müller glial cell reprogramming and progenitor formation in fish63,83,84,85,86,87. However, a gene that encodes CNTF has not been identified in zebrafish, perhaps suggesting that other interleukin-6 family cytokines contribute to retina regeneration in this species. Remarkably, some of these secreted factors have been found to stimulate reprogramming and proliferation of Müller glia in the uninjured retina83,85. In the injured retina, most of the genes encoding these secreted factors are induced in injury-responsive Müller glia and their products may contribute to Müller glial cell reprogramming and retina regeneration in an autocrine and paracrine manner (Fig. 3). Two factors that regulate Müller glial cell proliferation, ADP and Tnfα, are not only produced in Müller glia but also seem to be released by dying retinal neurons84,87; this suggests that they may have a role as injury signals that initiate a regenerative response by Müller glia. Tnfα contributes to injury-dependent induction of the expression of Ascl1a and signal transducer and activator of transcription 3 (Stat3)84; the expression of these two transcription factors is necessary for the generation of Müller glial cell-derived progenitors10,63,64,88. However, it is not yet clear whether the Tnfα that promotes this expression is released from dying cells and/or Müller glia. Finally, microglia and other immune-related cells respond to retinal injury by migrating to the injury site, where they may release factors that influence the reprogramming and proliferation of Müller glia89,90 (Fig. 3).

When considering additional mechanisms that may contribute to the transmission of injury signals to Müller glia, the fact that Müller glia make contact with neighbouring cells and can participate in phagocytosis (Fig. 3) should not be ignored23. Indeed, Müller glia phagocytose injured photoreceptors, and inhibitors of phagocytosis suppress progenitor formation in the injured zebrafish retina91. It can also be hypothesized that altered contact between Müller glia and their injured neighbours may be sensed by integrins, cadherins, Notch and other signalling components, which may contribute to initiating an injury response, although this has yet to be tested (Fig. 4).

Figure 4: Signalling cascades that contribute to Müller glial cell reprogramming and progenitor proliferation in zebrafish.
figure 4

Retina regeneration requires the activation of various signalling cascades. This diversity of signalling may reflect the variety of injuries and signalling molecules that stimulate retina regeneration. Signalling pathways that have been shown to regulate retina regeneration are indicated by solid lines, whereas those indirectly implicated or hypothesized to be involved are indicated by dashed lines. Secreted factors that regulate the proliferation of Müller glia are indicated outside the cell (those that affect the proliferation of Müller glia in birds and mammals49,50,51 but which have not yet been tested in zebrafish are annotated with a question mark). Wnts are secreted glycoproteins that bind to Frizzled family receptors to regulate β-catenin stabilization. Dickkopf (Dkk) is a secreted Wnt signalling antagonist. Dishevelled homologue 3 (Dvl3) is a cytoplasmic phosphoprotein that acts downstream of Wnt receptors. Glycogen synthase kinase 3β (Gsk3β) regulates β-catenin stabilization by phosphorylation. β-catenin regulates cell adhesion and gene expression. Insulin and insulin-like growth factor 1 (Igf1) are secreted proteins that bind to tyrosine kinase receptors that signal through insulin receptor substrate (Irs), an adaptor protein that couples insulin and Igf1 receptor to phosphoinositide 3-kinase (Pik3) and Akt (also known as protein kinase B) activation. Heparin-binding epidermal growth factor (EGF)-like growth factor (Hbegf) is a transmembrane protein that undergoes ectodomain shedding and acts through Egf receptors. Fibroblast growth factors (Fgfs) are secreted growth factors that bind to Fgf receptors. Egf and Fgf receptors are tyrosine kinase receptors that signal through mitogen-activated protein kinase (Mapk) and extracellular signal-regulated kinase (Erk). Cytokines signal through cytokine receptors to stimulate Janus kinase (Jak) activation. Jak proteins phosphorylate signal transducer and activator of transcription (Stat) proteins. Transforming growth factor-β (Tgfβ) is a secreted protein that signals through the Smad pathway to alter gene expression. The microRNA let-7 is a post-transcriptional regulator of RNA expression. Lin28 is an RNA-binding protein that regulates let-7 expression. Tumour necrosis factor-α (Tnfα) is a secreted cytokine that acts through Tnf receptors to regulate cell signalling and gene expression. Delta–Notch signalling is mediated by single-pass transmembrane proteins expressed on adjacent cells. The extracellular matrix (ECM) can signal through transmembrane integrin receptors to regulate cell function. GF, growth factor.

PowerPoint slide

Full size image

Signal transduction in injury-responsive Müller glia. The diversity of the signalling molecules that communicate with Müller glia in the injured retina (Fig. 3; Table 1) suggests that injuries activate multiple signalling cascades (Fig. 4). A major challenge in evaluating the roles of these signalling pathways is determining when and where they are activated after retinal injury. Pharmacological inhibition, knockdown strategies and genetic manipulations of gene expression often lack cellular resolution. In addition, even when signal transduction pathways can be identified in specific cells, one is limited by the sensitivity of the detection method. With these caveats in mind, pharmacological inhibition and genetic manipulations suggest that glycogen synthase kinase 3β (Gsk3β)–β-catenin, Notch, Mapk–Erk and Jak–Stat signalling pathways regulate zebrafish retina regeneration63,83,88,92 (Fig. 4). All of these pathways can couple extracellular events with the gene expression changes that drive Müller glial cell reprogramming. However, of these pathways, only β-catenin activation has been confirmed to occur in Müller glial cell-derived progenitors63,92.

In zebrafish, retinal injury results in wnt expression and β-catenin stabilization in Müller glial cell-derived progenitors63,92. β-catenin is a multifunctional protein that, in collaboration with T cell factor (Tcf) and lymphoid enhancer-binding factor (Lef) family members, links changes in Wnt and cadherin signalling on the cell surface to gene expression93. Inhibition of Wnt signalling or conditional expression of dominant-negative Tcf suppresses progenitor formation in the injured retina63,92. Furthermore, pharmacological activation of the β-catenin signalling pathway in the uninjured zebrafish retina with lithium or a Gsk3β inhibitor stimulates Müller glial cell reprogramming and progenitor formation63,92. These treatments bypass an inhibitory retinal environment that results, in part, from pan-retinal expression of the Wnt antagonist dickkopf (Dkk)63.

Although growth factors, such as Hbegf, can stimulate progenitor formation in the uninjured zebrafish retina83, their mechanism of action is poorly understood. Injury-dependent induction of Hbegfa seems to be necessary for retina regeneration following a mechanical injury83, but it may not be necessary for regeneration following photoreceptor damage84. In Hbegf-treated uninjured retinas or mechanically injured retinas, the generation of Müller glial cell-derived progenitors requires Egfr and Mapk–Erk signalling83 (Fig. 4).

The finding that cytokines, such as CNTF, can stimulate Müller glia to generate progenitors in the uninjured fish retina suggests that Jak–Stat signalling may also be involved in Müller glial cell reprogramming and retina regeneration85. Indeed, retinal injury stimulates Stat3 expression in both quiescent Müller glia and Müller glial cell-derived progenitors, and stat3 knockdown inhibits progenitor formation14,88. Taken together, the above studies suggest that it is the combinatorial action of cytokines, Wnts and growth factors that stimulate Müller glial cell reprogramming and retina regeneration in the injured zebrafish retina.

A Müller glial cell acquires the properties of a retinal stem cell by reprogramming its genome to express genes that enable it to generate multipotent progenitors for retinal repair. The above discussion suggests that activation of Mapk–Erk, Gsk3β–β-catenin and Jak–Stat signalling cascades may be crucial for this genomic reprogramming. However, changes in the levels of activated and/or stabilized β-catenin lag behind the earliest changes in gene expression noted after retinal injury and correlate best with the production of progenitors from reprogrammed Müller glia63. This discrepancy between β-catenin activation and Müller glial cell reprogramming may simply reflect the limits of immunofluorescence detection or may indicate that other signalling pathways act earlier, perhaps to control both Müller glial cell reprogramming and β-catenin stabilization.

Retina regeneration is not only driven by activation of signalling pathways that stimulate Müller glial cell reprogramming and progenitor formation but also by suppression of pathways that drive Müller glial cell differentiation and quiescence (Fig. 4). MicroRNA (miRNA) let-7 signalling and Dkk signalling are two such inhibitory pathways that help to maintain zebrafish Müller glia in a quiescent state10,63 (Fig. 4). TGFβ signalling inhibitors, TGFβ-induced factor 1 (Tgif1) and sine oculis homeobox homologue 3b (Six3b), enhance progenitor proliferation in the injured zebrafish retina94. However, these inhibitors are transcriptional co-repressors that have multiple targets, and their effect on TGFβ signalling in the injured zebrafish retina remains untested. Notch signalling also seems to have an inhibitory role during zebrafish retina regeneration (Fig. 4). However, unlike most inhibitory pathways, which are suppressed following retinal injury, Notch signalling components, such as deltaA, deltaB, deltaC and notch1, and Notch target genes, such as her4, are induced by injury83. Furthermore, unlike Notch's pro-proliferative effects in the chick retina48, Notch signalling in the injured fish retina suppresses the number of Müller glia in which an injury response is triggered83. In this way, Notch signalling seems to help to match the number of injury-responsive Müller glia with the extent of retinal damage83. It is likely that additional inhibitory pathways help to maintain Müller glia in a quiescent state in the uninjured retina, and their identification remains an important area of study.

Early response genes associated with reprogramming. By identifying the earliest changes in gene expression during retina regeneration, one gains insight into how Müller glia reprogramme from a differentiated support cell into one that produces progenitors for retinal repair. Microarray-based analysis of the zebrafish regeneration-associated transcriptome has identified over 1,500 genes that exhibit differential expression between Müller glia and Müller glial cell-derived progenitors12,14,89,95,96. This is probably an underestimate as these studies only assessed ∼60% or less of the zebrafish genome. Nonetheless, this analysis has already facilitated the identification and characterization of several regeneration-associated genes that contribute to a Müller glial cell's transition from a fully differentiated support cell to one with stem cell characteristics10,12,14,63,64,83,84,88,94,96,97,98,99,100,101.

One set of genes that are rapidly induced in zebrafish Müller glia following retinal injury are those encoding secreted growth factors and cytokines, such as hbegfa and tnfa83,84. Their expression by Müller glia suggests they may act in an autocrine and paracrine manner to stimulate progenitor formation and proliferation. The mechanism underlying injury-dependent induction of these genes has not been studied in detail and remains an important area of investigation.

Injury-dependent induction of ascl1a expression results in the suppression of genetic programmes that promote cellular differentiation and the activation of programmes that promote proliferation. Induction of ascl1a expression is under the control of Hbegfa in the mechanically injured zebrafish retina83. Ascl1a stimulates lin28 expression, which has been shown to contribute to both let-7 suppression and further Ascl1a induction10,88 (Figs 4,5). Lin28 is an RNA-binding protein that is highly expressed in embryonic stem cells and is associated with stem cell self-renewal102,103. It has been used to reprogramme somatic cells into induced pluripotent stem cells (iPSCs)104 and is an important regulator of tissue regeneration in mammals105. The let-7 miRNAs are small regulatory RNAs associated with cellular differentiation103. Lin28 and let-7 regulate each other's expression; as Lin28 levels rise, let-7 levels fall, and vice versa102,103,106. Because Lin28 and let-7 can regulate a large proportion of the cellular transcriptome103,107, they are important players in the reprogramming, proliferation and differentiation of Müller glia10. In addition to activating lin28 expression, Ascl1a also affects Müller glial cell reprogramming and proliferation by regulating the Wnt signalling pathway, in which it inhibits dkk expression and activates expression of Wnt genes63,96 (Fig. 5). Finally, Ascl1a regulates the expression of insulinoma-associated 1a (insm1a), a transcriptional repressor that affects both Müller glial cell reprogramming and progenitor cell cycle exit96 (Fig. 5).

Figure 5: Regeneration-associated transcriptional cascades in zebrafish.
figure 5

Growth factors, cytokines and Wnts seem to impinge on mitogen-activated protein kinase (Mapk)–extracellular signal-regulated kinase (Erk), glycogen synthase kinase 3β (Gsk3β)–β-catenin and Janus kinase (Jak)–signal transducer and activator of transcription (Stat) signalling pathways to stimulate reprogramming of Müller glia in response to retinal injury63,83,88,92. These pathways participate in injury-dependent achaete-scute homologue 1a (ascl1a) expression63,83,88. Ascl1a is a basic helix-loop-helix transcription factor that is involved in almost all aspects of retina regeneration. It regulates genes that are responsible for generating Müller glial cell-derived progenitors, such as those encoding Wnts, growth factors, Lin28, Myc, Apobec2b (apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like 2b), Insm1a (insulinoma-associated 1a) and Stat3 (Refs 10,63,64,83,88,96,97). Ascl1a also controls (directly or indirectly) the expression of proteins and microRNAs that inhibit progenitor formation and proliferation, such as Notch, dickkopf (Dkk), Insm1a (this protein contributes to both progenitor formation and differentiation), p57kip2 and let-7 (Refs 10,63,83,96). Gsk3β–β-catenin signalling stimulates paired box 6 (pax6b) expression in an Ascl1a-independent manner63. The regenerative steps outlined on the left-hand side, along with the purple gradient, illustrate the gradual transition of Müller glia to progenitors and their differentiation during retina regeneration. Cdk, cyclin-dependent kinase.

PowerPoint slide

Full size image

In the light-damaged zebrafish retina, stat3 expression may precede that of ascl1a88. Stat3 links membrane events with changes in gene expression. Following retinal injury, Stat3 levels are increased in both quiescent and proliferating Müller glia, whereas injury-dependent Ascl1a expression is restricted to reprogrammed Müller glia and Müller glial cell-derived progenitors10,64,88. Importantly, the retinal cell types that express phosphorylated (activated) Stat3 (pStat3) remain uncharacterized and further research may reveal a progenitor-specific action of this signalling molecule. Surprisingly, stat3 knockdown only reduced progenitor formation by ∼40%, which suggests that this protein has a modulatory role88. However, it may be that residual pStat3 remaining after knockdown was sufficient to drive progenitor formation. Alternatively, these findings may indicate that multiple Stat proteins are activated by injury and that knockdown of any individual Stat gene is insufficient to robustly suppress progenitor formation. Finally, it has been suggested that there is a pStat3–Ascl1a signalling loop: not only does stat3 knockdown reduce Ascl1a expression but ascl1a knockdown also suppresses Stat3 expression84,88.

Pax6b is necessary for the proliferation of Müller glial cell-derived progenitors in zebrafish, and its expression is induced just before Müller glia begin dividing10,98. Interestingly, pax6b expression is not controlled by Ascl1a expression but instead seems to be under the control of Gsk3β–β-catenin signalling63 (Fig. 5).

Epigenetic changes during retina regeneration. Injury-dependent reprogramming of Müller glia shares some features with the process of somatic cell reprogramming and the generation of iPSCs. Indeed, many of the genes that are used to stimulate pluripotency in somatic cells are induced in Müller glia as they reprogramme to a stem cell10. During iPSC formation, pluripotency genes undergo a demethylation event that enables their chromatin to assume a more 'open', accessible state that is permissive for gene expression108,109. Interestingly, forced hypomethylation of DNA in zebrafish Müller glial cell-derived progenitors with 5-aza-2′-deoxycytidine (5-dAza) stimulated expression of the transgenic reporter gene 1016 tuba1a:gfp, the expression of which reflects injury-dependent Müller glial cell reprogramming9. Furthermore, this hypomethylation reduced progenitor amplification, migration and differentiation9. These data are consistent with the idea that DNA demethylation contributes to reprogramming of Müller glia, whereas DNA methylation may be necessary for the migration and differentiation of Müller glial cell-derived progenitors. Indeed, reduced representation bisulphite sequencing identified a small proportion of the zebrafish Müller glial cell genome that changed its methylation pattern during retina regeneration9. Demethylation predominated early after injury when Müller glia were being reprogrammed to adopt the properties of retinal stem cells, and methylation regained prominence later when progenitors were migrating and differentiating9. Furthermore, this analysis revealed a correlation between regeneration-associated gene expression levels and DNA demethylation9. Importantly, the promoters of several regeneration-associated genes, including ascl1a, lin28, hbegfa and insm1a, exhibited a low basal level of methylation in Müller glia that remained unchanged in progenitors9. Interestingly, these same genes exhibit a similar low basal level of methylation in Müller glia from mammals9, which perhaps contributes to the noted progenitor-like characteristics and plasticity of these cells16,37,40,41,42,43,44,53,54,58.

DNA methylation reflects the capacity for gene expression, whereas histone modifications can distinguish active from repressed genes110. A striking feature of the chromatin present in iPSCs is the presence of bivalent domains that harbour histones with both active and repressive modifications111. These bivalent domains may indicate a transcriptionally poised state that can rapidly change under different cellular demands and thus increase cellular plasticity. It would not be too surprising if chromatin from zebrafish Müller glia exhibit histone modifications that contribute to their plasticity. Although this has not yet been studied in zebrafish, it is noteworthy that forced overexpression of ASCL1 in mouse Müller glia affects histone modification, gene expression and progenitor formation59.

Cell cycle exit and differentiation. Reprogrammed Müller glia in zebrafish divide asymmetrically near the ONL to generate a population of transient amplifying progenitors that contribute to retinal repair13 (Fig. 2). Gsk3β inhibition prevents this asymmetrical division and encourages a symmetrical division, resulting in depletion of the differentiated Müller glial cell pool92. Mapk–Erk, Gsk3β–β-catenin and Jak–Stat signalling cascades contribute to the formation of the progenitor pool63,83,88,92 (Fig. 5). Pax6b controls the earliest division of the first Müller glial cell-derived progenitor98, and Pax6a98, heat shock 60 kDa protein 1 (Hspd1)12 and many of the gene products described above contribute to their expansion (Table 1). Progenitors are born apically near the ONL and migrate into the INL in an N-cadherin-dependent manner13 (Fig. 2). Furthermore, monopolar spindle 1 (Mps1; also known as Ttk) may increase the proliferation of photoreceptor progenitors in the ONL12.

Although we know very little about the mechanisms that drive progenitors out of the cell cycle, the transcriptional repressor Insm1a seems to play an important part96 (Fig. 5). Insm1a drives cell cycle exit by inhibiting expression of cell cycle-associated genes and increasing the expression of p57kip2 (also known as cdkn1c), a gene encoding a cyclin-dependent kinase inhibitor96 (Fig. 5). Insm1a appears to stimulate p57kip2 expression by inhibiting the expression of the p57kip2 repressor Bcl11 (Ref. 96) (Fig. 5). Remarkably, Ascl1a also contributes to this regulation by increasing insm1a promoter activity96 (Fig. 5).

In zebrafish, Müller glial cell-derived progenitors can regenerate all major retinal cell types. This multipotency distinguishes them from rod precursors in fish and Müller glial cell-derived progenitors in birds and mammals, which have a severely limited ability to regenerate multiple cell types39,40,77. This difference in multipotency may reflect intrinsic differences in gene expression and extrinsic differences in the progenitors' environment. Although Müller glial cell-derived progenitors in zebrafish can regenerate all types of damaged retinal neurons, it is not clear whether the identity of the dying cells can influence progenitor differentiation in order to generate replacements. Interestingly, when Müller glia in the uninjured retina are forced to reprogramme and generate progenitors, these progenitors have the capacity to make all retinal cell types63,83,96. Experimental paradigms that result in damage to particular cell types (such as photoreceptors or bipolar cells) have demonstrated that progenitors can replace the lost cell types12,14,67,82,101,112,113,114. However, because these studies only used antibodies or transgenic reporter lines that detect the damaged cell type, it was not possible to determine whether progenitors also made other cell types that were only transiently maintained.

The mechanisms that control the differentiation of Müller glial cell-derived progenitors in the adult fish retina are poorly understood. In regeneration models in which photoreceptors are selectively damaged, Mps1 seems to promote proliferation of photoreceptor progenitors and controls their differentiation into cones12 (Table 2). It is not known whether Mps1 also affects the differentiation of other cell types. Interestingly, in the photoreceptor damage model, Fgf signalling and the galectin Drgal1-L2 (also known as Lgals2a) are necessary for the regeneration of rod photoreceptors but not cone photoreceptors99,115 (Table 2). In a mechanical injury model in which all retinal cell types are damaged, Notch activity seems to affect the differentiation of all cell types: Notch inhibition increases Müller glial cell differentiation and suppresses neuronal differentiation, whereas Notch intracellular domain overexpression stimulates photoreceptor differentiation at the expense of Müller glial, bipolar and ganglion cell differentiation83 (Table 2). Finally, cell adhesion or progenitor migration also affects progenitor differentiation, as inhibition of N-cadherin expression reduced progenitor migration into the INL and suppressed the regeneration of inner retinal neurons13 (Table 2). Understanding the mechanisms underlying progenitor differentiation and choices of cell fate in the adult retina may suggest ways of enhancing and directing the differentiation of progenitors in the adult mammalian retina.

Table 2 Factors affecting progenitor cell cycle exit and differentiation in zebrafish
Full size table

Future prospects

Some of the most pressing questions that remain concerning retina regeneration centre on the differences noted between fish and mammalian Müller glia. Why do zebrafish Müller glia readily reprogramme in response to injury, whereas those in mammals do not? Even when Müller glia are manipulated to divide in mammals, why do they only rarely regenerate neurons? There are several possible answers to these questions that span intrinsic differences between fish and mammalian Müller glia and progenitors to the different environments (niches) that nurture them.

High-throughput sequencing enables us to discern differences between the transcriptomes and epigenomes of zebrafish and mammalian Müller glia and Müller glia progenitors. Differences in the transcriptomes of human Müller glia in the retina and those that generate progenitors in culture may help to define signals that stimulate their conversion to multipotency. Zebrafish provide a convenient system for testing the significance of specific genes and regulatory events in regeneration. Those found to be important for regeneration can then be tested in mammals to determine whether regeneration can be increased.

Although injured cells seem to provide the initial stimulus that initiates Müller glial cell reprogramming and retina regeneration, it is not clear whether other cell types may also participate. In particular, microglia and infiltrating immune cells may have a role. These cells can respond to injury by migration, phagocytosis and release of factors that may act on Müller glia to initiate or increase their reprogramming and/or affect the proliferation and differentiation of Müller glial cell-derived progenitors. Microglia play an important part in zebrafish brain regeneration, and macrophages affect limb regeneration in salamanders116,117. Following retinal injury, microglia become activated and migrate to the injury site89,90,118, suggesting that they have a role in retina regeneration. Further analysis of their contribution to retina regeneration in zebrafish is warranted, and if they do have an important role, comparing their injury response with that of mammals may lead to the discovery of strategies for improving mammalian retina regeneration.

The variety of injury paradigms and secreted factors that stimulate zebrafish retina regeneration and converge on Mapk–Erk, Gsk3β–β-catenin and Jak–Stat signalling is remarkable and suggests that activation of these pathways is crucial for retina regeneration. However, it is still unclear whether these pathways drive all aspects of Müller glial cell reprogramming and progenitor proliferation or are more restricted in their action. Further analysis of their role in Müller glial cell reprogramming in fish is crucial for understanding their significance in controlling retina regeneration. These pathways have not been well characterized in the injured mammalian retina and may be good targets for enhancing retinal repair.

Even if mammalian Müller glia could be enabled to reprogramme, their environment may be hostile to progenitor formation and differentiation. Interestingly, ephrins, BMPs and secreted frizzled-related protein 2 (SFRP2) negatively regulate stem cells and are expressed in the adult mammalian retina119,120,121. Whether these signals interact with others to inhibit Müller glial cell reprogramming and progenitor formation remains to be determined. Retinal injury stimulates a series of events that overcome an environment that maintains Müller glial cell quiescence in the uninjured fish retina. Thus, zebrafish provide an ideal system for identifying these quiescence-promoting factors and for developing strategies for their suppression. A combination of neutralizing quiescence-promoting factors and stimulating activators of regeneration may be the most successful strategy for restoring a regenerative response in the mammalian retina.

The goal of scientists studying retina regeneration is to apply regenerative strategies to eye diseases and injuries that cause blindness. The ability to use endogenous stem cells for repair has a lot of appeal and avoids many concerns associated with prosthetic devices and cell transplants. In particular, endogenous repair will not stimulate an immune response and does not require cell infiltration. Because of their robust regenerative response and ease of genetic manipulation, fish have led the way in identifying mechanisms underlying retina regeneration. Recent progress in characterizing these mechanisms will enable the development of strategies for stimulating retinal repair in mammals. However, the problem is complex and many questions remain. Fish will continue to be an important model for uncovering the mechanisms that control retina regeneration, and other species such as birds and mice will serve as important models for testing these mechanisms and revealing others that restore a regenerative response in animals blinded by injury or disease.